Clinical Applications of AI in Sexually Transmitted Infection and Anogenital Dermatoses in Sexual Health: A Systematic Review and Meta-Analysis

Clinical Applications of AI in Sexually Transmitted Infection and Anogenital Dermatoses in Sexual Health: A Systematic Review and Meta-Analysis | 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 Article Clinical Applications of AI in Sexually Transmitted Infection and Anogenital Dermatoses in Sexual Health: A Systematic Review and Meta-Analysis Nyi Nyi Soe, Ingsun Isika Kusnandar, Phyu Mon Latt, Christopher K. Fairley, and 8 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6002285/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 Background Artificial intelligence (AI) excels in dermatology. However, its applications to sexually transmitted infections (STIs) remain unclear. We assessed the performance of AI algorithms and their applications in detecting STIs and anogenital dermatoses in sexual health. Methods We followed the PRISMA guidelines and searched six databases from January 1, 2010, to April 12, 2024, for studies using AI to identify STIs and anogenital dermatoses. We used a modified QUADAS-2 tool and the CLEAR Derm checklist for quality assessment. We conducted a bivariate random-effect meta-analysis to estimate the pooled sensitivity and specificity of AI applications for the conditions where sufficient data existed. Subgroup analysis and meta-regression were conducted to explore potential heterogeneity sources for mpox studies. Results Of 5,381 studies screened, 141 met the inclusion criteria. Most studies reported on mpox (111, 62.4%), while anogenital conditions, including syphilis, genital herpes, genital warts, scabies, psoriasis, lichenoid changes, and molluscum contagiosum, received less attention (each < 6.0% of the studies). Meta-analyses showed high performance of AI for mpox identification (pooled sensitivity: 0.96 [95% CI: 0.93–0.97], pooled specificity: 0.98 [0.97–0.99]), herpes simplex (0.91 [0.71–0.98], 0.97 [0.94–0.98]), genital warts (0.87 [0.67–0.96], 0.98 [0.95–0.99]), psoriasis (0.90 [0.78–0.95], 0.98 [0.96–0.99]), and scabies (0.89 [0.84–0.93], 0.98 [0.95–0.99]). We could not pool the sensitivity and specificity for other conditions due to insufficient data points. Meta-regression for mpox studies revealed higher pooled sensitivity in models with larger datasets (≥ 1000 images) and binary classification approaches compared to those with smaller datasets and multiclass predictions (p < 0.05). Study quality was variable and our assessment identified high risk of bias across the population selection (76.1%), reference standards (76.1%), and index tests (20.0%). Most studies relied on open-source datasets (87.8%), lacked external validation, and remained at the proof-of-concept stage without clinical implementation. Conclusions While AI shows potential promising performance for identifying STIs and anogenital dermatoses, significant research gaps exist. Future work should prioritise understudied STI and differential conditions, while improving data quality, conducting external validation, and validating in clinical settings. Clear policy guidance and standards are needed to determine how best to implement AI tools for diagnostic purposes and to provide clear performance criteria and frameworks for AI developers, healthcare providers, and clients. Health sciences/Diseases/Infectious diseases Health sciences/Signs and symptoms/Skin manifestations Biological sciences/Computational biology and bioinformatics/Machine learning Biological sciences/Computational biology and bioinformatics/Computational models Sexually Transmitted Infections Anogenital Dermatoses Artificial Intelligence Computer Vision Figures Figure 1 Figure 2 Figure 3 Figure 4 Introduction Sexually transmitted infections (STIs) profoundly affect sexual and reproductive health globally. 1 The World Health Organization (WHO) estimated 374 million new cases of STIs in 2020: chlamydia (129 million), gonorrhoea (82 million), syphilis (8 million) and trichomoniasis (156 million). 2 These estimates do not include mpox, for which evidence has recently emerged showing the important role of sexual transmission in mpox outbreaks. 3 Delayed diagnoses and untreated infections can result in serious complications, including a higher risk of HIV acquisition, infertility, and adverse pregnancy outcomes. 4 The recent increase in syphilis cases across populations and resurgence of congenital syphilis highlight the urgent need for improved early identification and control measures for STIs. 5 STIs can be asymptomatic or present with diverse clinical manifestations, including genitourinary symptoms and anogenital dermatological presentations. 6 Recent data from a sexual health clinic show that nearly half of the survey clients (n = 10,479) over three years presented with anogenital conditions such as lumps, spots, rashes, sores or itch. 7 These presentations can be infectious, such as painless chancres in syphilis or painful sores in genital herpes. These can also be non-infectious, like inflammatory conditions, benign lesions, and neoplasms. This complexity in clinical presentation poses a significant barrier for the healthcare workforce to timely diagnosis and management of STIs. Digital health interventions, particularly artificial intelligence (AI), have emerged as promising tools in sexual health to promote testing and health-seeking behaviours among symptomatic individuals and those at high risk of STIs. Machine learning approaches have been used to estimate HIV and STI risks based on demographic information and sexual health behaviours. 8 – 10 Machine learning and Bayesian-based symptom checkers have been developed to assess the different presentations of STIs among those with symptoms. 7 , 11 However, the user input data for these anogenital dermatoses could be subjective and limit the accuracy of the symptom checkers for dermatological conditions. 12 The distinctive patterns of anogenital lesions suggest potential value in deep learning-based image classification algorithms for STI diagnosis. 13 While AI has shown success in general dermatology, its application to STI-related dermatoses remains nascent, and its role in clinical implementation is unclear. 14 This systematic review and meta-analysis aim to: (1) evaluate the diagnostic performance of AI algorithms in identifying and classifying STIs and other anogenital dermatoses; (2) assess the methodological quality and clinical applicability of existing studies; (3) identify key technical and implementation challenges; and (4) provide evidence-based recommendations for advancing these technologies from research into clinical practice. We hypothesised that while AI shows promise for STIs and anogenital dermatoses, significant gaps exist in current research regarding clinical validation, demographic representation, and real-world implementation. Methods Study design and protocol We conducted this systematic review according to the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines. 15 We pre-registered our protocol in the International Prospective Register of Systematic Reviews (PROSPERO, CRD42024527420). 16 The study team included clinical experts in sexual health, specialists in medical AI applications, biostatisticians, and a medical librarian. Search strategy and selection criteria In consultation with a librarian (LR), we searched six databases (IEEE Xplore, Embase, Scopus, MEDLINE, Web of Science, and CINAHL) for studies from January 1, 2010, to April 12, 2024. We used three main concepts: “artificial intelligence”, “diagnosis”, and “sexually transmitted infections” for our search strategies. The detailed search strategies are presented in Table S1 . Our search included studies ranging from proof-of-concept to randomised controlled trials without language restrictions. We included studies that used artificial intelligence to identify or classify anogenital skin conditions using clinical images, with or without accompanying metadata. We excluded studies that did not involve AI-based approaches or did not target anogenital skin conditions related to sexual health. We also omitted review articles and studies that did not report model performance metrics or focused on dermatoscopic images and non-skin manifestations, such as cervical changes. Two reviewers (NS and IK) independently screened titles and abstracts against our eligibility criteria after removing duplicates using Covidence software. 17 We conducted pilot testing before formal screening and data extraction, with regular team meetings to ensure consistency. After screening 5,381 studies with 97% agreement in screening, two additional reviewers (PL and JJO) resolved any remaining conflicts through consensus. Data extraction Two reviewers (NS and IK) independently assessed 258 full-text articles and extracted data using a standardised Excel sheet. PL and JJO resolved any disagreements in data extraction through discussion. We collected data on the following aspects: (i) study characteristics, including authors, publication year, and study type; (ii) data characteristics, such as the image database, types of data input and predictions, reference conditions, and sample size; (iii) technical information, including model development, evaluation methods, subgroup analyses, model interpretability, and the best-performing algorithms; and (iv) performance metrics, such as ROC-AUC, accuracy, sensitivity, specificity, precision, F1 scores, and values from confusion matrices. We selected performance metrics from the best-performing model for studies reporting multiple models. We extracted true positives (TP), false negatives (FN), true negatives (TN), and false positives (FP) directly from 2x2 tables for binary outcome studies. We recalculated these values from the tables focusing on our target disease conditions in the studies with multiclass classifications. Quality assessment Two reviewers (NS and IK) used a modified QUADAS-2 critical appraisal tool 18 to independently assess the risk of bias and the applicability of included studies across the population, index test and reference standards. We rated each domain as low, high or uncertain risk based on the information provided in the studies. We also evaluated model fairness, reliability, and safety using the CLEAR Derm checklist, 19 scoring 25 items across data, technique, technical assessment, and application domains as present, partially present, absent, or not applicable. Disagreements were resolved through discussion among the review team (NS, IK, PL, JJO). Data analysis Due to incomplete reporting across studies, we conducted a descriptive analysis of available data on study characteristics, technical methodology, and model performance metrics. We used dupeGuru version 4.0.3 to identify duplicate images across public datasets. We conducted a bivariate random effect meta-analysis to estimate AUC-ROC, sensitivity and specificity using data from 31 mpox studies and four studies each for herpes simplex, genital warts, psoriasis, and scabies. We generated forest plots and summary receiver operating characteristic (SROC) graphs for each condition. We used funnel plots of diagnostic odds ratios and Deek’s regression test to assess publication bias. We conducted a meta-regression analysis on mpox studies using the Higgins I 2 statistic to assess heterogeneity based on sample size, number of reference skin conditions, type of model classification, and AI algorithm category. We measured the F-statistics and p-value for the significance of the regression models. We analysed all data using STATA version 17 (StataCorp, College Station, TX, USA). 20 Results Our initial search identified 6,035 studies. After removing duplicates, we screened 5,381 titles and abstracts and selected 258 studies for full-text review. After removing irrelevant studies, duplicates, and review articles, 141 studies were deemed eligible and included in the review (Fig. 1 ). Quality assessment The quality assessment revealed significant high level of bias across the included studies. Using the modified QUADAS-2 tool, we identified a high risk of bias in most studies regarding three domains: populations (76.1%), reference standard tests (76.1%) and index tests (20.0%). We also identified a high concern for applicability across these three domains (above 75.0%). In the CLEAR Derm checklist, most studies are inadequately reported across four domains: data, technique, technical assessment, and application. Detailed assessment scores for individual studies are presented in Fig. 2 and Table S3. Disease conditions Most studies (n = 111, 62.4%) reported on AI algorithms focused on mpox, while other conditions such as tinea cruris (n = 8), genital warts (n = 8), scabies (n = 8), herpes zoster (n = 8), psoriasis (n = 7), herpes simplex (n = 7), lichenoid changes (n = 6), molluscum contagiosum (n = 6) were reported in less than 6.0% of the studies. Less common conditions included folliculitis (n = 3), balanitis (n = 2), syphilis (n = 1), candidiasis (n = 1), penile cancer (n = 1), and combined STIs (n = 1) (Fig. 3 a). The 2022 mpox outbreak significantly increased mpox studies during 2022–2023 (Fig. 3 b). The types of target and reference conditions included in the studies ranged from 2 to 44. Most mpox studies (n = 96) focused on a limited number of differential diagnoses, primarily comparing mpox with conditions that are unrelated to common anogenital dermatoses typically seen in sexual health clinics (Table S2d). Among all included studies, only 18 incorporated one or more relevant differential conditions, while only six provided a broader comparison with multiple anogenital dermatoses. 21 – 26 Data Most studies used open-source data (n = 122, 87.8%), while a few used private databases (n = 12, 8.6%) or a combination of private and public datasets (n = 5, 3.6%) (Fig. 4 ). Nearly three-quarters of studies (n = 104, 73.8%) properly referenced the image databases. The most frequently used public databases for mpox studies were ‘MSLD’, ‘MSID’ and ‘MSCI’, 27 – 29 where 7.3% (84/1155) of images were identified as duplicates (Table S3). Private datasets originated mainly from Asia (28.6%; China = 2, India = 1, the Philippines = 1), followed by the Americas (21.4%; USA = 2, Peru = 1), with equal representation from Europe (Munich = 1, Sweden = 1) and Oceania (Australia = 2), and West Africa (7.1%). The sample size of the total images varied substantially across different datasets, ranging from 70 to 139,198. Studies showed considerable class imbalance, where the proportion of target condition images ranged from 0.1–71.8% of total images within datasets (median 36.2%, IQR 19.9%-44.7%). Most datasets (n = 133, 94.3%) that were used contained clinical images alone, while only eight (5.6%) contained additional patient metadata (sex, race, ethnicity, symptoms, etc.). 23 , 25 , 26 , 30 – 34 Eight studies reported skin tones in relation to disease conditions. 23 , 25 , 30 , 32 , 33 , 35 – 37 Most studies (n = 126, 89.4%) did not report technical details about image acquisition, such as image quality, camera type, or lighting conditions. Only 17 studies (12.1%) validated image diagnoses using laboratory tests, clinician-reviewed images, or both. Three studies standardised their image categories using International Classification of Diseases (ICD) codes. 23 , 25 , 30 Model development Over half of the studies (n = 78, 55.3%) focused on developing binary classification models (differentiating a target disease from other reference conditions), while 58 studies (41.1%) used multiclass classification approaches (classifying multiple conditions simultaneously). Only two studies (1.4%) employed multimodal approaches, incorporating images and patient metadata for model prediction, while all other studies used sorely image data input. 26 , 30 Most studies (n = 124, 87.9%) reported image preprocessing procedures such as resizing, normalisation, cropping, and rotation for model training. Studies addressed data scarcity and class imbalance through image augmentation, but none used synthetic images. Model evaluation To evaluate model performance, 127 studies (90.1%) reported fully or partially splitting their data into training and testing sets, while only 35 studies (24.8%) used k-fold cross-validation for more robust validation. Eight studies (5.6%) assessed model generalisability using external validation or prospectively collected image datasets, and only three studies (2.1%) used data from multiple vendors or clinics. 30 , 32 , 38 Patient-level metadata and the lesion site could influence the model performance. While six studies considered gender in their analyses, only one study specifically evaluated model performance for females. Only a few studies evaluated model performance across different subgroups: skin tone (n = 5), lesion site (n = 5), age (n = 4), and race and ethnicity (n = 2). To understand how models reached their decisions, 24 studies (17.0%) used interpretability techniques like Grad-CAM, LIME and SHAP to visualise important image features for model prediction. Model performance Due to diverse reporting approaches across studies, we summarised model performance metrics by target condition in Table 1 and Fig. 3 a. Studies commonly reported accuracy, sensitivity, specificity, positive predictive values (PPV), and F-1scores, while negative predictive values (NPV) and AUC-ROC scores were less frequently reported. Studies demonstrated high mean accuracy (> 70.0%) for syphilis, herpes simplex, genital warts, mpox, scabies, herpes zoster, and penile cancer, while other conditions showed lower accuracy (ranging from 42.0–65.0%). Most conditions achieved high specificity (> 95.0%), but other metrics showed considerable variation within conditions. For example, accuracy in herpes simplex models ranged from 10.0–97.0%, showing inconsistent model performance across studies. The highest performance was achieved with CNN-based architectures (n = 97, 68.8%), followed by hybrid or ensemble models (n = 29, 20.6%). Table 1 Summary of model performance across disease conditions Sensitivity Specificity PPV NPV AUC-ROC Accuracy F1-score Syphilis (n = 1) Mean ± sd 0.87 0.99 0.91 - - 0.94 0.95 # of studies 1 1 1 1 1 1 1 Herpes simplex (n = 7) Mean ± sd 0.58 ± 0.45 0.98 ± 0.02 0.63 ± 0.46 - 0.997 0.74 ± 0.30 0.66 ± 0.48 Median (IQR) 0.62 (0.20–0.96) 0.98 (0.97–0.99) 0.87 (0.10–0.93) - - 0.84 (0.69–0.94) 0.92 (0.10–0.96) Range 0.10–0.99 0.97–0.99 0.10–0.93 - - 0.1–0.97 0.10–0.96 # of studies 4 2 3 0 1 7 3 Genital warts (n = 8) Mean ± sd 0.82 ± 0.20 0.99 ± 0.02 0.69 ± 0.37 1.00 0.85 0.86 ± 0.15 0.96 ± 0.05 Median (IQR) 0.93 (0.57–0.96) 1.00 (0.96-1.00) 0.86 (0.57–0.94) - - 0.94 (0.69–0.97) 0.97 (0.91-1.00) Range 0.56-1.00 0.96-1.00 0.10-1 - - 0.65–0.98 0.91-1.00 # of studies 6 3 5 1 1 6 3 Mpox (n = 111) Mean ± sd 0.91 ± 0.09 0.93 ± 0.09 0.93 ± 0.06 0.94 ± 0.09 0.96 ± 0.06 0.93 ± 0.06 0.92 ± 0.08 Median (IQR) 0.94 (0.88-1.00) 0.97 (0.90–0.99) 0.95 (0.91-1.00) 0.98 (0.93–0.99) 0.98 (0.94-1.00) 0.95 (0.91-1.00) 0.94 (0.90-1.00) Range 0.58-1.00 0.61-1.00 0.70-1.00 0.79-1.00 0.74-1.00 0.94-1.00 0.67-1.00 # of models 115 37 103 5 37 128 108 Molluscum contagiosum (n = 6) Mean ± sd 0.53 ± 0.39 0.99 ± 0.01 0.28 ± 0.25 1.00 0.98 0.62 ± 0.36 0.52 ± 0.60 Median (IQR) 0.64 (0.10–0.85) 0.99 (0.99-1.00) 0.28 (0.10–0.45) - - 0.72 (0.40–0.84) 0.52 (0.10–0.94) Range 0.10–0.85 0.99-1.00 0.10–0.45 - - 0.10–0.92 0.10–0.94 # of studies 3 2 2 1 1 4 2 Tinea cruris (n = 8) Mean ± sd 0.55 ± 0.38 0.98 ± 0.02 0.59 ± 0.43 0.98 0.95 0.53 ± 0.35 0.51 ± 0.58 Median (IQR) 0.55 (0.23–0.92) 0.98 (0.97–0.99) 0.76 (0.10–0.92) - - 0.55 (0.25–0.80) 0.51 (0.10–0.92 Range 0.10–0.93 0.97–0.99 0.10–0.92 - - 0.10–0.92 0.10–0.92 # of studies 6 2 3 1 1 4 2 Lichenoid changes (n = 6) Mean ± sd 0.59 ± 0.30 0.97 ± 0.01 0.71 ± 0.26 0.99 0.88 0.65 ± 0.25 0.83 ± 0.24 Median (IQR) 0.57 (0.56–0.68) 0.97 (0.97–0.98) 0.64 (0.49-1.00) - - 0.69 (0.53–0.73) 0.83 (0.66-1.00) Range 0.16-1.00 0.97–0.98 - - 0.31–0.99 0.66-1.00 # of studies 5 2 3 1 1 5 2 Scabies (n = 8) Mean ± sd 0.83 ± 0.16 1.00 0.89 ± 0.13 - - 0.85 ± 0.10 - Median (IQR) 0.86 (0.71–0.95) - 0.89 (0.80–0.98) - - 0.85 (0.77–0.92) - Range 0.63–0.98 - 0.80–0.98 - - 0.69–0.99 - # of studies 4 1 2 0 0 7 0 Folliculitis (n = 3) Mean ± sd 0.36 ± 0.35 - - - - 0.42 ± 0.24 - Median (IQR) 0.36 (0.11–0.60) - - - - 0.55 (0.14–0.57) - Range 0.11–0.60 - - - - 0.14–0.57 - # of studies 2 0 0 0 0 3 0 Herpes zoster (n = 8) Mean ± sd 0.82 ± 0.13 0.98 ± 0.03 0.85 ± 0.19 1.00 0.94 ± 0.05 0.94 ± 0.03 0.95 ± 0.04 Median (IQR) 0.83 (0.71–0.91) 1.00 (0.94-1.00) 0.93 (0.74–0.96) - 0.93 (0.89–0.99) 0.95 (0.92–0.96) 0.95 (0.92–0.99) Range 0.67–0.96 0.94-1.00 0.57–0.97 - 0.89–0.99 0.90–0.97 0.92–0.99 # of studies 5 3 4 1 3 5 3 Psoriasis (n = 7) Mean ± sd 0.72 ± 0.35 0.96 0.65 ± 0.40 0.98 0.93 0.58 ± 0.41 0.55 ± 0.64 Median (IQR) 0.84 (0.78-1.00) - 0.75 (0.36-1.00) - - 0.62 (0.25–0.92) 0.55 (0.10-1.00) Range 0.10-1.00 - 0.10-1.00 - - 0.10–0.99 0.10-1.00 # of studies 5 1 4 1 1 4 2 Candidiasis (n = 1) Mean ± sd 0.41 1.00 0.41 1.00 0.81 - - # of studies 1 1 1 1 1 0 0 Balanitis (n = 2) Mean ± sd 0.88 1.00 0.97 - - 0.64 ± 0.43 0.98 Median (IQR) - - - - - 0.64 (0.33–0.94) - Range - - - - - 0.33–0.94 - # of studies 1 1 1 0 0 2 1 Penile cancer (n = 1) Mean ± sd 0.79 0.99 0.89 - - 0.94 0.93 # of studies 1 1 1 0 0 1 1 STIs (n = 1) Mean ± sd 0.95 0.62 0.43 - 0.89 0.69 - # of studies 1 1 1 0 1 1 1 AUC-ROC, Area Under the Curve-Receiver Operating Characteristic; IQR: Interquartile range; NPV: Negative Predictive Value; PPV: Positive Predictive Value; Sd: standard deviation; STI: sexually transmitted infection; Application Most studies remained at the proof-of-concept stage without publicly available models for external evaluation. They also did not specify their intended users (clinicians or the public) or the tool's purpose (such as triage, assisted diagnosis or population screening). Only one study tested a public-facing application (Skin Image Search™) with prospectively collected images. 