Importance of OCT-derived Biomarkers for the Recurrence of Central Serous Chorioretinopathy using Statistics and Predictive Modelling

Importance of OCT-derived Biomarkers for the Recurrence of Central Serous Chorioretinopathy using Statistics and Predictive Modelling | 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 Importance of OCT-derived Biomarkers for the Recurrence of Central Serous Chorioretinopathy using Statistics and Predictive Modelling Emilien Seiler, Léon Delachaux, Jennifer Cattaneo, Ali Garjani, and 8 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-4170618/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Central serous chorioretinopathy (CSCR) is a posterior segment disease characterized by accumulation of subretinal fluid that, in acute forms, resolves spontaneously. However, about a third of the cases experience recurrences that might cause severe and irreversible vision loss due to anatomical outer retinal and retinal pigment epithelium changes. This study aims to identify optical coherence tomography (OCT)-derived parameters linked to CSCR recurrence. Our dataset included 5211 OCTs from 344 eyes of 255 CSCR patients. After expert labeling, 178 eyes were identified as recurrent, 109 were non-recurrent, and 57 were excluded. We extracted parameters using artificial intelligence and computer vision. We used inferential statistics to assess differential distribution between the recurrent and non-recurrent groups, and we employed predictive modeling for feature importance analysis. We identified 9 predictive biomarkers for CSCR recurrence, including age, presence of subretinal fluid, intraretinal fluid and Pigment Epithelial detachments, as well as choroidal vascularity index, integrity of photoreceptors and RPE layer, thicknesses of choriocapillaris and choroidal stroma, and thinning of internal retinal layers (outer nuclear layer, and inner nuclear layer combined with and outer plexiform layer). These results can potentially enable future developments in automatic detection of CSCR recurrence, paving the way for translational medical applications. Artificial Intelligence and Machine Learning Ophthalmology Central Serous Chorioretinopathy Machine Learning biomarker importance Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 1. Introduction Central Serous Chorioretinopathy (CSCR) is an idiopathic chorioretinal disease characterized by localized and limited serous detachment of the neurosensory retina (shown in Fig. 1 ) and may be associated with focal retinal pigment epithelium detachments. The term 'central' refers to the form of the disease in which the serous detachment involves the macular area and causes visual symptoms accordingly. However, not all patients have macular involvement: some cases may present with extra macular serous detachments, as it is often observed in the contralateral eye of patients with active CSCR [ 1 ]. CSCR is the fourth most common retinopathy after age-related macular degeneration, diabetic retinopathy and branch retinal vein occlusion [ 2 ]. It affects 5.8 individuals per 100,000 and its incidence is 5 times higher in men [ 3 ]. CSCR typically occurs in males in their late 30s to 50s, even though it can occur at later ages, particularly in women and patients affected by atypical chronic forms [ 4 ]. The etiology of the disease is poorly understood, granting the need for further research into the factors influencing its manifestations. No underlying pathophysiologic mechanisms have been proven, but CSCR is thought to occur due to hyper-permeable choroidal capillaries which, in association with retinal pigment epithelium dysfunction, cause a serous detachment of the neurosensory retina [ 5 ]. Some of the known risk factors are the use of steroids [ 6 ], stress [ 7 ], psychopathology like “type A” personality pattern [ 8 ], hypertension, endocrine changes, drugs (sympathomimetic, MEK-inibitors) [ 9 ] and obstructive sleep apnea [ 10 ]. CSCR is classified into different forms according to the duration of the disease and clinical features. In the acute form of the disease, Subretinal Fluid (SRF) resolves spontaneously within 4 months of the onset of symptoms without anatomical changes [ 12 ]. Non-resolving (or "persistent") CSCR is characterized by persistence of SRF for more than 4 months, associated with elongated outer segments photoreceptor on B-scan SD-OCT [ 12 ]. Chronic CSCR (or "diffuse retinal epitheliopathy"), presents diffuse decompensation of the RPE with or without SRF. The chronic form of CSCR has a strong genetic determinant as 52% of patients have an affected relative [ 11 ]. CSCR is considered recurrent if there is a reappearance of SRF accumulation after a completely resolved acute episode. Conversely, it is classified as inactive if there is a history of acute CSCR but no SRF at the time of assessment. No treatment is generally required for the first episode, while for recurrent, non-resolving, and chronic forms the standard treatment is photodynamic therapy with verteporfin [ 12 ] or focal argon laser treatment [ 13 ]. After complete resolution, up to one-third of patients experience recurrence of the disease with persistence of SRF accumulation, which can lead to long-term visual impairment [ 3 ]. Retinal Pigment Epithelium (RPE) and changes in external layers are reported as indicative of future recurrences or poor visual prognosis [ 14 ][ 15 ][ 16 ][ 17 ]. Various observable parameters have been reported as potential biomarkers for CSCR, such as: choroidal thickness [ 18 ], Choroidal Vascularity Index (CVI, i.e., the ratio between vascular area in the choroid with the total area of the choroid, see Section 1 of SI and Fig. S1a) [ 19 ][ 20 ][ 21 ] and presence of pachy vessels [ 14 , 22 ] (i.e., vessels dilatation, Fig. S1b). Recent advances in Machine Learning (ML) have enabled the automatic processing of medical data, especially images. In particular, Deep Learning (DL) methods for image classification or segmentation have been found to be very effective in the analysis of medical images, such as brain MRI for Alzheimer diagnosis or CT scans for classification of lung tissue and lung diseases diagnosis [ 23 ]. These innovations can help clinicians in performing their daily practice when implemented in the form of Expert Support Systems. For certain diseases, ML and DL approaches have allowed computers to detect disease signs directly, as in case of diabetic retinopathy [ 24 ]. In the case of CSCR, research in automatic biomarker extraction and predictive modeling is still lacking. Prior research identified several biomarkers to differentiate between acute and chronic/recurrent CSCR using AI-based segmentation followed by statistical testing to determine if the extracted parameters were differently distributed between the two groups [ 25 ]. Other prior work attempted to automatically predict recurrence of CSCR directly from OCT Scans using DL [ 26 ], but the imbalance in the dataset makes it difficult to assess the results. Nowadays there is no effective way of predicting recurrence of CSCR, making it challenging to determine if and when a patient should be reevaluated after acute phase resolution. This research aims to identify predictive biomarkers associated with CSCR recurrence, paving the way towards automating predictions of the recurrence of this disease. 2. Methods 2.1 Data collection and biomarkers extraction We utilized longitudinal real-world data extracted from Jules-Gonin’s Eye Hospital’s Electronic Medical Record system (EMR), comprising all patients who visited the medical retina unit at Jules-Gonin Eye Hospital between March 2017 and September 2022 (Fig. 3 ). We extracted our dataset using an in-house software tool, Cohort Builder. First, Cohort Builder queried the EMRs to identify patients with a diagnosis of Central Serous Chorioretinopathy (CSCR). It then verified the hospital's General Consent database, and proceeded to the extraction of the raw fovea-centered OCT imaging dataset, consisting of 30° OCT cubes (98 B-scans, 10 ART) and Enhanced-depth imaging OCT line scans from the storage system (Heyex 2, Heidelberg Engineering, Heidelberg, Germany) of the OCT Scanners (SPECTRALIS®, Heidelberg Engineering, Heidelberg, Germany). Leveraging Cohort Builder, we uploaded the imaging dataset to an AI-powered generic image viewer for ophthalmology, Discovery® by RetinAI. This software is specialized in OCT-scan image analysis, specifically, in the estimation of image-derived biomarkers. Cohort Builder interacted with it automatically, via its API, extracting: volume of Subretinal Fluid (SRF), volume of Intraretinal Fluid (IRF), area of Pigment Epithelial Detachment (PED) in the central, pericentral and peripheral areas, and thicknesses of the following 6 layers (or combinations thereof) in the central area (C1): Chorio-Capillaris and Choroidal Stroma (CC + CS), Photo-Receptors and Retinal Pigment Epithelium (PR + RPE), Outer Nuclear Layer (ONL), Inner Nuclear Layer and Outer Plexiform Layer (INL + OPL), Ganglion Cell Layer and Inner Plexiform Layer (GCL + IPL), Retinal Nerve Fiber Layer (RNFL). We extracted three additional biomarkers using in-house algorithms, implemented in Cohort Builder (see Supplementary Methods): the Choroidal Vascularity Index (CVI), a score estimating disruption of the PR + RPE layer (DSCORE), and an estimation of the area of the largest Choroidal Pachyvessel (PV_AREA). 2.2 Data annotation and quality control We collected expert annotations using an in-house tool, Evolvision, thanks to which an ophthalmologist retina specialist (CME) labeled recurrent and non-recurrent cases. Evolvision is a GUI-based program that displays the results of the analysis performed by Cohort Builder, offering the annotator a comprehensive view of biomarker evolution over time, encompassing all biomarkers for each patient visit. If the annotator selects a specific visit, Evolvision shows the corresponding OCT line scan for visual inspection. In instances where certain patient time series prove challenging to annotate, Evolvision reports the Hospital PID so the retinal specialist may refer to the patient's medical records to evaluate the whole clinical case. The patient time series was assigned one of three labels (as shown in Fig. 2 ): "non-recurrent" indicated a single acute CSCR episode characterized by an accumulation of SRF that was resolved spontaneously; "recurrent" when the patient alternated periods characterized by SRF accumulation to periods when SRF would be complete resolution; finally, cases characterized by persistence of SRF were labeled as "inconclusive", warranting exclusion from further analysis. The acquisition and labeling process are schematised in Fig. 3 . Evolvision also allowed us to select a sub-portion of the time series: thanks to this feature, the annotator truncated "recurrent" cases to the initial acute CSCR episode, so that we could evaluate the link between a future recurrence of the disease and this first episode, which is a first step towards predictive modeling. 