Pub Date : 2024-08-26DOI: 10.1101/2024.08.25.24312563
Van Nguyen, Sreenidhi Iyengar, Haroon Rasheed, Galo Apolo, Zhiwei Li, Aniket Kumar, Hong Nguyen, Austin Bohner, Rahul Dhodapkar, Jiun Do, Andrew Duong, Jeffrey Gluckstein, Kendra Hong, Alanna James, Junhui Lee, Kent Nguyen, Brandon Wong, Jose-Luis Ambite, Carl Kesselman, Lauren Daskivich, Michael Pazzani, Benjamin Xu
Purpose: To develop and test a deep learning (DL) algorithm for detecting referable glaucoma in the Los Angeles County (LAC) Department of Health Services (DHS) teleretinal screening program. Methods: Fundus photographs and patient-level labels of referable glaucoma (defined as cup-to-disc ratio [CDR] <= 0.6) provided by 21 trained optometrist graders were obtained from the LAC DHS teleretinal screening program. A DL algorithm based on the VGG-19 architecture was trained using patient-level labels generalized to images from both eyes. Area under the receiver operating curve (AUC), sensitivity, and specificity were calculated to assess algorithm performance using an independent test set that was also graded by 13 clinicians with one to 15 years of experience. Algorithm performance was tested using reference labels provided by either LAC DHS optometrists or an expert panel of 3 glaucoma specialists. Results: 12,098 images from 5,616 patients (2,086 referable glaucoma, 3,530 non-glaucoma) were used to train the DL algorithm. In this dataset, mean age was 56.8 +/- 10.5 years with 54.8% females and 68.2% Latinos, 8.9% Blacks, 2.7% Caucasians, and 6.0% Asians. 1,000 images from 500 patients (250 referable glaucoma, 250 non-glaucoma) with similar demographics (p <= 0.57) were used to test the DL algorithm. Algorithm performance matched or exceeded that of all independent clinician graders in detecting patient-level referable glaucoma based on LAC DHS optometrist (AUC = 0.92) or expert panel (AUC = 0.93) reference labels. Clinician grader sensitivity (range: 0.33-0.99) and specificity (range: 0.68-0.98) ranged widely and did not correlate with years of experience (p <= 0.49). Algorithm performance (AUC = 0.93) also matched or exceeded the sensitivity (range: 0.78-1.00) and specificity (range: 0.32-0.87) of 6 LAC DHS optometrists in the subsets of the test dataset they graded based on expert panel reference labels. Conclusions: A DL algorithm for detecting referable glaucoma developed using patient-level data provided by trained LAC DHS optometrists approximates or exceeds performance by ophthalmologists and optometrists, who exhibit variable sensitivity and specificity unrelated to experience level. Implementation of this algorithm in screening workflows could help reallocate eye care resources and provide more reproducible and timely glaucoma care.
{"title":"Expert-Level Detection of Referable Glaucoma from Fundus Photographs in a Safety Net Population: The AI and Teleophthalmology in Los Angeles Initiative","authors":"Van Nguyen, Sreenidhi Iyengar, Haroon Rasheed, Galo Apolo, Zhiwei Li, Aniket Kumar, Hong Nguyen, Austin Bohner, Rahul Dhodapkar, Jiun Do, Andrew Duong, Jeffrey Gluckstein, Kendra Hong, Alanna James, Junhui Lee, Kent Nguyen, Brandon Wong, Jose-Luis Ambite, Carl Kesselman, Lauren Daskivich, Michael Pazzani, Benjamin Xu","doi":"10.1101/2024.08.25.24312563","DOIUrl":"https://doi.org/10.1101/2024.08.25.24312563","url":null,"abstract":"Purpose: To develop and test a deep learning (DL) algorithm for detecting referable glaucoma in the Los Angeles County (LAC) Department of Health Services (DHS) teleretinal screening program. Methods: Fundus photographs and patient-level labels of referable glaucoma (defined as cup-to-disc ratio [CDR] <= 0.6) provided by 21 trained optometrist graders were obtained from the LAC DHS teleretinal screening program. A DL algorithm based on the VGG-19 architecture was trained using patient-level labels generalized to images from both eyes. Area under the receiver operating curve (AUC), sensitivity, and specificity were calculated to assess algorithm performance using an independent test set that was also graded by 13 clinicians with one to 15 years of experience. Algorithm performance was tested using reference labels provided by either LAC DHS optometrists or an expert panel of 3 glaucoma specialists.