25 No studies evaluated model accuracy in clinical settings using randomised trials. Meta-analysis For meta-analysis, we included 31 mpox studies (34 contingency tables) and four studies each for herpes simplex, genital warts, psoriasis, and scabies (Fig. 1 ). We could not pool the sensitivity and specificity for other conditions due to insufficient data points. Models showed consistently high performance across conditions: mpox (pooled sensitivity: 0.96 [95% CI: 0.93–0.97], I 2 = 91.0%, pooled specificity: 0.98 [0.97–0.99], I 2 = 99.9%), herpes simplex (sensitivity: 0.91 [0.71–0.98], I 2 = 93.2%, specificity: 0.97 [0.94–0.98], I 2 = 72.6%), genital warts (sensitivity: 0.87 [0.67–0.96], I 2 = 87.4%, specificity: 0.98 [0.95–0.99], I 2 = 93.9%), psoriasis (sensitivity: 0.90 [0.78–0.95], I 2 = 92.8%, specificity: 0.98 [0.96–0.99], I 2 = 92.1%), and scabies (sensitivity: 0.89 [0.84–0.93], I 2 = 69.1%, specificity: 0.98 [0.95–0.99], I 2 = 83.8%). All results were statistically significant (p 0.05). Forest plots, SROC graphs and Deeks’s funnel plots are presented in Figure S1 . We identified the high heterogeneity in the meta-analysis for the above conditions (I 2 > 50.0%). We conducted subgroup analyses and meta-regression for the heterogeneity only for mpox studies as there were limited studies and insufficient data to perform subgroup analysis in other conditions (Table 2 ). Meta-regression revealed higher pooled sensitivity in models with larger datasets (≥ 1000 images) and binary classification approaches compared to those with smaller datasets and multiclass predictions (p = 0.049 and p = 0.028, respectively). Research gaps, limitations and recommendations Based on our systematic review findings, we summarised key research gaps and limitations in existing studies and formulated recommendations for future research, as presented in Table 3 . Table 2 Summary estimate of pooled performance of AI models for mpox Number of models Model F (4,29) = 2.53 (p-value: 0.062) Model F (4, 29) = 0.70 (p-value: 0.598) Sensitivity P-value* I 2 (95% CI) P-value** Specificity P-value* I 2 (95% CI) P-value** Overall 34 0.96 (0.93–0.97) 0.00 90.98 (88.74–93.22) 0.98 (0.97–0.99) 0.00 99.90 (99.90-99.91) # of lesion types included Less than 5 26 0.96 (0.93–0.98) 0.00 89.86 (86.87–92.85) 0.130 0.98 (0.97–0.99) 0.00 92.84 (90.93–94.74) 0.955 5 and above 8 0.93 (0.84–0.97) 0.00 93.47 (90.31–96.64) 0.97 (0.95–0.99) 0.00 99.95 (99.95–99.96) Sample size Less than 1000 23 0.95 (0.91–0.97) 0.00 88.49 (84.74–92.24) 0.049 0.98 (0.97–0.99) 0.00 91.29 (88.67–93.91) 0.532 100 and above 11 0.97 (0.92–0.99) 0.00 94.07 (91.73–96.42) 0.97 (0.95–0.99) 0.00 99.96 (99.96–99.97) Model classification Binary 17 0.97 (0.95–0.99) 0.00 87.87 (83.16–92.57) 0.028 0.98 (0.95–0.99) 0.00 92.43 (89.87–94.98) 0.197 Multiclass 17 0.93 (0.88–0.96) 0.00 89.02 (84.89–93.15) 0.98 (0.97–0.99) 0.00 99.95 (99.95–99.95) AI Algorithm CNN-based 26 0.95 (0.92–0.97) 0.00 91.31 (88.86–93.76) 0.537 0.98 (0.96–0.99) 0.00 99.91 (99.91–99.92) 0.396 Other 8 0.95 (0.89–0.98) 0.00 88.42 (81.81–95.04) 0.98 (0.97–0.99) 0.05 50.25 (10.19–90.31) ** P-value for heterogeneity between subgroups with meta-regression analysis *P-value for heterogeneity within each subgroup CI: Confidence Interval; I²: Higgins' I-squared statistic (a measure of heterogeneity) Table 3 Summary of research gaps and future recommendations for AI-based identification of STIs and anogenital dermatoses Research gaps/limitations Recommendations 1. Disease conditions 1.1 Imbalanced research focus : Studies predominantly focused on mpox (62.4%) due to the recent outbreak, while common STIs and anogenital conditions received limited attention (< 6.0% each). Balanced research agenda : Prioritise AI research into common STIs and anogenital conditions, while continuing to investigate emerging outbreaks like mpox, with an emphasis on WHO-priority infections such as syphilis, genital herpes, and genital warts. 1.2 Limited differential coverage : Studies lacked clinically relevant comparative conditions and comprehensive coverage of anogenital conditions as seen in sexual health practice. Diverse and representative data : Include a wide variety of anogenital reference conditions covering STIs, non-STIs, tumours, inflammatory diseases, and normal anatomical variants for robust differential diagnosis. 6 , 39 – 41 1.3 Poor disease standardisation : Lack of standardised disease definition undermined clinical relevance and hindered reproducibility (e.g. herpes simplex versus herpes zoster; genital warts versus flat warts). Adoption of International Classification (ICD) Systems : Use standardised ICD codes for labelling and disease categorisation in datasets. 19 2. Data 2.1 Data scarcity challenges : Models relied heavily on open-source datasets (87.8%), with small sample sizes, limited representation of target conditions, and absence of patient metadata. 42 Coordinated data infrastructure : Establish networks for standardised image collection and develop a centralised repository (similar to the IARC Cervical Cancer Image Bank) 43 for STIs and anogenital conditions. 2.2 Data quality concern : Duplicate images across public datasets compromised data quality. 44 , 45 Implement de-duplication processes : Use systematic methods or tools (e.g., DupeGuru software, difPy python package) to identify and remove duplicate images to ensure data quality. Transparent reporting : Document and transparently report all methods for image quality control. 46 2.3 Inadequate technical documentation : Most studies (89.4%) lacked essential details regarding image acquisition methods and quality standards. Technical standardisation : Adopt and adapt established technical guidelines to ensure consistent quality in clinical image acquisition. 47 Document and report for reproducibility : Provide detailed documentation of image acquisition methods, quality standards, and technical specifications. 47 2.4 Insufficient diagnostic validation : Limited validation of image diagnoses through laboratory tests or clinical reviews undermined data reliability. Diagnostic validation standards : Require validation of image diagnoses through laboratory confirmation and/or expert clinical review before inclusion in datasets. 6 3. Model development 3.1 Unclear clinical alignment : Model development lacked clear alignment with intended clinical applications. Purpose-driven development : Define clear clinical objectives (rule-in/rule-out) and model approach (binary/multiclass) based on intended clinical use (screening/triage/assisted-diagnosis). 48 Interdisciplinary collaboration : Ensure development teams include data scientists, AI experts, sexual health physicians, and end users. 48 , 49 3.2 Inadequate methodology reporting : Insufficient documentation of data splitting and cross-validation approaches, including crucial stratification methods with patient metadata (sex, lesion site, skin tone, etc.) and diagnoses. Data splitting integrity : Prevent data leakage by ensuring images from the same patient remain in the same dataset split. Stratified validation protocols : Implement stratified data splitting and cross-validation based on patient characteristics and diagnoses. 3.3 Underutilised multimodal approach : Most models used only image data, with minimal integration of patient metadata (1.4%) despite the potential benefits of multimodal approach. 26 , 30 Multimodal integration : Develop models that combine clinical images with relevant patient metadata to improve diagnostic accuracy. 4. Model evaluation 4.1 Limited generalisability : Model generalisability was limited by a lack of external validation and clinical trials, with most studies being single-centre evaluations. Model availability : Encourage that trained models are publicly accessible (e.g., through GitHub) to facilitate external validation. Multi-centre validation : Establish collaborative networks for prospective model testing across various clinical settings. 4.2 Limited subgroup assessment : Minimal evaluation across gender, skin tones, lesion sites, age, and ethnicity. Gender-specific models : Develop dedicated models specifically for females to address unique anatomical features and disease presentations. Diverse subgroup analysis : Evaluate model performance across different demographic groups, anatomical sites, and clinical presentations. 50 4.3 Unexplained model decisions : Limited reporting of model interpretability techniques (17%), leaving model decision processes unclear. Model transparency with visualisation : Implement modern interpretability techniques (e.g., Grad-CAM, LIME, SHAP) to explain the model’s decision-making processes. 5. Model performance 5.1 Inconsistent performance reporting : Inconsistent reporting performance metrics hindered model comparison. Standardised performance reporting : Document key performance metrics (AUC-ROC, sensitivity, specificity, PPV, F1-scores) and contingency tables on test datasets to enable meaningful comparisons. 5.2 Unclear threshold criteria : Binary classification studies lacked specified threshold selection criteria for sensitivity/specificity trade-offs based on intended use. Performance trade-off reporting : Specify how sensitivity/specificity thresholds were optimised for the intended clinical application. 5.3 Variable model performance : Performance varied widely due to differences in model architectures, purposes, and reference condition selections. Facilitate evidence synthesis : Conduct more studies on prioritised diseases using standardised reporting framework to generate robust evidence, enabling meta-analysis to derive stronger conclusions on model accuracy and clinical utility. 6. Application 6.1 Limited translation to practice : Studies remained at the proof-of-concept stage without defined clinical purpose (triage/screening) or target users (clinicians/public). 51 Bridge Research and Practice : Encourage collaborations between researchers, healthcare providers, and policymakers to translate proof-of-concept studies into clinically viable solutions. Develop implementation framework : Establish frameworks to guide from the development stage to integration of AI models into existing healthcare workflows with a multidisciplinary team. 48 , 52 6.2 Minimal clinical implementation : only one public-facing application was tested prospectively, with no randomised clinical trials conducted. Testing Beyond Conceptualisation : Advance AI models from the conceptual stage by conducting pilot testing and clinical trials to generate robust evidence on their safety, efficacy, and clinical utility. 53 6.3 Limited study on user's perspective : Limited evidence on the needs, expectations, and experience of end-users, such as clinicians and patients, in adopting AI models for implementation. Incorporate Usability Testing : Conduct feasibility, acceptability and usability studies to ensure models meet the practical requirements of their target users in clinical or public health settings. 54 – 56 Discussion Our systematic review identified 141 eligible studies that used AI algorithms to detect or classify STIs and anogenital dermatoses across six databases. Anogenital conditions like syphilis, genital herpes and other common differential dermatoses in sexual health have received limited attention. Meta-analysis demonstrated promising performance metrics, with pooled sensitivity above 87.0% and pooled specificity above 97% for conditions including mpox, herpes simplex, genital warts, psoriasis, and scabies. However, the high heterogeneity across studies and the limited number of studies for conditions other than mpox suggest that these results should be interpreted with caution and may not be generalisable beyond the settings of the studies included. Quality assessment using the modified QUADAS-2 tool indicated high concerns for risk of bias and applicability across studies, while the CLEAR Derm checklist highlighted incomplete reporting of data characteristics, methodological techniques, and clinical validation. 19 This finding aligns with similar AI models in dermatology conditions. 14 Moreover, insufficient external validation and the lack of prospective testing present significant barriers to translating these AI tools from proof-of-concept to clinical practice. In this review, we discuss these research gaps and provide recommendations for future studies to enhance the clinical utility of AI in STIs and anogenital dermatoses identification management. Although AI applications in dermatology have shown promising results, their development for STIs and anogenital dermatoses has remained limited. 14 Before 2022, few studies included anogenital dermatoses in their analyses, as either target or differential diagnoses. While the 2022 mpox outbreak sparked numerous research studies, other common STIs have received limited research attention, despite their increasing global prevalence in recent years. While AI holds promise for early identification of syphilis chancres, only two studies addressed this application with limited sample sizes until 2024. 26 , 34 The clinical relevance of current studies is further constrained by the comparative conditions they included. Most mpox studies, for instance, developed models to distinguish mpox from other conditions, such as chickenpox and measles, which are rarely seen in sexual health clinics. This limited coverage of relevant anogenital conditions hinders the practical utility of these models in clinical settings. The absence of normal anatomical variants (such as skin tags and Fordyce spots) in training datasets could lead models to misclassify these as pathological conditions like genital warts, potentially causing unnecessary concern. Therefore, future studies should prioritise clinically important STIs and incorporate a comprehensive range of relevant conditions in model development to enhance their clinical utility. 6 , 39 – 41 Data is essential for developing AI algorithms, and our review highlighted critical challenges regarding data scarcity, quality, and validity. The heavy reliance on limited open-source datasets, duplication of images, and, most importantly, the lack of clinical validation for image diagnoses raises concerns about the generalisability of reported model performance. Images of anogenital dermatoses are particularly scarce compared to other dermatological conditions. 42 Difficulty in extracting data from clinical notes, and privacy and anonymity concerns have resulted in most images lacking essential patient metadata, such as age and symptoms (pain, duration, etc.), which are crucial for clinical decision-making. Like the WHO's coordinated approach for the Cervical Cancer Image Bank, 43 establishing a centralised repository with a standardised image collection protocol for STIs and anogenital dermatoses could address the current challenges. Regarding model development and evaluation, most studies focused on technical aspects of data processing and algorithm design, yet overlooked two important areas. First, studies rarely defined their target users (public or healthcare providers) and use-case scenarios (self-symptom checking or clinical diagnosis support). Clear alignment between model design and clinical application could be achieved by a multidisciplinary team approach (data scientists, AI experts, sexual health physicians, and end users) and following the OPTICA tool. 48 Second, the predominant use of open-source or single-centre data without external validation or prospective testing raises concerns about model generalizability. The limited evaluation across different demographic groups, particularly regarding gender, skin tones, and anatomical sites, suggests potential performance disparities across diverse populations. 47 , 50 Future studies should prioritise external validation using diverse datasets, comprehensive demographic evaluation, and transparent reporting to improve the generalizability of AI models. Most studies remained at the conceptual stage with limited translation into real-world applications. However, this technical focus should not preclude the exploration of end-user perspectives, which are crucial for successful implementation. A few studies explored the public’s acceptability, feasibility, and preferences for the tool when it becomes available. Ly et al. found that nearly 40% of users were reluctant to share clinical images for AI-based healthcare tools, particularly genital images due to privacy concerns. 54 In contrast, Jakob et al. reported high interest in STI-related apps among dermatovenereological outpatients. 57 Soe et al. also found that sexual health clinic attendees were willing to use such apps and provide comprehensive information, including symptoms and sexual behaviours, along with anogenital lesion images, if the app was developed by a reputable organisation and demonstrated reliable accuracy. 