3.3 Feature-importance analysis using statistics and Machine Learning We performed a correlation analysis on the extracted biomarkers based on Pearson Correlation Coefficient (R). We subsequently bi-clustered the resulting correlation matrix (as shown in Fig. 4 ). We calculated the distributions of each biomarker based solely on the first visits of the patients, and obtained p-values (p) and the effect sizes (d) by comparing patient groups with a two-tailed Welch's t-test. We corrected for multiple hypotheses testing using a Benjamini-Hochberg procedure (BH), with a significance threshold (α) of 0.05. We performed an equivalent analysis leveraging the longitudinal nature of our data. We calculated biomarkers distributions based on the means of each biomarker across the truncated time series of each patient (i.e., not considering any visits after the recurrence of the disease). We performed a two-sample Kolmogorov-Smirnov test, obtaining p-values. We again corrected for multiple hypotheses using BH (as reported in Table 1 ). We computed the Cohen’s d effect sizes for the distributions based on first visits and then based on averaged patients (See Supplementary Methods). Finally, we delved into Machine Learning (ML) to determine the relative importance of each biomarker in building a predictive model for the recurrence of CSCR. We fitted a Logistic Regression (LR) model to distinguish between visits from recurrent and non-recurrent patients. We assessed the predictive power of each biomarker by extracting the p-value that estimates its statistical significance and analyzing the coefficient that defines its impact on the model output (as reported in Table 1 and Fig. 5 ). 3.3 Code and data availability A detailed experimental manual and the code that implements the experiments is available at https://github.com/JulesGoninRIO/cscr_stats_ml . To discuss access to the dataset (DTA) please contact the corresponding author. 3. Results 3.1 Baseline characteristics Following quality control measures, we analyzed 5211 OCT Scans from 344 eyes belonging to 255 patients. The mean age was 53 years (range 47, SD +/-12 years). At birth, 78% were men and 22% women. The mean number of visits per patient was 16 (+/-14 SD). Such a large variance is intrinsic in the nature of our dataset (as further reported in the Discussion section). The average duration of follow-up among the patients was 6 years (range 11, SD +/-3 years). 178 eyes were labeled as recurrent and 109 eyes as non-recurrent, while 57 were excluded because of poor quality of the OCT scans 3.2 Correlation among OCT-derived biomarkers A correlation analysis between the extracted biomarkers across all patients and all visits revealed several clusters (as shown in Fig. 4 ). A first cluster is composed of parameters related to the choroid: CC_CS thickness, CVI and PV_AREA. A second cluster includes thicknesses of layers of the inner retina (RNFL, GCL_IPL, and INL_OPL). The third cluster contains outer retinal layer thicknesses (ONL, PR + RPE), parameters related to their integrity (PED, DSCORE) and the presence of fluids (SRF and IRF). 3.3 Importance of OCT-derived biomarkers Our multi-step analysis incorporated increasingly more sophisticated approaches, which gradually included more information. In particular, we initially analyzed only the first visit, using inferential statistics; then, we extended our analysis to consider longitudinal data (averaging every biomarker over the truncated time series of the patient); finally, we applied predictive modeling techniques to the longitudinal data. The number of statistically significant parameters progressively increased (as shown in Fig. 5 ). Using inferential statistics on the single first visit, we identified the following biomarkers as statistically significant: age (p = 0.14E-1, d = 0.38), ONL (p = 0.20E-1, d=-0.36), DSCORE (p = 0.29E-2, d = 0.47). Extending the analysis to the longitudinal data set, uncovered additional statistically significant parameters, in addition to the ones already reported: age (p = 0.41E-1, d = 0.39), ONL (p = 0.30E-1, d=-0.45), PED (p = 0.2E-1, d = 0.22E-1), DSCORE (p = 0.21E-1, d = 0.43). Applying predictive modeling (i.e., Logistic Regression, LR), providing one biomarker at a time as a predictor, we identified the following as statistically significant: age (p = 0.10E-7, coef = 0.34), SRF (p = 0.99E-12, coef = 0.37), IRF (p = 0.30E-5, coef = 0.37), INL + OPL (p = 0.24E-1, coef=-0.13), ONL (p = 0.52E-12, coef=-0.43), CC + CS (p = 0.19E-1, coef = 0.13), DSCORE (p = 0.11E-10, coef = 0.39), CVI (p = 0.27E-4, coef=-0.25), and PED (p = 0.2E-8, coef = 0.36). 4. Discussion In this retrospective study we were able to identify biomarkers predictive of recurrence of CSCR by the means of AI and ML methods. This research represents the first investigation addressing a fundamental clinical inquiry: determining which patients are prone to experiencing recurrence following an initial, spontaneously resolved episode of CSCR. In fact, recurrence of SRF is a negative prognostic factor on long-term visual outcomes causing RPE and PR alterations [ 20 ]. The gold standard treatment of PDT with verteporfin showed good efficacy on fluid resolution and preservation of functional visual acuity [ 11 ]. Therefore, it would be helpful to be able to predict which patients will experience recurrences in order to optimize follow-up schedule and promptly arrange treatment. The current paper aims to identify optical coherence tomography (OCT)-derived parameters linked to CSCR recurrence, by means of Discovery® software (by RetinAI) for identifying and quantifying different SD-OCT biomarkers in CSRC. Our study identified nine OCT predictive biomarkers for the recurrence of CSCR, in particular, volume of Subretinal Fluid (SRF), volume of Intraretinal Fluid (IRF), area of Pigment Epithelial Detachment (PED) in the central, pericentral and peripheral areas, and thicknesses of the following 6 layers (or combinations thereof) in the central area (C1): Chorio-Capillaris and Choroidal Stroma (CC + CS), Photo-Receptors and Retinal Pigment Epithelium (PR + RPE), Outer Nuclear Layer (ONL), Inner Nuclear Layer and Outer Plexiform Layer (INL + OPL), Ganglion Cell Layer and Inner Plexiform Layer (GCL + IPL), Retinal Nerve Fiber Layer (RNFL). Three additional biomarkers were extracted: the Choroidal Vascularity Index (CVI), a score estimating disruption or the PR + RPE layer (DSCORE), and an estimation of the area of the largest Choroidal Pachyvessel (PV_AREA). The multivariate logistic regression to the longitudinal data, demonstrated that the following parameters were statistically significant: SRF, IRF, INL_OPL, CC_CS, CVI (as shown in Fig. 4 ). Therefore, we showed that all these parameters were related to CSCR recurrence. Recently, imaging biomarkers associated with acute and chronic CSCR were analyzed with an AI software [ 25 ]. The authors found a significant increase in thickness in both outer retinal layers (ONL, PR and RPE layer) and inner retinal layers (INL and GCL), and SRF in patients with acute CSCR compared to chronic CSCR at baseline. However, in our study we identified more parameters independently of the type of CSCR, which might better reflect the real-world practice. Also, Xu et al. analyzed recurrences in CSCR patients with six different models of AI and ML [ 26 ]. Imaging has become a key factor in clinical practice since the introduction of fast, high resolution machines. Therefore, the volume of data to analyze increased and new systems of automatic analysis are now available. The application of AI and ML models now becomes the next step forward in the optimization of data analysis mostly to predicting outcomes. Nowadays, by the word 'biomarker', scientists refer to morphological and structural changes that can provide important information on the stage of a disease. The search for new biomarkers in retinal diseases has been a field of great interest in recent years. This research has been driven by the need for earlier and more accurate diagnosis, better prognostication and the development of target therapies. Many researches have been carried out on biomarkers in retinal diseases such as age-related macular degeneration and diabetic macular edema. The predictive role of retinal biomarkers in CSCR has been reported in a recent study [ 27 ]. In particular, changes in central macular thickness and subfoveal choroidal thickness were associated to the resolution of CSCR, while the amount of SRF to the visual acuity outcomes. Moreover, the morphology of the retinal layers, such as ONL thinning, PR layer elongation, ELM and ellipsoid zone disruption was previously associated with longer duration of SRF and worse visual acuity [ 17 , 28 , 29 ]. The definition and validation of predictive biomarkers for central serous chorioretinopathy recurrence is crucial for physicians in daily clinical practice for several reasons. Firstly, predictive biomarkers may allow early diagnosis of individuals at risk of CSCR recurrence. Early diagnosis allows for early intervention and management, potentially preventing serious complications and preserving vision. Furthermore, biomarkers could help tailor treatment strategies to individual patient profiles. By identifying patients most likely to relapse, clinicians can offer customized treatment plans. By understanding the likelihood of recurrence, physicians can set realistic expectations and plan for long-term management, and by identifying high-risk patients, physicians can optimize the use of resources by defining proper follow-up in relation to the likelihood of recurrence. Biomarkers contribute to the advancement of research and development efforts in understanding the underlying mechanism and risk factors associated with CSCR recurrence. They can lead to the development of more effective therapies and preventive strategies. Despite the promising results linking OCT-derived biomarkers with a future recurrence of CSCR, our study was subject to several limitations that warrant discussion. One of the primary challenges was the variability in the number of visits of each patient, particularly pronounced among chronic or recurrent CSCR cases since recurrent patients are examined more often and for longer periods. Our dataset also presented a wide variability in terms of Subretinal Fluid (SRF) volumes. SRF can range from minimal (around 10 nL) to significant (up to 10,000 nL), complicating the identification and categorization of CSCR episodes. Despite a general definition of a CSCR episode as an accumulation of SRF which is followed by drying up of the fluid, the diverse manifestations based on the patient’s SRF range make it challenging to clearly distinguish between multiple occurrences and a single episode. This made it challenging even for medical professionals with decades of experience in the disease, underlining the complexity of the labeling task. In our study, a particularly significant limitation was the binary labeling system used to classify CSCR cases as either recurrent or non-recurrent. This classification framework, while straightforward and appropriate given the size of our dataset, proved insufficient handling the inherent complexities and ambiguities of certain patient cases. Especially challenging was the accurate categorization of chronic patients, which are characterized by fluctuating levels