Evaluation of the degree of agreement in the diagnosis of Diabetic Retinopathy between Ophthalmologists and EyeArt®

Evaluation of the degree of agreement in the diagnosis of Diabetic Retinopathy between Ophthalmologists and EyeArt® | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Evaluation of the degree of agreement in the diagnosis of Diabetic Retinopathy between Ophthalmologists and EyeArt ® Isabel Inmaculada Guedes Guedes, Pedro Saavedra, Francisco Cabrera López, and 3 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7284873/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 9 You are reading this latest preprint version Abstract Objective or Purpose: To evaluate the diagnostic performance and agreement of the EyeArt ® Artificial Intelligence (AI) system for detecting Diabetic Retinopathy (DR), comparing its results with ophthalmologists' assessments in a regional screening program. Design: Cross-sectional observational study. Subjects, Participants, and/or Controls: A total of 499 diabetic patients aged 18 years or older were enrolled between June and September 2023 through the Retisalud DR screening program in the Canary Islands. No separate control group was included. Methods: All participants underwent non-mydriatic fundus photography using the TRC-NW400 camera. Retinal images were analyzed by the EyeArt ® AI system (version 2.1.0), and results were compared with assessments by ophthalmologists based on the International Clinical Diabetic Retinopathy (ICDR) scale. Agreement was quantified using Cohen’s kappa coefficient. Additionally, mixed-effects logistic regression was used to explore associations between DR and clinical risk factors. Main Outcome Measures: Sensitivity, specificity, and agreement (Cohen’s kappa) of the AI system compared to clinical diagnosis; predictors of DR such as age, diabetes duration, presence of Diabetic Macular Edema (DME), and central retinal thickness (CRT). Results: The EyeArt® system achieved a binocular sensitivity of 100% (95% CI: 98.1–100) and a specificity of 93.5% (95% CI: 90.2–96.0). Agreement with ophthalmologist grading was excellent, with kappa values of 0.923 (right eye) and 0.949 (left eye). Younger age, longer diabetes duration, DME presence, and higher CRT were significantly associated with DR diagnosis. Conclusions: The EyeArt ® AI system showed excellent diagnostic accuracy and strong agreement with clinical evaluations in DR screening. Nonetheless, its tendency to overestimate DR severity indicates the need for further refinement of its grading algorithm. These findings support the potential integration of AI systems into large-scale diabetic retinopathy screening programs, pending further validation. Diabetic Retinopathy Artificial Intelligence Automated Diabetic Retinopathy Artificial Intelligence Detection of Diabetic Retinopathy Automated Retinal Image Figures Figure 1 1. INTRODUCTION Diabetic retinopathy (DR) is a prevalent microvascular complication of Diabetes Mellitus (DM), affecting an estimated one-third of all individuals with diabetes globally (1)(2)(3). As the prevalence of diabetes continues to increase worldwide, the burden of DR has escalated, making it the leading cause of blindness among working-age adults (4). In the United States alone, over 30 million adults live with DM, and studies indicate that more than 10% of these patients are at risk of vision-threatening DR (vtDR) if not monitored and treated promptly (1)(5). Untreated DR progresses asymptomatically, advancing through stages marked by retinal damage, which can lead to irreversible vision loss if early signs are not detected and managed (6)(7). The clinical importance of regular DR screening is underscored by the potential for early intervention to significantly reduce the progression of sight-threatening stages (8)(9). Effective treatments, such as laser photocoagulation and intravitreal injections, have proven successful in halting or slowing the progression of DR (10)(11)(12); however, their efficacy is maximized when applied at the onset of the disease (13). Consequently, clinical guidelines recommend annual eye examinations for patients with DM to detect DR early (14)(15). Yet, adherence to these recommendations is limited; in the United States, less than 60% of individuals with DM receive the recommended annual eye screening, with even lower rates in low-resource settings and among minority and underserved populations (16). Traditional DR screening programs rely on human graders to manually assess digital fundus photographs (17)(18). While effective, this approach is labor-intensive, expensive, and dependent on a limited pool of trained ophthalmic professionals (19). In regions with high demand and limited resources, such as rural areas and developing countries, barriers such as lack of access to eye care specialists, lengthy wait times, and high costs further limit the reach and efficacy of DR screening efforts (20)(21). Additionally, in low- and middle-income countries, the shortage of ophthalmologists and the logistical challenges of reaching remote populations exacerbate the gaps in screening and early detection (22)(23). In response to these limitations, Artificial Intelligence (AI) has emerged as a promising tool to address the challenges of DR screening (24). Recent advancements in AI-driven image analysis, powered by deep learning (DL) algorithms, allow automated detection of DR from retinal fundus photographs. AI-based screening systems, such as EyeArt ® , IDx-DR ® , and Retmarker ® , have shown comparable accuracy to human graders in detecting referral-warranted DR (rwDR) and vtDR, offering high sensitivity and specificity rates that make them viable alternatives to traditional manual grading (25)(26)(27). These systems can be deployed in primary care and telemedicine settings, making DR screening more accessible and potentially reducing the overall healthcare costs associated with vision loss prevention (28)(29). Among the available AI systems, EyeArt® was selected for this study due to practical considerations in our local healthcare context. Specifically, it was the system to which we had licensed access, allowing us to integrate it seamlessly into the existing Retisalud screening program infrastructure. Moreover, EyeArt® has demonstrated strong diagnostic performance in large population-based studies, further supporting its inclusion in this comparative analysis (30)(31). One of the major advantages of AI-based DR screening is the ability to implement point-of-care testing in primary and community healthcare settings without the need for specialized equipment or training in ophthalmology (32)(33). For example, studies have demonstrated that non-mydriatic cameras paired with AI algorithms can be operated effectively by general healthcare staff, facilitating DR screening in primary care practices and improving adherence to screening guidelines among underserved populations (34). Furthermore, the integration of AI into telemedicine frameworks has enabled remote DR screening in rural and isolated areas, providing timely detection and referral recommendations without requiring patients to travel long distances to access specialized eye care (35)(36). Economic evaluations of AI-based DR screening systems suggest that these tools can be cost-effective, especially when implemented at a large scale in settings with high diabetes prevalence (37). Studies show that by reducing the need for manual grading and minimizing unnecessary referrals, AI-driven DR screening can lower healthcare costs while maintaining high diagnostic accuracy (38)(39)(40)(41)(42). Moreover, AI systems have the potential to alleviate the workload on human graders, enabling ophthalmologists to focus on cases that require immediate intervention and enhancing the overall efficiency of DR screening programs (43)(44). The objective of this study is to evaluate the diagnostic performance of the EyeArt® AI system in detecting DR, by analyzing its concordance with ophthalmologist-determined diagnoses within a large-scale regional screening program. The study aims to determine the degree of agreement between both diagnostic approaches and assess the potential utility of EyeArt® as a reliable tool in early DR detection. 2. METHODS Design This is a cross-sectional study in which 499 diabetic patients over 18 years old were included. Patients were randomly recruited from the diabetic retinopathy screening program conducted in the Canary Islands ( Retisalud ) between June 1, 2023, and September 22, 2023. Once selected, they were scheduled for an appointment at the Ophthalmology Department of a tertiary-level hospital. During this consultation, patients were informed about the study, including the details of participation. All participants received detailed information about the study objectives and procedures and signed an informed consent form (Annex I) prior to enrollment. The study was reviewed and approved by the Ethics Committee of Las Palmas CEI/CEIm, under reference number 2022-376-1. All procedures complied with the Declaration of Helsinki and current data protection regulations. In the recruitment consultation, in addition to demographic data such as age and sex, information was collected regarding the type of diabetes mellitus (type 1 or type 2) and associated conditions that increase cardiovascular risk. These conditions included hypertension, dyslipidemia, cardiac pathology, stroke, obstructive sleep apnea síndrome, and renal pathology. Ophthalmological examination, included best-corrected visual acuity (BCVA), spherical equivalent (SE), crystalline lens status (assessing nuclear, cortical, and posterior subcapsular components), central retinal thickness measured by Optical Coherence Tomography (OCT) and vascular plexus density measured by angio-OCT. Following the examination and fundus evaluation, the degree of diabetic retinopathy was determined by ophthalmologist based on the International Clinical Diabetic Retinopathy (ICDR) scale (45). Eligible participants were screened using non-mydriatic fundus photographs taken with a TRC-NW400 (Topcon ® Corporation, Japan). The resultant images were subsequently analyzed using EyeArt 2.1.0 automated DR screening software (Eyenuk, Inc., Woodland Hills, CA). Images were evaluated for quality prior to analysis. Photographs with insufficient illumination, focus issues, or artifacts that impeded retinal evaluation were excluded from the final analysis. One participant was excluded due to poor image quality caused by advanced lens opacity (cataract), as the AI system could not process the data. The EyeArt ® software generated a report determining the degree of DR and provided a specific score based on the severity of the condition. The characteristics of the sample included in the study are presented in Table 1 . Sample size and power calculation The sample size was not determined based on a prior power analysis, but rather by operational limitations. A total of 500 patient licenses were made available by Eyenuk® (EyeArt®) for this research project, which defined the upper limit of included participants. Despite the absence of a formal power calculation, the sample of 499 patients provides a robust dataset for statistical modeling and concordance analysis. Statistical analysis Univariate analysis Categorical variables are expressed as frequencies and percentages and continuous as mean and standard deviation (SD) when data followed a normal distribution, or as median and interquartile range (IQR = 25 th – 75 th percentile) when distribution departed from normality. The percentages were compared, as appropriate, using the Chi-square ( ) test or the exact Fisher test, the means by the t-test, and the medians by the Wilcoxon test for independent data. In order to identify the factors that maintain independent association with the mutation, a multivariate logistic regression analysis was performed. Logistic mixed models for diabetic retinopathy To identify factors associated with DR for each of the diagnoses (OFT) and (AI) we used mixed-effects logistic regression models (observations for each eye are nested within each patient) (46). Model estimation is performed using the L1 penalty term, which imposes variable selection and shrinkage simultaneously. For the selected factors, the corresponding models were summed in p-values and odd-ratios, which were estimated using 95% confidence intervals. This analysis was carried out using the R glmmLasso statistical package. Kappa statistic for agreement between diagnoses In order to assess the agreement between the two diagnoses by each eye, we estimated Cohen’s kappa (47). The kappa statistis is defined as: Here, and denote the proportions of observed agreements between the two diagnoses and those expected under the random agreement hypothesis. Cohen suggested (48) the Kappa result be interpreted as follows: values ≤ 0 as indicating no agreement and 0.01–0.20 as none to slight, 0.21–0.40 as fair, 0.41– 0.60 as moderate, 0.61 0.80 as substantial, and 0.81–1.00 as almost perfect agreement. Statistical significance was set at . Data were analyzed using the R package, version 4.2.1. 3. RESULTS Five hundred adults with diabetes consented to participate in this prospective study. One participant was excluded from the análisis because the fundus images obtained were of poor quality due to advanced cataracts and could not be analyzed by the AI program. Four hundred and ninety-nine adults with diabetes were ultimately included in the study. Table 1 summarizes the patients characteristics, overall and DR (unilateral and bilateral). The median age was 65.1± 11.1. 53.5% of the patients were male, while 46.5% were female. The demographic analysis of the 499 patients included in the study revealed no statistically significant differences concerning age, gender, type of DM, or the coexistence of other conditions that increase cardiovascular risk, such as hypertension, dyslipidemia, Obstructive Sleep Apnea Sindrom (OSAS), Chronic Obstructive Pulmonary Disease (COPD), asthma, heart failure, acute myocardial infarction, among others. [Table 1. Patient characteristics: overall and by DR] Table 2 shows the logistic mixed models for DR (presence/absence) diagnosed by Ophalmologist (OFT) and Artificial Intelilgence (AI) respectively. The table 2.1 presents the clinical characteristics of right eyes of the sample and ophthalmological parameters obtained through OCT and other evaluation methods in the study sample, based on the degree of DR determined by the ophthalmologist. DME-OCT (Diabetic Macular Edema by OCT) showed statistically significant differences between groups (p < 0.001), indicating a higher presence of macular edema in more advanced stages of DR. Cortical Lens Status (Crystalline.C) and posterior subcapsular component of the lens (Crystalline.P) showed no significant differences between groups (p = 0.272 and p = 0.189 respectively), with most patients exhibiting no cortical and posterior subcapsular opacities. Nuclear Lens Status (Crystalline.N) showed statistically significant differences (p = 0.02), reflecting increased nuclear lens opacification in more advanced stages of the disease. BCVA demonstrated a significant difference between groups (p = 0.023), suggesting a reduction in visual acuity in patients with more severe DR. Superficial Vascular Plexus Density (VD A-OCT) in the temporal and inferior regions exhibited significant differences (p < 0.001), indicating reduced vascular density associated with greater severity of DR. Other parameters, such as central and nasal superficial plexus density, did not show significant differences (p = 0.189 and p = 0.245, respectively). These results suggest a correlation between the severity of DR and structural and functional changes in the retina, particularly evident in visual acuity and temporal and inferior plexus vascular density. [Table 2.1. Right eyes] Table 2.2 summarizes the characteristics of the left eyes of the patients included in the study, focusing on key clinical and ophthalmological parameters based on the degree of DR determined by the ophthalmologist. A statistically significant increase in the prevalence of macular edema (DME-OCT) was observed with disease severity (p < 0.001). Superficial vascular density (VD A-OCT) measurements revealed significant reductions in the superior plexus (p < 0.001), temporal plexus (p = 0.011) and nasal plexus (p = 0.001). Additionally, CRT - OCT (Central Retinal Thickness determined by Optical Coherence Tomography) scores varied significantly across the stages of retinopathy, highlighting the systemic impact of DR on ocular health (p<0.001). These results reinforce the importance of early detection and monitoring of macular edema, lens opacities, and vascular density changes in patients with DR. [Table 2.2 Left eyes] Table 2.3 provides a summary of the clinical and ophthalmological characteristics of the right eyes of the patients included in this study, based on the degree of DR determined by the AI. The results reveal significant differences in several parameters according to the severity of DR, while other variables remained unchanged. Regarding DME (DME - OCT), a statistically significant difference was observed between groups (p < 0.001), with a higher prevalence of macular edema as the severity of DR increased. The highest prevalence, 17.8%, was found in severe non-proliferative cases. In contrast, cortical lens opacity (Crystalline C) did not show significant differences between groups (p = 0.872). Most patients (93.4%) exhibited no cortical opacities, while a minority presented mild or moderate changes. Nuclear lens opacity (Crystalline N) exhibited a statistically significant increase in patients with greater disease severity (p = 0.024). Superficial vascular density (VD A-OCT) showed a significant reduction in the temporal, superior, inferior and nasal regions as the severity of DR increases. However, no significant differences were identified in the central region (p = 0.389) [Table 2.3 Right eyes] The table 2.4 provides a detailed breakdown of the clinical and ophthalmological characteristics of the patients’ eyes based on the degree of DR determined by the AI. Statistically significant differences were observed for DME (DME - OCT) across groups (p < 0.001), with a progressive increase in edema correlating strongly with the severity of DR. For lens opacities, no significant differences were found for cortical, nuclear and posterior opacities. BCVA declined significantly as DR severity increased (p = 0.004), highlighting the functional impact of disease progression. Regarding vascular density in the superficial plexus (VD A-OCT), the temporal, nasal, inferior a superior regions exhibited a significant reduction in density with disease progression, indicating its association with advanced retinopathy, while central region showed no significant changes (p = 0.688)), suggesting these region is less affected in early to moderate stages of the disease. Regarding the distribution of DR severity, ophthalmologist diagnoses classified 68.5% of eyes as having no DR, 19.4% as mild non-proliferative DR, 9.0% as moderate, and 3.0% as severe non-proliferative DR. In contrast, AI classified 65.1% as no DR, 11.4% as mild, 14.0% as moderate, and 9.0% as severe or proliferative DR, highlighting a shift toward higher severity categories in AI results. This disparity further underscores the tendency of the AI system to overclassify DR severity compared to clinical judgment. [Table 2.4 Left eyes] A stratified analysis of AI performance across different DR severity levels revealed that EyeArt ® tends to overestimate the disease severity, as demonstrated by the distribution of cases along the diagonal and above-diagonal in the scatter plot (Figure 1). The proportion of moderate and severe classifications made by AI was higher than those assigned by ophthalmologists, suggesting a shift toward conservative (over-referral) predictions. This trend was more pronounced in the moderate non-proliferative and severe non-proliferative categories, whereas underestimation of severity was not observed. These findings support the earlier observation that the binary classification (presence/absence of DR) yielded better agreement metrics compared to the multi-level severity classification, as evidenced by the lower kappa coefficients in the latter. Table 3 presents the results of a multivariable logistic regression model evaluating factors associated with DR. Age (per additional year) showed an inverse association with DR, with an odds ratio (OR) of 0.976 (95% CI: 0.963–0.990; p < 0.001). This indicates that for each additional year of life, the odds of developing retinopathy decrease by approximately 2.4%, assuming that all other variables remain constant. This finding may reflect that disease progression is influenced by factors related to the duration of diabetes in individuals diagnosed at younger ages. In other words, individuals diagnosed at older ages have a lower risk of developing complications associated with DR because they have fewer years to develop them. The duration of diabetes showed a positive and significant association with DR. The OR was 1.151 (95% CI: 1.109–1.194; p < 0.001). This indicates that for each additional year since diabetes diagnosis, the odds of developing DR increase by 15.1%, while keeping other variables in the model constant. This result highlights the cumulative impact of diabetes on the development of microvascular complications over time. Another statistically significant determinant was the presence of DME determined by OCT, as its presence increased the probability of developing DR by 32 times (p < 0.001), assuming all other variables remain constant. Additionally, spherical equivalent demonstrated a statistically significant association with the development of DR (p < 0.001), acting as a protective factor in this case. Specifically, for each unit increase in spherical equivalent, the probability of developing diabetic retinopathy decreased by 10.9%. [Table 3. Multivariate logistic regression for diabetic retinopathy. (*) per unit] Table 4 presents the estimates of bSen (95% CI) and bSp (95% CI). According to Perera et al(49) , we will conduct the evaluation of AI as a diagnostic marker for DR using the concepts of binocular sensitivity (bSen) and binocular specificity (bSp). bSen is the probability of obtaining at least one positive AI diagnosis in one eye, given that there is a diagnosis of DR by an ophthalmologist (OFT) in at least one eye. bSp is the probability of obtaining a negative AI diagnosis in both eyes, given that there is a negative DR diagnosis by an ophthalmologist in both eyes. The results presented in Table 6 indicate that the AI system demonstrated excellent diagnostic performance in detecting DR. The estimated binocular sensitivity (bSen) was 100% (95% CI: 98.1–100), meaning that the AI correctly identified all cases of DR in at least one eye, with no false negatives. This exceptionally