55 Future studies should explore the feasibility, acceptability, and usability to ensure that models meet the practical requirements of end users in clinical or public health settings. Our systematic review has a number of strengths. First, we conducted a comprehensive search across six major databases in consultation with a librarian to capture the full scope of AI applications in STIs and anogenital dermatoses. Second, we ensured the reliability of our findings through independent duplicate screening, data extraction, and quality assessment. Third, we conducted meta-analyses of AI performance for five conditions and explored sources of heterogeneity in mpox studies through meta-regression. Finally, we provided a structured framework of research gaps and practical recommendations for future studies and the use of AI in clinical settings. Our study has limitations. First, despite our comprehensive search strategy, we may have missed relevant studies, particularly those published in non-indexed journals or after our search date. Our review focused on peer-reviewed literature and may have overlooked potentially relevant commercial AI applications for STIs and anogenital dermatoses. Second, we modified the QUADAS-2 tool for the risk of bias assessment, which is not specifically designed for AI studies in dermatology. The modification might be subjective on the assessment scores. Therefore, we also used the CLEAR Derm checklist to check the quality of the studies. Third, the high heterogeneity in study methodologies and reporting made it challenging to synthesise findings and interpret meta-analysis results confidently. Thus, the pooled performance of the AI models may not reflect the actual performance in a local clinical setting. Fourth, while we included studies based on target diagnoses relevant to sexual health, many of these AI models were developed using images from both anogenital and non-anogenital body sites. Our findings may not accurately reflect the performance of AI models specifically designed for STIs and anogenital dermatoses. Finally, the AI models in our review were limited to identifying visible dermatological presentations and could not identify other important STI presentations such as discharge, bleeding and genitourinary symptoms. These AI models would need to be integrated with other approaches to assess a wider coverage of symptoms and risk factors for STIs. Conclusions While AI shows potential promising performance for identifying STIs and anogenital dermatoses, significant research gaps exist. Future work should prioritise understudied STI and differential conditions, while improving data quality, conducting external validation, and validating in clinical settings. Clear policy guidance and standards are needed to determine how best to implement AI tools for diagnostic purposes and to provide clear performance criteria and frameworks for AI developers, healthcare providers, and clients. Declarations Acknowledgement We thank Monash University and Melbourne Sexual Health Centre for NS's scholarship and Lorena Romero (LR), the librarian at Alfred Health's Ian Potter Library, for her expert guidance on the literature search strategy. Contributors NS, JJO and LZ conceptualised the study. NS and PL conducted the literature search. NS, IK and PL conducted screening and data extraction. NS conducted the statistical analysis and drafted the manuscript. JJO and PL verified the analyses. All authors contributed to interpreting the results, revised the manuscript, and approved it for publication submission. JJO supervised the study and served as the guarantor of the review. The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the World Health Organization. Declaration of interests None. Funding CKF is supported by the National Health and Medical Research Council (NHMRC) (GNT1172900). EPFC is supported by an NHMRC Leadership Investigator Grant (GNT2033299). JJO is supported by the NHMRC Emerging Leadership Investigator Grant (GNT1193955). Data availability Data can be requested from the corresponding author. References Centers for Disease Control and Prevention (CDC). Incidence, Prevalence, and Cost of Sexually Transmitted Infections in the United States. 2021. https://www.cdc.gov/nchhstp/newsroom/fact-sheets/std/STI-Incidence-Prevalence-Cost-Factsheet.html (accessed July 2023). World Health Organization. Implementing the global health sector strategies on HIV, viral hepatitis and sexually transmitted infections, 2022–2030, Report on progress and gaps, 2024. Laurenson-Schafer H, Sklenovská N, Hoxha A, et al. Description of the first global outbreak of mpox: an analysis of global surveillance data. Lancet Glob Health 2023; 11(7): e1012-e23. Fairley CK, Chow EPF, Hocking JS. Early presentation of symptomatic individuals is critical in controlling sexually transmissible infections. Sexual Health 2015; 12(3): 181. Centers for Disease Control and Prevention (CDC). STI Treatment Guidelines. July 22, 2021 2021. https://www.cdc.gov/std/treatment-guidelines/congenital-syphilis.htm 2024). Ong JJ, Bourne C, Dean JA, et al. Australian sexually transmitted infection (STI) management guidelines for use in primary care 2022 update. Sex Health 2023; 20(1): 1–8. Soe NN, Towns JM, Latt PM, et al. Accuracy of symptom checker for the diagnosis of sexually transmitted infections using machine learning and Bayesian network algorithms. BMC Infect Dis 2024; 24(1): 1408. He J, Li J, Jiang S, et al. Application of machine learning algorithms in predicting HIV infection among men who have sex with men: Model development and validation. Front Public Health 2022; 10: 967681. Xu X, Yu Z, Ge Z, et al. Web-Based Risk Prediction Tool for an Individual's Risk of HIV and Sexually Transmitted Infections Using Machine Learning Algorithms: Development and External Validation Study. Journal of Medical Internet Research 2022; 24(8): e37850. Phyu Mon L, Nyi Nyi S, Xianglong X, et al. Assessing disparity in the distribution of HIV and sexually transmitted infections in Australia: a retrospective cross-sectional study using Gini coefficients. BMJ Public Health 2023; 1(1): e000012. Nieuwenburg SA, Hoornenborg E, Davidovich U, de Vries HJC, Schim van der Loeff M. Developing a symptoms-based risk score for infectious syphilis among men who have sex with men. Sex Transm Infect 2023; 99(5): 324–9. Berry NA, Harvey JA, Pittelkow MR, Swanson DL, Yang YW. Online symptom checkers lack diagnostic accuracy for skin rashes. J Am Acad Dermatol 2023; 88(2): 487–8. Soe NN, Latt PM, Yu Z, et al. Clinical features-based machine learning models to separate sexually transmitted infections from other skin diagnoses. The Journal of infection 2024; 88(4): 106128. Venkatesh KP, Raza MM, Nickel G, Wang S, Kvedar JC. Deep learning models across the range of skin disease. NPJ Digital Medicine 2024; 7: 1–2. Page MJ, McKenzie JE, Bossuyt PM, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ 2021; 372: n71. Soe NN, Kusnandar II, Latt PM. Systematic review and meta-analysis of Artificial Intelligence for the diagnosis of anogenital skin lesions related to sexual health. 2024. https://www.crd.york.ac.uk/prospero/display_record.php?ID=CRD42024527420 . Covidence systematic review software. Veritas Health Innovation, Melbourne, Australia.; 2024. Whiting PF, Rutjes AWS, Westwood ME, et al. QUADAS-2: A Revised Tool for the Quality Assessment of Diagnostic Accuracy Studies. Annals of Internal Medicine 2011; 155(8): 529–36. Daneshjou R, Barata C, Betz-Stablein B, et al. Checklist for Evaluation of Image-Based Artificial Intelligence Reports in Dermatology: CLEAR Derm Consensus Guidelines From the International Skin Imaging Collaboration Artificial Intelligence Working Group. JAMA Dermatol 2022; 158(1): 90–6. StataCorp. 2024. Stata Statistical Software: Release 18. College Station, TX: StataCorp LLC. Gaffoor N, Soomro S. Skin Disease Detection and Classification Using ResNet-50 and Support Vector Machine: An Effective Approach for Dermatological Diagnosis. 2023; 2023. p. 140-5. Pangti R, Mathur J, Chouhan V, et al. A machine learning-based, decision support, mobile phone application for diagnosis of common dermatological diseases. Journal of the European Academy of Dermatology and Venereology 2021; 35(2): 536–45. Shen Y, Li H, Sun C, et al. Optimising skin disease diagnosis: harnessing online community data with contrastive learning and clustering techniques. npj Digital Medicine 2024; 7(1): 28. Soe NN, Yu Z, Latt PM, et al. Distinguishing monkeypox from common skin lesions using artificial intelligence in a sexual health clinic: a feasibility study. Journal of Medical Internet Research 2023; 20(5): XVI-XVI. Zaar O, Larson A, Polesie S, et al. Evaluation of the diagnostic accuracy of an online artificial intelligence application for skin disease diagnosis. Acta Dermato-Venereologica 2020; 100(16): 1–6. Soe NN, Yu Z, Latt PM, et al. Evaluation of artificial intelligence-powered screening for sexually transmitted infections-related skin lesions using clinical images and metadata. BMC Med 2024; 22(1): 296. Ali, Shams Nafisa, Ahmed, et al. Monkeypox Skin Lesion Detection Using Deep Learning Models: A Preliminary Feasibility Study. arXiv preprint arXiv:220703342 2022. 2022 International Conference on Computers and Artificial Intelligence Technologies, CAIT 2022. 2022; 2022. Campana MG, Colussi, M., Delmastro, F., Mascetti, S., & Pagani, E. Mpox Close Skin Images. Zenodo; 2023. Liu Y, Jain A, Eng C, et al. A deep learning system for differential diagnosis of skin diseases. Nature medicine 2020; 26(6): 900–8. Oztel I, Oztel GY, Sahin VH. Deep Learning-Based Skin Diseases Classification using Smartphones. ADVANCED INTELLIGENT SYSTEMS 2023; 5(12). Thieme AH, Zheng Y, Machiraju G, et al. A deep-learning algorithm to classify skin lesions from mpox virus infection. Nature medicine 2023; 29(3): 738–47. Yotsu RR, Ding Z, Hamm J, Blanton RE. Deep learning for AI-based diagnosis of skin-related neglected tropical diseases: A pilot study. PLoS neglected tropical diseases 2023; 17(8): e0011230. Allan-Blitz L-T, Ambepitiya S, Tirupathi R, Klausner JD. The Development and Performance of a Machine-Learning Based Mobile Platform for Visually Determining the Etiology of 5 Penile Diseases. Mayo Clinic Proceedings: Digital Health 2024; 2(2): 280-8. Campana MG, Colussi M, Delmastro F, Mascetti S, Pagani E. A Transfer Learning and Explainable Solution to Detect mpox from Smartphones images. PERVASIVE AND MOBILE COMPUTING 2024; 98. Casuayan De Goma J, Devaraj M. Recognising Common Skin Diseases in the Philippines Using Image Processing and Machine Learning Classification. 2020; 2020. p. 68–72. Thorat R, Gupta A. Transfer learning-enabled skin disease classification: the case of monkeypox detection. MULTIMEDIA TOOLS AND APPLICATIONS 2024. Back S, Lee S, Shin S, et al. Robust Skin Disease Classification by Distilling Deep Neural Network Ensemble for the Mobile Diagnosis of Herpes Zoster. IEEE ACCESS 2021; 9: 20156–69. Shumway NK. Skin Diseases of the Genitals and Perineum. In: Soutor C, Hordinsky MK, eds. Clinical Dermatology: Diagnosis and Management of Common Disorders, 2e. New York, NY: McGraw-Hill Education; 2022. DermNet. Genital skin problems. 2002. https://dermnetnz.org/topics/genital-skin-problems 2024). Stamm AW, Kobashi KC, Stefanovic KB. Urologic Dermatology: a Review. Curr Urol Rep 2017; 18(8): 62. Shobajo M, Abdulkadir S, Kraus CN, Shiu J. Anogenital skin of color images are underrepresented in dermatology, gynecology, and urology educational textbooks. Int J Womens Dermatol 2023; 9(3): e090. World Health Organization, International Agency for Research on Cancer. IARC Cervical Cancer Image Bank. https://screening.iarc.fr/cervicalimagebank.php . Asif S, Zhao M, Li YF, Tang FX, Khan SUR, Zhu YS. AI-Based Approaches for the Diagnosis of Mpox: Challenges and Future Prospects. ARCHIVES OF COMPUTATIONAL METHODS IN ENGINEERING 2024. Vega C, Schneider R, Satagopam V. Analysis: Flawed Datasets of Monkeypox Skin Images. Journal of medical systems 2023; 47(1): 37. Collins GS, Moons KGM, Dhiman P, et al. TRIPOD + AI statement: updated guidance for reporting clinical prediction models that use regression or machine learning methods. BMJ 2024; 385: e078378. Finnane A, Curiel-Lewandrowski C, Wimberley G, et al. Proposed Technical Guidelines for the Acquisition of Clinical Images of Skin-Related Conditions. JAMA Dermatol 2017; 153(5): 453–7. Dagan N, Devons-Sberro S, Paz Z, et al. Evaluation of AI Solutions in Health Care Organizations — The OPTICA Tool. Nejm Ai 2024; 1(9). Widner K, Virmani S, Krause J, et al. Lessons learned from translating AI from development to deployment in healthcare. Nat Med 2023; 29(6): 1304–6. Larrazabal AJ, Nieto N, Peterson V, Milone DH, Ferrante E. Gender imbalance in medical imaging datasets produces biased classifiers for computer-aided diagnosis. Proceedings of the National Academy of Sciences 2020; 117(23): 12592-4. Kelly CJ, Karthikesalingam A, Suleyman M, Corrado G, King D. Key challenges for delivering clinical impact with artificial intelligence. BMC Medicine 2019; 17(1). Callahan A, McElfresh D, Banda JM, et al. Standing on FURM Ground: A Framework for Evaluating Fair, Useful, and Reliable AI Models in Health Care Systems. NEJM Catalyst 2024; 5(10). Liu X, Cruz Rivera S, Moher D, et al. Reporting guidelines for clinical trial reports for interventions involving artificial intelligence: the CONSORT-AI extension. Nat Med 2020; 26(9): 1364–74. Ly S, Reyes-Hadsall S, Drake L, et al. Public Perceptions, Factors, and Incentives Influencing Patient Willingness to Share Clinical Images for Artificial Intelligence-Based Healthcare Tools. Dermatol Ther (Heidelb) 2023; 13(11): 2895–902. Soe NN, Latt PM, King A, et al. What Do People Want from an AI-Assisted Screening App for Sexually Transmitted Infection-Related Anogenital Lesions: A Discrete Choice Experiment. Patient 2024. Abraham E, Chow EP, Fairley CK, et al. eSexualHealth: Preferences to Use Technology to Promote Sexual Health Among Men Who Have Sex with Men and Trans and Gender Diverse People. Cold Spring Harbor Laboratory; 2022. Jakob L, Steeb T, Fiocco Z, et al. Patient Perception of Mobile Phone Apps for the Care and Prevention of Sexually Transmitted Diseases: Cross-Sectional Study. JMIR Mhealth Uhealth 2020; 8(11): e16517. Additional Declarations No competing interests reported. Supplementary Files 02SupplementaryInformation.docx 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-6002285","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":435067552,"identity":"407f5b97-dd47-41e2-9ec9-2e8f3810ab0a","order_by":0,"name":"Nyi Nyi Soe","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABDElEQVRIiWNgGAWjYDCCA0CcwMDAA2Z/qLCRAzMeEKmF8eCMM2nGYMEEQlqggPkwb9vhxAYGiCE4Ad/ts8ckHu44LCPf3nzgMM+Zw+nzww4/BNpiJ6fbgF2L5Lm8NInEM4d5DM4cSzg4pyI9d+PtNAOglmRjswPYtRic4TGTSGwDapHIMTjw5ox17sbZCSAtBxK3EdIiPyP/wwHeNuZ0w9npH4jTwnAjh+Egb5tzgrx0Dn5bJM/wGFsktqWD/GIACmTDDdI5BQcSDHD7he8Mj+HNn23W9sAQe/wBGJXy8rPTNwMZdnK4tAABiwQDQzOSU8EqDXAqBwHmDwwMdQiufANe1aNgFIyCUTACAQAV/2xnccyh7QAAAABJRU5ErkJggg==","orcid":"","institution":"Alfred Health","correspondingAuthor":true,"prefix":"","firstName":"Nyi","middleName":"Nyi","lastName":"Soe","suffix":""},{"id":435067553,"identity":"397b2360-01a6-4c90-bba7-a9fe7c259a21","order_by":1,"name":"Ingsun Isika Kusnandar","email":"","orcid":"","institution":"National School of Medicine, University of Notre Dame","correspondingAuthor":false,"prefix":"","firstName":"Ingsun","middleName":"Isika","lastName":"Kusnandar","suffix":""},{"id":435067554,"identity":"75a1b91c-8272-4f6e-a696-5ed75f07a389","order_by":2,"name":"Phyu Mon Latt","email":"","orcid":"","institution":"Alfred Health","correspondingAuthor":false,"prefix":"","firstName":"Phyu","middleName":"Mon","lastName":"Latt","suffix":""},{"id":435067555,"identity":"ec5281e8-dc9e-46c8-8ae9-c71701e0feb7","order_by":3,"name":"Christopher K. Fairley","email":"","orcid":"","institution":"Monash University","correspondingAuthor":false,"prefix":"","firstName":"Christopher","middleName":"K.","lastName":"Fairley","suffix":""},{"id":435067556,"identity":"ef38f9c0-1dfc-4be8-9662-e87f3cc1ddee","order_by":4,"name":"Eric P.F. Chow","email":"","orcid":"","institution":"Alfred Health","correspondingAuthor":false,"prefix":"","firstName":"Eric","middleName":"P.F.","lastName":"Chow","suffix":""},{"id":435067557,"identity":"39e3334a-db5b-47ef-ab34-549cd8a29cf4","order_by":5,"name":"Ismail Maatouk","email":"","orcid":"","institution":"World Health Organization","correspondingAuthor":false,"prefix":"","firstName":"Ismail","middleName":"","lastName":"Maatouk","suffix":""},{"id":435067559,"identity":"d5083d42-775f-4249-b24d-64207f2a64f3","order_by":6,"name":"Cheryl C. Johnson","email":"","orcid":"","institution":"World Health Organization","correspondingAuthor":false,"prefix":"","firstName":"Cheryl","middleName":"C.","lastName":"Johnson","suffix":""},{"id":435067560,"identity":"e8271cde-029b-4943-aac5-93ba59729bd3","order_by":7,"name":"Purvi Shah","email":"","orcid":"","institution":"World Health Organization","correspondingAuthor":false,"prefix":"","firstName":"Purvi","middleName":"","lastName":"Shah","suffix":""},{"id":435067562,"identity":"0ee76a68-741a-41d1-a046-c3d92f4a62fe","order_by":8,"name":"Remco Peters","email":"","orcid":"","institution":"World Health Organization","correspondingAuthor":false,"prefix":"","firstName":"Remco","middleName":"","lastName":"Peters","suffix":""},{"id":435067565,"identity":"e0792757-a0e1-4c76-b72e-f09d41e8deff","order_by":9,"name":"Lorenzo Subissi","email":"","orcid":"","institution":"World Health Organization","correspondingAuthor":false,"prefix":"","firstName":"Lorenzo","middleName":"","lastName":"Subissi","suffix":""},{"id":435067569,"identity":"86214c79-3df4-4ff8-83aa-b7417cb1b557","order_by":10,"name":"Lei Zhang","email":"","orcid":"","institution":"Alfred Health","correspondingAuthor":false,"prefix":"","firstName":"Lei","middleName":"","lastName":"Zhang","suffix":""},{"id":435067571,"identity":"09f0c2f3-ba61-4b96-97a4-146894762e8d","order_by":11,"name":"Jason J Ong","email":"","orcid":"","institution":"Alfred Health","correspondingAuthor":false,"prefix":"","firstName":"Jason","middleName":"J","lastName":"Ong","suffix":""}],"badges":[],"createdAt":"2025-02-10 22:23:08","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6002285/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6002285/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":80633347,"identity":"c425b4b0-aeaa-467e-8ab0-7dd755493c2f","added_by":"auto","created_at":"2025-04-15 11:58:53","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":618859,"visible":true,"origin":"","legend":"\u003cp\u003ePRISMA flow diagram of the study selection process\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-6002285/v1/c5664f0f927c39039f9e3d77.png"},{"id":80633346,"identity":"d069f4e1-a5c3-4334-ab1a-222f4e6dc5e3","added_by":"auto","created_at":"2025-04-15 11:58:53","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":392517,"visible":true,"origin":"","legend":"\u003cp\u003eQuality assessment by (a) CLEAR Derm Checklist and (b) modified QUADAS-2 tool\u003c/p\u003e","description":"","filename":"floatimage214.png","url":"https://assets-eu.researchsquare.com/files/rs-6002285/v1/70c37de176c7cc93bf4b225f.png"},{"id":80631764,"identity":"e43740bf-2ea0-4820-8fe1-ef75d7c42bbf","added_by":"auto","created_at":"2025-04-15 11:50:53","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":663385,"visible":true,"origin":"","legend":"\u003cp\u003eAI model accuracy for disease conditions and number of studies by year\u003c/p\u003e","description":"","filename":"floatimage48.png","url":"https://assets-eu.researchsquare.com/files/rs-6002285/v1/25c759bc9b7189c75dabde10.png"},{"id":80631768,"identity":"83ffda98-666d-4ccf-99e0-b158628999be","added_by":"auto","created_at":"2025-04-15 11:50:53","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":1583711,"visible":true,"origin":"","legend":"\u003cp\u003eCharacteristics of included studies\u003c/p\u003e","description":"","filename":"floatimage56.png","url":"https://assets-eu.researchsquare.com/files/rs-6002285/v1/8e41866f59f391202b140756.png"},{"id":80752963,"identity":"23516673-7dd9-41a3-a7be-3afc1b578137","added_by":"auto","created_at":"2025-04-16 16:55:27","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":5333834,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6002285/v1/39123cfe-fff8-4e7d-b846-8a206404497c.pdf"},{"id":80631777,"identity":"417650aa-bf48-497c-a492-4c397a79756b","added_by":"auto","created_at":"2025-04-15 11:50:53","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":6592417,"visible":true,"origin":"","legend":"","description":"","filename":"02SupplementaryInformation.docx","url":"https://assets-eu.researchsquare.com/files/rs-6002285/v1/9144652222945904b2b3e813.