of SRF. Applying our labeling system led to chronic patients being labeled as recurrent cases, despite the nuances in their condition that might suggest a more complex categorization. This subgroup, characterized by SRF volumes that persistently remain above zero, likely represents a distinct clinical picture. In a few cases, chronic patients were excluded from our analysis. Their exclusion, while necessary for the integrity of our current analysis, highlights the limitations of a binary classification system which future work could address via multi-label annotation. Additionally, an intrinsic limitation of our study is that some patients who were classified as non-recurrent (based on the data available at the time the dataset was collected), might have experienced a CSCR episode post-data collection. This potential for future occurrences not captured in our current dataset underscores the dynamic nature of CSCR and the challenges inherent in predicting its course. Looking at the correlation matrix we found that ONL thickness strongly correlated with D_SCORE, consistently with what is reported in the literature [ 30 , 31 ]. Fitting one biomarker at a time, the logistic regression results in a pseudo R-squared (~ proportion of variance explained) that is at maximum 3.0%. Then, the Logistic Regression (LR) model fitted with all our parameters at the same time explained only 8.7% of the variance among patients. Finally, our approach did not make explicit use of the time dimension inherent in our data. Leveraging methods able to consider the evolution of biomarkers over time might enable the prediction of CSCR recurrence and pave the way for translational medical applications. Our analysis revealed correlations between several biomarkers and CSCR recurrence. Applying logistic regression (LR) to all biomarkers explained 8.7% of the variance among patients. Despite the usefulness of these results in elucidating the importance of each biomarker, this suggests a limited capacity of these models to adequately capture the complexities of CSCR recurrence. Looking ahead, significant work remains before these models can be used in clinical applications. A critical unexplored aspect in our current approach is the incorporation of the time dimension in data analysis. Time-series analysis could significantly improve predictive accuracy, opening avenues for practical medical applications. However, advanced modeling techniques like deep learning, while promising enhanced predictive performance, often reduce the model's explainability, a critical factor in clinical settings. Therefore, our future research will aim to balance sophisticated modeling capabilities with the clarity and transparency essential for clinical applicability. Conclusion By applying inferential statistics and Machine Learning methods on a dataset of 5,211 OCT Scans from 344 eyes of 255 patients, we identified 9 predictive biomarkers for the recurrence of CSCR. Analyzing data from the first visit, we found that the following biomarkers were differentially distributed in patients with recurrent CSCR compared to non-recurrent ones: age (p = 0.14E-1, d = 0.38), Outer Nuclear Layer thickness (p = 0.20E-1, d=-0.36) and disruption of the photoreceptor and pigment retinal epithelium (p = 0.29E-2, d = 0.47). In addition to the parameters mentioned above, when analyzing the data longitudinally over the whole first episode of CSCR, we found that the volume of Pigmented Epithelium Detachment was also a statistically significant biomarker of recurrence (p = 0.2E-1, d = 0.22E-1). Finally using predictive modeling, we confirmed and extended the results obtained above using inferential statistics, establishing that the following 9 biomarkers were statistically significant predictors: age (p = 0.10E-08, coef = 0.34), volume of Subretinal Fluid (p = 0.99E-10, coef = 0.37), volume of Intraretinal Fluid (p = 0.30E-5, coef = 0.37), thickness of Inner Nuclear and Outer Plexiform Layers (p = 0.24E-1, coef=-0.13), thickness of Outer Nuclear Layer (p = 0.52E-12, coef=-0.43), thickness Choriocapillaris layer (p = 0.19E-1, coef = 0.13), disruption of Photoreceptor and Retinal Pigment Epithelium Layers (p = 0.11E-10, coef = 0.39), Choroidal Vascularity Index (p = 0.27E-4, coef=-0.25), and volume of Pigment Epithelium Detachment (p = 0.20E-08, coef = 0.36). These results have the potential to enable future developments in automatic prediction of CSCR recurrence, paving the way for translational medical applications. Declarations This study has obtained ethics approval from “Commission cantonale d'éthique de la recherche sur l'être humain (CER-VD)” (project PB_ 2017-00493, Canton of Vaud, Switzerland) Acknowledgements This work was supported by the Swiss Personalized Health Network (2018DRI13 to Thomas J. Wolfensberger) and by the Claire and Selma Kattenburg Foundation. References Liew G, Quin G, Gillies M, Fraser-Bell S. Central serous chorioretinopathy: a review of epidemiology and pathophysiology. Clin Experiment Ophthalmol. 2013;41: 201–214. Wang M, Munch IC, Hasler PW, Prünte C, Larsen M. Central serous chorioretinopathy. Acta Ophthalmol. 2008;86: 126–145. Kitzmann AS, Pulido JS, Diehl NN, Hodge DO, Burke JP. The incidence of central serous chorioretinopathy in Olmsted County, Minnesota, 1980-2002. Ophthalmology. 2008;115: 169–173. Sartini F, Figus M, Nardi M, Casini G, Posarelli C. Non-resolving, recurrent and chronic central serous chorioretinopathy: available treatment options. Eye . 2019;33: 1035–1043. Piccolino FC, Borgia L, Zinicola E, Zingirian M. Indocyanine green angiographic findings in central serous chorioretinopathy. Eye . 1995;9 ( Pt 3): 324–332. Corticosteroids and central serous chorioretinopathy. Ophthalmology. 2002;109: 1834–1837. Gelber GS, Schatz H. Loss of vision due to central serous chorioretinopathy following psychological stress. Am J Psychiatry. 1987;144: 46–50. Yannuzzi LA. TYPE-A BEHAVIOR AND CENTRAL SEROUS CHORIORETINOPATHY. Retina. 1987;7: 111. McCannel TA, Chmielowski B, Finn RS, Goldman J, Ribas A, Wainberg ZA, et al. Bilateral subfoveal neurosensory retinal detachment associated with MEK inhibitor use for metastatic cancer. JAMA Ophthalmol. 2014;132: 1005–1009. The Effect of Obstructive Sleep Apnea on Absolute Risk of Central Serous Chorioretinopathy. Am J Ophthalmol. 2020;218: 148–155. Weenink AC, Borsje RA, Oosterhuis JA. Familial chronic central serous chorioretinopathy. Ophthalmologica. 2001;215: 183–187. Piccolino FC, Eandi CM, Ventre L, Rigault de La Longrais RC, Grignolo FM. Photodynamic Therapy for Chronic Central Serous Chorioretinopathy. Retina. 2003;23: 752. Burumcek E, Mudun A, Karacorlu S, Arslan MO. Laser photocoagulation for persistent central serous retinopathy: results of long-term follow-up. Ophthalmology. 1997;104: 616–622. Zarnegar A, Ong J, Matsyaraja T, Arora S, Chhablani J. Pathomechanisms in central serous chorioretinopathy: A recent update. Int J Retina Vitreous. 2023;9: 3. Sahoo NK, Ong J, Selvam A, Maltsev D, Sacconi R, Venkatesh R, et al. Longitudinal follow-up and outcome analysis in central serous chorioretinopathy. Eye . 2023;37: 732–738. Mohabati D, Boon CJF, Yzer S. Risk of Recurrence and Transition to Chronic Disease in Acute Central Serous Chorioretinopathy. Clin Ophthalmol. 2020;14: 1165–1175. Piccolino FC, de la Longrais RR, Ravera G, Eandi CM, Ventre L, Abdollahi A, et al. The foveal photoreceptor layer and visual acuity loss in central serous chorioretinopathy. Am J Ophthalmol. 2005;139: 87–99. Kucukevcilioglu M, Durukan AH, Yumusak E, Ozgonul C. Choroidal Thickness Changes After Photodynamic Therapy and Recurrence of Chronic Central Serous Chorioretinopathy. American journal of ophthalmology. 2015. p. 841. Ruiz-Moreno JM, Gutiérrez-Bonet R, Chandra A, Vupparaboina KK, Chhablani J, Ruiz-Medrano J. Choroidal Vascularity Index versus Choroidal Thickness as Biomarkers of Acute Central Serous Chorioretinopathy. Ophthalmic Res. 2023;66: 627–635. Nkrumah G, Paez-Escamilla M, Singh SR, Rasheed MA, Maltsev D, Guduru A, et al. Biomarkers for central serous chorioretinopathy. Ther Adv Ophthalmol. 2020;12: 2515841420950846. Cardillo Piccolino F, Lupidi M, Cagini C, Fruttini D, Nicolò M, Eandi CM, et al. Choroidal Vascular Reactivity in Central Serous Chorioretinopathy. Invest Ophthalmol Vis Sci. 2018;59: 3897–3905. Imamura Y, Fujiwara T, Margolis R, Spaide RF. Enhanced depth imaging optical coherence tomography of the choroid in central serous chorioretinopathy. Retina. 2009;29: 1469–1473. Anwar SM, Majid M, Qayyum A, Awais M, Alnowami M, Khan MK. Medical Image Analysis using Convolutional Neural Networks: A Review. J Med Syst. 2018;42: 1–13. Lakshminarayanan V, Kheradfallah H, Sarkar A, Jothi Balaji J. Automated Detection and Diagnosis of Diabetic Retinopathy: A Comprehensive Survey. Journal of Imaging. 2021;7: 165. Desideri LF, Anguita R, Berger LE, Feenstra HMA, Scandella D, Sznitman R, et al. BASELINE SPECTRAL DOMAIN OPTICAL COHERENCE TOMOGRAPHIC RETINAL LAYER FEATURES IDENTIFIED BY ARTIFICIAL INTELLIGENCE PREDICT THE COURSE OF CENTRAL SEROUS CHORIORETINOPATHY. Retina. 2023. doi:10.1097/IAE.0000000000003965 Xu F, Wan C, Zhao L, You Q, Xiang Y, Zhou L, et al. Predicting Central Serous Chorioretinopathy Recurrence Using Machine Learning. Front Physiol. 2021;12: 649316. Singh SR, Iovino C, Zur D, Masarwa D, Iglicki M, Gujar R, et al. Central serous chorioretinopathy imaging biomarkers. Br J Ophthalmol. 2022;106: 553–558. van Rijssen TJ, Mohabati D, Dijkman G, Theelen T, de Jong EK, van Dijk EHC, et al. Correlation between redefined optical coherence tomography parameters and best-corrected visual acuity in non-resolving central serous chorioretinopathy treated with half-dose photodynamic therapy. PLoS One. 2018;13: e0202549. Deng K, Gui Y, Cai Y, Liang Z, Shi X, Sun Y, et al. Changes in the Foveal Outer Nuclear Layer of Central Serous Chorioretinopathy Patients Over the Disease Course and Their Response to Photodynamic Therapy. Front Med. 2021;8: 824239. Matsumoto H, Sato T, Kishi S. Outer nuclear layer thickness at the fovea determines visual outcomes in resolved central serous chorioretinopathy. Am J Ophthalmol. 2009;148: 105–10.e1. Oishi A, Fang PP, Thiele S, Holz FG, Krohne TU. LONGITUDINAL CHANGE OF OUTER NUCLEAR LAYER AFTER RETINAL PIGMENT EPITHELIAL TEAR SECONDARY TO AGE-RELATED MACULAR DEGENERATION. Retina. 