high sensitivity suggests that the system is highly effective for screening purposes. The estimated binocular specificity (bSp) was 93.5% (95% CI: 90.2–96.0), indicating that the AI correctly classified the absence of DR in both eyes in 93.5% of negative cases, with a false positive rate of 6.5%. [Table 4. Sensitiviy and specificity(*) ] The kappa statistics estimations, including their 95% CI, are detailed in Table 5 , which presents the observed and expected agreement proportions alongside kappa coefficients in the concordance analysis for both eyes. The degree of agreement was assessed considering the different stages of DR. For the right eye, the kappa coefficient was 0.650 (95% CI: 0.594–0.707), with an observed agreement proportion of 0.820 and an expected agreement of 0.484 under the random agreement hypothesis. For the left eye, the kappa coefficient was 0.693 (95% CI: 0.636–0.750), with an observed agreement proportion of 0.847 and an expected agreement of 0.503. These results highlight substantial agreement in both eyes. [Table 5. Proportion of agreements considering the different stages of DR (*) per unit. (*)] Table 6 summarizes the levels of agreement for a binary (Yes/No) diagnosis, including the observed and expected proportions of agreement, as well as the kappa coefficients with their 95% CI. For the right eye, the observed agreement proportion was 0.966, while the expected agreement under the random hypothesis was 0.556, resulting in a kappa coefficient of 0.923 (95% CI: 0.836–1.010). For the left eye, the observed agreement proportion was 0.978, with an expected agreement of 0.569, yielding a kappa coefficient of 0.949 (95% CI: 0.861–1.036). These findings demonstrate almost perfect agreement for both eyes in the binary diagnosis. [Table 6. Agreement between binary (Y/N) diagnosis(*).] Although the kappa coefficients indicate substantial to almost perfect agreement between AI and ophthalmologist diagnoses—particularly in binary classification—it is important to note that some confidence intervals marginally include or exceed the upper bound of 1.0, particularly in Table 6 . This statistical artifact may reflect a possible overestimation of agreement, likely influenced by the high prevalence of negative (No DR) classifications, which increases the observed agreement and may reduce the impact of discrepancies in moderate or severe cases. These wide CI should be interpreted with caution, as they may indicate variability or limitations in the estimation of concordance levels in smaller subgroups The Figure 1 consists of two scatter plots illustrating artificial intelligence (AI) and ophthalmologist diagnostic outcomes for right eyes (blue) and left eyes (red). X-axis (horizontal) represents the diagnosis performed by the ophthalmologist categorized as “No,” “Mild,” “Moderate,” and “Severe.” Y-axis (vertical) indicates the AI-generated diagnoses, classified using the same categories: “No,” “Mild,” “Moderate,” “Severe,” and an additional category, “Proliferative.” Each data point corresponds to a single AI prediction compared to its ground-truth diagnosis performed by the ophthalmologist. Right eye diagnoses are plotted in blue on the left panel and left eye diagnoses are plotted in red on the right panel. Points aligned along the diagonal (extending from the bottom-left to the top-right) indicate accurate AI predictions, where the AI diagnosis matches the diagnostis performed by the ophtalmologist. Points above the diagonal indicate instances where the AI overestimates the disease severity and points below the diagonal reflect cases where the AI underestimates the severity of the condition. The scatter plots for right and left eyes exhibit similar distributions of data points, suggesting consistent diagnostic performance by the AI across both eyes. Both panels demonstrate a higher concentration of points in the “No” and “Mild” categories, indicating a predominance of low-severity diagnoses in the dataset. [Figure 1. Scatter plots] 4. DISCUSSION The detection of DR is a coplex image-interpretation task and a key step in any successfull screening program. The findings of this study ( Tables 2 ) reinforce the strong association between the progression of DR and alterations in key ocular parameters. Specifically, changes in visual acuity, DME, nuclear lens opacification, and reductions in vascular density exhibited a significant correlation with disease severity. These results provide further insights into the systemic impact of DR on ocular health and highlight the relevance of these parameters as potential indicators of disease progression. Conversely, cortical lens opacity did not show significant differences among the studied groups, suggesting that its progression may not be directly linked to DR severity. These findings underscore the importance of a comprehensive ophthalmological assessment in patients with DR, emphasizing the need for thorough monitoring to detect and manage disease-related complications effectively. These insights highlight the critical role of early detection and continuous monitoring using OCT to assess disease progression. Implementing OCT as a routine tool for evaluating structural and vascular alterations in DR could enhance clinical decision-making, facilitating timely therapeutic interventions aimed at preserving visual function and mitigating disease-associated complications. Overall, these findings contribute to a deeper understanding of the ophthalmic manifestations of DR and reinforce the importance of integrating multimodal imaging techniques into routine clinical practice. Further studies with larger cohorts and longitudinal designs are warranted to validate these associations and explore their potential implications for personalized disease management strategies (38)(39)(50)(30)(31)(51). The analysis of sensitivity and specificity values highlights the robust diagnostic performance of the AI system in detecting DR. The high sensitivity observed suggests that the system is highly effective in identifying positive cases, reinforcing its utility as a screening tool for early disease detection. Similarly, the specificity is also high, although slightly lower than sensitivity, indicating a minor risk of overdiagnosis due to false-positive results. This could lead to unnecessary referrals for further ophthalmologic evaluation, which, while ensuring comprehensive patient care, may also contribute to an increased burden on healthcare systems. These high sensitivity and specificity values obtained are consistent with the results reported in other similar studies (52)(53) . The confidence intervals for both sensitivity and specificity demonstrate a high degree of reliability in these estimates, with slightly greater variability observed in specificity. This variability underscores the need for continuous refinement of AI algorithms to further optimize diagnostic accuracy and minimize unnecessary referrals. Nevertheless, these findings support the implementation of AI-assisted diagnostic methods as a valuable tool for DR detection, providing high sensitivity for disease identification and good specificity for distinguishing negative cases. The overall performance of the system reinforces its potential as an effective and reliable screening approach, particularly in large-scale screening programs aimed at early detection and timely intervention. Although EyeArt ® demonstrated high sensitivity, the moderately lower specificity implies a proportion of false positives, particularly in cases where severity was overestimated. From a clinical perspective, this may lead to increased referrals to ophthalmologists, which, while ensuring patient safety, may overload referral systems, especially in resource-limited settings. Such an increase in workload could affect waiting times and allocation of clinical resources, potentially reducing efficiency in managing truly severe cases. Hence, balancing sensitivity with specificity is essential to optimize screening workflows. The multivariate logistic regression analysis identified several significant predictors of DR, including age, diabetes duration, DME, and CRT detected by OCT. These findings underscore the multifactorial nature of DR progression and highlight the importance of comprehensive patient management. In particular, glycemic control, blood pressure regulation, and consideration of disease duration should be key components in the clinical approach to diabetic patients to mitigate the risk of retinopathy development and progression. The kappa coefficient values indicate a moderate level of agreement between diagnoses related to the severity of DR as determined by the ophthalmologist and the EyeArt ® AI system. However, when the agreement is assessed in the context of screening for a binary diagnosis—determining the presence or absence of DR without specifying its severity—the kappa coefficient approaches a value of one, indicating an almost perfect level of concordance between the ophthalmologist and the AI system. This high level of diagnostic accuracy supports the clinical and research applicability of AI-assisted methods, particularly in screening programs designed to differentiate between healthy individuals and those with DR. The strong agreement observed in both eyes further reinforces the reliability of this approach, highlighting its potential as an effective tool for large-scale DR detection and early intervention strategies. These results are consistent with those reported by Karabeg et al . (52) and Wintergest et al . (53) However, this study demonstrates significantly higher concordance values compared to those of Wang et al .(54), Rajalakshmi et al . (55), Kim et al . (56), Mokhanshi et al .(57), and Cicinelli et al (58). This improvement in concordance is likely attributable to the refinement of the artificial intelligence program and the quality of the images obtained with the device used. In the scatter plot, it is evident that the AI program tends to overestimate the severity of DR as determined by the ophthalmologist. Notably, there are no data points below the bisector, indicating the absence of underdiagnosis cases. However, several points are located above the bisector, demonstrating that the EyeArt ® system has a tendency to overestimate the severity of DR compared to the ophthalmologist’s assessment. As previously noted, the degree of agreement, as measured by the Kappa coefficient, was higher (0.923 for right eyes and 0.949 for left eyes) when comparing the binary classification between the ophthalmologist and the AI, i.e., when the diagnosis was limited to determining the presence or absence of DR, which is the primary objective of screening programs. However, when the Kappa coefficient was calculated based on the severity grading of DR, the level of agreement was lower (0.650 for right eyes and 0.693 for left eyes). The scatter plot further confirms that the lower agreement observed in DR grading is due to the AI system’s tendency to overestimate disease severity. This study has several limitations that should be acknowledged. First, a potential selection bias cannot be excluded, as participants were recruited from a single regional screening program and may not fully represent the broader diabetic population. Additionally, the exclusion of poor-quality fundus images—although necessary for analysis—may have introduced a spectrum bias, favoring patients with better media clarity and potentially underrepresenting more complex or advanced cases. Another limitation is the small number of cases diagnosed as proliferative DR (n=2), which precludes a robust evaluation of the AI system's performance in this critical category. It is important to note that, thanks to the existing screening programs, it is uncommon to encounter advanced stages such as proliferative DR, which explains the small sample size observed in this group. Future studies with larger samples in this subgroup are needed to validate diagnostic accuracy in more severe stages. The results of this study reinforce the role of EyeArt ® as a useful tool in DR screening. However, its integration into real-world healthcare settings must be carefully planned. Strategies such as a two-tiered screening approach, where positive AI cases are reviewed by a human grader before referral, could reduce unnecessary specialist consultations. In primary care or teleophthalmology environments, training non-specialist personnel in image acquisition, coupled with routine quality control protocols, can help ensure that the system functions optimally while maintaining diagnostic reliability. 5. CONCLUSIONS The EyeArt ® AI system has demonstrated high sensitivity and specificity in detecting DR, achieving an almost perfect agreement with ophthalmologists when using a binary diagnosis. However, its tendency to overestimate disease severity suggests the need for algorithmic improvements to enhance classification accuracy. These findings support the use of AI as a complementary tool in screening programs, enabling early detection and improved access to diagnosis in resource-limited settings, ultimately contributing to a reduction in the global burden of DR. However, its clinical implementation should be accompanied by further validations and refinements in severity classification, ensuring its effectiveness in medical decision-making and optimizing the management of DR patients. Declarations ACKNOWLEDGMENT/DISCLOSURE The authors declare that they have no conflicts of interest related to this study. No financial support was received from any commercial entity for the conduct of this research or the preparation of this manuscript. DECLARATION OF CONSENT TO PARTTICIPATE All procedures performed in this study involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. Written informed consent was obtained from all individual participants included in the study. Financial Support: None Conflict of Interest: No conflicting relationship exists for any autor Author Contribution I.I.G.G., A.G.H. and P.S.S. wrote the main manuscriptP.S.S. Prepared figures and prepared the statistical analysis.A.R.M., A.R.M. and F.C.L contributed to the revision of the manuscriptF.C.L. contacted with EyeNUK to acquire the licenses for the study. References Yau JWY, Rogers SL, Kawasaki R, Lamoureux EL, Kowalski JW, Bek T, et al. Global prevalence and major risk factors of diabetic retinopathy. Diabetes Care. 2012 Mar;35(3):556–64. Ogurtsova K, da Rocha Fernandes JD, Huang Y, Linnenkamp U, Guariguata L, Cho NH, et al. IDF Diabetes Atlas: Global estimates for the prevalence of diabetes for 2015 and 2040. Diabetes Res Clin Pract. 2017 Jun 1;128:40–50. Cho NH, Shaw JE, Karuranga S, Huang Y, da Rocha Fernandes JD, Ohlrogge AW, et al. IDF Diabetes Atlas: Global estimates of diabetes prevalence for 2017 and projections for 2045. Diabetes Res Clin Pract. 2018 Apr 1;138:271–81. Flaxman SR, Bourne RRA, Resnikoff S, Ackland P, Braithwaite T, Cicinelli M V., et al. Global causes of blindness and distance vision impairment 1990–2020: a systematic review and meta-analysis. Lancet Glob Health. 2017 Dec 1;5(12):e1221–34. Fox CR, Kronenberg K, Weiskopf ES. Using Telemedicine to Increase Eye Care Screening & Referral for People with Diabetes Article Details. J Pub Health Issue Pract JPHIP [Internet]. 2019;3:137. Available from: http://dx.doi.org/jphip/2019/137 Guariguata L, Whiting DR, Hambleton I, Beagley J, Linnenkamp U, Shaw JE. Global estimates of diabetes prevalence for 2013 and projections for 2035. Diabetes Res Clin Pract. 2014;103(2):137–49. Fleming AD, Goatman KA, Philip S, Prescott GJ, Sharp PF, Olson JA. Automated grading for diabetic retinopathy: a large-scale audit using arbitration by clinical experts. British Journal of Ophthalmology [Internet]. 2010;(12):1606. Available from: https://hal.science/hal-00578459v1 Corner S, Mohamed Q, Mark Gillies MC, Wong TY. CLINICAL, REVIEW Management of Diabetic Retinopathy A Systematic Review [Internet]. Available from: www.jama.com Lin S, Ramulu P, Lamoureux EL, Sabanayagam C. Addressing risk factors, screening, and preventative treatment for diabetic retinopathy in developing countries: a review. Vol. 44, Clinical and Experimental Ophthalmology. Blackwell Publishing; 2016. p. 300–20. Zhang W, Zhao G, Fan W, Zhao T. Panretinal photocoagulation after or prior to intravitreal conbercept injection for diabetic macular edema: a retrospective study. BMC Ophthalmol. 2021 Dec 1;21(1). Aiello LP, Gardner TW, King GL, Blankenship G, Cavallerano JD, Ferris FL, et al. Diabetic Retinopatiiy. Aljundi W, Suffo S, Munteanu C, Langenbucher A, Seitz B, Abdin AD. Intravitreal Injection for Diabetic Macular Edema as Adjunctive Therapy for Proliferative Diabetic Retinopathy: A Retrospective Study. Clinical Ophthalmology. 2022;16:135–43. Lopez-Ramos A, Gomez-Bastar PA, Lansingh VC, Rodriguez-Gomez JA, Vargas-Fragoso V, Soria-Arellano FA, et al. Rapid assessment of avoidable blindness: Prevalence of blindness, visual impairment and diabetes in nuevo leon, Mexico 2014. Ophthalmic Epidemiol. 2018 Nov 2;25(5–6):412–8. Murchison AP, Friedman DS, Gower EW, Haller JA, Lam BL, Lee DJ, et al. A multi-center diabetes eye screening study in community settings: Study design and methodology. Ophthalmic Epidemiol. 2016 Mar 3;23(2):109–15. Bek T, Andersen N, Andresen J, Grauslund J, Hajari J, Schmidt Laugesen C, et al. Individualised screening for diabetic retinopathy with proliferative retinopathy and macular oedema as separate end points. Acta Ophthalmol. 2024 Sep 1;102(6):627–34. Ting DSW, Cheung GCM, Wong TY. Diabetic retinopathy: global prevalence, major risk factors, screening practices and public health challenges: a review. Vol. 44, Clinical and Experimental Ophthalmology. Blackwell Publishing; 2016. p. 260–77. Ting DSW, Tan GSW. Telemedicine for diabetic retinopathy screening. Vol. 135, JAMA Ophthalmology. American Medical Association; 2017. p. 722–3. Mansberger SL, Sheppler C, Barker G, Gardiner SK, Demirel S, Wooten K, et al. Long-term comparative effectiveness of telemedicine in providing diabetic retinopathy screening examinations: A randomized clinical trial. JAMA Ophthalmol. 2015 May 1;133(5):518–25. Avidor D, Loewenstein A, Waisbourd M, Nutman A. Cost-effectiveness of diabetic retinopathy screening programs using telemedicine: A systematic review. Vol. 18, Cost Effectiveness and Resource Allocation. BioMed Central Ltd.; 2020. Ramasamy K, Mishra C, Kannan NB, Namperumalsamy P, Sen S. Telemedicine in diabetic retinopathy screening in India. Vol. 69, Indian journal of ophthalmology. NLM (Medline); 2021. p. 2977–86. Li X, Tan TE, Wong TY, Sun X. Diabetic retinopathy in China: Epidemiology, screening and treatment trends—A review. Vol. 51, Clinical and Experimental Ophthalmology. John Wiley and Sons Inc; 2023. p. 607–26. Zhang X, Beckles GL, Chou CF, Saaddine JB, Wilson MR, Lee PP, et al. Socioeconomic disparity in use of eye care services among US adults with age-related eye diseases: National health interview survey, 2002 and 2008. JAMA Ophthalmol. 2013 Sep;131(9):1198–206. Fathy C, Patel S, Sternberg P, Kohanim S. Disparities in Adherence to Screening Guidelines for Diabetic Retinopathy in the United States: A Comprehensive Review and Guide for Future Directions. Vol. 31, Seminars in Ophthalmology. Taylor and Francis Ltd; 2016. p. 364–77. Bellemo V, Lim G, Rim TH, Tan GSW, Cheung CY, Sadda SV, et al. Artificial Intelligence Screening for Diabetic Retinopathy: the Real-World Emerging Application. Vol. 19, Current Diabetes Reports. Current Medicine Group LLC 1; 2019. Bellemo V, Lim G, Rim TH, Tan GSW, Cheung CY, Sadda SV, et al. Artificial Intelligence Screening for Diabetic Retinopathy: the Real-World Emerging Application. Vol. 19, Current Diabetes Reports. Current Medicine Group LLC 1; 2019. Abràmoff MD, Roehrenbeck C, Trujillo S, Goldstein J, Graves AS, Repka MX, et al. A reimbursement framework for artificial intelligence in healthcare. Vol. 5, npj Digital Medicine. Nature Research; 2022. Nørgaard MF, Grauslund J. Automated screening for diabetic retinopathy - A systematic review. Ophthalmic Res. 2018 Jun 1;60(1):9–17. Kavitha DR, Student UG. DETECTION OF DIABETIC RETINOPATHY USING RETINAL FUNDUS PHOTOGRAPHS BASED ON DEEP LEARNING ALGORITHM. Turkish Journal of Physiotherapy and Rehabilitation [Internet]. 32(2). Available from: www.turkjphysiotherrehabil.org Wongchaisuwat N, Trinavarat A, Rodanant N, Thoongsuwan S, Phasukkijwatana N, Prakhunhungsit S, et al. In-person verification of deep learning algorithm for diabetic retinopathy screening using different techniques across fundus image devices. Transl Vis Sci Technol. 2021 Nov 1;10(13). Bhaskaranand M, Ramachandra C, Bhat S, Cuadros J, Nittala MG, Sadda SR, et al. The value of automated diabetic retinopathy screening with the EyeArt system: A study of more than 100,000 consecutive encounters from people with diabetes. Diabetes Technol Ther. 2019 Nov 1;21(11):635–43. Lim JI, Regillo CD, Sadda SVR, Ipp E, Bhaskaranand M, Ramachandra C, et al. Artificial Intelligence Detection of Diabetic Retinopathy: Subgroup Comparison of the EyeArt System with Ophthalmologists’ Dilated Examinations. Ophthalmology Science. 2023 Mar 1;3(1). Nielsen KB, Lautrup ML, Andersen JKH, Savarimuthu TR, Grauslund J. Deep Learning–Based Algorithms in Screening of Diabetic Retinopathy: A Systematic Review of Diagnostic Performance. Vol. 3, Ophthalmology Retina. Elsevier Inc; 2019. p. 294–304. Honasoge A, Nudleman E, Smith M, Rajagopal R. Emerging Insights and Interventions for Diabetic Retinopathy. Vol. 19, Current Diabetes Reports. Current Medicine Group LLC 1; 2019. De Fauw J, Ledsam JR, Romera-Paredes B, Nikolov S, Tomasev N, Blackwell S, et al. Clinically applicable deep learning for diagnosis and referral in retinal disease. Nat Med. 2018 Sep 1;24(9):1342–50. Huang X, Wang H, She C, Feng J, Liu X, Hu X, et al. Artificial intelligence promotes the diagnosis and screening of diabetic retinopathy. Vol. 13, Frontiers in Endocrinology. Frontiers Media S.A.; 2022. Wewetzer L, Held LA, Goetz K, Steinhäuser J. Determinants of the implementation of artificial intelligence-based screening for diabetic retinopathy—a cross-sectional study with general practitioners in Germany. Digit Health. 2023 Jan 1;9. Tufail A, Kapetanakis V V., Salas-Vega S, Egan C, Rudisill C, Owen CG, et al. An observational study to assess if automated diabetic retinopathy image assessment software can replace one or more steps of manual imaging grading and to determine their cost-effectiveness. Health Technol Assess (Rockv). 2016 Dec 1;20(92):1–72. Heydon P, Egan C, Bolter L, Chambers R, Anderson J, Aldington S, et al. Prospective evaluation of an artificial intelligence-enabled algorithm for automated diabetic retinopathy screening of 30 000 patients. Vol. 105, British Journal of Ophthalmology. BMJ Publishing Group; 2021. p. 723–8. Tufail A, Rudisill C, Egan C, Kapetanakis V V., Salas-Vega S, Owen CG, et al. Automated Diabetic Retinopathy Image Assessment Software: Diagnostic Accuracy and Cost-Effectiveness Compared with Human Graders. Ophthalmology. 