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Clinical Applications of AI in Sexually Transmitted Infection and Anogenital Dermatoses in Sexual Health: A Systematic Review and Meta-Analysis","fulltext":[{"header":"Introduction","content":"\u003cp\u003eSexually transmitted infections (STIs) profoundly affect sexual and reproductive health globally.\u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u003c/sup\u003e The World Health Organization (WHO) estimated 374\u0026nbsp;million new cases of STIs in 2020: chlamydia (129\u0026nbsp;million), gonorrhoea (82\u0026nbsp;million), syphilis (8\u0026nbsp;million) and trichomoniasis (156\u0026nbsp;million).\u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e These estimates do not include mpox, for which evidence has recently emerged showing the important role of sexual transmission in mpox outbreaks.\u003csup\u003e\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e Delayed diagnoses and untreated infections can result in serious complications, including a higher risk of HIV acquisition, infertility, and adverse pregnancy outcomes.\u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e The recent increase in syphilis cases across populations and resurgence of congenital syphilis highlight the urgent need for improved early identification and control measures for STIs.\u003csup\u003e\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e \u003cp\u003eSTIs can be asymptomatic or present with diverse clinical manifestations, including genitourinary symptoms and anogenital dermatological presentations.\u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u003c/sup\u003e Recent data from a sexual health clinic show that nearly half of the survey clients (n\u0026thinsp;=\u0026thinsp;10,479) over three years presented with anogenital conditions such as lumps, spots, rashes, sores or itch.\u003csup\u003e\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u003c/sup\u003e These presentations can be infectious, such as painless chancres in syphilis or painful sores in genital herpes. These can also be non-infectious, like inflammatory conditions, benign lesions, and neoplasms. This complexity in clinical presentation poses a significant barrier for the healthcare workforce to timely diagnosis and management of STIs.\u003c/p\u003e \u003cp\u003eDigital health interventions, particularly artificial intelligence (AI), have emerged as promising tools in sexual health to promote testing and health-seeking behaviours among symptomatic individuals and those at high risk of STIs. Machine learning approaches have been used to estimate HIV and STI risks based on demographic information and sexual health behaviours.\u003csup\u003e\u003cspan additionalcitationids=\"CR9\" citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u003c/sup\u003e Machine learning and Bayesian-based symptom checkers have been developed to assess the different presentations of STIs among those with symptoms.\u003csup\u003e\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e,\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e However, the user input data for these anogenital dermatoses could be subjective and limit the accuracy of the symptom checkers for dermatological conditions.\u003csup\u003e\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u003c/sup\u003e The distinctive patterns of anogenital lesions suggest potential value in deep learning-based image classification algorithms for STI diagnosis.\u003csup\u003e\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e\u003c/sup\u003e While AI has shown success in general dermatology, its application to STI-related dermatoses remains nascent, and its role in clinical implementation is unclear.\u003csup\u003e\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e \u003cp\u003eThis systematic review and meta-analysis aim to: (1) evaluate the diagnostic performance of AI algorithms in identifying and classifying STIs and other anogenital dermatoses; (2) assess the methodological quality and clinical applicability of existing studies; (3) identify key technical and implementation challenges; and (4) provide evidence-based recommendations for advancing these technologies from research into clinical practice. We hypothesised that while AI shows promise for STIs and anogenital dermatoses, significant gaps exist in current research regarding clinical validation, demographic representation, and real-world implementation.\u003c/p\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eStudy design and protocol\u003c/h2\u003e \u003cp\u003eWe conducted this systematic review according to the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines.\u003csup\u003e\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u003c/sup\u003e We pre-registered our protocol in the International Prospective Register of Systematic Reviews (PROSPERO, CRD42024527420).\u003csup\u003e\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u003c/sup\u003e The study team included clinical experts in sexual health, specialists in medical AI applications, biostatisticians, and a medical librarian.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eSearch strategy and selection criteria\u003c/h3\u003e\n\u003cp\u003eIn consultation with a librarian (LR), we searched six databases (IEEE Xplore, Embase, Scopus, MEDLINE, Web of Science, and CINAHL) for studies from January 1, 2010, to April 12, 2024. We used three main concepts: \u0026ldquo;artificial intelligence\u0026rdquo;, \u0026ldquo;diagnosis\u0026rdquo;, and \u0026ldquo;sexually transmitted infections\u0026rdquo; for our search strategies. The detailed search strategies are presented in Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e. Our search included studies ranging from proof-of-concept to randomised controlled trials without language restrictions. We included studies that used artificial intelligence to identify or classify anogenital skin conditions using clinical images, with or without accompanying metadata. We excluded studies that did not involve AI-based approaches or did not target anogenital skin conditions related to sexual health. We also omitted review articles and studies that did not report model performance metrics or focused on dermatoscopic images and non-skin manifestations, such as cervical changes. Two reviewers (NS and IK) independently screened titles and abstracts against our eligibility criteria after removing duplicates using Covidence software.\u003csup\u003e\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e\u003c/sup\u003e We conducted pilot testing before formal screening and data extraction, with regular team meetings to ensure consistency. After screening 5,381 studies with 97% agreement in screening, two additional reviewers (PL and JJO) resolved any remaining conflicts through consensus.\u003c/p\u003e\n\u003ch3\u003eData extraction\u003c/h3\u003e\n\u003cp\u003eTwo reviewers (NS and IK) independently assessed 258 full-text articles and extracted data using a standardised Excel sheet. PL and JJO resolved any disagreements in data extraction through discussion. We collected data on the following aspects: (i) study characteristics, including authors, publication year, and study type; (ii) data characteristics, such as the image database, types of data input and predictions, reference conditions, and sample size; (iii) technical information, including model development, evaluation methods, subgroup analyses, model interpretability, and the best-performing algorithms; and (iv) performance metrics, such as ROC-AUC, accuracy, sensitivity, specificity, precision, F1 scores, and values from confusion matrices. We selected performance metrics from the best-performing model for studies reporting multiple models. We extracted true positives (TP), false negatives (FN), true negatives (TN), and false positives (FP) directly from 2x2 tables for binary outcome studies. We recalculated these values from the tables focusing on our target disease conditions in the studies with multiclass classifications.\u003c/p\u003e\n\u003ch3\u003eQuality assessment\u003c/h3\u003e\n\u003cp\u003eTwo reviewers (NS and IK) used a modified QUADAS-2 critical appraisal tool\u003csup\u003e\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e\u003c/sup\u003e to independently assess the risk of bias and the applicability of included studies across the population, index test and reference standards. We rated each domain as low, high or uncertain risk based on the information provided in the studies. We also evaluated model fairness, reliability, and safety using the CLEAR Derm checklist,\u003csup\u003e\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e\u003c/sup\u003e scoring 25 items across data, technique, technical assessment, and application domains as present, partially present, absent, or not applicable. Disagreements were resolved through discussion among the review team (NS, IK, PL, JJO).\u003c/p\u003e \u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003eData analysis\u003c/h2\u003e \u003cp\u003eDue to incomplete reporting across studies, we conducted a descriptive analysis of available data on study characteristics, technical methodology, and model performance metrics. We used dupeGuru version 4.0.3 to identify duplicate images across public datasets. We conducted a bivariate random effect meta-analysis to estimate AUC-ROC, sensitivity and specificity using data from 31 mpox studies and four studies each for herpes simplex, genital warts, psoriasis, and scabies. We generated forest plots and summary receiver operating characteristic (SROC) graphs for each condition. We used funnel plots of diagnostic odds ratios and Deek\u0026rsquo;s regression test to assess publication bias. We conducted a meta-regression analysis on mpox studies using the Higgins I\u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e statistic to assess heterogeneity based on sample size, number of reference skin conditions, type of model classification, and AI algorithm category. We measured the F-statistics and p-value for the significance of the regression models. We analysed all data using STATA version 17 (StataCorp, College Station, TX, USA).\u003csup\u003e\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e \u003c/div\u003e"},{"header":"Results","content":"\u003cp\u003eOur initial search identified 6,035 studies. After removing duplicates, we screened 5,381 titles and abstracts and selected 258 studies for full-text review. After removing irrelevant studies, duplicates, and review articles, 141 studies were deemed eligible and included in the review (Fig. \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e).\u003c/p\u003e\n\u003ch3\u003eQuality assessment\u003c/h3\u003e\n\u003cp\u003eThe quality assessment revealed significant high level of bias across the included studies. Using the modified QUADAS-2 tool, we identified a high risk of bias in most studies regarding three domains: populations (76.1%), reference standard tests (76.1%) and index tests (20.0%). We also identified a high concern for applicability across these three domains (above 75.0%). In the CLEAR Derm checklist, most studies are inadequately reported across four domains: data, technique, technical assessment, and application. Detailed assessment scores for individual studies are presented in Fig. \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e and Table S3.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDisease conditions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eMost studies (n\u0026thinsp;=\u0026thinsp;111, 62.4%) reported on AI algorithms focused on mpox, while other conditions such as tinea cruris (n\u0026thinsp;=\u0026thinsp;8), genital warts (n\u0026thinsp;=\u0026thinsp;8), scabies (n\u0026thinsp;=\u0026thinsp;8), herpes zoster (n\u0026thinsp;=\u0026thinsp;8), psoriasis (n\u0026thinsp;=\u0026thinsp;7), herpes simplex (n\u0026thinsp;=\u0026thinsp;7), lichenoid changes (n\u0026thinsp;=\u0026thinsp;6), molluscum contagiosum (n\u0026thinsp;=\u0026thinsp;6) were reported in less than 6.0% of the studies. Less common conditions included folliculitis (n\u0026thinsp;=\u0026thinsp;3), balanitis (n\u0026thinsp;=\u0026thinsp;2), syphilis (n\u0026thinsp;=\u0026thinsp;1), candidiasis (n\u0026thinsp;=\u0026thinsp;1), penile cancer (n\u0026thinsp;=\u0026thinsp;1), and combined STIs (n\u0026thinsp;=\u0026thinsp;1) (Fig. \u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003ea). The 2022 mpox outbreak significantly increased mpox studies during 2022\u0026ndash;2023 (Fig. \u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003eb). The types of target and reference conditions included in the studies ranged from 2 to 44. Most mpox studies (n\u0026thinsp;=\u0026thinsp;96) focused on a limited number of differential diagnoses, primarily comparing mpox with conditions that are unrelated to common anogenital dermatoses typically seen in sexual health clinics (Table S2d). Among all included studies, only 18 incorporated one or more relevant differential conditions, while only six provided a broader comparison with multiple anogenital dermatoses.\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e21\u003c/span\u003e\u0026ndash;\u003cspan class=\"CitationRef\"\u003e26\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e\n\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\n \u003ch2\u003eData\u003c/h2\u003e\n \u003cp\u003eMost studies used open-source data (n\u0026thinsp;=\u0026thinsp;122, 87.8%), while a few used private databases (n\u0026thinsp;=\u0026thinsp;12, 8.6%) or a combination of private and public datasets (n\u0026thinsp;=\u0026thinsp;5, 3.6%) (Fig. \u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003e). Nearly three-quarters of studies (n\u0026thinsp;=\u0026thinsp;104, 73.8%) properly referenced the image databases. The most frequently used public databases for mpox studies were \u0026lsquo;MSLD\u0026rsquo;, \u0026lsquo;MSID\u0026rsquo; and \u0026lsquo;MSCI\u0026rsquo;,\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e27\u003c/span\u003e\u0026ndash;\u003cspan class=\"CitationRef\"\u003e29\u003c/span\u003e\u003c/sup\u003e where 7.3% (84/1155) of images were identified as duplicates (Table S3). Private datasets originated mainly from Asia (28.6%; China\u0026thinsp;=\u0026thinsp;2, India\u0026thinsp;=\u0026thinsp;1, the Philippines\u0026thinsp;=\u0026thinsp;1), followed by the Americas (21.4%; USA\u0026thinsp;=\u0026thinsp;2, Peru\u0026thinsp;=\u0026thinsp;1), with equal representation from Europe (Munich\u0026thinsp;=\u0026thinsp;1, Sweden\u0026thinsp;=\u0026thinsp;1) and Oceania (Australia\u0026thinsp;=\u0026thinsp;2), and West Africa (7.1%). The sample size of the total images varied substantially across different datasets, ranging from 70 to 139,198. Studies showed considerable class imbalance, where the proportion of target condition images ranged from 0.1\u0026ndash;71.8% of total images within datasets (median 36.2%, IQR 19.9%-44.7%).\u003c/p\u003e\n \u003cp\u003eMost datasets (n\u0026thinsp;=\u0026thinsp;133, 94.3%) that were used contained clinical images alone, while only eight (5.6%) contained additional patient metadata (sex, race, ethnicity, symptoms, etc.).\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e23\u003c/span\u003e,\u003cspan class=\"CitationRef\"\u003e25\u003c/span\u003e,\u003cspan class=\"CitationRef\"\u003e26\u003c/span\u003e,\u003cspan class=\"CitationRef\"\u003e30\u003c/span\u003e\u0026ndash;\u003cspan class=\"CitationRef\"\u003e34\u003c/span\u003e\u003c/sup\u003e Eight studies reported skin tones in relation to disease conditions.\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e23\u003c/span\u003e,\u003cspan class=\"CitationRef\"\u003e25\u003c/span\u003e,\u003cspan class=\"CitationRef\"\u003e30\u003c/span\u003e,\u003cspan class=\"CitationRef\"\u003e32\u003c/span\u003e,\u003cspan class=\"CitationRef\"\u003e33\u003c/span\u003e,\u003cspan class=\"CitationRef\"\u003e35\u003c/span\u003e\u0026ndash;\u003cspan class=\"CitationRef\"\u003e37\u003c/span\u003e\u003c/sup\u003e Most studies (n\u0026thinsp;=\u0026thinsp;126, 89.4%) did not report technical details about image acquisition, such as image quality, camera type, or lighting conditions. Only 17 studies (12.1%) validated image diagnoses using laboratory tests, clinician-reviewed images, or both. Three studies standardised their image categories using International Classification of Diseases (ICD) codes.\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e23\u003c/span\u003e,\u003cspan class=\"CitationRef\"\u003e25\u003c/span\u003e,\u003cspan class=\"CitationRef\"\u003e30\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\n \u003ch2\u003eModel development\u003c/h2\u003e\n \u003cp\u003eOver half of the studies (n\u0026thinsp;=\u0026thinsp;78, 55.3%) focused on developing binary classification models (differentiating a target disease from other reference conditions), while 58 studies (41.1%) used multiclass classification approaches (classifying multiple conditions simultaneously). Only two studies (1.4%) employed multimodal approaches, incorporating images and patient metadata for model prediction, while all other studies used sorely image data input.\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e26\u003c/span\u003e,\u003cspan class=\"CitationRef\"\u003e30\u003c/span\u003e\u003c/sup\u003e Most studies (n\u0026thinsp;=\u0026thinsp;124, 87.9%) reported image preprocessing procedures such as resizing, normalisation, cropping, and rotation for model training. Studies addressed data scarcity and class imbalance through image augmentation, but none used synthetic images.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\n \u003ch2\u003eModel evaluation\u003c/h2\u003e\n \u003cp\u003eTo evaluate model performance, 127 studies (90.1%) reported fully or partially splitting their data into training and testing sets, while only 35 studies (24.8%) used k-fold cross-validation for more robust validation. Eight studies (5.6%) assessed model generalisability using external validation or prospectively collected image datasets, and only three studies (2.1%) used data from multiple vendors or clinics.\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e30\u003c/span\u003e,\u003cspan class=\"CitationRef\"\u003e32\u003c/span\u003e,\u003cspan class=\"CitationRef\"\u003e38\u003c/span\u003e\u003c/sup\u003e Patient-level metadata and the lesion site could influence the model performance. While six studies considered gender in their analyses, only one study specifically evaluated model performance for females. Only a few studies evaluated model performance across different subgroups: skin tone (n\u0026thinsp;=\u0026thinsp;5), lesion site (n\u0026thinsp;=\u0026thinsp;5), age (n\u0026thinsp;=\u0026thinsp;4), and race and ethnicity (n\u0026thinsp;=\u0026thinsp;2). To understand how models reached their decisions, 24 studies (17.0%) used interpretability techniques like Grad-CAM, LIME and SHAP to visualise important image features for model prediction.