2018;38: 1331–1337. Agrawal R, Salman M, Tan K-A, Karampelas M, Sim DA, Keane PA, et al. Choroidal Vascularity Index (CVI)--A Novel Optical Coherence Tomography Parameter for Monitoring Patients with Panuveitis? PLoS One. 2016;11: e0146344. Table Table 1 is available in the Supplementary Files section. Additional Declarations The authors declare no competing interests. Supplementary Files CSCRstatsMLSupplementaryv1preprint.docx Importance of OCT-derived Biomarkers for the Recurrence of Central Serous Chorioretinopathy using Statistics and Predictive Modelling (Supplementary) Table1.png Table 1 | P-values for statistics and predictive modeling. With the most basic analysis method, a KS test performed on the first visit, only a limited number of OCT-derived parameters were statistically significant in distinguishing between recurrent and non-recurrent cases. The number of statistically significant parameters increased when considering longitudinal data (i.e., composed by multiple visits, capturing the evolution of the disease). Applying multivariate logistic regression to the longitudinal data, the following parameters were statistically significant: SRF, IRF, INL_OPL, CC_CS, CVI. P-values are marked with asterisk when significant: * if < 0.05, ** if < 0.01, *** if < 0.001. Cite Share Download PDF Status: Posted Version 1 posted 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-4170618","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":284153202,"identity":"0e88984e-4029-486d-94e9-83c59c65998e","order_by":0,"name":"Emilien Seiler","email":"","orcid":"","institution":"Platform for Research in Ocular Imaging, Department of Ophthalmology, University of Lausanne, Fondation Asile des Aveugles, Jules Gonin Eye Hospital, Lausanne, Switzerland","correspondingAuthor":false,"prefix":"","firstName":"Emilien","middleName":"","lastName":"Seiler","suffix":""},{"id":284155282,"identity":"f1ffc81a-6f04-4847-bf52-b656147fe68c","order_by":1,"name":"Léon Delachaux","email":"","orcid":"","institution":"Platform for Research in Ocular Imaging, Department of Ophthalmology, University of Lausanne, Fondation Asile des Aveugles, Jules Gonin Eye Hospital, Lausanne, Switzerland","correspondingAuthor":false,"prefix":"","firstName":"Léon","middleName":"","lastName":"Delachaux","suffix":""},{"id":284155283,"identity":"54eb612f-34e7-4c21-a161-c9898dd1f572","order_by":2,"name":"Jennifer Cattaneo","email":"","orcid":"","institution":"Department of Ophthalmology, University of Lausanne, Fondation Asile des Aveugles, Jules Gonin Eye Hospital, Lausanne, Switzerland","correspondingAuthor":false,"prefix":"","firstName":"Jennifer","middleName":"","lastName":"Cattaneo","suffix":""},{"id":284155284,"identity":"b7e79e7c-4724-4d12-866b-3a5f4e6fc987","order_by":3,"name":"Ali Garjani","email":"","orcid":"","institution":"Platform for Research in Ocular Imaging, Department of Ophthalmology, University of Lausanne, Fondation Asile des Aveugles, Jules Gonin Eye Hospital, Lausanne, Switzerland","correspondingAuthor":false,"prefix":"","firstName":"Ali","middleName":"","lastName":"Garjani","suffix":""},{"id":284155285,"identity":"ee78b47d-faa6-4f71-8f3d-88be68298afc","order_by":4,"name":"Alexia Duriez","email":"","orcid":"","institution":"Platform for Research in Ocular Imaging, Department of Ophthalmology, University of Lausanne, Fondation Asile des Aveugles, Jules Gonin Eye Hospital, Lausanne, Switzerland","correspondingAuthor":false,"prefix":"","firstName":"Alexia","middleName":"","lastName":"Duriez","suffix":""},{"id":284155286,"identity":"be70a7f3-a6a6-4e6f-9237-7b2c30f2cb74","order_by":5,"name":"Thibaud Martin","email":"","orcid":"","institution":"Platform for Research in Ocular Imaging, Department of Ophthalmology, University of Lausanne, Fondation Asile des Aveugles, Jules Gonin Eye Hospital, Lausanne, Switzerland","correspondingAuthor":false,"prefix":"","firstName":"Thibaud","middleName":"","lastName":"Martin","suffix":""},{"id":284155287,"identity":"6891d430-482d-48ba-a0ff-5a89d9f923b8","order_by":6,"name":"Jérémy Baffou","email":"","orcid":"","institution":"Platform for Research in Ocular Imaging, Department of Ophthalmology, University of Lausanne, Fondation Asile des Aveugles, Jules Gonin Eye Hospital, Lausanne, Switzerland","correspondingAuthor":false,"prefix":"","firstName":"Jérémy","middleName":"","lastName":"Baffou","suffix":""},{"id":284155288,"identity":"5ae0fc40-f729-4fda-a045-04c3a33fa1e4","order_by":7,"name":"Sepehr Mousavi","email":"","orcid":"","institution":"Platform for Research in Ocular Imaging, Department of Ophthalmology, University of Lausanne, Fondation Asile des Aveugles, Jules Gonin Eye Hospital, Lausanne, Switzerland","correspondingAuthor":false,"prefix":"","firstName":"Sepehr","middleName":"","lastName":"Mousavi","suffix":""},{"id":284155289,"identity":"95b82a05-74b2-4761-b851-87f542c6b584","order_by":8,"name":"Ilenia Meloni","email":"","orcid":"","institution":"Platform for Research in Ocular Imaging, Department of Ophthalmology, University of Lausanne, Fondation Asile des Aveugles, Jules Gonin Eye Hospital, Lausanne, Switzerland","correspondingAuthor":false,"prefix":"","firstName":"Ilenia","middleName":"","lastName":"Meloni","suffix":""},{"id":284155290,"identity":"ea8782b8-f443-4c04-b5ba-84ad1644e83d","order_by":9,"name":"Ciara Bergin","email":"","orcid":"","institution":"Department of Ophthalmology, University of Lausanne, Fondation Asile des Aveugles, Jules Gonin Eye Hospital, Lausanne, Switzerland.","correspondingAuthor":false,"prefix":"","firstName":"Ciara","middleName":"","lastName":"Bergin","suffix":""},{"id":284155291,"identity":"0dda5e12-040e-40f5-a17d-2b2559a94503","order_by":10,"name":"Mattia Tomasoni","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA9ElEQVRIie3OsYrCQBCA4QkBbVa2kwkBfQIhIgQLH2aDYBrF7qpDUq1NvPqCj2Fj58qBNgFbwSYSsDYIx9mIGzgJFm5ai/2rHdiPGQCd7l1jxRMb1cAUYASq/6YRSIIP0iGiwsoJFAS8UBBHSVrT3TZJPmFC6z/pOVt2/bAqHLguXxM37svDNoDWfOBaUYyjkDBmzGIFETmpADoH5po1jqMVMGEaXEF2qSS3nPiXiyQ+ocdATfZyi8dzMnRsSRjBPpSQtP3tfaEVzYcfVsSxHeIJ1jPlYV6S/f32KLX9xTnjkyah4yy5Ksh/+DyKUqDT6XQ6ZXdBj1CagO9JHQAAAABJRU5ErkJggg==","orcid":"https://orcid.org/0000-0001-8775-2384","institution":"Platform for Research in Ocular Imaging, Department of Ophthalmology, University of Lausanne, Fondation Asile des Aveugles, Jules Gonin Eye Hospital, Lausanne, Switzerland","correspondingAuthor":true,"prefix":"","firstName":"Mattia","middleName":"","lastName":"Tomasoni","suffix":""},{"id":284155292,"identity":"218f26b0-62b1-4628-a956-f16c143f34b4","order_by":11,"name":"Chiara M Eandi","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA1klEQVRIiWNgGAWjYDACZjgrsfFBBYMNA4MECVqaDc4wpBGhBQES2CTOMBwmrIW/nfnZhx8MNvLm7MltFQdqmBn4Zzfg1yJxmM14Zg9DmuHOnodtNw4cY2OQuHMAvxYDZgZjBh6Gw4wbbiS23f7AxsNgIJFASAv7Z8Y/DP/tQVoKDvyTIEYLjzEzD8OBRJAWhoNtBoS1SBzmKWaWMUhO3nDmYbPEwb4EHokbBLTw9x/fzPimws52w/H0hx8OfPsvxz+DgBao8xBMHmLUj4JRMApGwSggAAAmHkHEhznN+AAAAABJRU5ErkJggg==","orcid":"","institution":"Department of Ophthalmology, University of Lausanne, Fondation Asile des Aveugles, Jules Gonin Eye Hospital, Lausanne, Switzerland.","correspondingAuthor":true,"prefix":"","firstName":"Chiara","middleName":"M","lastName":"Eandi","suffix":""}],"badges":[],"createdAt":"2024-03-26 14:34:12","currentVersionCode":1,"declarations":{"humanSubjects":true,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":true,"humanSubjectConsent":true,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-4170618/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-4170618/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":53748462,"identity":"230e0a33-a705-4523-ba39-1d0bad4ea45f","added_by":"auto","created_at":"2024-03-29 18:23:51","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":472740,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eManifestations of Central serous chorioretinopathy (CSCR) signs in OCT-Line scans (OCT-EDI). \u003c/strong\u003eCSCR is a posterior segment disease characterized by accumulation of subretinal fluid that, in acute forms, resolves spontaneously. However, about a third of the cases experience recurrences that might cause severe and irreversible vision loss due to anatomical outer retinal and retinal pigment epithelium changes.\u003cstrong\u003e a\u003c/strong\u003e. Line scan showing pathological accumulation of fluid \u003cstrong\u003eb\u003c/strong\u003e. Line scan showing positive response to treatment and resolution of the pathological state.\u003c/p\u003e","description":"","filename":"Figure1.png","url":"https://assets-eu.researchsquare.com/files/rs-4170618/v1/2074b9836260505285df229d.png"},{"id":53748461,"identity":"41000e04-e967-4a6e-bf89-3b34426b88a4","added_by":"auto","created_at":"2024-03-29 18:23:51","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":212607,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eLongitudinal dataset comprising OCT-derived parameters in CSCR patients (N=337). \u003c/strong\u003eX Axis = time. Y Axis = accumulation of IRF and SRF. \u003cstrong\u003ea.\u003c/strong\u003e a non-recurrent case characterized by a single episode (i.e., accumulation of fluid), resolved by treatment. \u003cstrong\u003eb. \u003c/strong\u003erecurrent case: an acute episode following a previous episode, despite treatment.\u003c/p\u003e","description":"","filename":"Figure2.png","url":"https://assets-eu.researchsquare.com/files/rs-4170618/v1/18ecdfbed0f1de82cf8096d7.png"},{"id":53748467,"identity":"f0c4e97e-ba69-4662-a741-d407174c9f6d","added_by":"auto","created_at":"2024-03-29 18:23:52","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":178178,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eData extraction and processing pipeline\u003c/strong\u003e. We extracted the list of patients who visited the medical retina unit at Jules-Gonin Eye Hospital between March 2017 and September 2022. We crossed this list with information extracted from the electronic medical records (EMR) indicating a diagnosis of Central Serous Chorioretinopathy (CSCR) and extracted the corresponding OCT scans for the totality of their visits generating a longitudinal data set. We only considered those patients who had signed the General Consent for reuse of their data. We extracted parameters from the OCT scans using AI-based methods. An expert ophthalmologist labeled patients as recurrent or non-recurrent. Finally, we performed statistical tests and tested the efficacy of a machine learning model to study each parameter as well as its link to recurrence.\u003c/p\u003e","description":"","filename":"Figure3.png","url":"https://assets-eu.researchsquare.com/files/rs-4170618/v1/930637aeb682a56256cc8819.png"},{"id":53748464,"identity":"57a13203-ba8d-40dd-bd5f-cbc3cf061026","added_by":"auto","created_at":"2024-03-29 18:23:51","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":176501,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eCorrelation between OCT-derived parameters\u003c/strong\u003e. An expert ophthalmologist suggested a list of parameters that might play a role in the detection of CSCR recurrent events. We then extracted these parameters from the imaging dataset and correlated them among each other obtaining several clusters. A first cluster was composed of parameters related to the choroid: CC+CS thickness, CVI and PV_AREA. A second cluster was composed of thicknesses of layers of the inner retina (RNFL, GCL+IPL, and INL+OPL). The third cluster contained outer retinal layer thicknesses (ONL, PR+RPE), scores related to their integrity (PED, DSCORE) and the fluids (SRF and IRF).