2017 Mar 1;124(3):343–51. Nørgaard MF, Grauslund J. Automated screening for diabetic retinopathy - A systematic review. Ophthalmic Res. 2018 Jun 1;60(1):9–17. Poly TN, Islam MM, Walther BA, Lin MC, (Jack) Li YC. Artificial intelligence in diabetic retinopathy: Bibliometric analysis. Comput Methods Programs Biomed. 2023 Apr 1;231. Jones S, Edwards RT. Diabetic retinopathy screening: A systematic review of the economic evidence. Vol. 27, Diabetic Medicine. Blackwell Publishing Ltd; 2010. p. 249–56. Wong TY, Bressler NM. Artificial intelligence with deep learning technology looks into diabetic retinopathy screening. Vol. 316, JAMA - Journal of the American Medical Association. American Medical Association; 2016. p. 2366–7. Walton OB, Garoon RB, Weng CY, Gross J, Young AK, Camero KA, et al. Evaluation of automated teleretinal screening program for diabetic retinopathy. JAMA Ophthalmol. 2016 Feb 1;134(2):204–9. Esmaeilkhanian H, Liu H, Fasih-Ahmed S, Gnanaraj R, Verma A, Oncel D, et al. The relationship of diabetic retinopathy severity scales with frequency and surface area of diabetic retinopathy lesions. Graefe’s Archive for Clinical and Experimental Ophthalmology. 2023 Nov 1;261(11):3165–76. Groll A, Tutz G. Variable selection for generalized linear mixed models by L1-penalized estimation. Stat Comput. 2014 Mar;24(2):137–54. Frontmatter. In 2003. Bakeman R. KappaAcc: A program for assessing the adequacy of kappa. Behav Res Methods. 2023 Feb 1;55(2):633–8. Perera Y, Ren M, Punzalan JRB, Rudnisky CJ, de Leon AR. Binocular sensitivity and specificity of screening tests in cross-sectional diagnostic studies of paired organs. Stat Med. 2017 May 20;36(11):1754–66. Bhaskaranand M, Ramachandra C, Bhat S, Cuadros J, Nittala MG, Sadda S, et al. Automated Diabetic Retinopathy Screening and Monitoring Using Retinal Fundus Image Analysis. J Diabetes Sci Technol. 2016 Mar 1;10(2):254–61. Olvera-Barrios A, Heeren TFC, Balaskas K, Chambers R, Bolter L, Egan C, et al. Diagnostic accuracy of diabetic retinopathy grading by an artificial intelligence-enabled algorithm compared with a human standard for wide-field true-colour confocal scanning and standard digital retinal images. British Journal of Ophthalmology. 2021 Feb 1;105(2):265–70. Karabeg M, Petrovski G, Hertzberg SNW, Erke MG, Fosmark DS, Russell G, et al. A pilot cost-analysis study comparing AI-based EyeArt® and ophthalmologist assessment of diabetic retinopathy in minority women in Oslo, Norway. Int J Retina Vitreous. 2024 Dec 1;10(1). Wintergerst MWM, Bejan V, Hartmann V, Schnorrenberg M, Bleckwenn M, Weckbecker K, et al. Telemedical Diabetic Retinopathy Screening in a Primary Care Setting: Quality of Retinal Photographs and Accuracy of Automated Image Analysis. Ophthalmic Epidemiol. 2022;29(3):286–95. Wang K, Jayadev C, Nittala MG, Velaga SB, Ramachandra CA, Bhaskaranand M, et al. Automated detection of diabetic retinopathy lesions on ultrawidefield pseudocolour images. Acta Ophthalmol. 2018 Mar 1;96(2):e168–73. Rajalakshmi R, Subashini R, Anjana RM, Mohan V. Automated diabetic retinopathy detection in smartphone-based fundus photography using artificial intelligence. Eye (Basingstoke). 2018 Jun 1;32(6):1138–44. Kim TN, Aaberg MT, Li P, Davila JR, Bhaskaranand M, Bhat S, et al. Comparison of automated and expert human grading of diabetic retinopathy using smartphone-based retinal photography. Eye (Basingstoke). 2021 Jan 1;35(1):334–42. Mokhashi N, Grachevskaya J, Cheng L, Yu D, Lu X, Zhang Y, et al. A Comparison of Artificial Intelligence and Human Diabetic Retinal Image Interpretation in an Urban Health System. J Diabetes Sci Technol. 2022 Jul 1;16(4):1003–7. Cicinelli MV, Gravina S, Rutigliani C, Checchin L, La Franca L, Lattanzio R, et al. Assessing Diabetic Retinopathy Staging With AI: A Comparative Analysis Between Pseudocolor and LED Imaging. Transl Vis Sci Technol. 2024 Mar 1;13(3). Tables Diabetic retinopathy Overall N = 499 No N = 310 Unilateral N = 70 Bilateral N = 119 P-value Age (years) 65.1 ± 11.1 65.8 ± 10.6 65.5 ± 11.6 63.0 ± 11.7 0.062 Sex male 267 (53.5) 160 (51.6) 37 (52.9) 70 (58.8) 0.404 Arterial hypertension 368 (73.8) 228 (73.5) 53 (75.7) 87 (73.1) 0.918 Diabetes mellitus Type 1 14 (2.8) 5 (1.6) 2 (2.9) 7 (5.9) 0.055 Type 2 484 (97.0) 304 (98.1) 68 (97.1) 112 (94.1) 0.081 Dyslipidemia 237 (47.5) 147 (47.4) 37 (52.9) 53 (44.5) 0.542 Asthma 8 (1.6) 6 (1.9) 1 (1.4) 1 (0.8) 0.874 AIM 16 (3.2) 11 (3.5) 1 (1.4) 4 (3.4) 0.869 COPD 25 (5.0) 15 (4.8) 3 (4.3) 7 (5.9) 0.834 HF 4 (0.8) 3 (1.0) 0 1 (0.8) 1 CKD 13 (2.6) 10 (3.2) 0 3 (2.5) 0.427 OSAS 6 (1.2) 5 (1.6) 0 1 (0.8) 0.849 Years of diabetes mellitus 8 (3; 10) 6 (2; 9) 9 (5; 10) 9 (6; 12) < 0.001 Table 1. Patient characteristics: overall and by DR. Data are means ± SD, frequencies (%) and medians (IQR). AIM: Acute Myocardial Infarction. COPD: Chronic Obstructive Pulmonary Disease. HF: Heart Failure. CKD: Chronic Kidney Disease. OSAS: Obstructive Sleep Apnea Syndrome. Diabetic retinopathy diagnosed by OFT Overall N = 499 No N = 342 Mild non-proliferative N = 97 Moderate non-proliferative N = 45 Severe non-proliferative N = 15 P-value CRT - OCT 254 (235; 275) 252 (234; 273) 254 (235; 270) 269 (253; 292) 292 (266; 298) < 0.001 DME - OCT 16 (3.2) 0 5 (5.2) 4 (8.9) 7 (46.7) < 0.001 DME - AI 20 (4.0) 1 (0.3) 6 (6.2) 6 (13.3) 7 (46.7) < 0.001 Crystalline N 4 (2; 11) 3 (2; 7) 3 (1; 6) 4 (3; 11) 5 (3; 11) 0.02 Crystalline C 0.272 0 466 (93.4) 318 (93.0) 94 (96.9) 41 (91.1) 13 (86.7) 1 23 (4.6) 17 (5.0) 1 (1.0) 3 (6.7) 2 (13.3) 2 8 (1.6) 6 (1.8) 1 (1.0) 1 (2.2) 0 3 2 (0.4) 1 (0.3) 1 (1.0) 0 0 Crystalline P 0.189 0 492 (98.6) 337 (98.5) 95 (97.9) 45 (100) 15 (100.0) 1 5 (1.0) 5 (1.5) 0 0 0 2 2 (0.4) 0 2 (2.1) 0 0 BCVA 0.8 (0.5; 0.8) 0.8 (0.5; 0.8) 0.8 (0.6; 0.9) 0.6 (0.4; 0.8) 0.7 (0.6; 0.8) 0.023 Spherical.equivalent 0.38 (-0.38;1.38) 0.50 (-0.25;1.50) 0.12 (-0.75; 1) 0.5 (-0.25; 1.75) 0 (-0.69; 0.38) 0.004 VD A-OCT SUPERFICIAL. PLEXUS CENTRAL 16 (13; 19) 16 (13; 19) 17 (13; 20) 16 (13; 19) 13 (11; 16) 0.189 VD A-OCT SUPERFICIAL. PLEXUS SUPERIOR 46 (43; 48) 46 (43; 48) 45 (42; 47) 44 (42; 46) 43 (40; 46) 0.003 VD A-OCT SUPERFICIAL .PLEXUS.TEMPORAL 46 (44; 49) 47 (45; 49) 46 (44; 48) 45 (42; 47) 41 (40; 46) < 0.001 VD A-OCT SUPERFICIAL. PLEXUS INFERIOR 46 (42; 48) 47 (44; 48) 46 (42; 48) 43 (40; 45) 41 (40; 44) < 0.001 VD A-OCT SUPERFICIAL. PLEXUS.NASAL 45 (42; 47) 45 (43; 47) 44 (42; 47) 45 (42; 47) 43 (42; 45) 0.245 Table 2.1. Right eyes. Summary of the characteristics of the right eyes of the patients included in the study. Data are frequencies (%) and medians (IQR). BCVA: Best Corrected Visual Acuity. Cristaline C: Cortical component of the lens. Cristaline N: Nuclear component of the lens. Cristaline P: Posterior component of the lens. CRT OCT: Central retinal thickness by Optical Coherence Tomography. DME- AI: Diabetic Macular Edema by Artificial Intelligence. DME-OCT: Diabetic Macular Edema by Optical Coherence Tomography. VD A-OCT: Vascular Density by Angio-Optical Coherence Tomography. Diabetic retinopathy OFT Overall N = 499 No N = 342 Mild non-proliferative N = 97 Moderate non-proliferative N = 45 Severe non-proliferative N = 15 P-value CRT - OCT 260 (238; 278) 256 (234; 271) 262 (246; 278) 270 (250; 294) 279 (268; 301) < 0.001 DME - OCT 22 (4.4) 2 (0.6) 3 (3.6) 8 (17.0) 9 (45.0) < 0.001 DME - AI 26 (5.2) 1 (0.3) 1 (1.2) 14 (29.8) 10 (50.0) < 0.001 Crystalline N 3 (2; 11) 3 (2; 6) 3 (1; 11) 4 (2; 11) 6 (3; 11) 0.083 Crystalline C 0.558 0 466 (93.4) 325 (93.4) 77 (91.7) 46 (97.9) 18 (90.0) 1 21 (4.2) 16 (4.6) 3 (3.6) 1 (2.1) 1 (5.0) 2 11 (2.2) 6 (1.7) 4 (4.8) 0 1 (5.0) 3 1 (0.2) 1 (0.3) 0 0 0 Crystalline P 0.662 0 496 (99.4) 346 (99.4) 83 (98.8) 47 (100.0) 20 (100.0) 1 1 (0.2) 1 (0.3) 0 0 0 2 1 (0.2) 0 1 (1.2) 0 0 3 1 (0.2) 1 (0.3) 0 0 0 BCVA 0.8 (0.6; 0.9) 0.8 (0.6; 0.93) 0.8 (0.6; 1) 0.7 (0.6; 0.8) 0.7 (0.5; 0.8) 0.076 Spherical.equivalent 0.50 (-0.25; 1.62) 0.62 (-0.12; 1.75) 0 (-1.12; 1.12) 0.25 (-0.75; 0.88) 0.31 (-0.16; 0.81) 0.001 VD A-OCT SUPERFICIAL PLEXUS CENTRAL 16 (13; 19) 16 (13; 20) 17 (15; 20) 16 (12; 19) 14 (14; 19) 0.417 VD A-OCT SUPERFICIAL PLEXUS SUPERIOR 46 (43; 48) 47 (44; 48) 47 (43; 49) 44 (42; 46) 44 (41; 46) < 0.001 VD A-OCT SUPERFICIAL PLEXUS TEMPORAL 46 (44; 49) 47 (45; 49) 47 (44; 48) 45 (43; 47) 45 (41; 47) 0.011 VD A-OCT SUPERFICIAL PLEXUS INFERIOR 46 (43; 48) 46 (43; 48) 45 (42; 47) 45 (42; 48) 45 (43; 47) 0.099 VD A-OCT SUPERFICIAL. PLEXUS NASAL 45 (42; 47) 45 (43; 47) 45 (42; 47) 43 (40; 45) 42 (40; 45) 0.001 Table 2.2 Left eyes. Summary of the characteristics of the left eyes of the patients included in the study. Data are frequencies (%) and medians (IQR). BCVA: Best Corrected Visual Acuity. Cristaline C: Cortical component of the lens. Cristaline N: Nuclear component of the lens. Cristaline P: Posterior component of the lens. CRT OCT: Central retinal thickness by Optical Coherence Tomography. DME- AI: Diabetic Macular Edema by Artificial Intelligence. DME-OCT: Diabetic Macular Edema by Optical Coherence Tomography. VD A-OCT: Vascular Density by Angio-Optical Coherence Tomography. Diabetic retinopathy AI Overall N = 499 No N = 325 Mild non-proliferative N = 57 Moderate non-proliferative N = 70 Severe non-proliferative N = 45 Proliferative N = 2 P-value CRT - OCT 254 (235; 275) 252 (235; 274) 253 (229; 265) 257 (238; 274) 279 (256; 297) 233 (216; 250) < 0.001 DME - OCT 16 (3.2) 0 0 8 (11.4) 8 (17.8) 0 < 0.001 DME - .IA 20 (4.0) 0 0 9 (12.9) 9 (20.0) 2 (100.0) < 0.001 Crystalline.N 4 (2; 11) 3 (2; 6) 3 (2; 6) 4 (1; 5) 5 (3; 11) 8 (6; 9) 0.024 Crystalline C 0.872 0 466 (93.4) 302 (92.9) 55 (96.5) 65 (92.9) 42 (93.3) 2 (100.0) 1 23 (4.6) 17 (5.2) 1 (1.8) 3 (4.3) 2 (4.4) 0 2 8 (1.6) 5 (1.5) 1 (1.8) 1 (1.4) 1 (2.2) 0 3 2 (0.4) 1 (0.3) 0 1 (1.4) 0 0 Crystalline P 0.236 0 492 (98.6) 320 (98.5) 56 (98.2) 69 (98.6) 45 (100.0) 2 (100.0) 1 5 (1.0) 5 (1.5) 0 0 0 0 2 2 (0.4) 0 1 (1.8) 1 (1.4) 0 0 BCVA 0.8 (0.5; 0.8) 0.8 (0.6; 0.8) 0.8 (0.5; 0.8) 0.8 (0.5; 0.88) 0.6 (0.5; 0.8) 0.35 (0.28; 0.42) 0.26 Spherical.equivalent 0.38 (-0.38; 1.38) 0.5 (-0.25; 1.62) 0 (-1.38; 0.88) 0.12 (-0.59; 1.12) 0.38 (-0.50; 1.12) 0.56 (-0.09; 1.22) 0.073 VD A-OCT SUPERFICIAL. PLEXUS CENTRAL 16 (13; 19) 16 (13; 19) 16 (13; 20) 16 (13; 20) 15 (11; 18) 13 (12; 13) 0.389 VD A-OCT SUPERFICIAL. PLEXUS SUPERIOR 46 (43; 48) 46 (43; 48) 46 (43; 48) 45 (41; 47) 44 (40; 47) 42 (41; 43) < 0.001 VD A-OCT SUPERFICIAL. PLEXUS TEMPORAL 46 (44; 49) 47 (45; 49) 46 (45; 49) 45 (43; 48) 44 (40; 46) 41 (39; 43) < 0.001 VD A-OCT SUPERFICIAL. PLEXUSINFERIOR 46 (42; 48) 47 (44; 48) 47 (44; 48) 45 (40; 47) 42 (40; 44) 36 (35; 37) < 0.001 VD A -OCT SUPERFICIAL. PLEXUS NASAL 45 (42; 47) 45 (43; 47) 45 (42; 47) 44 (41; 47) 43 (42; 45) 37 (36; 38) 0.008 Table 2.3 Right eyes. Summary of the characteristics of the left eyes of the patients included in the study. Data are frequencies (%) and medians (IQR). BCVA: Best Corrected Visual Acuity. Cristaline C: Cortical component of the lens. Cristaline N: Nuclear component of the lens. Cristaline P: Posterior component of the lens. CRT - OCT: Central retinal thickness by Optical Coherence Tomography. DME- AI: Diabetic Macular Edema by Artificial Intelligence. DME - OCT: Diabetic Macular Edema by Optical Coherence Tomography. VD A-OCT: Vascular Density by Angio-Optical Coherence Tomography. Diabetic retinopathy AI Overall N = 498* No N = 336 Mild non-proliferative N = 51 Moderate non-proliferative N = 65 Severe non-proliferative N = 43 Proliferative N = 3 P-value CRT - OCT 260 (238; 278) 257 (234; 272) 259 (242; 273) 266 (249; 288) 272 (256; 295) 310 (306; 404) < 0.001 DME - OCT 22 (4.4) 1 (0.3) 1 (2.0) 7 (10.8) 12 (27.9) 1 (33.3) < 0.001 DME – IA 26 (5.2) 0 0 11 (16.9) 14 (32.6) 1 (33.3) < 0.001 Crystalline N 3 (2; 11) 3 (2; 6) 4 (1; 8) 4 (1; 11) 4 (2; 11) 11 (8; 11) 0.138 Crystalline C 0.623 0 465 (93.4) 314 (93.5) 49 (96.1) 58 (89.2) 41 (95.3) 3 (100) 1 21 (4.2) 16 (4.8) 1 (2.0) 3 (4.6) 1 (2.3) 0 2 11 (2.2) 5 (1.5) 1 (2.0) 4 (6.2) 1 (2.3) 0 3 1 (0.2) 1 (0.3) 0 0 0 0 Crystalline P 0.694 0 495 (99.4) 334 (99.4) 51 (100.0) 64 (98.5) 43 (100.0) 3 (100) 1 1 (0.2) 1 (0.3) 0 0 0 0 2 1 (0.2) 0 0 1 (1.5) 0 0 3 1 (0.2) 1 (0.3) 0 0 0 0 BCVA 0.8 (0.6; 0.9) 0.8 (0.6; 0.9) 0.8 (0.6; 1) 0.8 (0.5; 1.) 0.7 (0.5; 0.8) 0.1 (0.08; 0.2) 0.004 Spherical.equivalent 0.50 (-0.25; 1.62) 0.62 (-0.12; 1.75) 0 (-2; 0.88) 0.25 (-0.75; 1) 0.25 (-0.38; 0.88) 0.75 (0.31; 0.88) 0.002 VD A- OCT SUPERFICIAL. PLEXUS CENTRAL 16 (13; 19) 16 (13; 20) 17 (15; 20) 16 (13; 19) 16 (14; 19) 19 (14; 30) 0.688 VD A- OCT SUPERFICIAL. PLEXUS SUPERIOR 46 (43; 48) 47 (44; 48) 47 (44; 50) 45 (42; 48) 43 (42; 46) 43 (38; 44) < 0.001 VD A-OCT SUPERFICIAL. PLEXUS TEMPORAL 46 (44; 49) 47 (45; 49) 47 (45; 49) 46 (43; 48) 45 (42; 47) 46 (43; 49) 0.001 VD A-OCT SUPERFICIAL. PLEXUS INFERIOR 46 (43; 48) 46 (43; 48) 45 (42; 47) 45 (42; 47) 45 (41; 47) 46 (46; 54) 0.041 VD A-OCT SUPERFICIAL. PLEXUS NASAL 45 (42; 47) 45 (43; 47) 45 (42; 47) 44 (40; 46) 42 (40; 44) 45 (44; 51) < 0.001 Table 2.4 Left eyes. Summary of the characteristics of the left eyes of the patients included in the study. Data are frequencies (%) and medians (IQR). BCVA: Best Corrected Visual Acuity. Cristaline C: Cortical component of the lens. Cristaline N: Nuclear component of the lens. Cristaline P: Posterior component of the lens. CRT - OCT: Central retinal thickness by Optical Coherence Tomography. DME - AI: Diabetic Macular Edema by Artificial Intelligence. DME-OCT: Diabetic Macular Edema by Optical Coherence Tomography. VD A-OCT: Vascular Density by Angio-Optical Coherence Tomography. Diabetic retinopathy Diagnosed by OFT Diagnosed by AI P-value Odd-Ratio (95% CI) P-value Odd-Ratio (95% CI) Age, per year < .001 0.976 (0.963; 0.990) - - Years DM, per year < .001 1.151 (1.109; 1.194) < .001 1.120 (1.081; 1.160) Type-1 diabetes mellitus 0.009 3.168 (1.339; 7.494) DME - OCT* < .001 32 (7.26; 142) < .001 62.7 (8.38; 469) CRT - OCT* 0.004 1.006 (1.002; 1.011) 0.047 1.004 (1.000; 1.009) Spherical equivalent* < .001 0.891 (0.832; 0.954) < .001 0.880 (0.823; 0.942) VD_A.OCT.SUPERFICIAL* PLEXUS.TEMPORAL < .001 0.951 (0.925; 0.978) 0.001 0.957 (0.932; 0.983) Table 3. Multivariate logistic regression for diabetic retinopathy. (*) per unit. CRT- OCT: Central retinal thickness by Optical Coherence Tomography. DME - OCT: Diabetic Macular Edema by Optical Coherence Tomography. Diagnosis of DR in at least one eye Binocular sensitivity and specificity (%) AI Yes No bSen [CI - 95%] bSp [CI - 95%] (+) * 189 20 100 [98,1 – 100] 93,5 [90,2 – 96,0] (-) 0 289 Table 4. Sensitiviy and specificity(*) The AI is considered positive (+) when it is positive in at least one eye. Proportion of agreements Observed Expected* kappa (95%CI) Right eye 0.820 0.484 0.650 (0.594; 0.707) Left eye 0.847 0.503 0.693 (0.636; 0.750) Table 5. Proportion of agreements considering the different stages of DR (*) per unit. (*) Under the hypothesis of no agreement. Random agreement hypothesis Proportion of agreements Observed Expected* kappa (95%CI) Right eye 0.966 0.556 0.923 (0.836; 1.010) Left eye 0.978 0.569 0.949 (0.861; 1.036) Table 6. Agreement between binary (Y/N) diagnosis(*). Under the hypothesis of no agreement. Random agreement hypothesis Additional Declarations No competing interests reported. Cite Share Download PDF Status: Under Review Version 1 posted Editorial decision: Revision requested 15 Sep, 2025 Reviews received at journal 15 Sep, 2025 Reviews received at journal 05 Sep, 2025 Reviewers agreed at journal 05 Sep, 2025 Reviewers agreed at journal 01 Sep, 2025 Reviewers invited by journal 11 Aug, 2025 Editor assigned by journal 10 Aug, 2025 Submission checks completed at journal 06 Aug, 2025 First submitted to journal 03 Aug, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7284873","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":500722852,"identity":"f8f0c791-6ce9-418d-abcc-424b91c44ab6","order_by":0,"name":"Isabel Inmaculada Guedes Guedes","email":"data:image/png;base64,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","orcid":"","institution":"University of Las Palmas de Gran Canaria (ULPGC) - Las Palmas de Gran Canaria","correspondingAuthor":true,"prefix":"","firstName":"Isabel","middleName":"Inmaculada Guedes","lastName":"Guedes","suffix":""},{"id":500722854,"identity":"b8067361-673a-40bc-ad0b-e2553aff2a80","order_by":1,"name":"Pedro Saavedra","email":"","orcid":"","institution":"University of Las Palmas de Gran Canaria (ULPGC) - Las Palmas de Gran Canaria","correspondingAuthor":false,"prefix":"","firstName":"Pedro","middleName":"","lastName":"Saavedra","suffix":""},{"id":500722856,"identity":"7658bdeb-58b6-4f83-ac70-97b7f6c416e0","order_by":2,"name":"Francisco Cabrera López","email":"","orcid":"","institution":"University of Las Palmas de Gran Canaria (ULPGC) - Las Palmas de Gran Canaria","correspondingAuthor":false,"prefix":"","firstName":"Francisco","middleName":"Cabrera","lastName":"López","suffix":""},{"id":500722858,"identity":"4dc2c26d-408c-4d66-80c8-f84e8bf258b4","order_by":3,"name":"Ángel Ramos de Miguel","email":"","orcid":"","institution":"Doctor of Medicine. Otolaryngology. - Complejo Hospitalario Universitario Insular Materno Infantil de Gran Canaria. - University of Las Palmas de Gran Canaria (ULPGC) - Las Palmas de Gran Canaria","correspondingAuthor":false,"prefix":"","firstName":"Ángel","middleName":"Ramos","lastName":"de Miguel","suffix":""},{"id":500722860,"identity":"2ae507d2-f5f9-4658-a6b0-768d51554073","order_by":4,"name":"Ángel Ramos Miguel","email":"","orcid":"","institution":"University of Las Palmas de Gran Canaria (ULPGC) - Las Palmas de Gran Canaria","correspondingAuthor":false,"prefix":"","firstName":"Ángel","middleName":"Ramos","lastName":"Miguel","suffix":""},{"id":500722862,"identity":"18e577e2-e425-4c15-8544-591e17a3385c","order_by":5,"name":"Ayoze González Hernández","email":"","orcid":"","institution":"Universidad Fernando Pessoa Canarias","correspondingAuthor":false,"prefix":"","firstName":"Ayoze","middleName":"González","lastName":"Hernández","suffix":""}],"badges":[],"createdAt":"2025-08-03 17:53:11","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7284873/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7284873/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":89276360,"identity":"911d7e64-f957-4ae7-b7e3-38db609a8d53","added_by":"auto","created_at":"2025-08-18 09:31:58","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":30328,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003e\u003cstrong\u003eScatter plots.\u003c/strong\u003e\u003c/em\u003e Agreement between both diagnoses. The dots have been fluctuated to improve the impact of the frequencies.\u003c/p\u003e","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7284873/v1/f0dc06af17377ba2bb78cdfe.png"},{"id":89278457,"identity":"a74ced8f-db61-4510-8ee3-0fd2cc7749cf","added_by":"auto","created_at":"2025-08-18 09:55:59","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1469957,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7284873/v1/4a1ca1b4-d61a-41e3-b4d4-43e95c579def.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"\u003cp\u003eEvaluation of the degree of agreement in the diagnosis of Diabetic Retinopathy between Ophthalmologists and EyeArt\u003csup\u003e®\u003c/sup\u003e\u003c/p\u003e","fulltext":[{"header":"1.\tINTRODUCTION","content":"\u003cp\u003eDiabetic retinopathy (DR) is a prevalent microvascular complication of Diabetes Mellitus (DM), affecting an estimated one-third of all individuals with diabetes globally (1)(2)(3). As the prevalence of diabetes continues to increase worldwide, the burden of DR has escalated, making it the leading cause of blindness among working-age adults (4). In the United States alone, over 30 million adults live with DM, and studies indicate that more than 10% of these patients are at risk of vision-threatening DR (vtDR) if not monitored and treated promptly\u0026nbsp;(1)(5). Untreated DR progresses asymptomatically, advancing through stages marked by retinal damage, which can lead to irreversible vision loss if early signs are not detected and managed (6)(7).\u003c/p\u003e\n\u003cp\u003eThe clinical importance of regular DR screening is underscored by the potential for early intervention to significantly reduce the progression of sight-threatening stages (8)(9). Effective treatments, such as laser photocoagulation and intravitreal injections, have proven successful in halting or slowing the progression of DR (10)(11)(12); however, their efficacy is maximized when applied at the onset of the disease (13). Consequently, clinical guidelines recommend annual eye examinations for patients with DM to detect DR early (14)(15). Yet, adherence to these recommendations is limited; in the United States, less than 60% of individuals with DM receive the recommended annual eye screening, with even lower rates in low-resource settings and among minority and underserved populations (16).\u003c/p\u003e\n\u003cp\u003eTraditional DR screening programs rely on human graders to manually assess digital fundus photographs (17)(18). While effective, this approach is labor-intensive, expensive, and dependent on a limited pool of trained ophthalmic professionals (19). In regions with high demand and limited resources, such as rural areas and developing countries, barriers such as lack of access to eye care specialists, lengthy wait times, and high costs further limit the reach and efficacy of DR screening efforts (20)(21). Additionally, in low- and middle-income countries, the shortage of ophthalmologists and the logistical challenges of reaching remote populations exacerbate the gaps in screening and early detection (22)(23).\u003c/p\u003e\n\u003cp\u003eIn response to these limitations, Artificial Intelligence (AI) has emerged as a promising tool to address the challenges of DR screening (24). Recent advancements in AI-driven image analysis, powered by deep learning (DL) algorithms, allow automated detection of DR from retinal fundus photographs. AI-based screening systems, such as EyeArt\u003csup\u003e\u0026reg;\u003c/sup\u003e, IDx-DR\u003csup\u003e\u0026reg;\u003c/sup\u003e, and Retmarker\u003csup\u003e\u0026reg;\u003c/sup\u003e, have shown comparable accuracy to human graders in detecting referral-warranted DR (rwDR) and vtDR, offering high sensitivity and specificity rates that make them viable alternatives to traditional manual grading (25)(26)(27).\u0026nbsp;These systems can be deployed in primary care and telemedicine settings, making DR screening more accessible and potentially reducing the overall healthcare costs associated with vision loss prevention (28)(29).\u003c/p\u003e\n\u003cp\u003eAmong the available AI systems, EyeArt\u0026reg; was selected for this study due to practical considerations in our local healthcare context. Specifically, it was the system to which we had licensed access, allowing us to integrate it seamlessly into the existing Retisalud screening program infrastructure. Moreover, EyeArt\u0026reg; has demonstrated strong diagnostic performance in large population-based studies, further supporting its inclusion in this comparative analysis (30)(31).