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\n \u003ch2\u003eModel performance\u003c/h2\u003e\n \u003cp\u003eDue to diverse reporting approaches across studies, we summarised model performance metrics by target condition in Table \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e and Fig. \u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003ea. Studies commonly reported accuracy, sensitivity, specificity, positive predictive values (PPV), and F-1scores, while negative predictive values (NPV) and AUC-ROC scores were less frequently reported. Studies demonstrated high mean accuracy (\u0026gt;\u0026thinsp;70.0%) for syphilis, herpes simplex, genital warts, mpox, scabies, herpes zoster, and penile cancer, while other conditions showed lower accuracy (ranging from 42.0\u0026ndash;65.0%). Most conditions achieved high specificity (\u0026gt;\u0026thinsp;95.0%), but other metrics showed considerable variation within conditions. For example, accuracy in herpes simplex models ranged from 10.0\u0026ndash;97.0%, showing inconsistent model performance across studies. The highest performance was achieved with CNN-based architectures (n\u0026thinsp;=\u0026thinsp;97, 68.8%), followed by hybrid or ensemble models (n\u0026thinsp;=\u0026thinsp;29, 20.6%).\u003c/p\u003e\n \u003cdiv class=\"gridtable\"\u003e\u0026nbsp;\u003ctable id=\"Tab1\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eSummary of model performance across disease conditions\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003ccolgroup cols=\"8\"\u003e\u003c/colgroup\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\u0026nbsp;\u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eSensitivity\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eSpecificity\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003ePPV\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eNPV\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eAUC-ROC\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eAccuracy\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eF1-score\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003cth align=\"left\" colspan=\"8\"\u003e\n \u003cp\u003eSyphilis (n\u0026thinsp;=\u0026thinsp;1)\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMean\u0026thinsp;\u0026plusmn;\u0026thinsp;sd\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.87\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.99\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.91\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.94\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.95\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e# of studies\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colspan=\"8\"\u003e\n \u003cp\u003e\u003cstrong\u003eHerpes simplex (n\u0026thinsp;=\u0026thinsp;7)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMean\u0026thinsp;\u0026plusmn;\u0026thinsp;sd\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.58\u0026thinsp;\u0026plusmn;\u0026thinsp;0.45\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.98\u0026thinsp;\u0026plusmn;\u0026thinsp;0.02\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.63\u0026thinsp;\u0026plusmn;\u0026thinsp;0.46\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.997\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.74\u0026thinsp;\u0026plusmn;\u0026thinsp;0.30\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.66\u0026thinsp;\u0026plusmn;\u0026thinsp;0.48\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMedian (IQR)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.62\u003c/p\u003e\n \u003cp\u003e(0.20\u0026ndash;0.96)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.98\u003c/p\u003e\n \u003cp\u003e(0.97\u0026ndash;0.99)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.87\u003c/p\u003e\n \u003cp\u003e(0.10\u0026ndash;0.93)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.84\u003c/p\u003e\n \u003cp\u003e(0.69\u0026ndash;0.94)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.92 (0.10\u0026ndash;0.96)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eRange\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.10\u0026ndash;0.99\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.97\u0026ndash;0.99\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.10\u0026ndash;0.93\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.1\u0026ndash;0.97\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.10\u0026ndash;0.96\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e# of studies\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colspan=\"8\"\u003e\n \u003cp\u003e\u003cstrong\u003eGenital warts (n\u0026thinsp;=\u0026thinsp;8)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMean\u0026thinsp;\u0026plusmn;\u0026thinsp;sd\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.82\u0026thinsp;\u0026plusmn;\u0026thinsp;0.20\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.99\u0026thinsp;\u0026plusmn;\u0026thinsp;0.02\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.69\u0026thinsp;\u0026plusmn;\u0026thinsp;0.37\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.85\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.86\u0026thinsp;\u0026plusmn;\u0026thinsp;0.15\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.96\u0026thinsp;\u0026plusmn;\u0026thinsp;0.05\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMedian (IQR)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.93\u003c/p\u003e\n \u003cp\u003e(0.57\u0026ndash;0.96)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.00\u003c/p\u003e\n \u003cp\u003e(0.96-1.00)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.86\u003c/p\u003e\n \u003cp\u003e(0.57\u0026ndash;0.94)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.94\u003c/p\u003e\n \u003cp\u003e(0.69\u0026ndash;0.97)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.97\u003c/p\u003e\n \u003cp\u003e(0.91-1.00)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eRange\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.56-1.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.96-1.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.10-1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.65\u0026ndash;0.98\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.91-1.00\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e# of studies\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colspan=\"8\"\u003e\n \u003cp\u003e\u003cstrong\u003eMpox (n\u0026thinsp;=\u0026thinsp;111)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMean\u0026thinsp;\u0026plusmn;\u0026thinsp;sd\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.91\u0026thinsp;\u0026plusmn;\u0026thinsp;0.09\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.93\u0026thinsp;\u0026plusmn;\u0026thinsp;0.09\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.93\u0026thinsp;\u0026plusmn;\u0026thinsp;0.06\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.94\u0026thinsp;\u0026plusmn;\u0026thinsp;0.09\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.96\u0026thinsp;\u0026plusmn;\u0026thinsp;0.06\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.93\u0026thinsp;\u0026plusmn;\u0026thinsp;0.06\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.92\u0026thinsp;\u0026plusmn;\u0026thinsp;0.08\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMedian (IQR)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.94\u003c/p\u003e\n \u003cp\u003e(0.88-1.00)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.97\u003c/p\u003e\n \u003cp\u003e(0.90\u0026ndash;0.99)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.95\u003c/p\u003e\n \u003cp\u003e(0.91-1.00)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.98\u003c/p\u003e\n \u003cp\u003e(0.93\u0026ndash;0.99)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.98\u003c/p\u003e\n \u003cp\u003e(0.94-1.00)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.95\u003c/p\u003e\n \u003cp\u003e(0.91-1.00)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.94\u003c/p\u003e\n \u003cp\u003e(0.90-1.00)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eRange\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.58-1.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.61-1.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.70-1.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.79-1.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.74-1.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.94-1.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.67-1.00\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e# of models\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e115\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e37\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e103\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e37\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e128\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e108\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colspan=\"8\"\u003e\n \u003cp\u003e\u003cstrong\u003eMolluscum contagiosum (n\u0026thinsp;=\u0026thinsp;6)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMean\u0026thinsp;\u0026plusmn;\u0026thinsp;sd\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.53\u0026thinsp;\u0026plusmn;\u0026thinsp;0.39\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.99\u0026thinsp;\u0026plusmn;\u0026thinsp;0.01\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.28\u0026thinsp;\u0026plusmn;\u0026thinsp;0.25\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.98\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.62\u0026thinsp;\u0026plusmn;\u0026thinsp;0.36\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.52\u0026thinsp;\u0026plusmn;\u0026thinsp;0.60\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMedian (IQR)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.64\u003c/p\u003e\n \u003cp\u003e(0.10\u0026ndash;0.85)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.99\u003c/p\u003e\n \u003cp\u003e(0.99-1.00)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.28\u003c/p\u003e\n \u003cp\u003e(0.10\u0026ndash;0.45)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.72\u003c/p\u003e\n \u003cp\u003e(0.40\u0026ndash;0.84)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.52\u003c/p\u003e\n \u003cp\u003e(0.10\u0026ndash;0.94)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eRange\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.10\u0026ndash;0.85\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.99-1.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.10\u0026ndash;0.45\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.10\u0026ndash;0.92\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.10\u0026ndash;0.94\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e# of studies\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colspan=\"8\"\u003e\n \u003cp\u003e\u003cstrong\u003eTinea cruris (n\u0026thinsp;=\u0026thinsp;8)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMean\u0026thinsp;\u0026plusmn;\u0026thinsp;sd\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.55\u0026thinsp;\u0026plusmn;\u0026thinsp;0.38\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.98\u0026thinsp;\u0026plusmn;\u0026thinsp;0.02\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.59\u0026thinsp;\u0026plusmn;\u0026thinsp;0.43\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.98\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.95\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.53\u0026thinsp;\u0026plusmn;\u0026thinsp;0.35\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.51\u0026thinsp;\u0026plusmn;\u0026thinsp;0.58\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMedian (IQR)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.55\u003c/p\u003e\n \u003cp\u003e(0.23\u0026ndash;0.92)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.98\u003c/p\u003e\n \u003cp\u003e(0.97\u0026ndash;0.99)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.76\u003c/p\u003e\n \u003cp\u003e(0.10\u0026ndash;0.92)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.55\u003c/p\u003e\n \u003cp\u003e(0.25\u0026ndash;0.80)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.51\u003c/p\u003e\n \u003cp\u003e(0.10\u0026ndash;0.92\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eRange\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.10\u0026ndash;0.93\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.97\u0026ndash;0.99\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.10\u0026ndash;0.92\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.10\u0026ndash;0.92\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.10\u0026ndash;0.92\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e# of studies\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colspan=\"8\"\u003e\n \u003cp\u003e\u003cstrong\u003eLichenoid changes (n\u0026thinsp;=\u0026thinsp;6)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMean\u0026thinsp;\u0026plusmn;\u0026thinsp;sd\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.59\u0026thinsp;\u0026plusmn;\u0026thinsp;0.30\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.97\u0026thinsp;\u0026plusmn;\u0026thinsp;0.01\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.71\u0026thinsp;\u0026plusmn;\u0026thinsp;0.26\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.99\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.88\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.65\u0026thinsp;\u0026plusmn;\u0026thinsp;0.25\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.83\u0026thinsp;\u0026plusmn;\u0026thinsp;0.24\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMedian (IQR)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.57\u003c/p\u003e\n \u003cp\u003e(0.56\u0026ndash;0.68)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.97\u003c/p\u003e\n \u003cp\u003e(0.97\u0026ndash;0.98)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.64\u003c/p\u003e\n \u003cp\u003e(0.49-1.00)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.69\u003c/p\u003e\n \u003cp\u003e(0.53\u0026ndash;0.73)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.83\u003c/p\u003e\n \u003cp\u003e(0.66-1.00)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eRange\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.16-1.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.97\u0026ndash;0.98\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.31\u0026ndash;0.99\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.66-1.00\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e# of studies\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colspan=\"8\"\u003e\n \u003cp\u003e\u003cstrong\u003eScabies (n\u0026thinsp;=\u0026thinsp;8)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMean\u0026thinsp;\u0026plusmn;\u0026thinsp;sd\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.83\u0026thinsp;\u0026plusmn;\u0026thinsp;0.16\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.89\u0026thinsp;\u0026plusmn;\u0026thinsp;0.13\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.85\u0026thinsp;\u0026plusmn;\u0026thinsp;0.10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMedian (IQR)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.86\u003c/p\u003e\n \u003cp\u003e(0.71\u0026ndash;0.95)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.89\u003c/p\u003e\n \u003cp\u003e(0.80\u0026ndash;0.98)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.85\u003c/p\u003e\n \u003cp\u003e(0.77\u0026ndash;0.92)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eRange\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.63\u0026ndash;0.98\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.80\u0026ndash;0.98\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.69\u0026ndash;0.99\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e# of studies\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colspan=\"8\"\u003e\n \u003cp\u003e\u003cstrong\u003eFolliculitis (n\u0026thinsp;=\u0026thinsp;3)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMean\u0026thinsp;\u0026plusmn;\u0026thinsp;sd\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.36\u0026thinsp;\u0026plusmn;\u0026thinsp;0.35\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.42\u0026thinsp;\u0026plusmn;\u0026thinsp;0.24\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMedian (IQR)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.36\u003c/p\u003e\n \u003cp\u003e(0.11\u0026ndash;0.60)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.55\u003c/p\u003e\n \u003cp\u003e(0.14\u0026ndash;0.57)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eRange\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.11\u0026ndash;0.60\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.14\u0026ndash;0.57\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e# of studies\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colspan=\"8\"\u003e\n \u003cp\u003e\u003cstrong\u003eHerpes zoster (n\u0026thinsp;=\u0026thinsp;8)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMean\u0026thinsp;\u0026plusmn;\u0026thinsp;sd\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.82\u0026thinsp;\u0026plusmn;\u0026thinsp;0.13\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.98\u0026thinsp;\u0026plusmn;\u0026thinsp;0.03\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.85\u0026thinsp;\u0026plusmn;\u0026thinsp;0.19\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.94\u0026thinsp;\u0026plusmn;\u0026thinsp;0.05\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.94\u0026thinsp;\u0026plusmn;\u0026thinsp;0.03\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.95\u0026thinsp;\u0026plusmn;\u0026thinsp;0.04\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMedian (IQR)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.83\u003c/p\u003e\n \u003cp\u003e(0.71\u0026ndash;0.91)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.00\u003c/p\u003e\n \u003cp\u003e(0.94-1.00)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.93\u003c/p\u003e\n \u003cp\u003e(0.74\u0026ndash;0.96)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.93\u003c/p\u003e\n \u003cp\u003e(0.89\u0026ndash;0.99)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.95\u003c/p\u003e\n \u003cp\u003e(0.92\u0026ndash;0.96)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.95\u003c/p\u003e\n \u003cp\u003e(0.92\u0026ndash;0.99)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eRange\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.67\u0026ndash;0.96\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.94-1.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.57\u0026ndash;0.97\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.89\u0026ndash;0.99\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.90\u0026ndash;0.97\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.92\u0026ndash;0.99\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e# of studies\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colspan=\"8\"\u003e\n \u003cp\u003e\u003cstrong\u003ePsoriasis (n\u0026thinsp;=\u0026thinsp;7)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMean\u0026thinsp;\u0026plusmn;\u0026thinsp;sd\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.72\u0026thinsp;\u0026plusmn;\u0026thinsp;0.35\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.96\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.65\u0026thinsp;\u0026plusmn;\u0026thinsp;0.40\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.98\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.93\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.58\u0026thinsp;\u0026plusmn;\u0026thinsp;0.41\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.55\u0026thinsp;\u0026plusmn;\u0026thinsp;0.64\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMedian (IQR)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.84\u003c/p\u003e\n \u003cp\u003e(0.78-1.00)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.75\u003c/p\u003e\n \u003cp\u003e(0.36-1.00)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.62\u003c/p\u003e\n \u003cp\u003e(0.25\u0026ndash;0.92)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.55\u003c/p\u003e\n \u003cp\u003e(0.10-1.00)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eRange\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.10-1.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.10-1.