\u003c/p\u003e","description":"","filename":"Figure4.png","url":"https://assets-eu.researchsquare.com/files/rs-4170618/v1/014abfcbf472e1840c54c9bf.png"},{"id":53748466,"identity":"4f5f81ee-246c-4baf-b66f-18982d978528","added_by":"auto","created_at":"2024-03-29 18:23:51","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":129792,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eImportance of predictive biomarkers for CSCR, across analysis methods\u003c/strong\u003e. The effects of each biomarker either with the Cohen’s d effect size or simply the coefficient in the case of predictive modeling. \u003cstrong\u003ea. \u003c/strong\u003eCohen’s d effect sizes obtained from the statistical analysis (KS test) of biomarkers extracted from images collected during the first visit \u003cstrong\u003eb. \u003c/strong\u003eCohen’s d effect sizes obtained from the statistical analysis (KS test) of longitudinal data (i.e. averaging all visits of the initial acute CSCR episode) \u003cstrong\u003ec. \u003c/strong\u003efeaturing importance derived from predictive modeling, specifically, the coefficients of the logistic regression model obtained fitting one biomarker at a time.\u003c/p\u003e","description":"","filename":"Figure5.png","url":"https://assets-eu.researchsquare.com/files/rs-4170618/v1/ca17421dfda1f1d807954684.png"},{"id":53750438,"identity":"bf9eba33-e92c-4770-9ea4-92d676d7f38a","added_by":"auto","created_at":"2024-03-29 18:39:52","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1187115,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4170618/v1/e6305c1e-f587-4088-a60e-98434a90a444.pdf"},{"id":53748460,"identity":"9777f162-a8a9-4351-ba08-1fbd90435fe3","added_by":"auto","created_at":"2024-03-29 18:23:51","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":252579,"visible":true,"origin":"","legend":"\u003cp\u003eImportance of OCT-derived Biomarkers for the Recurrence of Central Serous Chorioretinopathy using Statistics and Predictive Modelling (Supplementary)\u003c/p\u003e","description":"","filename":"CSCRstatsMLSupplementaryv1preprint.docx","url":"https://assets-eu.researchsquare.com/files/rs-4170618/v1/08ebb0398c5653c206b54eb1.docx"},{"id":53749357,"identity":"575b8a35-27e2-4f8d-b874-f1841f6d0ae5","added_by":"auto","created_at":"2024-03-29 18:31:51","extension":"png","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":259759,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eTable 1 | P-values for statistics and predictive modeling. \u003c/strong\u003eWith the most basic analysis method, a KS test performed on the first visit, only a limited number of OCT-derived parameters were statistically significant in distinguishing between recurrent and non-recurrent cases. The number of statistically significant parameters increased when considering longitudinal data (i.e., composed by multiple visits, capturing the evolution of the disease). Applying multivariate logistic regression to the longitudinal data, the following parameters were statistically significant: SRF, IRF, INL_OPL, CC_CS, CVI. \u003cem\u003eP-values are marked with asterisk when significant: * if \u0026lt; 0.05, ** if \u0026lt; 0.01, *** if \u0026lt; 0.001.\u0026nbsp;\u003c/em\u003e\u003c/p\u003e","description":"","filename":"Table1.png","url":"https://assets-eu.researchsquare.com/files/rs-4170618/v1/e22f871bec8c74f343077f20.png"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003eImportance of OCT-derived Biomarkers for the Recurrence of Central Serous Chorioretinopathy using Statistics and Predictive Modelling\u003c/p\u003e","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eCentral Serous Chorioretinopathy (CSCR) is an idiopathic chorioretinal disease characterized by localized and limited serous detachment of the neurosensory retina (shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e) and may be associated with focal retinal pigment epithelium detachments.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThe term 'central' refers to the form of the disease in which the serous detachment involves the macular area and causes visual symptoms accordingly. However, not all patients have macular involvement: some cases may present with extra macular serous detachments, as it is often observed in the contralateral eye of patients with active CSCR [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eCSCR is the fourth most common retinopathy after age-related macular degeneration, diabetic retinopathy and branch retinal vein occlusion [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eIt affects 5.8 individuals per 100,000 and its incidence is 5 times higher in men [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. CSCR typically occurs in males in their late 30s to 50s, even though it can occur at later ages, particularly in women and patients affected by atypical chronic forms [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eThe etiology of the disease is poorly understood, granting the need for further research into the factors influencing its manifestations. No underlying pathophysiologic mechanisms have been proven, but CSCR is thought to occur due to hyper-permeable choroidal capillaries which, in association with retinal pigment epithelium dysfunction, cause a serous detachment of the neurosensory retina [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eSome of the known risk factors are the use of steroids [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e], stress [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e], psychopathology like \u0026ldquo;type A\u0026rdquo; personality pattern [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e], hypertension, endocrine changes, drugs (sympathomimetic, MEK-inibitors) [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e] and obstructive sleep apnea [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eCSCR is classified into different forms according to the duration of the disease and clinical features. In the acute form of the disease, Subretinal Fluid (SRF) resolves spontaneously within 4 months of the onset of symptoms without anatomical changes [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]. Non-resolving (or \"persistent\") CSCR is characterized by persistence of SRF for more than 4 months, associated with elongated outer segments photoreceptor on B-scan SD-OCT [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]. Chronic CSCR (or \"diffuse retinal epitheliopathy\"), presents diffuse decompensation of the RPE with or without SRF. The chronic form of CSCR has a strong genetic determinant as 52% of patients have an affected relative [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. CSCR is considered recurrent if there is a reappearance of SRF accumulation after a completely resolved acute episode. Conversely, it is classified as inactive if there is a history of acute CSCR but no SRF at the time of assessment.\u003c/p\u003e \u003cp\u003eNo treatment is generally required for the first episode, while for recurrent, non-resolving, and chronic forms the standard treatment is photodynamic therapy with verteporfin [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e] or focal argon laser treatment [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]. After complete resolution, up to one-third of patients experience recurrence of the disease with persistence of SRF accumulation, which can lead to long-term visual impairment [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. Retinal Pigment Epithelium (RPE) and changes in external layers are reported as indicative of future recurrences or poor visual prognosis [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e][\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e][\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e][\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eVarious observable parameters have been reported as potential biomarkers for CSCR, such as: choroidal thickness [\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e], Choroidal Vascularity Index (CVI, i.e., the ratio between vascular area in the choroid with the total area of the choroid, see Section \u003cspan refid=\"Sec1\" class=\"InternalRef\"\u003e1\u003c/span\u003e of SI and Fig. S1a) [\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e][\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e][\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e] and presence of pachy vessels [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e] (i.e., vessels dilatation, Fig. S1b).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eRecent advances in Machine Learning (ML) have enabled the automatic processing of medical data, especially images. In particular, Deep Learning (DL) methods for image classification or segmentation have been found to be very effective in the analysis of medical images, such as brain MRI for Alzheimer diagnosis or CT scans for classification of lung tissue and lung diseases diagnosis [\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e]. These innovations can help clinicians in performing their daily practice when implemented in the form of Expert Support Systems. For certain diseases, ML and DL approaches have allowed computers to detect disease signs directly, as in case of diabetic retinopathy [\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e]. In the case of CSCR, research in automatic biomarker extraction and predictive modeling is still lacking.\u003c/p\u003e \u003cp\u003ePrior research identified several biomarkers to differentiate between acute and chronic/recurrent CSCR using AI-based segmentation followed by statistical testing to determine if the extracted parameters were differently distributed between the two groups [\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e]. Other prior work attempted to automatically predict recurrence of CSCR directly from OCT Scans using DL [\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e], but the imbalance in the dataset makes it difficult to assess the results.\u003c/p\u003e \u003cp\u003eNowadays there is no effective way of predicting recurrence of CSCR, making it challenging to determine if and when a patient should be reevaluated after acute phase resolution. This research aims to identify predictive biomarkers associated with CSCR recurrence, paving the way towards automating predictions of the recurrence of this disease.\u003c/p\u003e"},{"header":"2. Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\n\u003ch2\u003e2.1 Data collection and biomarkers extraction\u003c/h2\u003e\n\u003cp\u003eWe utilized longitudinal real-world data extracted from Jules-Gonin\u0026rsquo;s Eye Hospital\u0026rsquo;s Electronic Medical Record system (EMR), comprising all patients who visited the medical retina unit at Jules-Gonin Eye Hospital between March 2017 and September 2022 (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003e).\u003c/p\u003e\n\u003cp\u003eWe extracted our dataset using an in-house software tool, Cohort Builder. First, Cohort Builder queried the EMRs to identify patients with a diagnosis of Central Serous Chorioretinopathy (CSCR). It then verified the hospital's General Consent database, and proceeded to the extraction of the raw fovea-centered OCT imaging dataset, consisting of 30\u0026deg; OCT cubes (98 B-scans, 10 ART) and Enhanced-depth imaging OCT line scans from the storage system (Heyex 2, Heidelberg Engineering, Heidelberg, Germany) of the OCT Scanners (SPECTRALIS\u0026reg;, Heidelberg Engineering, Heidelberg, Germany).