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003eOne of the major advantages of AI-based DR screening is the ability to implement point-of-care testing in primary and community healthcare settings without the need for specialized equipment or training in ophthalmology (32)(33). \u0026nbsp;For example, studies have demonstrated that non-mydriatic cameras paired with AI algorithms can be operated effectively by general healthcare staff, facilitating DR screening in primary care practices and improving adherence to screening guidelines among underserved populations (34). Furthermore, the integration of AI into telemedicine frameworks has enabled remote DR screening in rural and isolated areas, providing timely detection and referral recommendations without requiring patients to travel long distances to access specialized eye care (35)(36).\u003c/p\u003e\n\u003cp\u003eEconomic evaluations of AI-based DR screening systems suggest that these tools can be cost-effective, especially when implemented at a large scale in settings with high diabetes prevalence (37). Studies show that by reducing the need for manual grading and minimizing unnecessary referrals, AI-driven DR screening can lower healthcare costs while maintaining high diagnostic accuracy (38)(39)(40)(41)(42). Moreover, AI systems have the potential to alleviate the workload on human graders, enabling ophthalmologists to focus on cases that require immediate intervention and enhancing the overall efficiency of DR screening programs (43)(44).\u003c/p\u003e\n\u003cp\u003eThe objective of this study is to evaluate the diagnostic performance of the EyeArt\u0026reg; AI system in detecting DR, by analyzing its concordance with ophthalmologist-determined diagnoses within a large-scale regional screening program. The study aims to determine the degree of agreement between both diagnostic approaches and assess the potential utility of EyeArt\u0026reg; as a reliable tool in early DR detection.\u003c/p\u003e"},{"header":"2.\tMETHODS","content":"\u003cp\u003e\u003cstrong\u003eDesign\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis is a cross-sectional study in which 499 diabetic patients over 18 years old were included. Patients were randomly recruited from the diabetic retinopathy screening program conducted in the Canary Islands (\u003cem\u003eRetisalud\u003c/em\u003e) between June 1, 2023, and September 22, 2023. Once selected, they were scheduled for an appointment at the Ophthalmology Department of a tertiary-level hospital. During this consultation, patients were informed about the study, including the details of participation. All participants received detailed information about the study objectives and procedures and signed an informed consent form (Annex I) prior to enrollment. The study was reviewed and approved by the Ethics Committee of Las Palmas CEI/CEIm, under reference number 2022-376-1. All procedures complied with the Declaration of Helsinki and current data protection regulations.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eIn the recruitment consultation, in addition to demographic data such as age and sex, information was collected regarding the type of diabetes mellitus (type 1 or type 2) and associated conditions that increase cardiovascular risk. These conditions included hypertension, dyslipidemia, cardiac pathology, stroke, obstructive sleep apnea s\u0026iacute;ndrome, and renal pathology.\u003c/p\u003e\n\u003cp\u003eOphthalmological examination, included best-corrected visual acuity (BCVA), spherical equivalent (SE), crystalline lens status (assessing nuclear, cortical, and posterior subcapsular components), central retinal thickness measured by Optical Coherence Tomography (OCT) and vascular plexus density measured by angio-OCT. Following the examination and fundus evaluation, the degree of diabetic retinopathy was determined by ophthalmologist based on the International Clinical Diabetic Retinopathy (ICDR) scale (45).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eEligible participants were screened using non-mydriatic fundus photographs taken with a TRC-NW400 (Topcon\u003csup\u003e\u0026reg;\u0026nbsp;\u003c/sup\u003eCorporation, Japan). The resultant images were subsequently analyzed using EyeArt 2.1.0 automated DR screening software (Eyenuk, Inc., Woodland Hills, CA). \u0026nbsp;Images were evaluated for quality prior to analysis. Photographs with insufficient illumination, focus issues, or artifacts that impeded retinal evaluation were excluded from the final analysis. One participant was excluded due to poor image quality caused by advanced lens opacity (cataract), as the AI system could not process the data.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe EyeArt\u003csup\u003e\u0026reg;\u003c/sup\u003e software generated a report determining the degree of DR and provided a specific score based on the severity of the condition. The characteristics of the sample included in the study are presented in \u003cem\u003eTable 1\u003c/em\u003e.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSample size and power calculation\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe sample size was not determined based on a prior power analysis, but rather by operational limitations. A total of 500 patient licenses were made available by Eyenuk\u0026reg; (EyeArt\u0026reg;) for this research project, which defined the upper limit of included participants. Despite the absence of a formal power calculation, the sample of 499 patients provides a robust dataset for statistical modeling and concordance analysis.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eStatistical analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eUnivariate analysis\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eCategorical variables are expressed as frequencies and percentages and continuous as mean and standard deviation (SD) when data followed a normal distribution, or as median and interquartile range (IQR = 25\u003csup\u003eth\u003c/sup\u003e \u0026ndash; 75\u003csup\u003eth\u003c/sup\u003e percentile) when distribution departed from normality. The percentages were compared, as appropriate, using the Chi-square (\u003cimg width=\"16\" height=\"29\" src=\"data:image/png;base64,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\" alt=\"image\"\u003e) test or the exact Fisher test, the means by the t-test, and the medians by the Wilcoxon test for independent data. In order to identify the factors that maintain independent association with the mutation, a multivariate logistic regression analysis was performed.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eLogistic mixed models for diabetic retinopathy\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eTo identify factors associated with DR for each of the diagnoses (OFT) and (AI) we used mixed-effects logistic regression models (observations for each eye are nested within each patient) (46). Model estimation is performed using the L1 penalty term, which imposes variable selection and shrinkage simultaneously. For the selected factors, the corresponding models were summed in p-values and odd-ratios, which were estimated using 95% confidence intervals. This analysis was carried out using the R glmmLasso statistical package.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eKappa statistic for agreement between diagnoses\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eIn order to assess the agreement between the two diagnoses by each eye, we estimated Cohen\u0026rsquo;s kappa (47). The kappa statistis is defined as:\u003c/p\u003e\n\u003cp\u003e\u003cimg width=\"71\" height=\"37\" src=\"data:image/png;base64,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\" alt=\"image\"\u003e\u003c/p\u003e\n\u003cp\u003eHere,\u0026nbsp;\u003cimg width=\"13\" height=\"29\" src=\"data:image/png;base64,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\" alt=\"image\"\u003e\u0026nbsp;and\u0026nbsp;\u003cimg width=\"13\" height=\"29\" src=\"data:image/png;base64,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\" alt=\"image\"\u003e\u0026nbsp;denote the proportions of observed agreements between the two diagnoses and those expected under the random agreement hypothesis.\u003c/p\u003e\n\u003cp\u003eCohen suggested (48) the Kappa result be interpreted as follows: values \u0026le; 0 as indicating no agreement and 0.01\u0026ndash;0.20 as none to slight, 0.21\u0026ndash;0.40 as fair, 0.41\u0026ndash; 0.60 as moderate, 0.61 0.80 as substantial, and 0.81\u0026ndash;1.00 as almost perfect agreement.\u003c/p\u003e\n\u003cp\u003eStatistical significance was set at\u0026nbsp;\u003cimg width=\"56\" height=\"19\" src=\"data:image/png;base64,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\" alt=\"image\"\u003e. Data were analyzed using the R package, version 4.2.1.\u003c/p\u003e"},{"header":"3.\tRESULTS","content":"\u003cp\u003eFive hundred adults with diabetes consented to participate in this prospective study. One participant was excluded from the an\u0026aacute;lisis because the fundus images obtained were of poor quality due to advanced cataracts and could not be analyzed by the AI program. Four hundred and ninety-nine adults with diabetes were ultimately included in the study.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eTable 1\u003c/em\u003e summarizes the patients characteristics, overall and DR (unilateral and bilateral). The median age was 65.1\u0026plusmn; 11.1.\u0026nbsp;53.5% of the patients were male, while 46.5% were female. The demographic analysis of the 499 patients included in the study revealed no statistically significant differences concerning age, gender, type of DM, or the coexistence of other conditions that increase cardiovascular risk, such as hypertension, dyslipidemia, Obstructive Sleep Apnea Sindrom (OSAS), Chronic Obstructive Pulmonary Disease (COPD), asthma, heart failure, acute myocardial infarction, among others.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003e[Table 1.\u0026nbsp;\u003c/em\u003e\u003c/strong\u003e\u003cstrong\u003e\u003cem\u003ePatient characteristics: overall and by DR]\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eTable 2\u003c/em\u003e shows the logistic mixed models for DR (presence/absence) diagnosed by Ophalmologist (OFT) and Artificial Intelilgence (AI) respectively. The table 2.1 presents the clinical characteristics of right eyes of the sample and ophthalmological parameters obtained through OCT and other evaluation methods in the study sample, based on the degree of DR determined by the ophthalmologist. DME-OCT (Diabetic Macular Edema by OCT) showed statistically significant differences between groups (p \u0026lt; 0.001), indicating a higher presence of macular edema in more advanced stages of DR. Cortical Lens Status (Crystalline.C) and posterior subcapsular component of the lens (Crystalline.P) showed no significant differences between groups (p = 0.272 and p = 0.189 respectively), with most patients exhibiting no cortical and posterior subcapsular opacities. Nuclear Lens Status (Crystalline.N) showed statistically significant differences (p = 0.02), reflecting increased nuclear lens opacification in more advanced stages of the disease. BCVA demonstrated a significant difference between groups (p = 0.023), suggesting a reduction in visual acuity in patients with more severe DR. Superficial Vascular Plexus Density (VD A-OCT) in the temporal and inferior regions exhibited significant differences (p \u0026lt; 0.001), indicating reduced vascular density associated with greater severity of DR. Other parameters, such as central and nasal superficial plexus density, did not show significant differences (p = 0.189 and p = 0.245, respectively). These results suggest a correlation between the severity of DR and structural and functional changes in the retina, particularly evident in visual acuity and temporal and inferior plexus vascular density.\u003c/p\u003e\n\u003ch3\u003e\u003cstrong\u003e\u003cem\u003e[Table 2.1. Right eyes]\u003c/em\u003e\u003c/strong\u003e\u0026nbsp;\u003c/h3\u003e\n\u003cp\u003eTable 2.2 summarizes the characteristics of the left eyes of the patients included in the study, focusing on key clinical and ophthalmological parameters based on the degree of DR determined by the ophthalmologist. A statistically significant increase in the prevalence of macular edema (DME-OCT) was observed with disease severity (p \u0026lt; 0.001). Superficial vascular density (VD A-OCT) measurements revealed significant reductions in the superior plexus (p \u0026lt; 0.001), temporal plexus (p = 0.011) and nasal plexus (p = 0.001). Additionally, CRT - OCT (Central Retinal Thickness determined by Optical Coherence Tomography) scores varied significantly across the stages of retinopathy, highlighting the systemic impact of DR on ocular health (p\u0026lt;0.001). These results reinforce the importance of early detection and monitoring of macular edema, lens opacities, and vascular density changes in patients with DR.\u0026nbsp;\u003c/p\u003e\n\u003ch3\u003e\u003cstrong\u003e\u003cem\u003e[Table 2.2 Left eyes]\u003c/em\u003e\u003c/strong\u003e\u003c/h3\u003e\n\u003cp\u003eTable 2.3 provides a summary of the clinical and ophthalmological characteristics of the right eyes of the patients included in this study, based on the degree of DR determined by the AI. The results reveal significant differences in several parameters according to the severity of DR, while other variables remained unchanged. Regarding DME (DME - OCT), a statistically significant difference was observed between groups (p \u0026lt; 0.001), with a higher prevalence of macular edema as the severity of DR increased. The highest prevalence, 17.8%, was found in severe non-proliferative cases. In contrast, cortical lens opacity (Crystalline C) did not show significant differences between groups (p = 0.872). Most patients (93.4%) exhibited no cortical opacities, while a minority presented mild or moderate changes. Nuclear lens opacity (Crystalline N) exhibited a statistically significant increase in patients with greater disease severity (p = 0.024). Superficial vascular density (VD A-OCT) showed a significant reduction in the temporal, superior, inferior and nasal regions as the severity of DR increases. However, no significant differences were identified in the central region (p = 0.389)\u003c/p\u003e\n\u003ch3\u003e\u003cstrong\u003e\u003cem\u003e[Table 2.3 Right eyes]\u003c/em\u003e\u003c/strong\u003e\u003c/h3\u003e\n\u003cp\u003eThe table 2.4 provides a detailed breakdown of the clinical and ophthalmological characteristics of the patients\u0026rsquo; eyes based on the degree of DR determined by the AI. Statistically significant differences were observed for DME (DME - OCT) across groups (p \u0026lt; 0.001), with a progressive increase in edema correlating strongly with the severity of DR. For lens opacities, no significant differences were found for cortical, nuclear and posterior opacities. BCVA declined significantly as DR severity increased (p = 0.004), highlighting the functional impact of disease progression. Regarding vascular density in the superficial plexus (VD A-OCT), the temporal, nasal, inferior a superior regions exhibited a significant reduction in density with disease progression, indicating its association with advanced retinopathy, while central region showed no significant changes (p = 0.688)), suggesting these region is less affected in early to moderate stages of the disease.\u003c/p\u003e\n\u003cp\u003eRegarding the distribution of DR severity, ophthalmologist diagnoses classified 68.5% of eyes as having no DR, 19.4% as mild non-proliferative DR, 9.0% as moderate, and 3.0% as severe non-proliferative DR. In contrast, AI classified 65.1% as no DR, 11.4% as mild, 14.0% as moderate, and 9.0% as severe or proliferative DR, highlighting a shift toward higher severity categories in AI results. This disparity further underscores the tendency of the AI system to overclassify DR severity compared to clinical judgment.\u0026nbsp;\u003c/p\u003e\n\u003ch3\u003e\u003cstrong\u003e\u003cem\u003e[Table 2.4 Left eyes]\u003c/em\u003e\u003c/strong\u003e\u003c/h3\u003e\n\u003cp\u003eA stratified analysis of AI performance across different DR severity levels revealed that EyeArt\u003csup\u003e\u0026reg;\u003c/sup\u003e tends to overestimate the disease severity, as demonstrated by the distribution of cases along the diagonal and above-diagonal in the scatter plot (Figure 1). The proportion of moderate and severe classifications made by AI was higher than those assigned by ophthalmologists, suggesting a shift toward conservative (over-referral) predictions. This trend was more pronounced in the moderate non-proliferative and severe non-proliferative categories, whereas underestimation of severity was not observed. These findings support the earlier observation that the binary classification (presence/absence of DR) yielded better agreement metrics compared to the multi-level severity classification, as evidenced by the lower kappa coefficients in the latter.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTable 3 presents the results of a multivariable logistic regression model evaluating factors associated with DR. Age (per additional year) showed an inverse association with DR, with an odds ratio (OR) of 0.976 (95% CI: 0.963\u0026ndash;0.990; p \u0026lt; 0.001). This indicates that for each additional year of life, the odds of developing retinopathy decrease by approximately 2.4%, assuming that all other variables remain constant. This finding may reflect that disease progression is influenced by factors related to the duration of diabetes in individuals diagnosed at younger ages. In other words, individuals diagnosed at older ages have a lower risk of developing complications associated with DR because they have fewer years to develop them.\u003c/p\u003e\n\u003cp\u003eThe duration of diabetes showed a positive and significant association with DR. The OR was 1.151 (95% CI: 1.109\u0026ndash;1.194; p \u0026lt; 0.001). This indicates that for each additional year since diabetes diagnosis, the odds of developing DR increase by 15.1%, while keeping other variables in the model constant. This result highlights the cumulative impact of diabetes on the development of microvascular complications over time. Another statistically significant determinant was the presence of DME determined by OCT, as its presence increased the probability of developing DR by 32 times (p \u0026lt; 0.001), assuming all other variables remain constant. Additionally, spherical equivalent demonstrated a statistically significant association with the development of DR (p \u0026lt; 0.001), acting as a protective factor in this case. Specifically, for each unit increase in spherical equivalent, the probability of developing diabetic retinopathy decreased by 10.9%.\u003c/p\u003e\n\u003ch3\u003e\u003cstrong\u003e\u003cem\u003e[Table 3. Multivariate logistic regression for diabetic retinopathy. (*) per unit]\u003c/em\u003e\u003c/strong\u003e\u003c/h3\u003e\n\u003cp\u003e\u003cem\u003eTable 4\u003c/em\u003e presents the estimates of bSen (95% CI) and bSp (95% CI). According to Perera \u003cem\u003eet al(49)\u003c/em\u003e, we will conduct the evaluation of AI as a diagnostic marker for DR using the concepts of binocular sensitivity (bSen) and binocular specificity (bSp). bSen is the probability of obtaining at least one positive AI diagnosis in one eye, given that there is a diagnosis of DR by an ophthalmologist (OFT) in at least one eye. bSp is the probability of obtaining a negative AI diagnosis in both eyes, given that there is a negative DR diagnosis by an ophthalmologist in both eyes.\u003c/p\u003e\n\u003cp\u003eThe results presented in \u003cem\u003eTable 6\u003c/em\u003e indicate that the AI system demonstrated excellent diagnostic performance in detecting DR. The estimated binocular sensitivity (bSen) was 100% (95% CI: 98.1\u0026ndash;100), meaning that the AI correctly identified all cases of DR in at least one eye, with no false negatives. This exceptionally high sensitivity suggests that the system is highly effective for screening purposes. The estimated binocular specificity (bSp) was 93.5% (95% CI: 90.2\u0026ndash;96.0), indicating that the AI correctly classified the absence of DR in both eyes in 93.5% of negative cases, with a false positive rate of 6.5%.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003e[Table 4. Sensitiviy and specificity(*)\u003c/em\u003e\u003c/strong\u003e ]\u003c/p\u003e\n\u003cp\u003eThe kappa statistics estimations, including their 95% CI, are detailed in \u003cem\u003eTable 5\u003c/em\u003e, which presents the observed and expected agreement proportions alongside kappa coefficients in the concordance analysis for both eyes. The degree of agreement was assessed considering the different stages of DR. For the right eye, the kappa coefficient was 0.650 (95% CI: 0.594\u0026ndash;0.707), with an observed agreement proportion of 0.820 and an expected agreement of 0.484 under the random agreement hypothesis. For the left eye, the kappa coefficient was 0.693 (95% CI: 0.636\u0026ndash;0.750), with an observed agreement proportion of 0.847 and an expected agreement of 0.503. These results highlight substantial agreement in both eyes.\u003c/p\u003e\n\u003ch3\u003e\u003cstrong\u003e\u003cem\u003e[Table 5. Proportion of agreements considering the different stages of DR\u003c/em\u003e\u003c/strong\u003e\u003cstrong\u003e\u003cem\u003e\u0026nbsp;\u003c/em\u003e\u003c/strong\u003e\u003cstrong\u003e\u003cem\u003e(*) per unit.\u0026nbsp;\u003c/em\u003e\u003c/strong\u003e(*)]\u003c/h3\u003e\n\u003cp\u003e\u003cem\u003eTable 6\u003c/em\u003e summarizes the levels of agreement for a binary (Yes/No) diagnosis, including the observed and expected proportions of agreement, as well as the kappa coefficients with their 95% CI. For the right eye, the observed agreement proportion was 0.966, while the expected agreement under the random hypothesis was 0.556, resulting in a kappa coefficient of 0.923 (95% CI: 0.836\u0026ndash;1.010). For the left eye, the observed agreement proportion was 0.978, with an expected agreement of 0.569, yielding a kappa coefficient of 0.949 (95% CI: 0.861\u0026ndash;1.036). These findings demonstrate almost perfect agreement for both eyes in the binary diagnosis.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e[Table 6. Agreement between binary (Y/N) diagnosis(*).]