\u0000Results: 12,098 images from 5,616 patients (2,086 referable glaucoma, 3,530 non-glaucoma) were used to train the DL algorithm. In this dataset, mean age was 56.8 +/- 10.5 years with 54.8% females and 68.2% Latinos, 8.9% Blacks, 2.7% Caucasians, and 6.0% Asians. 1,000 images from 500 patients (250 referable glaucoma, 250 non-glaucoma) with similar demographics (p <= 0.57) were used to test the DL algorithm. Algorithm performance matched or exceeded that of all independent clinician graders in detecting patient-level referable glaucoma based on LAC DHS optometrist (AUC = 0.92) or expert panel (AUC = 0.93) reference labels. Clinician grader sensitivity (range: 0.33-0.99) and specificity (range: 0.68-0.98) ranged widely and did not correlate with years of experience (p <= 0.49). Algorithm performance (AUC = 0.93) also matched or exceeded the sensitivity (range: 0.78-1.00) and specificity (range: 0.32-0.87) of 6 LAC DHS optometrists in the subsets of the test dataset they graded based on expert panel reference labels.\u0000Conclusions: A DL algorithm for detecting referable glaucoma developed using patient-level data provided by trained LAC DHS optometrists approximates or exceeds performance by ophthalmologists and optometrists, who exhibit variable sensitivity and specificity unrelated to experience level. Implementation of this algorithm in screening workflows could help reallocate eye care resources and provide more reproducible and timely glaucoma care.","PeriodicalId":501390,"journal":{"name":"medRxiv - Ophthalmology","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-08-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142217429","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Purpose: To train and evaluate a segmentation-free 3D convolutional neural network (3DCNN) model for estimating visual field (VF) from optical coherence tomography (OCT) images and to compare the residual variability of OCT-based estimated VF (OCT-VF) with that of Humphrey Field Analyzer (HFA) measurements in a diverse clinical population. Design: Retrospective cross-sectional study. Participants: 5,351 patients (9,564 eyes) who underwent macular OCT imaging and Humphrey Field Analyzer (HFA) tests (24-2 or 10-2 test patterns) at a university hospital from 2006 to 2023. The dataset included 47,653 paired OCT-VF data points, including various ocular conditions. Methods: We trained a segmentation-free 3DCNN model based on the EfficientNet3D-b0 architecture on a comprehensive OCT dataset to estimate VF. We evaluated the model's performance using Pearson's correlation coefficient and Bland‒Altman analysis. We assessed residual variability using a jackknife resampling approach and compared OCT-VF and HFA datasets using generalized estimating equations (GEE), adjusting the number of VF tests, follow-up duration, age, and clustering by eye and patient. Main Outcome Measures: Correlations between estimated and measured VF thresholds and mean deviations (MDs), and residual variability of OCT-VF and HFA. Results: We observed strong correlations between the estimated and measured VF parameters (Pearson's r: 24-2 thresholds 0.893, MD 0.932; 10-2 thresholds 0.902, MD 0.945; all p < 0.001). Bland‒Altman analysis showed good agreement between the estimated and measured MD, with a slight proportional bias. GEE analysis demonstrated significantly lower residual variability for OCT-VF than for HFA (24-2 thresholds: 1.10 vs. 2.48 dB; 10-2 thresholds: 1.20 vs. 2.48 dB; all p < 0.001, Bonferroni-corrected), with lower variability across all test points, severities, and ages, thus highlighting the robustness of the segmentation-free 3DCNN approach in a heterogeneous clinical sample. Conclusions: A segmentation-free 3DCNN model objectively estimated VF from OCT images with high accuracy and significantly lower residual variability than subjective HFA measurements in a heterogeneous clinical sample, including patients with glaucoma and individuals with other ocular diseases. The improved reliability, lower variability, and objective nature of OCT-VF highlight its value for enhancing VF assessment and monitoring of various ocular conditions, potentially facilitating earlier detection of progression and more efficient disease management.