\u003c/p\u003e\n\u003cp\u003eLeveraging Cohort Builder, we uploaded the imaging dataset to an AI-powered generic image viewer for ophthalmology, Discovery\u0026reg; by RetinAI. This software is specialized in OCT-scan image analysis, specifically, in the estimation of image-derived biomarkers. Cohort Builder interacted with it automatically, via its API, extracting: volume of Subretinal Fluid (SRF), volume of Intraretinal Fluid (IRF), area of Pigment Epithelial Detachment (PED) in the central, pericentral and peripheral areas, and thicknesses of the following 6 layers (or combinations thereof) in the central area (C1): Chorio-Capillaris and Choroidal Stroma (CC\u0026thinsp;+\u0026thinsp;CS), Photo-Receptors and Retinal Pigment Epithelium (PR\u0026thinsp;+\u0026thinsp;RPE), Outer Nuclear Layer (ONL), Inner Nuclear Layer and Outer Plexiform Layer (INL\u0026thinsp;+\u0026thinsp;OPL), Ganglion Cell Layer and Inner Plexiform Layer (GCL\u0026thinsp;+\u0026thinsp;IPL), Retinal Nerve Fiber Layer (RNFL).\u003c/p\u003e\n\u003cp\u003eWe extracted three additional biomarkers using in-house algorithms, implemented in Cohort Builder (see Supplementary Methods): the Choroidal Vascularity Index (CVI), a score estimating disruption of the PR\u0026thinsp;+\u0026thinsp;RPE layer (DSCORE), and an estimation of the area of the largest Choroidal Pachyvessel (PV_AREA).\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec4\" class=\"Section2\"\u003e\n\u003ch2\u003e2.2 Data annotation and quality control\u003c/h2\u003e\n\u003cp\u003eWe collected expert annotations using an in-house tool, Evolvision, thanks to which an ophthalmologist retina specialist (CME) labeled recurrent and non-recurrent cases. Evolvision is a GUI-based program that displays the results of the analysis performed by Cohort Builder, offering the annotator a comprehensive view of biomarker evolution over time, encompassing all biomarkers for each patient visit. If the annotator selects a specific visit, Evolvision shows the corresponding OCT line scan for visual inspection. In instances where certain patient time series prove challenging to annotate, Evolvision reports the Hospital PID so the retinal specialist may refer to the patient's medical records to evaluate the whole clinical case.\u003c/p\u003e\n\u003cp\u003eThe patient time series was assigned one of three labels (as shown in Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e): \"non-recurrent\" indicated a single acute CSCR episode characterized by an accumulation of SRF that was resolved spontaneously; \"recurrent\" when the patient alternated periods characterized by SRF accumulation to periods when SRF would be complete resolution; finally, cases characterized by persistence of SRF were labeled as \"inconclusive\", warranting exclusion from further analysis. The acquisition and labeling process are schematised in Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003e.\u003c/p\u003e\n\u003cp\u003eEvolvision also allowed us to select a sub-portion of the time series: thanks to this feature, the annotator truncated \"recurrent\" cases to the initial acute CSCR episode, so that we could evaluate the link between a future recurrence of the disease and this first episode, which is a first step towards predictive modeling.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec5\" class=\"Section2\"\u003e\n\u003ch2\u003e3.3 Feature-importance analysis using statistics and Machine Learning\u003c/h2\u003e\n\u003cp\u003eWe performed a correlation analysis on the extracted biomarkers based on Pearson Correlation Coefficient (R). We subsequently bi-clustered the resulting correlation matrix (as shown in Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003e).\u003c/p\u003e\n\u003cp\u003eWe calculated the distributions of each biomarker based solely on the first visits of the patients, and obtained p-values (p) and the effect sizes (d) by comparing patient groups with a two-tailed Welch's t-test. We corrected for multiple hypotheses testing using a Benjamini-Hochberg procedure (BH), with a significance threshold (\u0026alpha;) of 0.05.\u003c/p\u003e\n\u003cp\u003eWe performed an equivalent analysis leveraging the longitudinal nature of our data. We calculated biomarkers distributions based on the means of each biomarker across the truncated time series of each patient (i.e., not considering any visits after the recurrence of the disease). We performed a two-sample Kolmogorov-Smirnov test, obtaining p-values. We again corrected for multiple hypotheses using BH (as reported in Table\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e).\u003c/p\u003e\n\u003cp\u003eWe computed the Cohen\u0026rsquo;s d effect sizes for the distributions based on first visits and then based on averaged patients (See Supplementary Methods).\u003c/p\u003e\n\u003cp\u003eFinally, we delved into Machine Learning (ML) to determine the relative importance of each biomarker in building a predictive model for the recurrence of CSCR. We fitted a Logistic Regression (LR) model to distinguish between visits from recurrent and non-recurrent patients. We assessed the predictive power of each biomarker by extracting the p-value that estimates its statistical significance and analyzing the coefficient that defines its impact on the model output (as reported in Table\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e and Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e5\u003c/span\u003e).\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec6\" class=\"Section2\"\u003e\n\u003ch2\u003e3.3 Code and data availability\u003c/h2\u003e\n\u003cp\u003eA detailed experimental manual and the code that implements the experiments is available at \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://github.com/JulesGoninRIO/cscr_stats_ml\u003c/span\u003e\u003c/span\u003e. To discuss access to the dataset (DTA) please contact the corresponding author.\u003c/p\u003e\n\u003c/div\u003e"},{"header":"3. Results","content":"\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003e3.1 Baseline characteristics\u003c/h2\u003e \u003cp\u003eFollowing quality control measures, we analyzed 5211 OCT Scans from 344 eyes belonging to 255 patients. The mean age was 53 years (range 47, SD +/-12 years). At birth, 78% were men and 22% women. The mean number of visits per patient was 16 (+/-14 SD). Such a large variance is intrinsic in the nature of our dataset (as further reported in the \u003cspan refid=\"Sec11\" class=\"InternalRef\"\u003eDiscussion\u003c/span\u003e section). The average duration of follow-up among the patients was 6 years (range 11, SD +/-3 years). 178 eyes were labeled as recurrent and 109 eyes as non-recurrent, while 57 were excluded because of poor quality of the OCT scans\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003e3.2 Correlation among OCT-derived biomarkers\u003c/h2\u003e \u003cp\u003eA correlation analysis between the extracted biomarkers across all patients and all visits revealed several clusters (as shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e4\u003c/span\u003e). A first cluster is composed of parameters related to the choroid: CC_CS thickness, CVI and PV_AREA. A second cluster includes thicknesses of layers of the inner retina (RNFL, GCL_IPL, and INL_OPL). The third cluster contains outer retinal layer thicknesses (ONL, PR\u0026thinsp;+\u0026thinsp;RPE), parameters related to their integrity (PED, DSCORE) and the presence of fluids (SRF and IRF).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec10\" class=\"Section2\"\u003e \u003ch2\u003e3.3 Importance of OCT-derived biomarkers\u003c/h2\u003e \u003cp\u003eOur multi-step analysis incorporated increasingly more sophisticated approaches, which gradually included more information. In particular, we initially analyzed only the first visit, using inferential statistics; then, we extended our analysis to consider longitudinal data (averaging every biomarker over the truncated time series of the patient); finally, we applied predictive modeling techniques to the longitudinal data. The number of statistically significant parameters progressively increased (as shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e5\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eUsing inferential statistics on the single first visit, we identified the following biomarkers as statistically significant: age (p\u0026thinsp;=\u0026thinsp;0.14E-1, d\u0026thinsp;=\u0026thinsp;0.38), ONL (p\u0026thinsp;=\u0026thinsp;0.20E-1, d=-0.36), DSCORE (p\u0026thinsp;=\u0026thinsp;0.29E-2, d\u0026thinsp;=\u0026thinsp;0.47).\u003c/p\u003e \u003cp\u003eExtending the analysis to the longitudinal data set, uncovered additional statistically significant parameters, in addition to the ones already reported: age (p\u0026thinsp;=\u0026thinsp;0.41E-1, d\u0026thinsp;=\u0026thinsp;0.39), ONL (p\u0026thinsp;=\u0026thinsp;0.30E-1, d=-0.45), PED (p\u0026thinsp;=\u0026thinsp;0.2E-1, d\u0026thinsp;=\u0026thinsp;0.22E-1), DSCORE (p\u0026thinsp;=\u0026thinsp;0.21E-1, d\u0026thinsp;=\u0026thinsp;0.43).\u003c/p\u003e \u003cp\u003eApplying predictive modeling (i.e., Logistic Regression, LR), providing one biomarker at a time as a predictor, we identified the following as statistically significant: age (p\u0026thinsp;=\u0026thinsp;0.10E-7, coef\u0026thinsp;=\u0026thinsp;0.34), SRF (p\u0026thinsp;=\u0026thinsp;0.99E-12, coef\u0026thinsp;=\u0026thinsp;0.37), IRF (p\u0026thinsp;=\u0026thinsp;0.30E-5, coef\u0026thinsp;=\u0026thinsp;0.37), INL\u0026thinsp;+\u0026thinsp;OPL (p\u0026thinsp;=\u0026thinsp;0.24E-1, coef=-0.13), ONL (p\u0026thinsp;=\u0026thinsp;0.52E-12, coef=-0.43), CC\u0026thinsp;+\u0026thinsp;CS (p\u0026thinsp;=\u0026thinsp;0.19E-1, coef\u0026thinsp;=\u0026thinsp;0.13), DSCORE (p\u0026thinsp;=\u0026thinsp;0.11E-10, coef\u0026thinsp;=\u0026thinsp;0.39), CVI (p\u0026thinsp;=\u0026thinsp;0.27E-4, coef=-0.25), and PED (p\u0026thinsp;=\u0026thinsp;0.2E-8, coef\u0026thinsp;=\u0026thinsp;0.36).\u003c/p\u003e \u003c/div\u003e"},{"header":"4. Discussion","content":"\u003cp\u003eIn this retrospective study we were able to identify biomarkers predictive of recurrence of CSCR by the means of AI and ML methods. This research represents the first investigation addressing a fundamental clinical inquiry: determining which patients are prone to experiencing recurrence following an initial, spontaneously resolved episode of CSCR. In fact, recurrence of SRF is a negative prognostic factor on long-term visual outcomes causing RPE and PR alterations [\u003cspan class=\"CitationRef\"\u003e20\u003c/span\u003e]. The gold standard treatment of PDT with verteporfin showed good efficacy on fluid resolution and preservation of functional visual acuity [\u003cspan class=\"CitationRef\"\u003e11\u003c/span\u003e]. Therefore, it would be helpful to be able to predict which patients will experience recurrences in order to optimize follow-up schedule and promptly arrange treatment.\u003c/p\u003e\n\u003cp\u003eThe current paper aims to identify optical coherence tomography (OCT)-derived parameters linked to CSCR recurrence, by means of Discovery\u0026reg; software (by RetinAI) for identifying and quantifying different SD-OCT biomarkers in CSRC.