\u003c/strong\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAlthough the kappa coefficients indicate substantial to almost perfect agreement between AI and ophthalmologist diagnoses\u0026mdash;particularly in binary classification\u0026mdash;it is important to note that some confidence intervals marginally include or exceed the upper bound of 1.0, particularly in \u003cem\u003eTable 6\u003c/em\u003e. This statistical artifact may reflect a possible overestimation of agreement, likely influenced by the high prevalence of negative (No DR) classifications, which increases the observed agreement and may reduce the impact of discrepancies in moderate or severe cases. These wide CI should be interpreted with caution, as they may indicate variability or limitations in the estimation of concordance levels in smaller subgroups\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe \u003cem\u003eFigure 1\u003c/em\u003e consists of two scatter plots illustrating artificial intelligence (AI) and ophthalmologist diagnostic outcomes for right eyes (blue) and left eyes (red).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eX-axis (horizontal)\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003erepresents the diagnosis performed by the ophthalmologist categorized as \u0026ldquo;No,\u0026rdquo; \u0026ldquo;Mild,\u0026rdquo; \u0026ldquo;Moderate,\u0026rdquo; and \u0026ldquo;Severe.\u0026rdquo; Y-axis (vertical)\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003eindicates the AI-generated diagnoses, classified using the same categories: \u0026ldquo;No,\u0026rdquo; \u0026ldquo;Mild,\u0026rdquo; \u0026ldquo;Moderate,\u0026rdquo; \u0026ldquo;Severe,\u0026rdquo; and an additional category, \u0026ldquo;Proliferative.\u0026rdquo; Each data point corresponds to a single AI prediction compared to its ground-truth diagnosis performed by the ophthalmologist. Right eye diagnoses are plotted in blue on the left panel and left eye diagnoses are plotted in red on the right panel. Points aligned along the diagonal (extending from the bottom-left to the top-right) indicate accurate AI predictions, where the AI diagnosis matches the diagnostis performed by the ophtalmologist. \u0026nbsp; Points above the diagonal indicate instances where the AI overestimates the disease severity and points below the diagonal\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003ereflect cases where the AI underestimates the severity of the condition.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe scatter plots for right and left eyes exhibit similar distributions of data points, suggesting consistent diagnostic performance by the AI across both eyes. Both panels demonstrate a higher concentration of points in the \u0026ldquo;No\u0026rdquo; and \u0026ldquo;Mild\u0026rdquo; categories, indicating a predominance of low-severity diagnoses in the dataset.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003e[Figure 1. Scatter plots]\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e"},{"header":"4.\tDISCUSSION","content":"\u003cp\u003eThe detection of DR is a coplex image-interpretation task and a key step in any successfull screening program.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe findings of this study (\u003cem\u003eTables 2\u003c/em\u003e) reinforce the strong association between the progression of DR and alterations in key ocular parameters. Specifically, changes in visual acuity, DME, nuclear lens opacification, and reductions in vascular density exhibited a significant correlation with disease severity. These results provide further insights into the systemic impact of DR on ocular health and highlight the relevance of these parameters as potential indicators of disease progression. Conversely, cortical lens opacity did not show significant differences among the studied groups, suggesting that its progression may not be directly linked to DR severity. These findings underscore the importance of a comprehensive ophthalmological assessment in patients with DR, emphasizing the need for thorough monitoring to detect and manage disease-related complications effectively. These insights highlight the critical role of early detection and continuous monitoring using OCT to assess disease progression. Implementing OCT as a routine tool for evaluating structural and vascular alterations in DR could enhance clinical decision-making, facilitating timely therapeutic interventions aimed at preserving visual function and mitigating disease-associated complications. Overall, these findings contribute to a deeper understanding of the ophthalmic manifestations of DR and reinforce the importance of integrating multimodal imaging techniques into routine clinical practice. Further studies with larger cohorts and longitudinal designs are warranted to validate these associations and explore their potential implications for personalized disease management strategies (38)(39)(50)(30)(31)(51).\u003c/p\u003e\n\u003cp\u003eThe analysis of sensitivity and specificity values highlights the robust diagnostic performance of the AI system in detecting DR. The high sensitivity observed suggests that the system is highly effective in identifying positive cases, reinforcing its utility as a screening tool for early disease detection. Similarly, the specificity is also high, although slightly lower than sensitivity, indicating a minor risk of overdiagnosis due to false-positive results. This could lead to unnecessary referrals for further ophthalmologic evaluation, which, while ensuring comprehensive patient care, may also contribute to an increased burden on healthcare systems. These high sensitivity and specificity values obtained are consistent with the results reported in other similar studies\u0026nbsp;(52)(53)\u0026nbsp;.\u003c/p\u003e\n\u003cp\u003eThe confidence intervals for both sensitivity and specificity demonstrate a high degree of reliability in these estimates, with slightly greater variability observed in specificity. This variability underscores the need for continuous refinement of AI algorithms to further optimize diagnostic accuracy and minimize unnecessary referrals. Nevertheless, these findings support the implementation of AI-assisted diagnostic methods as a valuable tool for DR detection, providing high sensitivity for disease identification and good specificity for distinguishing negative cases. The overall performance of the system reinforces its potential as an effective and reliable screening approach, particularly in large-scale screening programs aimed at early detection and timely intervention.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAlthough EyeArt\u003csup\u003e\u0026reg;\u003c/sup\u003e demonstrated high sensitivity, the moderately lower specificity implies a proportion of false positives, particularly in cases where severity was overestimated. From a clinical perspective, this may lead to increased referrals to ophthalmologists, which, while ensuring patient safety, may overload referral systems, especially in resource-limited settings. Such an increase in workload could affect waiting times and allocation of clinical resources, potentially reducing efficiency in managing truly severe cases. Hence, balancing sensitivity with specificity is essential to optimize screening workflows.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe multivariate logistic regression analysis identified several significant predictors of DR, including age, diabetes duration, DME, and CRT detected by OCT. These findings underscore the multifactorial nature of DR progression and highlight the importance of comprehensive patient management. In particular, glycemic control, blood pressure regulation, and consideration of disease duration should be key components in the clinical approach to diabetic patients to mitigate the risk of retinopathy development and progression.\u003c/p\u003e\n\u003cp\u003eThe kappa coefficient values indicate a moderate level of agreement between diagnoses related to the severity of DR as determined by the ophthalmologist and the EyeArt\u003csup\u003e\u0026reg;\u003c/sup\u003e AI system. However, when the agreement is assessed in the context of screening for a binary diagnosis\u0026mdash;determining the presence or absence of DR without specifying its severity\u0026mdash;the kappa coefficient approaches a value of one, indicating an almost perfect level of concordance between the ophthalmologist and the AI system. This high level of diagnostic accuracy supports the clinical and research applicability of AI-assisted methods, particularly in screening programs designed to differentiate between healthy individuals and those with DR. The strong agreement observed in both eyes further reinforces the reliability of this approach, highlighting its potential as an effective tool for large-scale DR detection and early intervention strategies. These results are consistent with those reported by Karabeg \u003cem\u003eet al\u003c/em\u003e. (52) and Wintergest \u003cem\u003eet al\u003c/em\u003e. (53) However, this study demonstrates significantly higher concordance values compared to those of Wang \u003cem\u003eet al\u003c/em\u003e.(54), Rajalakshmi \u003cem\u003eet al\u003c/em\u003e. (55), Kim \u003cem\u003eet al\u003c/em\u003e. (56), Mokhanshi \u003cem\u003eet al\u003c/em\u003e.(57), and Cicinelli \u003cem\u003eet al\u003c/em\u003e (58). This improvement in concordance is likely attributable to the refinement of the artificial intelligence program and the quality of the images obtained with the device used.\u003c/p\u003e\n\u003cp\u003eIn the scatter plot, it is evident that the AI program tends to overestimate the severity of DR as determined by the ophthalmologist. Notably, there are no data points below the bisector, indicating the absence of underdiagnosis cases. However, several points are located above the bisector, demonstrating that the EyeArt\u003csup\u003e\u0026reg;\u003c/sup\u003e system has a tendency to overestimate the severity of DR compared to the ophthalmologist\u0026rsquo;s assessment. As previously noted, the degree of agreement, as measured by the Kappa coefficient, was higher (0.923 for right eyes and 0.949 for left eyes) when comparing the binary classification between the ophthalmologist and the AI, i.e., when the diagnosis was limited to determining the presence or absence of DR, which is the primary objective of screening programs. However, when the Kappa coefficient was calculated based on the severity grading of DR, the level of agreement was lower (0.650 for right eyes and 0.693 for left eyes). The scatter plot further confirms that the lower agreement observed in DR grading is due to the AI system\u0026rsquo;s tendency to overestimate disease severity.\u003c/p\u003e\n\u003cp\u003eThis study has several limitations that should be acknowledged. First, a potential selection bias cannot be excluded, as participants were recruited from a single regional screening program and may not fully represent the broader diabetic population. Additionally, the exclusion of poor-quality fundus images\u0026mdash;although necessary for analysis\u0026mdash;may have introduced a spectrum bias, favoring patients with better media clarity and potentially underrepresenting more complex or advanced cases.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAnother limitation is the small number of cases diagnosed as proliferative DR (n=2), which precludes a robust evaluation of the AI system\u0026apos;s performance in this critical category. It is important to note that, thanks to the existing screening programs, it is uncommon to encounter advanced stages such as proliferative DR, which explains the small sample size observed in this group. Future studies with larger samples in this subgroup are needed to validate diagnostic accuracy in more severe stages.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe results of this study reinforce the role of EyeArt\u003csup\u003e\u0026reg;\u003c/sup\u003e as a useful tool in DR screening. However, its integration into real-world healthcare settings must be carefully planned. Strategies such as a two-tiered screening approach, where positive AI cases are reviewed by a human grader before referral, could reduce unnecessary specialist consultations. In primary care or teleophthalmology environments, training non-specialist personnel in image acquisition, coupled with routine quality control protocols, can help ensure that the system functions optimally while maintaining diagnostic reliability.\u0026nbsp;\u003c/p\u003e"},{"header":"5.\tCONCLUSIONS","content":"\u003cp\u003eThe EyeArt\u003csup\u003e\u0026reg;\u003c/sup\u003e AI system has demonstrated high sensitivity and specificity in detecting DR, achieving an almost perfect agreement with ophthalmologists when using a binary diagnosis. However, its tendency to overestimate disease severity suggests the need for algorithmic improvements to enhance classification accuracy.\u003c/p\u003e\n\u003cp\u003eThese findings support the use of AI as a complementary tool in screening programs, enabling early detection and improved access to diagnosis in resource-limited settings, ultimately contributing to a reduction in the global burden of DR. However, its clinical implementation should be accompanied by further validations and refinements in severity classification, ensuring its effectiveness in medical decision-making and optimizing the management of DR patients.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eACKNOWLEDGMENT/DISCLOSURE\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare that they have no conflicts of interest related to this study. No financial support was received from any commercial entity for the conduct of this research or the preparation of this manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDECLARATION OF CONSENT TO PARTTICIPATE\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll procedures performed in this study involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. Written informed consent was obtained from all individual participants included in the study.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFinancial Support:\u003c/strong\u003e None\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConflict of Interest:\u0026nbsp;\u003c/strong\u003eNo conflicting relationship exists for any autor\u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eI.I.G.G., A.G.H. and P.S.S. wrote the main manuscriptP.S.S. Prepared figures and prepared the statistical analysis.A.R.M., A.R.M. and F.C.L contributed to the revision of the manuscriptF.C.L. contacted with EyeNUK to acquire the licenses for the study.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eYau JWY, Rogers SL, Kawasaki R, Lamoureux EL, Kowalski JW, Bek T, et al. Global prevalence and major risk factors of diabetic retinopathy. Diabetes Care. 2012 Mar;35(3):556\u0026ndash;64. \u003c/li\u003e\n\u003cli\u003eOgurtsova K, da Rocha Fernandes JD, Huang Y, Linnenkamp U, Guariguata L, Cho NH, et al. IDF Diabetes Atlas: Global estimates for the prevalence of diabetes for 2015 and 2040. Diabetes Res Clin Pract. 2017 Jun 1;128:40\u0026ndash;50. \u003c/li\u003e\n\u003cli\u003eCho NH, Shaw JE, Karuranga S, Huang Y, da Rocha Fernandes JD, Ohlrogge AW, et al. IDF Diabetes Atlas: Global estimates of diabetes prevalence for 2017 and projections for 2045. Diabetes Res Clin Pract. 2018 Apr 1;138:271\u0026ndash;81. \u003c/li\u003e\n\u003cli\u003eFlaxman SR, Bourne RRA, Resnikoff S, Ackland P, Braithwaite T, Cicinelli M V., et al. Global causes of blindness and distance vision impairment 1990\u0026ndash;2020: a systematic review and meta-analysis. Lancet Glob Health. 2017 Dec 1;5(12):e1221\u0026ndash;34. \u003c/li\u003e\n\u003cli\u003eFox CR, Kronenberg K, Weiskopf ES. Using Telemedicine to Increase Eye Care Screening \u0026amp; Referral for People with Diabetes Article Details. J Pub Health Issue Pract JPHIP [Internet]. 2019;3:137. Available from: http://dx.doi.org/jphip/2019/137\u003c/li\u003e\n\u003cli\u003eGuariguata L, Whiting DR, Hambleton I, Beagley J, Linnenkamp U, Shaw JE. Global estimates of diabetes prevalence for 2013 and projections for 2035. Diabetes Res Clin Pract. 2014;103(2):137\u0026ndash;49. \u003c/li\u003e\n\u003cli\u003eFleming AD, Goatman KA, Philip S, Prescott GJ, Sharp PF, Olson JA. Automated grading for diabetic retinopathy: a large-scale audit using arbitration by clinical experts. British Journal of Ophthalmology [Internet]. 2010;(12):1606. Available from: https://hal.science/hal-00578459v1\u003c/li\u003e\n\u003cli\u003eCorner S, Mohamed Q, Mark Gillies MC, Wong TY. CLINICAL, REVIEW Management of Diabetic Retinopathy A Systematic Review [Internet]. Available from: www.jama.com\u003c/li\u003e\n\u003cli\u003eLin S, Ramulu P, Lamoureux EL, Sabanayagam C. Addressing risk factors, screening, and preventative treatment for diabetic retinopathy in developing countries: a review. Vol. 44, Clinical and Experimental Ophthalmology. Blackwell Publishing; 2016. p. 300\u0026ndash;20. \u003c/li\u003e\n\u003cli\u003eZhang W, Zhao G, Fan W, Zhao T. Panretinal photocoagulation after or prior to intravitreal conbercept injection for diabetic macular edema: a retrospective study. BMC Ophthalmol. 2021 Dec 1;21(1). \u003c/li\u003e\n\u003cli\u003eAiello LP, Gardner TW, King GL, Blankenship G, Cavallerano JD, Ferris FL, et al. Diabetic Retinopatiiy. \u003c/li\u003e\n\u003cli\u003eAljundi W, Suffo S, Munteanu C, Langenbucher A, Seitz B, Abdin AD. Intravitreal Injection for Diabetic Macular Edema as Adjunctive Therapy for Proliferative Diabetic Retinopathy: A Retrospective Study. Clinical Ophthalmology. 2022;16:135\u0026ndash;43. \u003c/li\u003e\n\u003cli\u003eLopez-Ramos A, Gomez-Bastar PA, Lansingh VC, Rodriguez-Gomez JA, Vargas-Fragoso V, Soria-Arellano FA, et al. Rapid assessment of avoidable blindness: Prevalence of blindness, visual impairment and diabetes in nuevo leon, Mexico 2014. Ophthalmic Epidemiol. 2018 Nov 2;25(5\u0026ndash;6):412\u0026ndash;8. \u003c/li\u003e\n\u003cli\u003eMurchison AP, Friedman DS, Gower EW, Haller JA, Lam BL, Lee DJ, et al. A multi-center diabetes eye screening study in community settings: Study design and methodology. Ophthalmic Epidemiol. 2016 Mar 3;23(2):109\u0026ndash;15. \u003c/li\u003e\n\u003cli\u003eBek T, Andersen N, Andresen J, Grauslund J, Hajari J, Schmidt Laugesen C, et al. Individualised screening for diabetic retinopathy with proliferative retinopathy and macular oedema as separate end points. Acta Ophthalmol. 2024 Sep 1;102(6):627\u0026ndash;34. \u003c/li\u003e\n\u003cli\u003eTing DSW, Cheung GCM, Wong TY. Diabetic retinopathy: global prevalence, major risk factors, screening practices and public health challenges: a review. Vol. 44, Clinical and Experimental Ophthalmology. Blackwell Publishing; 2016. p. 260\u0026ndash;77. \u003c/li\u003e\n\u003cli\u003eTing DSW, Tan GSW. Telemedicine for diabetic retinopathy screening. Vol. 135, JAMA Ophthalmology. American Medical Association; 2017. p. 722\u0026ndash;3. \u003c/li\u003e\n\u003cli\u003eMansberger SL, Sheppler C, Barker G, Gardiner SK, Demirel S, Wooten K, et al. Long-term comparative effectiveness of telemedicine in providing diabetic retinopathy screening examinations: A randomized clinical trial. JAMA Ophthalmol. 2015 May 1;133(5):518\u0026ndash;25. \u003c/li\u003e\n\u003cli\u003eAvidor D, Loewenstein A, Waisbourd M, Nutman A. Cost-effectiveness of diabetic retinopathy screening programs using telemedicine: A systematic review. Vol. 18, Cost Effectiveness and Resource Allocation. BioMed Central Ltd.; 2020. \u003c/li\u003e\n\u003cli\u003eRamasamy K, Mishra C, Kannan NB, Namperumalsamy P, Sen S. Telemedicine in diabetic retinopathy screening in India. Vol. 69, Indian journal of ophthalmology. NLM (Medline); 2021. p. 2977\u0026ndash;86. \u003c/li\u003e\n\u003cli\u003eLi X, Tan TE, Wong TY, Sun X. Diabetic retinopathy in China: Epidemiology, screening and treatment trends\u0026mdash;A review. Vol. 51, Clinical and Experimental Ophthalmology. John Wiley and Sons Inc; 2023. p. 607\u0026ndash;26. \u003c/li\u003e\n\u003cli\u003eZhang X, Beckles GL, Chou CF, Saaddine JB, Wilson MR, Lee PP, et al. Socioeconomic disparity in use of eye care services among US adults with age-related eye diseases: National health interview survey, 2002 and 2008. JAMA Ophthalmol. 2013 Sep;131(9):1198\u0026ndash;206. \u003c/li\u003e\n\u003cli\u003eFathy C, Patel S, Sternberg P, Kohanim S. Disparities in Adherence to Screening Guidelines for Diabetic Retinopathy in the United States: A Comprehensive Review and Guide for Future Directions. Vol. 31, Seminars in Ophthalmology. Taylor and Francis Ltd; 2016. p. 364\u0026ndash;77. \u003c/li\u003e\n\u003cli\u003eBellemo V, Lim G, Rim TH, Tan GSW, Cheung CY, Sadda SV, et al. Artificial Intelligence Screening for Diabetic Retinopathy: the Real-World Emerging Application. Vol. 19, Current Diabetes Reports. Current Medicine Group LLC 1; 2019. \u003c/li\u003e\n\u003cli\u003eBellemo V, Lim G, Rim TH, Tan GSW, Cheung CY, Sadda SV, et al. Artificial Intelligence Screening for Diabetic Retinopathy: the Real-World Emerging Application. Vol. 19, Current Diabetes Reports. Current Medicine Group LLC 1; 2019. \u003c/li\u003e\n\u003cli\u003eAbr\u0026agrave;moff MD, Roehrenbeck C, Trujillo S, Goldstein J, Graves AS, Repka MX, et al. A reimbursement framework for artificial intelligence in healthcare. Vol. 5, npj Digital Medicine. Nature Research; 2022. \u003c/li\u003e\n\u003cli\u003eN\u0026oslash;rgaard MF, Grauslund J. Automated screening for diabetic retinopathy - A systematic review. Ophthalmic Res. 2018 Jun 1;60(1):9\u0026ndash;17. \u003c/li\u003e\n\u003cli\u003eKavitha DR, Student UG. DETECTION OF DIABETIC RETINOPATHY USING RETINAL FUNDUS PHOTOGRAPHS BASED ON DEEP LEARNING ALGORITHM. Turkish Journal of Physiotherapy and Rehabilitation [Internet]. 32(2). Available from: www.turkjphysiotherrehabil.org\u003c/li\u003e\n\u003cli\u003eWongchaisuwat N, Trinavarat A, Rodanant N, Thoongsuwan S, Phasukkijwatana N, Prakhunhungsit S, et al. In-person verification of deep learning algorithm for diabetic retinopathy screening using different techniques across fundus image devices. Transl Vis Sci Technol. 2021 Nov 1;10(13). \u003c/li\u003e\n\u003cli\u003eBhaskaranand M, Ramachandra C, Bhat S, Cuadros J, Nittala MG, Sadda SR, et al. The value of automated diabetic retinopathy screening with the EyeArt system: A study of more than 100,000 consecutive encounters from people with diabetes. Diabetes Technol Ther. 2019 Nov 1;21(11):635\u0026ndash;43. \u003c/li\u003e\n\u003cli\u003eLim JI, Regillo CD, Sadda SVR, Ipp E, Bhaskaranand M, Ramachandra C, et al. Artificial Intelligence Detection of Diabetic Retinopathy: Subgroup Comparison of the EyeArt System with Ophthalmologists\u0026rsquo; Dilated Examinations. Ophthalmology Science. 2023 Mar 1;3(1). \u003c/li\u003e\n\u003cli\u003eNielsen KB, Lautrup ML, Andersen JKH, Savarimuthu TR, Grauslund J. Deep Learning\u0026ndash;Based Algorithms in Screening of Diabetic Retinopathy: A Systematic Review of Diagnostic Performance. Vol. 3, Ophthalmology Retina. Elsevier Inc; 2019. p. 294\u0026ndash;304. \u003c/li\u003e\n\u003cli\u003eHonasoge A, Nudleman E, Smith M, Rajagopal R. Emerging Insights and Interventions for Diabetic Retinopathy. Vol. 19, Current Diabetes Reports. Current Medicine Group LLC 1; 2019. \u003c/li\u003e\n\u003cli\u003eDe Fauw J, Ledsam JR, Romera-Paredes B, Nikolov S, Tomasev N, Blackwell S, et al. Clinically applicable deep learning for diagnosis and referral in retinal disease. Nat Med. 2018 Sep 1;24(9):1342\u0026ndash;50. \u003c/li\u003e\n\u003cli\u003eHuang X, Wang H, She C, Feng J, Liu X, Hu X, et al. Artificial intelligence promotes the diagnosis and screening of diabetic retinopathy. Vol. 13, Frontiers in Endocrinology. Frontiers Media S.A.; 2022. \u003c/li\u003e\n\u003cli\u003eWewetzer L, Held LA, Goetz K, Steinh\u0026auml;user J. Determinants of the implementation of artificial intelligence-based screening for diabetic retinopathy\u0026mdash;a cross-sectional study with general practitioners in Germany. Digit Health. 