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.10\u0026ndash;0.99\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.10-1.00\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e# of studies\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colspan=\"8\"\u003e\n \u003cp\u003e\u003cstrong\u003eCandidiasis (n\u0026thinsp;=\u0026thinsp;1)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMean\u0026thinsp;\u0026plusmn;\u0026thinsp;sd\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.41\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.41\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.81\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e# of studies\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colspan=\"8\"\u003e\n \u003cp\u003e\u003cstrong\u003eBalanitis (n\u0026thinsp;=\u0026thinsp;2)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMean\u0026thinsp;\u0026plusmn;\u0026thinsp;sd\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.88\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.97\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.64\u0026thinsp;\u0026plusmn;\u0026thinsp;0.43\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.98\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMedian (IQR)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.64\u003c/p\u003e\n \u003cp\u003e(0.33\u0026ndash;0.94)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eRange\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.33\u0026ndash;0.94\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e# of studies\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colspan=\"8\"\u003e\n \u003cp\u003e\u003cstrong\u003ePenile cancer (n\u0026thinsp;=\u0026thinsp;1)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMean\u0026thinsp;\u0026plusmn;\u0026thinsp;sd\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.79\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.99\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.89\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.94\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.93\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e# of studies\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colspan=\"8\"\u003e\n \u003cp\u003e\u003cstrong\u003eSTIs (n\u0026thinsp;=\u0026thinsp;1)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMean\u0026thinsp;\u0026plusmn;\u0026thinsp;sd\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.95\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.62\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.43\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.89\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.69\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e# of studies\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003ctfoot\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"8\"\u003e\u003cem\u003eAUC-ROC, Area Under the Curve-Receiver Operating Characteristic; IQR: Interquartile range; NPV: Negative Predictive Value; PPV: Positive Predictive Value; Sd: standard deviation; STI: sexually transmitted infection;\u003c/em\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tfoot\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n \u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e\n \u003ch2\u003eApplication\u003c/h2\u003e\n \u003cp\u003eMost studies remained at the proof-of-concept stage without publicly available models for external evaluation. They also did not specify their intended users (clinicians or the public) or the tool\u0026apos;s purpose (such as triage, assisted diagnosis or population screening). Only one study tested a public-facing application (Skin Image Search\u0026trade;) with prospectively collected images.\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e25\u003c/span\u003e\u003c/sup\u003e No studies evaluated model accuracy in clinical settings using randomised trials.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec16\" class=\"Section2\"\u003e\n \u003ch2\u003eMeta-analysis\u003c/h2\u003e\n \u003cp\u003eFor meta-analysis, we included 31 mpox studies (34 contingency tables) and four studies each for herpes simplex, genital warts, psoriasis, and scabies (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e). We could not pool the sensitivity and specificity for other conditions due to insufficient data points. Models showed consistently high performance across conditions: mpox (pooled sensitivity: 0.96 [95% CI: 0.93\u0026ndash;0.97], I\u003csup\u003e2\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;91.0%, pooled specificity: 0.98 [0.97\u0026ndash;0.99], I\u003csup\u003e2\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;99.9%), herpes simplex (sensitivity: 0.91 [0.71\u0026ndash;0.98], I\u003csup\u003e2\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;93.2%, specificity: 0.97 [0.94\u0026ndash;0.98], I\u003csup\u003e2\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;72.6%), genital warts (sensitivity: 0.87 [0.67\u0026ndash;0.96], I\u003csup\u003e2\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;87.4%, specificity: 0.98 [0.95\u0026ndash;0.99], I\u003csup\u003e2\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;93.9%), psoriasis (sensitivity: 0.90 [0.78\u0026ndash;0.95], I\u003csup\u003e2\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;92.8%, specificity: 0.98 [0.96\u0026ndash;0.99], I\u003csup\u003e2\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;92.1%), and scabies (sensitivity: 0.89 [0.84\u0026ndash;0.93], I\u003csup\u003e2\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;69.1%, specificity: 0.98 [0.95\u0026ndash;0.99], I\u003csup\u003e2\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;83.8%). All results were statistically significant (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05). There was no evidence of publication bias (Deeks\u0026rsquo;s funnel plot asymmetry test, p\u0026thinsp;\u0026gt;\u0026thinsp;0.05). Forest plots, SROC graphs and Deeks\u0026rsquo;s funnel plots are presented in Figure \u003cspan class=\"InternalRef\"\u003eS1\u003c/span\u003e.\u003c/p\u003e\n \u003cp\u003eWe identified the high heterogeneity in the meta-analysis for the above conditions (I\u003csup\u003e2\u003c/sup\u003e\u0026thinsp;\u0026gt;\u0026thinsp;50.0%). We conducted subgroup analyses and meta-regression for the heterogeneity only for mpox studies as there were limited studies and insufficient data to perform subgroup analysis in other conditions (Table\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e). Meta-regression revealed higher pooled sensitivity in models with larger datasets (\u0026ge;\u0026thinsp;1000 images) and binary classification approaches compared to those with smaller datasets and multiclass predictions (p\u0026thinsp;=\u0026thinsp;0.049 and p\u0026thinsp;=\u0026thinsp;0.028, respectively).\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec17\" class=\"Section2\"\u003e\n \u003ch2\u003eResearch gaps, limitations and recommendations\u003c/h2\u003e\n \u003cp\u003eBased on our systematic review findings, we summarised key research gaps and limitations in existing studies and formulated recommendations for future research, as presented in Table \u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003e.\u003c/p\u003e\n \u003cdiv class=\"gridtable\"\u003e\u0026nbsp;\u003ctable id=\"Tab2\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eSummary estimate of pooled performance of AI models for mpox\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003ccolgroup cols=\"10\"\u003e\u003c/colgroup\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\u0026nbsp;\u003c/th\u003e\n \u003cth align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003eNumber of models\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colspan=\"4\"\u003e\n \u003cp\u003eModel F (4,29)\u0026thinsp;=\u0026thinsp;2.53 (p-value: 0.062)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colspan=\"4\"\u003e\n \u003cp\u003eModel F (4, 29)\u0026thinsp;=\u0026thinsp;0.70 (p-value: 0.598)\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\u0026nbsp;\u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eSensitivity\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eP-value*\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eI\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e (95% CI)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eP-value**\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eSpecificity\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eP-value*\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eI\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e (95% CI)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eP-value**\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eOverall\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e34\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.96 (0.93\u0026ndash;0.97)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e90.98 (88.74\u0026ndash;93.22)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.98 (0.97\u0026ndash;0.99)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e99.90 (99.90-99.91)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colspan=\"10\"\u003e\n \u003cp\u003e\u003cstrong\u003e# of lesion types included\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eLess than 5\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e26\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.96 (0.93\u0026ndash;0.98)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e89.86 (86.87\u0026ndash;92.85)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.130\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.98 (0.97\u0026ndash;0.99)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e92.84 (90.93\u0026ndash;94.74)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.955\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003e5 and above\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.93 (0.84\u0026ndash;0.97)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e93.47 (90.31\u0026ndash;96.64)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.97 (0.95\u0026ndash;0.99)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e99.95 (99.95\u0026ndash;99.96)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colspan=\"10\"\u003e\n \u003cp\u003e\u003cstrong\u003eSample size\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eLess than 1000\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e23\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.95 (0.91\u0026ndash;0.97)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e88.49 (84.74\u0026ndash;92.24)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.049\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.98 (0.97\u0026ndash;0.99)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e91.29 (88.67\u0026ndash;93.91)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.532\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003e100 and above\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e11\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.97 (0.92\u0026ndash;0.99)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e94.07 (91.73\u0026ndash;96.42)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.97 (0.95\u0026ndash;0.99)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e99.96 (99.96\u0026ndash;99.97)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colspan=\"10\"\u003e\n \u003cp\u003e\u003cstrong\u003eModel classification\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eBinary\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e17\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.97 (0.95\u0026ndash;0.99)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e87.87 (83.16\u0026ndash;92.57)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.028\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.98 (0.95\u0026ndash;0.99)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e92.43 (89.87\u0026ndash;94.98)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.197\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eMulticlass\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e17\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.93 (0.88\u0026ndash;0.96)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e89.02 (84.89\u0026ndash;93.15)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.98 (0.97\u0026ndash;0.99)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e99.95 (99.95\u0026ndash;99.95)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colspan=\"10\"\u003e\n \u003cp\u003e\u003cstrong\u003eAI Algorithm\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eCNN-based\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e26\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.95 (0.92\u0026ndash;0.97)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e91.31 (88.86\u0026ndash;93.76)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.537\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.98 (0.96\u0026ndash;0.99)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e99.91 (99.91\u0026ndash;99.92)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.396\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eOther\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.95 (0.89\u0026ndash;0.98)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e88.42 (81.81\u0026ndash;95.04)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.98 (0.97\u0026ndash;0.99)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.05\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e50.25 (10.19\u0026ndash;90.31)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec18\" class=\"Section2\"\u003e\n \u003cp\u003e** P-value for heterogeneity between subgroups with meta-regression analysis\u003c/p\u003e\n \u003cdiv id=\"Sec19\" class=\"Section3\"\u003e\n \u003cp\u003e*P-value for heterogeneity within each subgroup\u003c/p\u003e\n \u003cdiv id=\"Sec20\" class=\"Section4\"\u003e\n \u003cp\u003eCI: Confidence Interval; I\u0026sup2;: Higgins\u0026apos; I-squared statistic (a measure of heterogeneity)\u003c/p\u003e\n \u003cdiv class=\"gridtable\"\u003e\n \u003cdiv align=\"left\" class=\"colspec\"\u003e\u003cbr\u003e\u003c/div\u003e\u0026nbsp;\u003ctable id=\"Tab3\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eSummary of research gaps and future recommendations for AI-based identification of STIs and anogenital dermatoses\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003ccolgroup cols=\"3\"\u003e\u003c/colgroup\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\" colspan=\"2\"\u003e\n \u003cp\u003eResearch gaps/limitations\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eRecommendations\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003cth align=\"left\" colspan=\"3\"\u003e\n \u003cp\u003e1. Disease conditions\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003e1.1\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eImbalanced research focus\u003c/strong\u003e: Studies predominantly focused on mpox (62.4%) due to the recent outbreak, while common STIs and anogenital conditions received limited attention (\u0026lt;\u0026thinsp;6.0% each).\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eBalanced research agenda\u003c/strong\u003e: Prioritise AI research into common STIs and anogenital conditions, while continuing to investigate emerging outbreaks like mpox, with an emphasis on WHO-priority infections such as syphilis, genital herpes, and genital warts.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003e1.2\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eLimited differential coverage\u003c/strong\u003e: Studies lacked clinically relevant comparative conditions and comprehensive coverage of anogenital conditions as seen in sexual health practice.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eDiverse and representative data\u003c/strong\u003e: Include a wide variety of anogenital reference conditions covering STIs, non-STIs, tumours, inflammatory diseases, and normal anatomical variants for robust differential diagnosis.\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e6\u003c/span\u003e,\u003cspan class=\"CitationRef\"\u003e39\u003c/span\u003e\u0026ndash;\u003cspan class=\"CitationRef\"\u003e41\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003e1.3\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003ePoor disease standardisation\u003c/strong\u003e: Lack of standardised disease definition undermined clinical relevance and hindered reproducibility (e.g. herpes simplex versus herpes zoster; genital warts versus flat warts).\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eAdoption of International Classification (ICD) Systems\u003c/strong\u003e: Use standardised ICD codes for labelling and disease categorisation in datasets.\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e19\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colspan=\"3\"\u003e\n \u003cp\u003e\u003cstrong\u003e2. Data\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003e2.1\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eData scarcity challenges\u003c/strong\u003e: Models relied heavily on open-source datasets (87.8%), with small sample sizes, limited representation of target conditions, and absence of patient metadata.\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e42\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eCoordinated data infrastructure\u003c/strong\u003e: Establish networks for standardised image collection and develop a centralised repository (similar to the IARC Cervical Cancer Image Bank)\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e43\u003c/span\u003e\u003c/sup\u003e for STIs and anogenital conditions.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003e\u003cstrong\u003e2.2\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003e\u003cstrong\u003eData quality concern\u003c/strong\u003e: Duplicate images across public datasets compromised data quality.\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e44\u003c/span\u003e,\u003cspan class=\"CitationRef\"\u003e45\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eImplement de-duplication processes\u003c/strong\u003e: Use systematic methods or tools (e.g., DupeGuru software, difPy python package) to identify and remove duplicate images to ensure data quality.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eTransparent reporting\u003c/strong\u003e: Document and transparently report all methods for image quality control.\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e46\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003e\u003cstrong\u003e2.3\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003e\u003cstrong\u003eInadequate technical documentation\u003c/strong\u003e: Most studies (89.4%) lacked essential details regarding image acquisition methods and quality standards.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eTechnical standardisation\u003c/strong\u003e: Adopt and adapt established technical guidelines to ensure consistent quality in clinical image acquisition.\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e47\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eDocument and report for reproducibility\u003c/strong\u003e: Provide detailed documentation of image acquisition methods, quality standards, and technical specifications.\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e47\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003e2.4\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eInsufficient diagnostic validation\u003c/strong\u003e: Limited validation of image diagnoses through laboratory tests or clinical reviews undermined data reliability.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eDiagnostic validation standards\u003c/strong\u003e: Require validation of image diagnoses through laboratory confirmation and/or expert clinical review before inclusion in datasets.