\u003c/p\u003e\n\u003cp\u003eOur study identified nine OCT predictive biomarkers for the recurrence of CSCR, in particular, volume of Subretinal Fluid (SRF), volume of Intraretinal Fluid (IRF), area of Pigment Epithelial Detachment (PED) in the central, pericentral and peripheral areas, and thicknesses of the following 6 layers (or combinations thereof) in the central area (C1): Chorio-Capillaris and Choroidal Stroma (CC\u0026thinsp;+\u0026thinsp;CS), Photo-Receptors and Retinal Pigment Epithelium (PR\u0026thinsp;+\u0026thinsp;RPE), Outer Nuclear Layer (ONL), Inner Nuclear Layer and Outer Plexiform Layer (INL\u0026thinsp;+\u0026thinsp;OPL), Ganglion Cell Layer and Inner Plexiform Layer (GCL\u0026thinsp;+\u0026thinsp;IPL), Retinal Nerve Fiber Layer (RNFL).\u003c/p\u003e\n\u003cp\u003eThree additional biomarkers were extracted: the Choroidal Vascularity Index (CVI), a score estimating disruption or the PR\u0026thinsp;+\u0026thinsp;RPE layer (DSCORE), and an estimation of the area of the largest Choroidal Pachyvessel (PV_AREA).\u003c/p\u003e\n\u003cp\u003eThe multivariate logistic regression to the longitudinal data, demonstrated that the following parameters were statistically significant: SRF, IRF, INL_OPL, CC_CS, CVI (as shown in Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003e). Therefore, we showed that all these parameters were related to CSCR recurrence.\u003c/p\u003e\n\u003cp\u003eRecently, imaging biomarkers associated with acute and chronic CSCR were analyzed with an AI software [\u003cspan class=\"CitationRef\"\u003e25\u003c/span\u003e]. The authors found a significant increase in thickness in both outer retinal layers (ONL, PR and RPE layer) and inner retinal layers (INL and GCL), and SRF in patients with acute CSCR compared to chronic CSCR at baseline. However, in our study we identified more parameters independently of the type of CSCR, which might better reflect the real-world practice. Also, Xu et al. analyzed recurrences in CSCR patients with six different models of AI and ML [\u003cspan class=\"CitationRef\"\u003e26\u003c/span\u003e].\u003c/p\u003e\n\u003cp\u003eImaging has become a key factor in clinical practice since the introduction of fast, high resolution machines. Therefore, the volume of data to analyze increased and new systems of automatic analysis are now available. The application of AI and ML models now becomes the next step forward in the optimization of data analysis mostly to predicting outcomes.\u003c/p\u003e\n\u003cp\u003eNowadays, by the word 'biomarker', scientists refer to morphological and structural changes that can provide important information on the stage of a disease. The search for new biomarkers in retinal diseases has been a field of great interest in recent years. This research has been driven by the need for earlier and more accurate diagnosis, better prognostication and the development of target therapies.\u003c/p\u003e\n\u003cp\u003eMany researches have been carried out on biomarkers in retinal diseases such as age-related macular degeneration and diabetic macular edema. The predictive role of retinal biomarkers in CSCR has been reported in a recent study [\u003cspan class=\"CitationRef\"\u003e27\u003c/span\u003e]. In particular, changes in central macular thickness and subfoveal choroidal thickness were associated to the resolution of CSCR, while the amount of SRF to the visual acuity outcomes. Moreover, the morphology of the retinal layers, such as ONL thinning, PR layer elongation, ELM and ellipsoid zone disruption was previously associated with longer duration of SRF and worse visual acuity [\u003cspan class=\"CitationRef\"\u003e17\u003c/span\u003e, \u003cspan class=\"CitationRef\"\u003e28\u003c/span\u003e, \u003cspan class=\"CitationRef\"\u003e29\u003c/span\u003e].\u003c/p\u003e\n\u003cp\u003eThe definition and validation of predictive biomarkers for central serous chorioretinopathy recurrence is crucial for physicians in daily clinical practice for several reasons. Firstly, predictive biomarkers may allow early diagnosis of individuals at risk of CSCR recurrence. Early diagnosis allows for early intervention and management, potentially preventing serious complications and preserving vision. Furthermore, biomarkers could help tailor treatment strategies to individual patient profiles.\u003c/p\u003e\n\u003cp\u003eBy identifying patients most likely to relapse, clinicians can offer customized treatment plans.\u003c/p\u003e\n\u003cp\u003eBy understanding the likelihood of recurrence, physicians can set realistic expectations and plan for long-term management, and by identifying high-risk patients, physicians can optimize the use of resources by defining proper follow-up in relation to the likelihood of recurrence.\u003c/p\u003e\n\u003cp\u003eBiomarkers contribute to the advancement of research and development efforts in understanding the underlying mechanism and risk factors associated with CSCR recurrence. They can lead to the development of more effective therapies and preventive strategies.\u003c/p\u003e\n\u003cp\u003eDespite the promising results linking OCT-derived biomarkers with a future recurrence of CSCR, our study was subject to several limitations that warrant discussion.\u003c/p\u003e\n\u003cp\u003eOne of the primary challenges was the variability in the number of visits of each patient, particularly pronounced among chronic or recurrent CSCR cases since recurrent patients are examined more often and for longer periods. Our dataset also presented a wide variability in terms of Subretinal Fluid (SRF) volumes. SRF can range from minimal (around 10 nL) to significant (up to 10,000 nL), complicating the identification and categorization of CSCR episodes. Despite a general definition of a CSCR episode as an accumulation of SRF which is followed by drying up of the fluid, the diverse manifestations based on the patient\u0026rsquo;s SRF range make it challenging to clearly distinguish between multiple occurrences and a single episode. This made it challenging even for medical professionals with decades of experience in the disease, underlining the complexity of the labeling task.\u003c/p\u003e\n\u003cp\u003eIn our study, a particularly significant limitation was the binary labeling system used to classify CSCR cases as either recurrent or non-recurrent. This classification framework, while straightforward and appropriate given the size of our dataset, proved insufficient handling the inherent complexities and ambiguities of certain patient cases. Especially challenging was the accurate categorization of chronic patients, which are characterized by fluctuating levels of SRF. Applying our labeling system led to chronic patients being labeled as recurrent cases, despite the nuances in their condition that might suggest a more complex categorization. This subgroup, characterized by SRF volumes that persistently remain above zero, likely represents a distinct clinical picture. In a few cases, chronic patients were excluded from our analysis. Their exclusion, while necessary for the integrity of our current analysis, highlights the limitations of a binary classification system which future work could address via multi-label annotation.\u003c/p\u003e\n\u003cp\u003eAdditionally, an intrinsic limitation of our study is that some patients who were classified as non-recurrent (based on the data available at the time the dataset was collected), might have experienced a CSCR episode post-data collection. This potential for future occurrences not captured in our current dataset underscores the dynamic nature of CSCR and the challenges inherent in predicting its course.\u003c/p\u003e\n\u003cp\u003eLooking at the correlation matrix we found that ONL thickness strongly correlated with D_SCORE, consistently with what is reported in the literature [\u003cspan class=\"CitationRef\"\u003e30\u003c/span\u003e, \u003cspan class=\"CitationRef\"\u003e31\u003c/span\u003e].\u003c/p\u003e\n\u003cp\u003eFitting one biomarker at a time, the logistic regression results in a pseudo R-squared (~\u0026thinsp;proportion of variance explained) that is at maximum 3.0%. Then, the Logistic Regression (LR) model fitted with all our parameters at the same time explained only 8.7% of the variance among patients.\u003c/p\u003e\n\u003cp\u003eFinally, our approach did not make explicit use of the time dimension inherent in our data. Leveraging methods able to consider the evolution of biomarkers over time might enable the prediction of CSCR recurrence and pave the way for translational medical applications.\u003c/p\u003e\n\u003cp\u003eOur analysis revealed correlations between several biomarkers and CSCR recurrence. Applying logistic regression (LR) to all biomarkers explained 8.7% of the variance among patients. Despite the usefulness of these results in elucidating the importance of each biomarker, this suggests a limited capacity of these models to adequately capture the complexities of CSCR recurrence.\u003c/p\u003e\n\u003cp\u003eLooking ahead, significant work remains before these models can be used in clinical applications. A critical unexplored aspect in our current approach is the incorporation of the time dimension in data analysis. Time-series analysis could significantly improve predictive accuracy, opening avenues for practical medical applications. However, advanced modeling techniques like deep learning, while promising enhanced predictive performance, often reduce the model's explainability, a critical factor in clinical settings. Therefore, our future research will aim to balance sophisticated modeling capabilities with the clarity and transparency essential for clinical applicability.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eBy applying inferential statistics and Machine Learning methods on a dataset of 5,211 OCT Scans from 344 eyes of 255 patients, we identified 9 predictive biomarkers for the recurrence of CSCR.\u003c/p\u003e\n\u003cp\u003eAnalyzing data from the first visit, we found that the following biomarkers were differentially distributed in patients with recurrent CSCR compared to non-recurrent ones: age (p\u0026thinsp;=\u0026thinsp;0.14E-1, d\u0026thinsp;=\u0026thinsp;0.38), Outer Nuclear Layer thickness (p\u0026thinsp;=\u0026thinsp;0.20E-1, d=-0.36) and disruption of the photoreceptor and pigment retinal epithelium (p\u0026thinsp;=\u0026thinsp;0.29E-2, d\u0026thinsp;=\u0026thinsp;0.47).\u003c/p\u003e\n\u003cp\u003eIn addition to the parameters mentioned above, when analyzing the data longitudinally over the whole first episode of CSCR, we found that the volume of Pigmented Epithelium Detachment was also a statistically significant biomarker of recurrence (p\u0026thinsp;=\u0026thinsp;0.2E-1, d\u0026thinsp;=\u0026thinsp;0.22E-1).\u003c/p\u003e\n\u003cp\u003eFinally using predictive modeling, we confirmed and extended the results obtained above using inferential statistics, establishing that the following 9 biomarkers were statistically significant predictors: age (p\u0026thinsp;=\u0026thinsp;0.10E-08, coef\u0026thinsp;=\u0026thinsp;0.34), volume of Subretinal Fluid (p\u0026thinsp;=\u0026thinsp;0.99E-10, coef\u0026thinsp;=\u0026thinsp;0.37), volume of Intraretinal Fluid (p\u0026thinsp;=\u0026thinsp;0.30E-5, coef\u0026thinsp;=\u0026thinsp;0.37), thickness of Inner Nuclear and Outer Plexiform Layers (p\u0026thinsp;=\u0026thinsp;0.24E-1, coef=-0.13), thickness of Outer Nuclear Layer (p\u0026thinsp;=\u0026thinsp;0.52E-12, coef=-0.43), thickness Choriocapillaris layer (p\u0026thinsp;=\u0026thinsp;0.19E-1, coef\u0026thinsp;=\u0026thinsp;0.13), disruption of Photoreceptor and Retinal Pigment Epithelium Layers (p\u0026thinsp;=\u0026thinsp;0.11E-10, coef\u0026thinsp;=\u0026thinsp;0.39), Choroidal Vascularity Index (p\u0026thinsp;=\u0026thinsp;0.27E-4, coef=-0.25), and volume of Pigment Epithelium Detachment (p\u0026thinsp;=\u0026thinsp;0.20E-08, coef\u0026thinsp;=\u0026thinsp;0.36).