2023 Jan 1;9. \u003c/li\u003e\n\u003cli\u003eTufail A, Kapetanakis V V., Salas-Vega S, Egan C, Rudisill C, Owen CG, et al. An observational study to assess if automated diabetic retinopathy image assessment software can replace one or more steps of manual imaging grading and to determine their cost-effectiveness. Health Technol Assess (Rockv). 2016 Dec 1;20(92):1\u0026ndash;72. \u003c/li\u003e\n\u003cli\u003eHeydon P, Egan C, Bolter L, Chambers R, Anderson J, Aldington S, et al. Prospective evaluation of an artificial intelligence-enabled algorithm for automated diabetic retinopathy screening of 30 000 patients. Vol. 105, British Journal of Ophthalmology. BMJ Publishing Group; 2021. p. 723\u0026ndash;8. \u003c/li\u003e\n\u003cli\u003eTufail A, Rudisill C, Egan C, Kapetanakis V V., Salas-Vega S, Owen CG, et al. Automated Diabetic Retinopathy Image Assessment Software: Diagnostic Accuracy and Cost-Effectiveness Compared with Human Graders. Ophthalmology. 2017 Mar 1;124(3):343\u0026ndash;51. \u003c/li\u003e\n\u003cli\u003eN\u0026oslash;rgaard MF, Grauslund J. Automated screening for diabetic retinopathy - A systematic review. Ophthalmic Res. 2018 Jun 1;60(1):9\u0026ndash;17. \u003c/li\u003e\n\u003cli\u003ePoly TN, Islam MM, Walther BA, Lin MC, (Jack) Li YC. Artificial intelligence in diabetic retinopathy: Bibliometric analysis. Comput Methods Programs Biomed. 2023 Apr 1;231. \u003c/li\u003e\n\u003cli\u003eJones S, Edwards RT. Diabetic retinopathy screening: A systematic review of the economic evidence. Vol. 27, Diabetic Medicine. Blackwell Publishing Ltd; 2010. p. 249\u0026ndash;56. \u003c/li\u003e\n\u003cli\u003eWong TY, Bressler NM. Artificial intelligence with deep learning technology looks into diabetic retinopathy screening. Vol. 316, JAMA - Journal of the American Medical Association. American Medical Association; 2016. p. 2366\u0026ndash;7. \u003c/li\u003e\n\u003cli\u003eWalton OB, Garoon RB, Weng CY, Gross J, Young AK, Camero KA, et al. Evaluation of automated teleretinal screening program for diabetic retinopathy. JAMA Ophthalmol. 2016 Feb 1;134(2):204\u0026ndash;9. \u003c/li\u003e\n\u003cli\u003eEsmaeilkhanian H, Liu H, Fasih-Ahmed S, Gnanaraj R, Verma A, Oncel D, et al. The relationship of diabetic retinopathy severity scales with frequency and surface area of diabetic retinopathy lesions. Graefe\u0026rsquo;s Archive for Clinical and Experimental Ophthalmology. 2023 Nov 1;261(11):3165\u0026ndash;76. \u003c/li\u003e\n\u003cli\u003eGroll A, Tutz G. Variable selection for generalized linear mixed models by L1-penalized estimation. Stat Comput. 2014 Mar;24(2):137\u0026ndash;54. \u003c/li\u003e\n\u003cli\u003eFrontmatter. In 2003. \u003c/li\u003e\n\u003cli\u003eBakeman R. KappaAcc: A program for assessing the adequacy of kappa. Behav Res Methods. 2023 Feb 1;55(2):633\u0026ndash;8. \u003c/li\u003e\n\u003cli\u003ePerera Y, Ren M, Punzalan JRB, Rudnisky CJ, de Leon AR. Binocular sensitivity and specificity of screening tests in cross-sectional diagnostic studies of paired organs. Stat Med. 2017 May 20;36(11):1754\u0026ndash;66. \u003c/li\u003e\n\u003cli\u003eBhaskaranand M, Ramachandra C, Bhat S, Cuadros J, Nittala MG, Sadda S, et al. Automated Diabetic Retinopathy Screening and Monitoring Using Retinal Fundus Image Analysis. J Diabetes Sci Technol. 2016 Mar 1;10(2):254\u0026ndash;61. \u003c/li\u003e\n\u003cli\u003eOlvera-Barrios A, Heeren TFC, Balaskas K, Chambers R, Bolter L, Egan C, et al. Diagnostic accuracy of diabetic retinopathy grading by an artificial intelligence-enabled algorithm compared with a human standard for wide-field true-colour confocal scanning and standard digital retinal images. British Journal of Ophthalmology. 2021 Feb 1;105(2):265\u0026ndash;70. \u003c/li\u003e\n\u003cli\u003eKarabeg M, Petrovski G, Hertzberg SNW, Erke MG, Fosmark DS, Russell G, et al. A pilot cost-analysis study comparing AI-based EyeArt\u0026reg; and ophthalmologist assessment of diabetic retinopathy in minority women in Oslo, Norway. Int J Retina Vitreous. 2024 Dec 1;10(1). \u003c/li\u003e\n\u003cli\u003eWintergerst MWM, Bejan V, Hartmann V, Schnorrenberg M, Bleckwenn M, Weckbecker K, et al. Telemedical Diabetic Retinopathy Screening in a Primary Care Setting: Quality of Retinal Photographs and Accuracy of Automated Image Analysis. Ophthalmic Epidemiol. 2022;29(3):286\u0026ndash;95. \u003c/li\u003e\n\u003cli\u003eWang K, Jayadev C, Nittala MG, Velaga SB, Ramachandra CA, Bhaskaranand M, et al. Automated detection of diabetic retinopathy lesions on ultrawidefield pseudocolour images. Acta Ophthalmol. 2018 Mar 1;96(2):e168\u0026ndash;73. \u003c/li\u003e\n\u003cli\u003eRajalakshmi R, Subashini R, Anjana RM, Mohan V. Automated diabetic retinopathy detection in smartphone-based fundus photography using artificial intelligence. Eye (Basingstoke). 2018 Jun 1;32(6):1138\u0026ndash;44. \u003c/li\u003e\n\u003cli\u003eKim TN, Aaberg MT, Li P, Davila JR, Bhaskaranand M, Bhat S, et al. Comparison of automated and expert human grading of diabetic retinopathy using smartphone-based retinal photography. Eye (Basingstoke). 2021 Jan 1;35(1):334\u0026ndash;42. \u003c/li\u003e\n\u003cli\u003eMokhashi N, Grachevskaya J, Cheng L, Yu D, Lu X, Zhang Y, et al. A Comparison of Artificial Intelligence and Human Diabetic Retinal Image Interpretation in an Urban Health System. J Diabetes Sci Technol. 2022 Jul 1;16(4):1003\u0026ndash;7. \u003c/li\u003e\n\u003cli\u003eCicinelli MV, Gravina S, Rutigliani C, Checchin L, La Franca L, Lattanzio R, et al. Assessing Diabetic Retinopathy Staging With AI: A Comparative Analysis Between Pseudocolor and LED Imaging. Transl Vis Sci Technol. 2024 Mar 1;13(3). \u003c/li\u003e\n\u003c/ol\u003e"},{"header":"Tables","content":"\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"100%\"\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"bottom\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"3\" valign=\"bottom\"\u003e\u003cstrong\u003e\u003cem\u003eDiabetic retinopathy\u003c/em\u003e\u003c/strong\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"bottom\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"bottom\"\u003eOverall\u0026nbsp;\u003cbr\u003eN = 499\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"bottom\"\u003eNo\u0026nbsp;\u003cbr\u003eN = 310\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"bottom\"\u003eUnilateral\u0026nbsp;\u003cbr\u003eN = 70\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"bottom\"\u003eBilateral\u0026nbsp;\u003cbr\u003eN = 119\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"bottom\"\u003eP-value\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eAge (years)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e65.1 \u0026plusmn; 11.1\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e65.8 \u0026plusmn; 10.6\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e65.5 \u0026plusmn; 11.6\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e63.0 \u0026plusmn; 11.7\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.062\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eSex male\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e267 (53.5)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e160 (51.6)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e37 (52.9)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e70 (58.8)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.404\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eArterial hypertension\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e368 (73.8)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e228 (73.5)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e53 (75.7)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e87 (73.1)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.918\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eDiabetes mellitus\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eType 1\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e14 (2.8)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e5 (1.6)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e2 (2.9)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e7 (5.9)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.055\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eType 2\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e484 (97.0)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e304 (98.1)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e68 (97.1)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e112 (94.1)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.081\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eDyslipidemia\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e237 (47.5)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e147 (47.4)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e37 (52.9)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e53 (44.5)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.542\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eAsthma\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e8 (1.6)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e6 (1.9)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e1 (1.4)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e1 (0.8)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.874\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eAIM\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e16 (3.2)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e11 (3.5)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e1 (1.4)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e4 (3.4)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.869\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eCOPD\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e25 (5.0)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e15 (4.8)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e3 (4.3)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e7 (5.9)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.834\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eHF\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e4 (0.8)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e3 (1.0)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e1 (0.8)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e1\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eCKD\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e13 (2.6)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e10 (3.2)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e3 (2.5)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.427\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eOSAS\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e6 (1.2)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e5 (1.6)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e1 (0.8)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.849\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eYears of diabetes mellitus\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e8 (3; 10)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e6 (2; 9)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e9 (5; 10)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e9 (6; 12)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026lt; 0.001\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eTable 1.\u0026nbsp;\u003c/em\u003e\u003c/strong\u003e\u003cstrong\u003e\u003cem\u003ePatient characteristics: overall and by DR.\u003c/em\u003e\u003c/strong\u003e\u0026nbsp; \u0026nbsp;Data are means \u0026plusmn; SD, frequencies (%) and medians (IQR). AIM: Acute Myocardial Infarction. COPD: Chronic Obstructive Pulmonary Disease. HF: Heart Failure. CKD: Chronic Kidney Disease. OSAS: Obstructive Sleep Apnea Syndrome.\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"101%\"\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"bottom\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"4\" valign=\"bottom\"\u003e\u003cstrong\u003e\u003cem\u003eDiabetic retinopathy diagnosed by OFT\u003c/em\u003e\u003c/strong\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"bottom\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"bottom\"\u003eOverall\u0026nbsp;\u003cbr\u003eN = 499\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"bottom\"\u003eNo\u0026nbsp;\u003cbr\u003eN = 342\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"bottom\"\u003eMild non-proliferative\u003cbr\u003eN = 97\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"bottom\"\u003eModerate non-proliferative\u003cbr\u003eN = 45\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"bottom\"\u003eSevere non-proliferative\u003cbr\u003eN = 15\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"bottom\"\u003eP-value\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eCRT - OCT\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e254\u0026nbsp;\u003cbr\u003e(235; 275)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e252\u0026nbsp;\u003cbr\u003e(234; 273)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e254\u0026nbsp;\u003cbr\u003e(235; 270)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e269\u0026nbsp;\u003cbr\u003e(253; 292)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e292\u0026nbsp;\u003cbr\u003e(266; 298)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026lt; 0.001\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eDME - OCT\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e16 (3.2)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e5 (5.2)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e4 (8.9)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e7 (46.7)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026lt; 0.001\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eDME - AI\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e20 (4.0)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e1 (0.3)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e6 (6.2)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e6 (13.3)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e7 (46.7)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026lt; 0.001\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eCrystalline N\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e4 (2; 11)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e3 (2; 7)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e3 (1; 6)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e4 (3; 11)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e5 (3; 11)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.02\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eCrystalline C\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.272\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e466 (93.4)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e318 (93.0)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e94 (96.9)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e41 (91.1)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e13 (86.7)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e1\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e23 (4.6)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e17 (5.0)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e1 (1.0)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e3 (6.7)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e2 (13.3)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e2\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e8 (1.6)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e6 (1.8)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e1 (1.0)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e1 (2.2)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e3\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e2 (0.4)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e1 (0.3)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e1 (1.0)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eCrystalline P\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.189\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e492 (98.6)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e337 (98.5)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e95 (97.9)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e45 (100)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e15 (100.0)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e1\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e5 (1.0)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e5 (1.5)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e2\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e2 (0.4)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e2 (2.1)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eBCVA\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.8 (0.5; 0.8)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.8 (0.5; 0.8)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.8 (0.6; 0.9)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.6 (0.4; 0.8)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.7 (0.6; 0.8)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.023\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eSpherical.equivalent\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.38\u0026nbsp;\u003cbr\u003e(-0.38;1.38)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.50\u0026nbsp;\u003cbr\u003e(-0.25;1.50)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.12\u0026nbsp;\u003cbr\u003e(-0.75; 1)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.5\u0026nbsp;\u003cbr\u003e(-0.25; 1.75)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0\u0026nbsp;\u003cbr\u003e(-0.69; 0.38)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.004\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eVD A-OCT SUPERFICIAL.\u003cbr\u003ePLEXUS CENTRAL\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e16 (13; 19)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e16 (13; 19)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e17 (13; 20)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e16 (13; 19)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e13 (11; 16)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.189\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eVD A-OCT SUPERFICIAL.\u003cbr\u003ePLEXUS SUPERIOR\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e46 (43; 48)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e46 (43; 48)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e45 (42; 47)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e44 (42; 46)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e43 (40; 46)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.003\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eVD A-OCT SUPERFICIAL\u003cbr\u003e.PLEXUS.TEMPORAL\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e46 (44; 49)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e47 (45; 49)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e46 (44; 48)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e45 (42; 47)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e41 (40; 46)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026lt; 0.001\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eVD A-OCT SUPERFICIAL.\u003cbr\u003ePLEXUS INFERIOR\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e46 (42; 48)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e47 (44; 48)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e46 (42; 48)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e43 (40; 45)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e41 (40; 44)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026lt; 0.001\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eVD A-OCT SUPERFICIAL.\u003cbr\u003ePLEXUS.NASAL\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e45 (42; 47)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e45 (43; 47)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e44 (42; 47)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e45 (42; 47)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e43 (42; 45)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.245\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eTable 2.1. Right eyes.