\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e6\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colspan=\"3\"\u003e\n \u003cp\u003e\u003cstrong\u003e3. Model development\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003e\u003cstrong\u003e3.1\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003e\u003cstrong\u003eUnclear clinical alignment\u003c/strong\u003e: Model development lacked clear alignment with intended clinical applications.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003ePurpose-driven development\u003c/strong\u003e: Define clear clinical objectives (rule-in/rule-out) and model approach (binary/multiclass) based on intended clinical use (screening/triage/assisted-diagnosis).\u003csup\u003e48\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eInterdisciplinary collaboration\u003c/strong\u003e: Ensure development teams include data scientists, AI experts, sexual health physicians, and end users.\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e48\u003c/span\u003e,\u003cspan class=\"CitationRef\"\u003e49\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003e\u003cstrong\u003e3.2\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003e\u003cstrong\u003eInadequate methodology reporting\u003c/strong\u003e: Insufficient documentation of data splitting and cross-validation approaches, including crucial stratification methods with patient metadata (sex, lesion site, skin tone, etc.) and diagnoses.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eData splitting integrity\u003c/strong\u003e: Prevent data leakage by ensuring images from the same patient remain in the same dataset split.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eStratified validation protocols\u003c/strong\u003e: Implement stratified data splitting and cross-validation based on patient characteristics and diagnoses.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003e3.3\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eUnderutilised multimodal approach\u003c/strong\u003e: Most models used only image data, with minimal integration of patient metadata (1.4%) despite the potential benefits of multimodal approach.\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e26\u003c/span\u003e,\u003cspan class=\"CitationRef\"\u003e30\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eMultimodal integration\u003c/strong\u003e: Develop models that combine clinical images with relevant patient metadata to improve diagnostic accuracy.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colspan=\"3\"\u003e\n \u003cp\u003e\u003cstrong\u003e4. Model evaluation\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003e\u003cstrong\u003e4.1\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003e\u003cstrong\u003eLimited generalisability\u003c/strong\u003e: Model generalisability was limited by a lack of external validation and clinical trials, with most studies being single-centre evaluations.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eModel availability\u003c/strong\u003e: Encourage that trained models are publicly accessible (e.g., through GitHub) to facilitate external validation.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eMulti-centre validation\u003c/strong\u003e: Establish collaborative networks for prospective model testing across various clinical settings.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003e\u003cstrong\u003e4.2\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003e\u003cstrong\u003eLimited subgroup assessment\u003c/strong\u003e: Minimal evaluation across gender, skin tones, lesion sites, age, and ethnicity.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eGender-specific models\u003c/strong\u003e: Develop dedicated models specifically for females to address unique anatomical features and disease presentations.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eDiverse subgroup analysis\u003c/strong\u003e: Evaluate model performance across different demographic groups, anatomical sites, and clinical presentations.\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e50\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003e4.3\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eUnexplained model decisions\u003c/strong\u003e: Limited reporting of model interpretability techniques (17%), leaving model decision processes unclear.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eModel transparency with visualisation\u003c/strong\u003e: Implement modern interpretability techniques (e.g., Grad-CAM, LIME, SHAP) to explain the model\u0026rsquo;s decision-making processes.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colspan=\"3\"\u003e\n \u003cp\u003e\u003cstrong\u003e5. Model performance\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003e5.1\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eInconsistent performance reporting\u003c/strong\u003e: Inconsistent reporting performance metrics hindered model comparison.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eStandardised performance reporting\u003c/strong\u003e: Document key performance metrics (AUC-ROC, sensitivity, specificity, PPV, F1-scores) and contingency tables on test datasets to enable meaningful comparisons.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003e5.2\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eUnclear threshold criteria\u003c/strong\u003e: Binary classification studies lacked specified threshold selection criteria for sensitivity/specificity trade-offs based on intended use.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003ePerformance trade-off reporting\u003c/strong\u003e: Specify how sensitivity/specificity thresholds were optimised for the intended clinical application.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003e5.3\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eVariable model performance\u003c/strong\u003e: Performance varied widely due to differences in model architectures, purposes, and reference condition selections.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eFacilitate evidence synthesis\u003c/strong\u003e: Conduct more studies on prioritised diseases using standardised reporting framework to generate robust evidence, enabling meta-analysis to derive stronger conclusions on model accuracy and clinical utility.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" colspan=\"3\"\u003e\n \u003cp\u003e\u003cstrong\u003e6. Application\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003e\u003cstrong\u003e6.1\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003e\u003cstrong\u003eLimited translation to practice\u003c/strong\u003e: Studies remained at the proof-of-concept stage without defined clinical purpose (triage/screening) or target users (clinicians/public).\u003csup\u003e51\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eBridge Research and Practice\u003c/strong\u003e: Encourage collaborations between researchers, healthcare providers, and policymakers to translate proof-of-concept studies into clinically viable solutions.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eDevelop implementation framework\u003c/strong\u003e: Establish frameworks to guide from the development stage to integration of AI models into existing healthcare workflows with a multidisciplinary team.\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e48\u003c/span\u003e,\u003cspan class=\"CitationRef\"\u003e52\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003e6.2\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eMinimal clinical implementation\u003c/strong\u003e: only one public-facing application was tested prospectively, with no randomised clinical trials conducted.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eTesting Beyond Conceptualisation\u003c/strong\u003e: Advance AI models from the conceptual stage by conducting pilot testing and clinical trials to generate robust evidence on their safety, efficacy, and clinical utility.\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e53\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003e6.3\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eLimited study on user\u0026apos;s perspective\u003c/strong\u003e: Limited evidence on the needs, expectations, and experience of end-users, such as clinicians and patients, in adopting AI models for implementation.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eIncorporate Usability Testing\u003c/strong\u003e: Conduct feasibility, acceptability and usability studies to ensure models meet the practical requirements of their target users in clinical or public health settings.\u003csup\u003e\u003cspan class=\"CitationRef\"\u003e54\u003c/span\u003e\u0026ndash;\u003cspan class=\"CitationRef\"\u003e56\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n \u003c/div\u003e\n \u003c/div\u003e\n\u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003e Our systematic review identified 141 eligible studies that used AI algorithms to detect or classify STIs and anogenital dermatoses across six databases. Anogenital conditions like syphilis, genital herpes and other common differential dermatoses in sexual health have received limited attention. Meta-analysis demonstrated promising performance metrics, with pooled sensitivity above 87.0% and pooled specificity above 97% for conditions including mpox, herpes simplex, genital warts, psoriasis, and scabies. However, the high heterogeneity across studies and the limited number of studies for conditions other than mpox suggest that these results should be interpreted with caution and may not be generalisable beyond the settings of the studies included. Quality assessment using the modified QUADAS-2 tool indicated high concerns for risk of bias and applicability across studies, while the CLEAR Derm checklist highlighted incomplete reporting of data characteristics, methodological techniques, and clinical validation.\u003csup\u003e\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e\u003c/sup\u003e This finding aligns with similar AI models in dermatology conditions.\u003csup\u003e\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e\u003c/sup\u003e Moreover, insufficient external validation and the lack of prospective testing present significant barriers to translating these AI tools from proof-of-concept to clinical practice. In this review, we discuss these research gaps and provide recommendations for future studies to enhance the clinical utility of AI in STIs and anogenital dermatoses identification management.\u003c/p\u003e \u003cp\u003eAlthough AI applications in dermatology have shown promising results, their development for STIs and anogenital dermatoses has remained limited.\u003csup\u003e\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e\u003c/sup\u003e Before 2022, few studies included anogenital dermatoses in their analyses, as either target or differential diagnoses. While the 2022 mpox outbreak sparked numerous research studies, other common STIs have received limited research attention, despite their increasing global prevalence in recent years. While AI holds promise for early identification of syphilis chancres, only two studies addressed this application with limited sample sizes until 2024.\u003csup\u003e\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e,\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e\u003c/sup\u003e The clinical relevance of current studies is further constrained by the comparative conditions they included. Most mpox studies, for instance, developed models to distinguish mpox from other conditions, such as chickenpox and measles, which are rarely seen in sexual health clinics. This limited coverage of relevant anogenital conditions hinders the practical utility of these models in clinical settings. The absence of normal anatomical variants (such as skin tags and Fordyce spots) in training datasets could lead models to misclassify these as pathological conditions like genital warts, potentially causing unnecessary concern. Therefore, future studies should prioritise clinically important STIs and incorporate a comprehensive range of relevant conditions in model development to enhance their clinical utility.\u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e,\u003cspan additionalcitationids=\"CR40\" citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e \u003cp\u003eData is essential for developing AI algorithms, and our review highlighted critical challenges regarding data scarcity, quality, and validity. The heavy reliance on limited open-source datasets, duplication of images, and, most importantly, the lack of clinical validation for image diagnoses raises concerns about the generalisability of reported model performance. Images of anogenital dermatoses are particularly scarce compared to other dermatological conditions.\u003csup\u003e\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e\u003c/sup\u003e Difficulty in extracting data from clinical notes, and privacy and anonymity concerns have resulted in most images lacking essential patient metadata, such as age and symptoms (pain, duration, etc.), which are crucial for clinical decision-making. Like the WHO's coordinated approach for the Cervical Cancer Image Bank,\u003csup\u003e\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e\u003c/sup\u003e establishing a centralised repository with a standardised image collection protocol for STIs and anogenital dermatoses could address the current challenges.\u003c/p\u003e \u003cp\u003eRegarding model development and evaluation, most studies focused on technical aspects of data processing and algorithm design, yet overlooked two important areas. First, studies rarely defined their target users (public or healthcare providers) and use-case scenarios (self-symptom checking or clinical diagnosis support). Clear alignment between model design and clinical application could be achieved by a multidisciplinary team approach (data scientists, AI experts, sexual health physicians, and end users) and following the OPTICA tool.\u003csup\u003e\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e\u003c/sup\u003e Second, the predominant use of open-source or single-centre data without external validation or prospective testing raises concerns about model generalizability. The limited evaluation across different demographic groups, particularly regarding gender, skin tones, and anatomical sites, suggests potential performance disparities across diverse populations.\u003csup\u003e\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e,\u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e\u003c/sup\u003e Future studies should prioritise external validation using diverse datasets, comprehensive demographic evaluation, and transparent reporting to improve the generalizability of AI models.\u003c/p\u003e \u003cp\u003eMost studies remained at the conceptual stage with limited translation into real-world applications. However, this technical focus should not preclude the exploration of end-user perspectives, which are crucial for successful implementation. A few studies explored the public\u0026rsquo;s acceptability, feasibility, and preferences for the tool when it becomes available. Ly \u003cem\u003eet al.\u003c/em\u003e found that nearly 40% of users were reluctant to share clinical images for AI-based healthcare tools, particularly genital images due to privacy concerns.\u003csup\u003e\u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e54\u003c/span\u003e\u003c/sup\u003e In contrast, Jakob \u003cem\u003eet al.\u003c/em\u003e reported high interest in STI-related apps among dermatovenereological outpatients.\u003csup\u003e\u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e57\u003c/span\u003e\u003c/sup\u003e Soe \u003cem\u003eet al.\u003c/em\u003e also found that sexual health clinic attendees were willing to use such apps and provide comprehensive information, including symptoms and sexual behaviours, along with anogenital lesion images, if the app was developed by a reputable organisation and demonstrated reliable accuracy.\u003csup\u003e\u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e55\u003c/span\u003e\u003c/sup\u003e Future studies should explore the feasibility, acceptability, and usability to ensure that models meet the practical requirements of end users in clinical or public health settings.\u003c/p\u003e \u003cp\u003e Our systematic review has a number of strengths. First, we conducted a comprehensive search across six major databases in consultation with a librarian to capture the full scope of AI applications in STIs and anogenital dermatoses. Second, we ensured the reliability of our findings through independent duplicate screening, data extraction, and quality assessment. Third, we conducted meta-analyses of AI performance for five conditions and explored sources of heterogeneity in mpox studies through meta-regression. Finally, we provided a structured framework of research gaps and practical recommendations for future studies and the use of AI in clinical settings.\u003c/p\u003e \u003cp\u003eOur study has limitations. First, despite our comprehensive search strategy, we may have missed relevant studies, particularly those published in non-indexed journals or after our search date. Our review focused on peer-reviewed literature and may have overlooked potentially relevant commercial AI applications for STIs and anogenital dermatoses. Second, we modified the QUADAS-2 tool for the risk of bias assessment, which is not specifically designed for AI studies in dermatology. The modification might be subjective on the assessment scores. Therefore, we also used the CLEAR Derm checklist to check the quality of the studies. Third, the high heterogeneity in study methodologies and reporting made it challenging to synthesise findings and interpret meta-analysis results confidently. Thus, the pooled performance of the AI models may not reflect the actual performance in a local clinical setting. Fourth, while we included studies based on target diagnoses relevant to sexual health, many of these AI models were developed using images from both anogenital and non-anogenital body sites. Our findings may not accurately reflect the performance of AI models specifically designed for STIs and anogenital dermatoses. Finally, the AI models in our review were limited to identifying visible dermatological presentations and could not identify other important STI presentations such as discharge, bleeding and genitourinary symptoms. These AI models would need to be integrated with other approaches to assess a wider coverage of symptoms and risk factors for STIs.\u003c/p\u003e"},{"header":"Conclusions","content":"\u003cp\u003eWhile AI shows potential promising performance for identifying STIs and anogenital dermatoses, significant research gaps exist. Future work should prioritise understudied STI and differential conditions, while improving data quality, conducting external validation, and validating in clinical settings. Clear policy guidance and standards are needed to determine how best to implement AI tools for diagnostic purposes and to provide clear performance criteria and frameworks for AI developers, healthcare providers, and clients.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch3\u003eAcknowledgement\u003c/h3\u003e\n\u003cp\u003eWe thank Monash University and Melbourne Sexual Health Centre for NS\u0026apos;s scholarship and Lorena Romero (LR), the librarian at Alfred Health\u0026apos;s Ian Potter Library, for her expert guidance on the literature search strategy.