\u003c/p\u003e\n\u003cp\u003eThese results have the potential to enable future developments in automatic prediction of CSCR recurrence, paving the way for translational medical applications.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003eThis study has obtained ethics approval from \u0026ldquo;Commission cantonale d'\u0026eacute;thique de la recherche sur l'\u0026ecirc;tre humain (CER-VD)\u0026rdquo; (project PB_ 2017-00493, Canton of Vaud, Switzerland)\u003c/p\u003e\u003ch2\u003eAcknowledgements\u003c/h2\u003e\n\u003cp\u003eThis work was supported by the Swiss Personalized Health Network (2018DRI13 to Thomas J. Wolfensberger) and by the Claire and Selma Kattenburg Foundation.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eLiew G, Quin G, Gillies M, Fraser-Bell S. Central serous chorioretinopathy: a review of epidemiology and pathophysiology. Clin Experiment Ophthalmol. 2013;41: 201\u0026ndash;214.\u003c/li\u003e\n\u003cli\u003eWang M, Munch IC, Hasler PW, Pr\u0026uuml;nte C, Larsen M. Central serous chorioretinopathy. Acta Ophthalmol. 2008;86: 126\u0026ndash;145.\u003c/li\u003e\n\u003cli\u003eKitzmann AS, Pulido JS, Diehl NN, Hodge DO, Burke JP. The incidence of central serous chorioretinopathy in Olmsted County, Minnesota, 1980-2002. Ophthalmology. 2008;115: 169\u0026ndash;173.\u003c/li\u003e\n\u003cli\u003eSartini F, Figus M, Nardi M, Casini G, Posarelli C. Non-resolving, recurrent and chronic central serous chorioretinopathy: available treatment options. Eye . 2019;33: 1035\u0026ndash;1043.\u003c/li\u003e\n\u003cli\u003ePiccolino FC, Borgia L, Zinicola E, Zingirian M. Indocyanine green angiographic findings in central serous chorioretinopathy. Eye . 1995;9 ( Pt 3): 324\u0026ndash;332.\u003c/li\u003e\n\u003cli\u003eCorticosteroids and central serous chorioretinopathy. Ophthalmology. 2002;109: 1834\u0026ndash;1837.\u003c/li\u003e\n\u003cli\u003eGelber GS, Schatz H. Loss of vision due to central serous chorioretinopathy following psychological stress. Am J Psychiatry. 1987;144: 46\u0026ndash;50.\u003c/li\u003e\n\u003cli\u003eYannuzzi LA. TYPE-A BEHAVIOR AND CENTRAL SEROUS CHORIORETINOPATHY. Retina. 1987;7: 111.\u003c/li\u003e\n\u003cli\u003eMcCannel TA, Chmielowski B, Finn RS, Goldman J, Ribas A, Wainberg ZA, et al. Bilateral subfoveal neurosensory retinal detachment associated with MEK inhibitor use for metastatic cancer. JAMA Ophthalmol. 2014;132: 1005\u0026ndash;1009.\u003c/li\u003e\n\u003cli\u003eThe Effect of Obstructive Sleep Apnea on Absolute Risk of Central Serous Chorioretinopathy. Am J Ophthalmol. 2020;218: 148\u0026ndash;155.\u003c/li\u003e\n\u003cli\u003eWeenink AC, Borsje RA, Oosterhuis JA. Familial chronic central serous chorioretinopathy. Ophthalmologica. 2001;215: 183\u0026ndash;187.\u003c/li\u003e\n\u003cli\u003ePiccolino FC, Eandi CM, Ventre L, Rigault de La Longrais RC, Grignolo FM. Photodynamic Therapy for Chronic Central Serous Chorioretinopathy. Retina. 2003;23: 752.\u003c/li\u003e\n\u003cli\u003eBurumcek E, Mudun A, Karacorlu S, Arslan MO. Laser photocoagulation for persistent central serous retinopathy: results of long-term follow-up. Ophthalmology. 1997;104: 616\u0026ndash;622.\u003c/li\u003e\n\u003cli\u003eZarnegar A, Ong J, Matsyaraja T, Arora S, Chhablani J. Pathomechanisms in central serous chorioretinopathy: A recent update. Int J Retina Vitreous. 2023;9: 3.\u003c/li\u003e\n\u003cli\u003eSahoo NK, Ong J, Selvam A, Maltsev D, Sacconi R, Venkatesh R, et al. Longitudinal follow-up and outcome analysis in central serous chorioretinopathy. Eye . 2023;37: 732\u0026ndash;738.\u003c/li\u003e\n\u003cli\u003eMohabati D, Boon CJF, Yzer S. Risk of Recurrence and Transition to Chronic Disease in Acute Central Serous Chorioretinopathy. Clin Ophthalmol. 2020;14: 1165\u0026ndash;1175.\u003c/li\u003e\n\u003cli\u003ePiccolino FC, de la Longrais RR, Ravera G, Eandi CM, Ventre L, Abdollahi A, et al. The foveal photoreceptor layer and visual acuity loss in central serous chorioretinopathy. Am J Ophthalmol. 2005;139: 87\u0026ndash;99.\u003c/li\u003e\n\u003cli\u003eKucukevcilioglu M, Durukan AH, Yumusak E, Ozgonul C. Choroidal Thickness Changes After Photodynamic Therapy and Recurrence of Chronic Central Serous Chorioretinopathy. American journal of ophthalmology. 2015. p. 841.\u003c/li\u003e\n\u003cli\u003eRuiz-Moreno JM, Guti\u0026eacute;rrez-Bonet R, Chandra A, Vupparaboina KK, Chhablani J, Ruiz-Medrano J. Choroidal Vascularity Index versus Choroidal Thickness as Biomarkers of Acute Central Serous Chorioretinopathy. Ophthalmic Res. 2023;66: 627\u0026ndash;635.\u003c/li\u003e\n\u003cli\u003eNkrumah G, Paez-Escamilla M, Singh SR, Rasheed MA, Maltsev D, Guduru A, et al. Biomarkers for central serous chorioretinopathy. Ther Adv Ophthalmol. 2020;12: 2515841420950846.\u003c/li\u003e\n\u003cli\u003eCardillo Piccolino F, Lupidi M, Cagini C, Fruttini D, Nicol\u0026ograve; M, Eandi CM, et al. Choroidal Vascular Reactivity in Central Serous Chorioretinopathy. Invest Ophthalmol Vis Sci. 2018;59: 3897\u0026ndash;3905.\u003c/li\u003e\n\u003cli\u003eImamura Y, Fujiwara T, Margolis R, Spaide RF. Enhanced depth imaging optical coherence tomography of the choroid in central serous chorioretinopathy. Retina. 2009;29: 1469\u0026ndash;1473.\u003c/li\u003e\n\u003cli\u003eAnwar SM, Majid M, Qayyum A, Awais M, Alnowami M, Khan MK. Medical Image Analysis using Convolutional Neural Networks: A Review. J Med Syst. 2018;42: 1\u0026ndash;13.\u003c/li\u003e\n\u003cli\u003eLakshminarayanan V, Kheradfallah H, Sarkar A, Jothi Balaji J. Automated Detection and Diagnosis of Diabetic Retinopathy: A Comprehensive Survey. Journal of Imaging. 2021;7: 165.\u003c/li\u003e\n\u003cli\u003eDesideri LF, Anguita R, Berger LE, Feenstra HMA, Scandella D, Sznitman R, et al. BASELINE SPECTRAL DOMAIN OPTICAL COHERENCE TOMOGRAPHIC RETINAL LAYER FEATURES IDENTIFIED BY ARTIFICIAL INTELLIGENCE PREDICT THE COURSE OF CENTRAL SEROUS CHORIORETINOPATHY. Retina. 2023. doi:10.1097/IAE.0000000000003965\u003c/li\u003e\n\u003cli\u003eXu F, Wan C, Zhao L, You Q, Xiang Y, Zhou L, et al. Predicting Central Serous Chorioretinopathy Recurrence Using Machine Learning. Front Physiol. 2021;12: 649316.\u003c/li\u003e\n\u003cli\u003eSingh SR, Iovino C, Zur D, Masarwa D, Iglicki M, Gujar R, et al. Central serous chorioretinopathy imaging biomarkers. Br J Ophthalmol. 2022;106: 553\u0026ndash;558.\u003c/li\u003e\n\u003cli\u003evan Rijssen TJ, Mohabati D, Dijkman G, Theelen T, de Jong EK, van Dijk EHC, et al. Correlation between redefined optical coherence tomography parameters and best-corrected visual acuity in non-resolving central serous chorioretinopathy treated with half-dose photodynamic therapy. PLoS One. 2018;13: e0202549.\u003c/li\u003e\n\u003cli\u003eDeng K, Gui Y, Cai Y, Liang Z, Shi X, Sun Y, et al. Changes in the Foveal Outer Nuclear Layer of Central Serous Chorioretinopathy Patients Over the Disease Course and Their Response to Photodynamic Therapy. Front Med. 2021;8: 824239.\u003c/li\u003e\n\u003cli\u003eMatsumoto H, Sato T, Kishi S. Outer nuclear layer thickness at the fovea determines visual outcomes in resolved central serous chorioretinopathy. Am J Ophthalmol. 2009;148: 105\u0026ndash;10.e1.\u003c/li\u003e\n\u003cli\u003eOishi A, Fang PP, Thiele S, Holz FG, Krohne TU. LONGITUDINAL CHANGE OF OUTER NUCLEAR LAYER AFTER RETINAL PIGMENT EPITHELIAL TEAR SECONDARY TO AGE-RELATED MACULAR DEGENERATION. Retina. 2018;38: 1331\u0026ndash;1337.\u003c/li\u003e\n\u003cli\u003eAgrawal R, Salman M, Tan K-A, Karampelas M, Sim DA, Keane PA, et al. Choroidal Vascularity Index (CVI)--A Novel Optical Coherence Tomography Parameter for Monitoring Patients with Panuveitis? PLoS One. 2016;11: e0146344.\u003c/li\u003e\n\u003c/ol\u003e"},{"header":"Table","content":"\u003cp\u003eTable 1 is available in the Supplementary Files section.\u003c/p\u003e "}],"fulltextSource":"","fullText":"","funders":[{"identity":"bae85a5a-a715-424e-bdad-881bee897009","identifier":"10.13039/501100001711","name":"Schweizerischer Nationalfonds zur Förderung der Wissenschaftlichen Forschung","awardNumber":"2018DRI13 ","order_by":0}],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"Hôpital Ophtalmique Jules-Gonin","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Central Serous Chorioretinopathy, Machine Learning, biomarker importance","lastPublishedDoi":"10.21203/rs.3.rs-4170618/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-4170618/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eCentral serous chorioretinopathy (CSCR) is a posterior segment disease characterized by accumulation of subretinal fluid that, in acute forms, resolves spontaneously. However, about a third of the cases experience recurrences that might cause severe and irreversible vision loss due to anatomical outer retinal and retinal pigment epithelium changes.\u003c/p\u003e \u003cp\u003eThis study aims to identify optical coherence tomography (OCT)-derived parameters linked to CSCR recurrence. Our dataset included 5211 OCTs from 344 eyes of 255 CSCR patients. After expert labeling, 178 eyes were identified as recurrent, 109 were non-recurrent, and 57 were excluded. We extracted parameters using artificial intelligence and computer vision. We used inferential statistics to assess differential distribution between the recurrent and non-recurrent groups, and we employed predictive modeling for feature importance analysis.\u003c/p\u003e \u003cp\u003eWe identified 9 predictive biomarkers for CSCR recurrence, including age, presence of subretinal fluid, intraretinal fluid and Pigment Epithelial detachments, as well as choroidal vascularity index, integrity of photoreceptors and RPE layer, thicknesses of choriocapillaris and choroidal stroma, and thinning of internal retinal layers (outer nuclear layer, and inner nuclear layer combined with and outer plexiform layer).\u003c/p\u003e \u003cp\u003eThese results can potentially enable future developments in automatic detection of CSCR recurrence, paving the way for translational medical applications.\u003c/p\u003e","manuscriptTitle":"Importance of OCT-derived Biomarkers for the Recurrence of Central Serous Chorioretinopathy using Statistics and Predictive Modelling","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-03-29 18:23:46","doi":"10.21203/rs.3.rs-4170618/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"14d63ade-d141-4b81-ae8c-b98871122bf7","owner":[],"postedDate":"March 29th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":29913626,"name":"Artificial Intelligence and Machine Learning"},{"id":29913627,"name":"Ophthalmology"}],"tags":[],"updatedAt":"2024-07-24T04:11:49+00:00","versionOfRecord":[],"versionCreatedAt":"2024-03-29 18:23:46","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-4170618","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-4170618","identity":"rs-4170618","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}