\u003c/em\u003e\u003c/strong\u003e Summary of the characteristics of the right eyes of the patients included in the study. Data are frequencies (%) and medians (IQR). BCVA: Best Corrected Visual Acuity. Cristaline C: Cortical component of the lens. Cristaline N: Nuclear component of the lens. Cristaline P: Posterior component of the lens. CRT OCT: Central retinal thickness by Optical Coherence Tomography. DME- AI: Diabetic Macular Edema by Artificial Intelligence. DME-OCT: Diabetic Macular Edema by Optical Coherence Tomography. VD A-OCT: Vascular Density by Angio-Optical Coherence Tomography. \u0026nbsp;\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"101%\"\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"bottom\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"4\" valign=\"bottom\"\u003e\u003cstrong\u003e\u003cem\u003eDiabetic retinopathy OFT\u003c/em\u003e\u003c/strong\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"bottom\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"bottom\"\u003eOverall\u0026nbsp;\u003cbr\u003eN = 499\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"bottom\"\u003eNo\u0026nbsp;\u003cbr\u003eN = 342\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"bottom\"\u003eMild non-proliferative\u003cbr\u003eN = 97\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"bottom\"\u003eModerate non-proliferative\u003cbr\u003eN = 45\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"bottom\"\u003eSevere non-proliferative\u003cbr\u003eN = 15\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"bottom\"\u003eP-value\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eCRT - OCT\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e260 (238; 278)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e256 (234; 271)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e262 (246; 278)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e270 (250; 294)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e279 (268; 301)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026lt; 0.001\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eDME - OCT\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e22 (4.4)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e2 (0.6)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e3 (3.6)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e8 (17.0)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e9 (45.0)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026lt; 0.001\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eDME - AI\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e26 (5.2)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e1 (0.3)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e1 (1.2)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e14 (29.8)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e10 (50.0)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026lt; 0.001\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eCrystalline N\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e3 (2; 11)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e3 (2; 6)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e3 (1; 11)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e4 (2; 11)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e6 (3; 11)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.083\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eCrystalline C\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.558\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e466 (93.4)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e325 (93.4)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e77 (91.7)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e46 (97.9)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e18 (90.0)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e1\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e21 (4.2)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e16 (4.6)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e3 (3.6)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e1 (2.1)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e1 (5.0)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e2\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e11 (2.2)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e6 (1.7)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e4 (4.8)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e1 (5.0)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e3\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e1 (0.2)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e1 (0.3)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eCrystalline P\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.662\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e496 (99.4)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e346 (99.4)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e83 (98.8)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e47 (100.0)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e20 (100.0)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e1\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e1 (0.2)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e1 (0.3)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e2\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e1 (0.2)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e1 (1.2)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e3\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e1 (0.2)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e1 (0.3)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eBCVA\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.8 (0.6; 0.9)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.8 (0.6; 0.93)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.8 (0.6; 1)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.7 (0.6; 0.8)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.7 (0.5; 0.8)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.076\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eSpherical.equivalent\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.50\u0026nbsp;\u003cbr\u003e(-0.25; 1.62)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.62\u0026nbsp;\u003cbr\u003e(-0.12; 1.75)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0\u0026nbsp;\u003cbr\u003e(-1.12; 1.12)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.25\u0026nbsp;\u003cbr\u003e(-0.75; 0.88)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.31\u003cbr\u003e\u0026nbsp;(-0.16; 0.81)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.001\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eVD A-OCT SUPERFICIAL\u003cbr\u003ePLEXUS CENTRAL\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e16 (13; 19)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e16 (13; 20)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e17 (15; 20)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e16 (12; 19)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e14 (14; 19)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.417\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eVD A-OCT SUPERFICIAL\u003cbr\u003ePLEXUS SUPERIOR\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e46 (43; 48)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e47 (44; 48)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e47 (43; 49)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e44 (42; 46)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e44 (41; 46)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026lt; 0.001\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eVD A-OCT SUPERFICIAL\u003cbr\u003ePLEXUS TEMPORAL\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e46 (44; 49)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e47 (45; 49)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e47 (44; 48)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e45 (43; 47)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e45 (41; 47)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.011\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eVD A-OCT SUPERFICIAL\u003cbr\u003ePLEXUS INFERIOR\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e46 (43; 48)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e46 (43; 48)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e45 (42; 47)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e45 (42; 48)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e45 (43; 47)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.099\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eVD A-OCT SUPERFICIAL.\u003cbr\u003ePLEXUS NASAL\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e45 (42; 47)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e45 (43; 47)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e45 (42; 47)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e43 (40; 45)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e42 (40; 45)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.001\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eTable 2.2 Left eyes.\u003c/em\u003e\u003c/strong\u003e Summary of the characteristics of the left eyes of the patients included in the study. Data are frequencies (%) and medians (IQR). BCVA: Best Corrected Visual Acuity. Cristaline C: Cortical component of the lens. Cristaline N: Nuclear component of the lens. Cristaline P: Posterior component of the lens. CRT OCT: Central retinal thickness by Optical Coherence Tomography. DME- AI: Diabetic Macular Edema by Artificial Intelligence. DME-OCT: Diabetic Macular Edema by Optical Coherence Tomography. VD A-OCT: Vascular Density by Angio-Optical Coherence Tomography.\u0026nbsp;\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"110%\"\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"bottom\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"9\" valign=\"bottom\"\u003e\u003cstrong\u003e\u003cem\u003eDiabetic retinopathy AI\u003c/em\u003e\u003c/strong\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"bottom\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"bottom\"\u003eOverall\u0026nbsp;\u003cbr\u003eN = 499\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"bottom\"\u003eNo\u0026nbsp;\u003cbr\u003eN = 325\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"bottom\"\u003eMild non-proliferative\u0026nbsp;\u003cbr\u003eN = 57\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"3\" valign=\"bottom\"\u003eModerate non-proliferative\u003cbr\u003eN = 70\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"bottom\"\u003eSevere non-proliferative\u003cbr\u003eN = 45\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"bottom\"\u003eProliferative\u003cbr\u003eN = 2\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"bottom\"\u003eP-value\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eCRT - OCT\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e254\u0026nbsp;\u003cbr\u003e(235; 275)\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"top\"\u003e252\u0026nbsp;\u003cbr\u003e(235; 274)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e253\u0026nbsp;\u003cbr\u003e(229; 265)\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"3\" valign=\"top\"\u003e257\u0026nbsp;\u003cbr\u003e(238; 274)\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"top\"\u003e279\u0026nbsp;\u003cbr\u003e(256; 297)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e233\u0026nbsp;\u003cbr\u003e(216; 250)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026lt; 0.001\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eDME - OCT\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e16 (3.2)\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"top\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"top\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"3\" valign=\"top\"\u003e8 (11.4)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e8 (17.8)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026lt; 0.001\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eDME - .IA\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e20 (4.0)\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"top\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"top\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"3\" valign=\"top\"\u003e9 (12.9)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e9 (20.0)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e2 (100.0)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026lt; 0.001\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eCrystalline.N\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"top\"\u003e4 (2; 11)\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"3\" valign=\"top\"\u003e3 (2; 6)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e3 (2; 6)\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"top\"\u003e4 (1; 5)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e5 (3; 11)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e8 (6; 9)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.024\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eCrystalline C\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"3\" valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.872\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e466 (93.4)\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"top\"\u003e302 (92.9)\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"top\"\u003e55 (96.5)\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"3\" valign=\"top\"\u003e65 (92.9)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e42 (93.3)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e2 (100.0)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e1\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e23 (4.6)\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"top\"\u003e17 (5.2)\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"top\"\u003e1 (1.8)\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"3\" valign=\"top\"\u003e3 (4.3)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e2 (4.4)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e2\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e8 (1.6)\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"top\"\u003e5 (1.5)\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"top\"\u003e1 (1.8)\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"3\" valign=\"top\"\u003e1 (1.4)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e1 (2.2)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e3\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e2 (0.4)\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"top\"\u003e1 (0.3)\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"top\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"3\" valign=\"top\"\u003e1 (1.4)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eCrystalline P\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"3\" valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.236\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e492 (98.6)\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"top\"\u003e320 (98.5)\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"top\"\u003e56 (98.2)\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"3\" valign=\"top\"\u003e69 (98.6)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e45 (100.0)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e2 (100.0)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e1\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e5 (1.0)\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"top\"\u003e5 (1.5)\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"top\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"3\" valign=\"top\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e2\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e2 (0.4)\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"top\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"top\"\u003e1 (1.8)\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"3\" valign=\"top\"\u003e1 (1.4)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eBCVA\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.8 (0.5; 0.8)\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"top\"\u003e0.8 (0.6; 0.8)\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"top\"\u003e0.8 (0.5; 0.8)\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"3\" valign=\"top\"\u003e0.8 (0.5; 0.88)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.6 (0.5; 0.8)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.35 (0.28; 0.42)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.26\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eSpherical.equivalent\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.38\u0026nbsp;\u003cbr\u003e(-0.38; 1.38)\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"top\"\u003e0.5\u0026nbsp;\u003cbr\u003e(-0.25; 1.62)\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"top\"\u003e0\u003cbr\u003e(-1.38; 0.88)\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"3\" valign=\"top\"\u003e0.12\u0026nbsp;\u003cbr\u003e(-0.59; 1.12)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.38\u0026nbsp;\u003cbr\u003e(-0.50; 1.12)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.56\u0026nbsp;\u003cbr\u003e(-0.09; 1.22)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.073\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eVD A-OCT SUPERFICIAL.\u003cbr\u003ePLEXUS CENTRAL\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e16 (13; 19)\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"top\"\u003e16 (13; 19)\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"top\"\u003e16 (13; 20)\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"3\" valign=\"top\"\u003e16 (13; 20)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e15 (11; 18)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e13 (12; 13)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.389\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eVD A-OCT SUPERFICIAL.\u003cbr\u003ePLEXUS SUPERIOR\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e46 (43; 48)\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"top\"\u003e46 (43; 48)\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"top\"\u003e46 (43; 48)\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"3\" valign=\"top\"\u003e45 (41; 47)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e44 (40; 47)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e42 (41; 43)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026lt; 0.001\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eVD A-OCT SUPERFICIAL.\u003cbr\u003ePLEXUS TEMPORAL\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e46 (44; 49)\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"top\"\u003e47 (45; 49)\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"top\"\u003e46 (45; 49)\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"3\" valign=\"top\"\u003e45 (43; 48)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e44 (40; 46)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e41 (39; 43)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026lt; 0.001\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eVD A-OCT SUPERFICIAL.\u003cbr\u003ePLEXUSINFERIOR\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e46 (42; 48)\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"top\"\u003e47 (44; 48)\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"top\"\u003e47 (44; 48)\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"3\" valign=\"top\"\u003e45 (40; 47)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e42 (40; 44)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e36 (35; 37)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026lt; 0.001\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eVD A -OCT SUPERFICIAL.\u003cbr\u003ePLEXUS NASAL\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e45 (42; 47)\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"top\"\u003e45 (43; 47)\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"top\"\u003e45 (42; 47)\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"3\" valign=\"top\"\u003e44 (41; 47)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e43 (42; 45)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e37 (36; 38)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.008\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eTable 2.3 Right eyes.\u003c/em\u003e\u003c/strong\u003e Summary of the characteristics of the left eyes of the patients included in the study. Data are frequencies (%) and medians (IQR). BCVA: Best Corrected Visual Acuity. Cristaline C: Cortical component of the lens. Cristaline N: Nuclear component of the lens. Cristaline P: Posterior component of the lens. CRT - OCT: Central retinal thickness by Optical Coherence Tomography. DME- AI: Diabetic Macular Edema by Artificial Intelligence. DME - OCT: Diabetic Macular Edema by Optical Coherence Tomography. VD A-OCT: Vascular Density by Angio-Optical Coherence Tomography.\u0026nbsp;\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"112%\"\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 126px;\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 80px;\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"5\" valign=\"top\" style=\"width: 378px;\"\u003e\u003cstrong\u003e\u003cem\u003eDiabetic retinopathy AI\u003c/em\u003e\u003c/strong\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 51px;\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 126px;\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 80px;\"\u003eOverall\u0026nbsp;\u003cbr\u003eN = 498*\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003eNo\u0026nbsp;\u003cbr\u003eN = 336\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003eMild non-proliferative\u0026nbsp;\u003cbr\u003eN = 51\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003eModerate non-proliferative\u0026nbsp;\u003cbr\u003eN = 65\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 72px;\"\u003eSevere non-proliferative\u0026nbsp;\u003cbr\u003eN = 43\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 75px;\"\u003eProliferative\u0026nbsp;\u003cbr\u003eN = 3\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 51px;\"\u003eP-value\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 126px;\"\u003eCRT - OCT\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 80px;\"\u003e260 (238; 278)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e257 (234; 272)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e259 (242; 273)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e266 (249; 288)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 72px;\"\u003e272 (256; 295)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 75px;\"\u003e310 (306; 404)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 51px;\"\u003e\u0026lt; 0.001\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 126px;\"\u003eDME - OCT\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 80px;\"\u003e22 (4.4)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e1 (0.3)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e1 (2.0)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e7 (10.8)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 72px;\"\u003e12 (27.9)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 75px;\"\u003e1 (33.3)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 51px;\"\u003e\u0026lt; 0.001\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 126px;\"\u003eDME \u0026ndash; IA\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 80px;\"\u003e26 (5.2)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e11 (16.9)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 72px;\"\u003e14 (32.6)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 75px;\"\u003e1 (33.3)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 51px;\"\u003e\u0026lt; 0.001\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 126px;\"\u003eCrystalline N\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 80px;\"\u003e3 (2; 11)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e3 (2; 6)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e4 (1; 8)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e4 (1; 11)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 72px;\"\u003e4 (2; 11)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 75px;\"\u003e11 (8; 11)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 51px;\"\u003e0.138\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 126px;\"\u003eCrystalline C\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 80px;\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 72px;\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 75px;\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 51px;\"\u003e0.623\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 126px;\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 80px;\"\u003e465 (93.4)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e314 (93.5)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e49 (96.1)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e58 (89.2)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 72px;\"\u003e41 (95.3)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 75px;\"\u003e3 (100)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 51px;\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 126px;\"\u003e1\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 80px;\"\u003e21 (4.2)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e16 (4.8)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e1 (2.0)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e3 (4.6)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 72px;\"\u003e1 (2.3)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 75px;\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 51px;\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 126px;\"\u003e2\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 80px;\"\u003e11 (2.2)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e5 (1.5)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e1 (2.0)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e4 (6.2)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 72px;\"\u003e1 (2.3)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 75px;\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 51px;\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 126px;\"\u003e3\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 80px;\"\u003e1 (0.2)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e1 (0.3)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 72px;\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 75px;\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 51px;\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 126px;\"\u003eCrystalline P\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 80px;\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 72px;\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 75px;\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 51px;\"\u003e0.694\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 126px;\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 80px;\"\u003e495 (99.4)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e334 (99.4)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e51 (100.0)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e64 (98.5)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 72px;\"\u003e43 (100.0)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 75px;\"\u003e3 (100)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 51px;\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 126px;\"\u003e1\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 80px;\"\u003e1 (0.2)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e1 (0.3)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 72px;\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 75px;\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 51px;\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 126px;\"\u003e2\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 80px;\"\u003e1 (0.2)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e1 (1.5)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 72px;\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 75px;\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 51px;\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 126px;\"\u003e3\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 80px;\"\u003e1 (0.2)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e1 (0.3)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 72px;\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 75px;\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 51px;\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 126px;\"\u003eBCVA\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 80px;\"\u003e0.8 (0.6; 0.9)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e0.8 (0.6; 0.9)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e0.8 (0.6; 1)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e0.8 (0.5; 1.)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 72px;\"\u003e0.7 (0.5; 0.8)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 75px;\"\u003e0.1 (0.08; 0.2)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 51px;\"\u003e0.004\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 126px;\"\u003eSpherical.equivalent\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 80px;\"\u003e0.50\u0026nbsp;\u003cbr\u003e(-0.25; 1.62)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e0.62\u0026nbsp;\u003cbr\u003e(-0.12; 1.75)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e0\u003cbr\u003e\u0026nbsp;(-2; 0.88)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e0.25\u0026nbsp;\u003cbr\u003e(-0.75; 1)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 72px;\"\u003e0.25\u0026nbsp;\u003cbr\u003e(-0.38; 0.88)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 75px;\"\u003e0.75\u0026nbsp;\u003cbr\u003e(0.31; 0.88)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 51px;\"\u003e0.002\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 126px;\"\u003eVD A- OCT SUPERFICIAL.\u003cbr\u003ePLEXUS CENTRAL\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 80px;\"\u003e16 (13; 19)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e16 (13; 20)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e17 (15; 20)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e16 (13; 19)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 72px;\"\u003e16 (14; 19)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 75px;\"\u003e19 (14; 30)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 51px;\"\u003e0.688\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 126px;\"\u003eVD A- OCT SUPERFICIAL.\u003cbr\u003ePLEXUS SUPERIOR\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 80px;\"\u003e46 (43; 48)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e47 (44; 48)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e47 (44; 50)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e45 (42; 48)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 72px;\"\u003e43 (42; 46)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 75px;\"\u003e43 (38; 44)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 51px;\"\u003e\u0026lt; 0.001\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 126px;\"\u003eVD A-OCT SUPERFICIAL.\u003cbr\u003ePLEXUS TEMPORAL\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 80px;\"\u003e46 (44; 49)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e47 (45; 49)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e47 (45; 49)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e46 (43; 48)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 72px;\"\u003e45 (42; 47)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 75px;\"\u003e46 (43; 49)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 51px;\"\u003e0.001\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 126px;\"\u003eVD A-OCT SUPERFICIAL.\u003cbr\u003ePLEXUS INFERIOR\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 80px;\"\u003e46 (43; 48)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e46 (43; 48)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e45 (42; 47)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e45 (42; 47)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 72px;\"\u003e45 (41; 47)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 75px;\"\u003e46 (46; 54)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 51px;\"\u003e0.041\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 126px;\"\u003eVD A-OCT SUPERFICIAL.\u003cbr\u003ePLEXUS NASAL\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 80px;\"\u003e45 (42; 47)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e45 (43; 47)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e45 (42; 47)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 77px;\"\u003e44 (40; 46)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 72px;\"\u003e42 (40; 44)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 75px;\"\u003e45 (44; 51)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 51px;\"\u003e\u0026lt; 0.001\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eTable 2.4 Left eyes.\u003c/em\u003e\u003c/strong\u003e Summary of the characteristics of the left eyes of the patients included in the study. Data are frequencies (%) and medians (IQR). BCVA: Best Corrected Visual Acuity. Cristaline C: Cortical component of the lens. Cristaline N: Nuclear component of the lens. Cristaline P: Posterior component of the lens. CRT - OCT: Central retinal thickness by Optical Coherence Tomography. DME - AI: Diabetic Macular Edema by Artificial Intelligence. DME-OCT: Diabetic Macular Edema by Optical Coherence Tomography. VD A-OCT: Vascular Density by Angio-Optical Coherence Tomography.\u0026nbsp;\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"100%\"\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 30px;\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"5\" valign=\"top\" style=\"width: 69px;\"\u003e\u003cstrong\u003e\u003cem\u003eDiabetic retinopathy\u003c/em\u003e\u003c/strong\u003e\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 30px;\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"top\" style=\"width: 32px;\"\u003eDiagnosed by OFT\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 6px;\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"top\" style=\"width: 30px;\"\u003eDiagnosed by AI\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 30px;\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003eP-value\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003eOdd-Ratio (95% CI)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 6px;\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003eP-value\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 21px;\"\u003eOdd-Ratio (95% CI)\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 30px;\"\u003eAge, per year\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\u0026lt; .001\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e0.976 (0.963; 0.990)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 6px;\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e-\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 21px;\"\u003e-\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 30px;\"\u003eYears DM, per year\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\u0026lt; .001\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e1.151 (1.109; 1.194)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 6px;\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\u0026lt; .001\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 21px;\"\u003e1.120 (1.081; 1.160)\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 30px;\"\u003eType-1 diabetes mellitus\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 6px;\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e0.009\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 21px;\"\u003e3.168 (1.339; 7.494)\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 30px;\"\u003eDME - OCT*\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\u0026lt; .001\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e32 (7.26; 142)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 6px;\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\u0026lt; .001\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 21px;\"\u003e62.7 (8.38; 469)\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 30px;\"\u003eCRT - OCT*\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e0.004\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e1.006 (1.002; 1.011)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 6px;\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e0.047\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 21px;\"\u003e1.004 (1.000; 1.009)\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 30px;\"\u003eSpherical equivalent*\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\u0026lt; .001\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e0.891 (0.832; 0.954)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 6px;\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e\u0026lt; .001\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 21px;\"\u003e0.880 (0.823; 0.942)\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 30px;\"\u003eVD_A.OCT.SUPERFICIAL*\u003cbr\u003ePLEXUS.TEMPORAL\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\u0026lt; .001\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 22px;\"\u003e0.951 (0.925; 0.978)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 6px;\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 8px;\"\u003e0.001\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 21px;\"\u003e0.957 (0.932; 0.983)\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eTable 3. Multivariate logistic regression for diabetic retinopathy. (*) per unit.\u0026nbsp;\u003c/em\u003e\u003c/strong\u003eCRT- OCT: Central retinal thickness by Optical Coherence Tomography. DME - OCT: Diabetic Macular Edema by Optical Coherence Tomography.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"601\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 47px;\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"top\" style=\"width: 246px;\"\u003e\u003cem\u003eDiagnosis of DR in at least one eye\u003c/em\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 19px;\"\u003e\u003cem\u003e\u0026nbsp;\u003c/em\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"top\" style=\"width: 289px;\"\u003e\u003cem\u003eBinocular sensitivity and specificity (%)\u003c/em\u003e\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 47px;\"\u003eAI\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003eYes\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003eNo\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 19px;\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 151px;\"\u003e\u003cem\u003ebSen\u003c/em\u003e [CI - 95%]\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 138px;\"\u003e\u003cem\u003ebSp\u0026nbsp;\u003c/em\u003e[CI - 95%]\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 47px;\"\u003e(+) *\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e189\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e20\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 19px;\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 151px;\"\u003e100 [98,1 \u0026ndash; 100]\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 138px;\"\u003e93,5 [90,2 \u0026ndash; 96,0]\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 47px;\"\u003e(-)\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e0\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 123px;\"\u003e289\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 19px;\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 151px;\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 138px;\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eTable 4. Sensitiviy and specificity(*)\u003c/em\u003e\u003c/strong\u003e The AI is considered positive (+) when it is positive in at least one eye.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"100%\"\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"top\"\u003e\u003cstrong\u003e\u003cem\u003eProportion of agreements\u003c/em\u003e\u003c/strong\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003eObserved\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003eExpected*\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u003cem\u003ekappa\u003c/em\u003e (95%CI)\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eRight eye\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.820\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.484\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.650 (0.594; 0.707)\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eLeft eye\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.847\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.503\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.693 (0.636; 0.750)\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eTable 5. Proportion of agreements considering the different stages of DR\u003c/em\u003e\u003c/strong\u003e\u003cstrong\u003e\u003cem\u003e\u0026nbsp;\u003c/em\u003e\u003c/strong\u003e\u003cstrong\u003e\u003cem\u003e(*) per unit.\u0026nbsp;\u003c/em\u003e\u003c/strong\u003e(*) Under the hypothesis of no agreement. Random agreement hypothesis\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"100%\"\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd colspan=\"2\" valign=\"top\"\u003e\u003cstrong\u003e\u003cem\u003eProportion of agreements\u003c/em\u003e\u003c/strong\u003e\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\u0026nbsp;\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003eObserved\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003eExpected*\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\u003cem\u003ekappa\u003c/em\u003e (95%CI)\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eRight eye\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.966\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.556\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.923 (0.836; 1.010)\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003eLeft eye\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.978\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.569\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\"\u003e0.949 (0.861; 1.036)\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cstrong\u003eTable 6. Agreement between binary (Y/N) diagnosis(*).\u003c/strong\u003e Under the hypothesis of no agreement. Random agreement hypothesis\u003c/p\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"international-journal-of-retina-and-vitreous","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"IJRV","sideBox":"Learn more about [International Journal of Retina and Vitreous](https://jneurodevdisorders.biomedcentral.com/)","snPcode":"40942","submissionUrl":"https://submission.nature.com/new-submission/40942/3","title":"International Journal of Retina and Vitreous","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Diabetic Retinopathy, Artificial Intelligence, Automated Diabetic Retinopathy, Artificial Intelligence Detection of Diabetic Retinopathy, Automated Retinal Image","lastPublishedDoi":"10.21203/rs.3.rs-7284873/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7284873/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003eObjective or Purpose: \u003c/strong\u003eTo evaluate the diagnostic performance and agreement of the EyeArt\u003csup\u003e®\u003c/sup\u003e Artificial Intelligence (AI) system for detecting Diabetic Retinopathy (DR), comparing its results with ophthalmologists' assessments in a regional screening program.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDesign: \u003c/strong\u003eCross-sectional observational study.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSubjects, Participants, and/or Controls: \u003c/strong\u003eA total of 499 diabetic patients aged 18 years or older were enrolled between June and September 2023 through the Retisalud DR screening program in the Canary Islands. No separate control group was included.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMethods: \u003c/strong\u003eAll participants underwent non-mydriatic fundus photography using the TRC-NW400 camera. Retinal images were analyzed by the EyeArt\u003csup\u003e®\u003c/sup\u003e AI system (version 2.1.0), and results were compared with assessments by ophthalmologists based on the International Clinical Diabetic Retinopathy (ICDR) scale. Agreement was quantified using Cohen’s kappa coefficient. Additionally, mixed-effects logistic regression was used to explore associations between DR and clinical risk factors.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMain Outcome Measures: \u003c/strong\u003eSensitivity, specificity, and agreement (Cohen’s kappa) of the AI system compared to clinical diagnosis; predictors of DR such as age, diabetes duration, presence of Diabetic Macular Edema (DME), and central retinal thickness (CRT).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eResults: \u003c/strong\u003eThe EyeArt® system achieved a binocular sensitivity of 100% (95% CI: 98.1–100) and a specificity of 93.5% (95% CI: 90.2–96.0). Agreement with ophthalmologist grading was excellent, with kappa values of 0.923 (right eye) and 0.949 (left eye). Younger age, longer diabetes duration, DME presence, and higher CRT were significantly associated with DR diagnosis.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConclusions: \u003c/strong\u003eThe EyeArt\u003csup\u003e®\u003c/sup\u003e AI system showed excellent diagnostic accuracy and strong agreement with clinical evaluations in DR screening. Nonetheless, its tendency to overestimate DR severity indicates the need for further refinement of its grading algorithm. These findings support the potential integration of AI systems into large-scale diabetic retinopathy screening programs, pending further validation.\u003c/p\u003e","manuscriptTitle":"Evaluation of the degree of agreement in the diagnosis of Diabetic Retinopathy between Ophthalmologists and EyeArt®","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-08-18 09:31:54","doi":"10.21203/rs.3.rs-7284873/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-09-15T22:53:26+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-09-15T18:46:23+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-09-05T12:29:01+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"80233277436789626514403883580323219891","date":"2025-09-05T07:05:18+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"4926015051350894721874531908934961510","date":"2025-09-02T02:33:51+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-08-11T05:32:21+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-08-10T17:38:10+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-08-06T12:56:33+00:00","index":"","fulltext":""},{"type":"submitted","content":"International Journal of Retina and Vitreous","date":"2025-08-03T17:41:47+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"international-journal-of-retina-and-vitreous","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"IJRV","sideBox":"Learn more about [International Journal of Retina and Vitreous](https://jneurodevdisorders.biomedcentral.com/)","snPcode":"40942","submissionUrl":"https://submission.nature.com/new-submission/40942/3","title":"International Journal of Retina and Vitreous","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"98c78733-0089-4028-8cc5-ee8a5a5430d8","owner":[],"postedDate":"August 18th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2025-10-20T10:23:44+00:00","versionOfRecord":[],"versionCreatedAt":"2025-08-18 09:31:54","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7284873","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7284873","identity":"rs-7284873","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}