\u003c/p\u003e\n\u003ch3\u003eContributors\u003c/h3\u003e\n\u003cp\u003eNS, JJO and LZ conceptualised the study. NS and PL conducted the literature search. NS, IK and PL conducted screening and data extraction. NS conducted the statistical analysis and drafted the manuscript. JJO and PL verified the analyses. All authors contributed to interpreting the results, revised the manuscript, and approved it for publication submission. JJO supervised the study and served as the guarantor of the review. The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the World Health Organization.\u003c/p\u003e\n\u003ch3\u003eDeclaration of interests\u003c/h3\u003e\n\u003cp\u003eNone.\u003c/p\u003e\n\u003ch3\u003eFunding\u003c/h3\u003e\n\u003cp\u003eCKF is supported by the National Health and Medical Research Council (NHMRC) (GNT1172900). EPFC is supported by an NHMRC Leadership Investigator Grant (GNT2033299). JJO is supported by the NHMRC Emerging Leadership Investigator Grant (GNT1193955).\u0026nbsp;\u003c/p\u003e\n\u003ch3\u003eData availability\u003c/h3\u003e\n\u003cp\u003eData can be requested from the corresponding author.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eCenters for Disease Control and Prevention (CDC). Incidence, Prevalence, and Cost of Sexually Transmitted Infections in the United States. 2021. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.cdc.gov/nchhstp/newsroom/fact-sheets/std/STI-Incidence-Prevalence-Cost-Factsheet.html\u003c/span\u003e\u003cspan address=\"https://www.cdc.gov/nchhstp/newsroom/fact-sheets/std/STI-Incidence-Prevalence-Cost-Factsheet.html\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (accessed July 2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWorld Health Organization. Implementing the global health sector strategies on HIV, viral hepatitis and sexually transmitted infections, 2022\u0026ndash;2030, Report on progress and gaps, 2024.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLaurenson-Schafer H, Sklenovsk\u0026aacute; N, Hoxha A, et al. Description of the first global outbreak of mpox: an analysis of global surveillance data. Lancet Glob Health 2023; 11(7): e1012-e23.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFairley CK, Chow EPF, Hocking JS. Early presentation of symptomatic individuals is critical in controlling sexually transmissible infections. Sexual Health 2015; 12(3): 181.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCenters for Disease Control and Prevention (CDC). STI Treatment Guidelines. July 22, 2021 2021. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.cdc.gov/std/treatment-guidelines/congenital-syphilis.htm\u003c/span\u003e\u003cspan address=\"https://www.cdc.gov/std/treatment-guidelines/congenital-syphilis.htm\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOng JJ, Bourne C, Dean JA, et al. Australian sexually transmitted infection (STI) management guidelines for use in primary care 2022 update. Sex Health 2023; 20(1): 1\u0026ndash;8.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSoe NN, Towns JM, Latt PM, et al. Accuracy of symptom checker for the diagnosis of sexually transmitted infections using machine learning and Bayesian network algorithms. BMC Infect Dis 2024; 24(1): 1408.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHe J, Li J, Jiang S, et al. Application of machine learning algorithms in predicting HIV infection among men who have sex with men: Model development and validation. Front Public Health 2022; 10: 967681.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXu X, Yu Z, Ge Z, et al. Web-Based Risk Prediction Tool for an Individual's Risk of HIV and Sexually Transmitted Infections Using Machine Learning Algorithms: Development and External Validation Study. Journal of Medical Internet Research 2022; 24(8): e37850.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePhyu Mon L, Nyi Nyi S, Xianglong X, et al. Assessing disparity in the distribution of HIV and sexually transmitted infections in Australia: a retrospective cross-sectional study using Gini coefficients. BMJ Public Health 2023; 1(1): e000012.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNieuwenburg SA, Hoornenborg E, Davidovich U, de Vries HJC, Schim van der Loeff M. Developing a symptoms-based risk score for infectious syphilis among men who have sex with men. Sex Transm Infect 2023; 99(5): 324\u0026ndash;9.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBerry NA, Harvey JA, Pittelkow MR, Swanson DL, Yang YW. Online symptom checkers lack diagnostic accuracy for skin rashes. J Am Acad Dermatol 2023; 88(2): 487\u0026ndash;8.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSoe NN, Latt PM, Yu Z, et al. Clinical features-based machine learning models to separate sexually transmitted infections from other skin diagnoses. The Journal of infection 2024; 88(4): 106128.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVenkatesh KP, Raza MM, Nickel G, Wang S, Kvedar JC. Deep learning models across the range of skin disease. NPJ Digital Medicine 2024; 7: 1\u0026ndash;2.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePage MJ, McKenzie JE, Bossuyt PM, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ 2021; 372: n71.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSoe NN, Kusnandar II, Latt PM. Systematic review and meta-analysis of Artificial Intelligence for the diagnosis of anogenital skin lesions related to sexual health. 2024. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.crd.york.ac.uk/prospero/display_record.php?ID=CRD42024527420\u003c/span\u003e\u003cspan address=\"https://www.crd.york.ac.uk/prospero/display_record.php?ID=CRD42024527420\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCovidence systematic review software. Veritas Health Innovation, Melbourne, Australia.; 2024.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWhiting PF, Rutjes AWS, Westwood ME, et al. QUADAS-2: A Revised Tool for the Quality Assessment of Diagnostic Accuracy Studies. Annals of Internal Medicine 2011; 155(8): 529\u0026ndash;36.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDaneshjou R, Barata C, Betz-Stablein B, et al. Checklist for Evaluation of Image-Based Artificial Intelligence Reports in Dermatology: CLEAR Derm Consensus Guidelines From the International Skin Imaging Collaboration Artificial Intelligence Working Group. JAMA Dermatol 2022; 158(1): 90\u0026ndash;6.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStataCorp. 2024. Stata Statistical Software: Release 18. College Station, TX: StataCorp LLC.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGaffoor N, Soomro S. Skin Disease Detection and Classification Using ResNet-50 and Support Vector Machine: An Effective Approach for Dermatological Diagnosis. 2023; 2023. p. 140-5.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePangti R, Mathur J, Chouhan V, et al. A machine learning-based, decision support, mobile phone application for diagnosis of common dermatological diseases. Journal of the European Academy of Dermatology and Venereology 2021; 35(2): 536\u0026ndash;45.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShen Y, Li H, Sun C, et al. Optimising skin disease diagnosis: harnessing online community data with contrastive learning and clustering techniques. npj Digital Medicine 2024; 7(1): 28.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSoe NN, Yu Z, Latt PM, et al. Distinguishing monkeypox from common skin lesions using artificial intelligence in a sexual health clinic: a feasibility study. \u003cem\u003eJournal of Medical Internet Research\u003c/em\u003e 2023; 20(5): XVI-XVI.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZaar O, Larson A, Polesie S, et al. Evaluation of the diagnostic accuracy of an online artificial intelligence application for skin disease diagnosis. Acta Dermato-Venereologica 2020; 100(16): 1\u0026ndash;6.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSoe NN, Yu Z, Latt PM, et al. Evaluation of artificial intelligence-powered screening for sexually transmitted infections-related skin lesions using clinical images and metadata. BMC Med 2024; 22(1): 296.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAli, Shams Nafisa, Ahmed, et al. Monkeypox Skin Lesion Detection Using Deep Learning Models: A Preliminary Feasibility Study. \u003cem\u003earXiv preprint arXiv:220703342\u003c/em\u003e 2022.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003e2022 International Conference on Computers and Artificial Intelligence Technologies, CAIT 2022. 2022; 2022.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCampana MG, Colussi, M., Delmastro, F., Mascetti, S., \u0026amp; Pagani, E. Mpox Close Skin Images. Zenodo; 2023.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiu Y, Jain A, Eng C, et al. A deep learning system for differential diagnosis of skin diseases. Nature medicine 2020; 26(6): 900\u0026ndash;8.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOztel I, Oztel GY, Sahin VH. Deep Learning-Based Skin Diseases Classification using Smartphones. ADVANCED INTELLIGENT SYSTEMS 2023; 5(12).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eThieme AH, Zheng Y, Machiraju G, et al. A deep-learning algorithm to classify skin lesions from mpox virus infection. Nature medicine 2023; 29(3): 738\u0026ndash;47.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYotsu RR, Ding Z, Hamm J, Blanton RE. Deep learning for AI-based diagnosis of skin-related neglected tropical diseases: A pilot study. PLoS neglected tropical diseases 2023; 17(8): e0011230.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAllan-Blitz L-T, Ambepitiya S, Tirupathi R, Klausner JD. The Development and Performance of a Machine-Learning Based Mobile Platform for Visually Determining the Etiology of 5 Penile Diseases. \u003cem\u003eMayo Clinic Proceedings: Digital Health\u003c/em\u003e 2024; 2(2): 280-8.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCampana MG, Colussi M, Delmastro F, Mascetti S, Pagani E. A Transfer Learning and Explainable Solution to Detect mpox from Smartphones images. PERVASIVE AND MOBILE COMPUTING 2024; 98.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCasuayan De Goma J, Devaraj M. Recognising Common Skin Diseases in the Philippines Using Image Processing and Machine Learning Classification. 2020; 2020. p. 68\u0026ndash;72.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eThorat R, Gupta A. Transfer learning-enabled skin disease classification: the case of monkeypox detection. \u003cem\u003eMULTIMEDIA TOOLS AND APPLICATIONS\u003c/em\u003e 2024.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBack S, Lee S, Shin S, et al. Robust Skin Disease Classification by Distilling Deep Neural Network Ensemble for the Mobile Diagnosis of Herpes Zoster. IEEE ACCESS 2021; 9: 20156\u0026ndash;69.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShumway NK. Skin Diseases of the Genitals and Perineum. In: Soutor C, Hordinsky MK, eds. Clinical Dermatology: Diagnosis and Management of Common Disorders, 2e. New York, NY: McGraw-Hill Education; 2022.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDermNet. Genital skin problems. 2002. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://dermnetnz.org/topics/genital-skin-problems\u003c/span\u003e\u003cspan address=\"https://dermnetnz.org/topics/genital-skin-problems\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStamm AW, Kobashi KC, Stefanovic KB. Urologic Dermatology: a Review. Curr Urol Rep 2017; 18(8): 62.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShobajo M, Abdulkadir S, Kraus CN, Shiu J. Anogenital skin of color images are underrepresented in dermatology, gynecology, and urology educational textbooks. Int J Womens Dermatol 2023; 9(3): e090.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWorld Health Organization, International Agency for Research on Cancer. IARC Cervical Cancer Image Bank. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://screening.iarc.fr/cervicalimagebank.php\u003c/span\u003e\u003cspan address=\"https://screening.iarc.fr/cervicalimagebank.php\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAsif S, Zhao M, Li YF, Tang FX, Khan SUR, Zhu YS. AI-Based Approaches for the Diagnosis of Mpox: Challenges and Future Prospects. \u003cem\u003eARCHIVES OF COMPUTATIONAL METHODS IN ENGINEERING\u003c/em\u003e 2024.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVega C, Schneider R, Satagopam V. Analysis: Flawed Datasets of Monkeypox Skin Images. Journal of medical systems 2023; 47(1): 37.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCollins GS, Moons KGM, Dhiman P, et al. TRIPOD\u0026thinsp;+\u0026thinsp;AI statement: updated guidance for reporting clinical prediction models that use regression or machine learning methods. BMJ 2024; 385: e078378.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFinnane A, Curiel-Lewandrowski C, Wimberley G, et al. Proposed Technical Guidelines for the Acquisition of Clinical Images of Skin-Related Conditions. JAMA Dermatol 2017; 153(5): 453\u0026ndash;7.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDagan N, Devons-Sberro S, Paz Z, et al. Evaluation of AI Solutions in Health Care Organizations \u0026mdash; The OPTICA Tool. Nejm Ai 2024; 1(9).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWidner K, Virmani S, Krause J, et al. Lessons learned from translating AI from development to deployment in healthcare. Nat Med 2023; 29(6): 1304\u0026ndash;6.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLarrazabal AJ, Nieto N, Peterson V, Milone DH, Ferrante E. Gender imbalance in medical imaging datasets produces biased classifiers for computer-aided diagnosis. \u003cem\u003eProceedings of the National Academy of Sciences\u003c/em\u003e 2020; 117(23): 12592-4.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKelly CJ, Karthikesalingam A, Suleyman M, Corrado G, King D. Key challenges for delivering clinical impact with artificial intelligence. BMC Medicine 2019; 17(1).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCallahan A, McElfresh D, Banda JM, et al. Standing on FURM Ground: A Framework for Evaluating Fair, Useful, and Reliable AI Models in Health Care Systems. NEJM Catalyst 2024; 5(10).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiu X, Cruz Rivera S, Moher D, et al. Reporting guidelines for clinical trial reports for interventions involving artificial intelligence: the CONSORT-AI extension. Nat Med 2020; 26(9): 1364\u0026ndash;74.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLy S, Reyes-Hadsall S, Drake L, et al. Public Perceptions, Factors, and Incentives Influencing Patient Willingness to Share Clinical Images for Artificial Intelligence-Based Healthcare Tools. Dermatol Ther (Heidelb) 2023; 13(11): 2895\u0026ndash;902.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSoe NN, Latt PM, King A, et al. What Do People Want from an AI-Assisted Screening App for Sexually Transmitted Infection-Related Anogenital Lesions: A Discrete Choice Experiment. \u003cem\u003ePatient\u003c/em\u003e 2024.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAbraham E, Chow EP, Fairley CK, et al. eSexualHealth: Preferences to Use Technology to Promote Sexual Health Among Men Who Have Sex with Men and Trans and Gender Diverse People. Cold Spring Harbor Laboratory; 2022.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJakob L, Steeb T, Fiocco Z, et al. Patient Perception of Mobile Phone Apps for the Care and Prevention of Sexually Transmitted Diseases: Cross-Sectional Study. JMIR Mhealth Uhealth 2020; 8(11): e16517.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","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":"Sexually Transmitted Infections, Anogenital Dermatoses, Artificial Intelligence, Computer Vision","lastPublishedDoi":"10.21203/rs.3.rs-6002285/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6002285/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e \u003cp\u003eArtificial intelligence (AI) excels in dermatology. However, its applications to sexually transmitted infections (STIs) remain unclear. We assessed the performance of AI algorithms and their applications in detecting STIs and anogenital dermatoses in sexual health.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003e We followed the PRISMA guidelines and searched six databases from January 1, 2010, to April 12, 2024, for studies using AI to identify STIs and anogenital dermatoses. We used a modified QUADAS-2 tool and the CLEAR Derm checklist for quality assessment. We conducted a bivariate random-effect meta-analysis to estimate the pooled sensitivity and specificity of AI applications for the conditions where sufficient data existed. Subgroup analysis and meta-regression were conducted to explore potential heterogeneity sources for mpox studies.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eOf 5,381 studies screened, 141 met the inclusion criteria. Most studies reported on mpox (111, 62.4%), while anogenital conditions, including syphilis, genital herpes, genital warts, scabies, psoriasis, lichenoid changes, and molluscum contagiosum, received less attention (each \u0026lt;\u0026thinsp;6.0% of the studies). Meta-analyses showed high performance of AI for mpox identification (pooled sensitivity: 0.96 [95% CI: 0.93\u0026ndash;0.97], pooled specificity: 0.98 [0.97\u0026ndash;0.99]), herpes simplex (0.91 [0.71\u0026ndash;0.98], 0.97 [0.94\u0026ndash;0.98]), genital warts (0.87 [0.67\u0026ndash;0.96], 0.98 [0.95\u0026ndash;0.99]), psoriasis (0.90 [0.78\u0026ndash;0.95], 0.98 [0.96\u0026ndash;0.99]), and scabies (0.89 [0.84\u0026ndash;0.93], 0.98 [0.95\u0026ndash;0.99]). We could not pool the sensitivity and specificity for other conditions due to insufficient data points. Meta-regression for mpox studies revealed higher pooled sensitivity in models with larger datasets (\u0026ge;\u0026thinsp;1000 images) and binary classification approaches compared to those with smaller datasets and multiclass predictions (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05). Study quality was variable and our assessment identified high risk of bias across the population selection (76.1%), reference standards (76.1%), and index tests (20.0%). Most studies relied on open-source datasets (87.8%), lacked external validation, and remained at the proof-of-concept stage without clinical implementation.\u003c/p\u003e\u003ch2\u003eConclusions\u003c/h2\u003e \u003cp\u003eWhile AI shows potential promising performance for identifying STIs and anogenital dermatoses, significant research gaps exist. Future work should prioritise understudied STI and differential conditions, while improving data quality, conducting external validation, and validating in clinical settings. Clear policy guidance and standards are needed to determine how best to implement AI tools for diagnostic purposes and to provide clear performance criteria and frameworks for AI developers, healthcare providers, and clients.\u003c/p\u003e","manuscriptTitle":"Clinical Applications of AI in Sexually Transmitted Infection and Anogenital Dermatoses in Sexual Health: A Systematic Review and Meta-Analysis","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-04-15 11:50:48","doi":"10.21203/rs.3.rs-6002285/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":"f832bb13-972f-409a-89a9-e902b251bdd0","owner":[],"postedDate":"April 15th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":46328412,"name":"Health sciences/Diseases/Infectious diseases"},{"id":46328413,"name":"Health sciences/Signs and symptoms/Skin manifestations"},{"id":46328414,"name":"Biological sciences/Computational biology and bioinformatics/Machine learning"},{"id":46328415,"name":"Biological sciences/Computational biology and bioinformatics/Computational models"}],"tags":[],"updatedAt":"2025-04-15T11:50:51+00:00","versionOfRecord":[],"versionCreatedAt":"2025-04-15 11:50:48","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6002285","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6002285","identity":"rs-6002285","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}