目的：训练和评估一种无分割三维卷积神经网络（3DCNN）模型，用于从光学相干断层扫描（OCT）图像中估计视野（VF），并在不同临床人群中比较基于OCT估计的VF（OCT-VF）与汉弗莱视野分析仪（HFA）测量的残差：设计：回顾性横断面研究：5351名患者（9564只眼睛）于2006年至2023年期间在一家大学医院接受了黄斑OCT成像和汉弗莱视场分析仪（HFA）测试（24-2或10-2测试模式）。数据集包括 47,653 个成对的 OCT-VF 数据点，其中包括各种眼部状况：我们在综合 OCT 数据集上训练了一个基于 EfficientNet3D-b0 架构的无分割 3DCNN 模型，以估计 VF。我们使用皮尔逊相关系数和 Bland-Altman 分析评估了模型的性能。我们使用杰克刀重采样方法评估了残余变异性，并使用广义估计方程（GEE）比较了OCT-VF和HFA数据集，调整了VF测试次数、随访时间、年龄以及按眼睛和患者分组的情况：主要结果指标：估计VF阈值和测量VF阈值与平均偏差（MD）之间的相关性，以及OCT-VF和HFA的残差：我们观察到估计和测量的 VF 参数之间存在很强的相关性（Pearson's r：24-2 阈值 0.893，MD 0.932；10-2 阈值 0.902，MD 0.945；所有 p < 0.001）。Bland-Altman 分析表明，估计值与测量值之间的一致性良好，但存在轻微的比例偏差。GEE 分析表明，OCT-VF 的残差明显低于 HFA（24-2 阈值：1.10 vs. 2.48）：1.10 vs. 2.48 dB; 10-2 thresholds：1.20 vs. 2.48 dB; all p < 0.001, Bonferroni-corrected），所有测试点、严重程度和年龄的变异性都更低，从而突出了无分割 3DCNN 方法在异质性临床样本中的稳健性：结论：在包括青光眼患者和其他眼部疾病患者在内的异质性临床样本中，无分割 3DCNN 模型能从 OCT 图像中客观估计 VF，与主观 HFA 测量值相比，准确性高，残余变异性低得多。OCT-VF 可靠性更高、变异性更低，而且具有客观性，这凸显了它在加强 VF 评估和监测各种眼部疾病方面的价值，可能有助于更早地发现病情发展和更有效地进行疾病管理。
{"title":"OCT-based Visual Field Estimation Using Segmentation-free 3D CNN Shows Lower Variability than Subjective Standard Automated Perimetry","authors":"Makoto Koyama, Satoru Inoda, Yuta Ueno, Yoshikazu Ito, Tetsuro Oshika, Masaki Tanito","doi":"10.1101/2024.08.17.24312150","DOIUrl":"https://doi.org/10.1101/2024.08.17.24312150","url":null,"abstract":"Purpose: To train and evaluate a segmentation-free 3D convolutional neural network (3DCNN) model for estimating visual field (VF) from optical coherence tomography (OCT) images and to compare the residual variability of OCT-based estimated VF (OCT-VF) with that of Humphrey Field Analyzer (HFA) measurements in a diverse clinical population.\u0000Design: Retrospective cross-sectional study.\u0000Participants: 5,351 patients (9,564 eyes) who underwent macular OCT imaging and Humphrey Field Analyzer (HFA) tests (24-2 or 10-2 test patterns) at a university hospital from 2006 to 2023. The dataset included 47,653 paired OCT-VF data points, including various ocular conditions.\u0000Methods: We trained a segmentation-free 3DCNN model based on the EfficientNet3D-b0 architecture on a comprehensive OCT dataset to estimate VF. We evaluated the model's performance using Pearson's correlation coefficient and Bland‒Altman analysis. We assessed residual variability using a jackknife resampling approach and compared OCT-VF and HFA datasets using generalized estimating equations (GEE), adjusting the number of VF tests, follow-up duration, age, and clustering by eye and patient.\u0000Main Outcome Measures: Correlations between estimated and measured VF thresholds and mean deviations (MDs), and residual variability of OCT-VF and HFA.\u0000Results: We observed strong correlations between the estimated and measured VF parameters (Pearson's r: 24-2 thresholds 0.893, MD 0.932; 10-2 thresholds 0.902, MD 0.945; all p < 0.001). Bland‒Altman analysis showed good agreement between the estimated and measured MD, with a slight proportional bias. GEE analysis demonstrated significantly lower residual variability for OCT-VF than for HFA (24-2 thresholds: 1.10 vs. 2.48 dB; 10-2 thresholds: 1.20 vs. 2.48 dB; all p < 0.001, Bonferroni-corrected), with lower variability across all test points, severities, and ages, thus highlighting the robustness of the segmentation-free 3DCNN approach in a heterogeneous clinical sample.\u0000Conclusions: A segmentation-free 3DCNN model objectively estimated VF from OCT images with high accuracy and significantly lower residual variability than subjective HFA measurements in a heterogeneous clinical sample, including patients with glaucoma and individuals with other ocular diseases. The improved reliability, lower variability, and objective nature of OCT-VF highlight its value for enhancing VF assessment and monitoring of various ocular conditions, potentially facilitating earlier detection of progression and more efficient disease management.","PeriodicalId":501390,"journal":{"name":"medRxiv - Ophthalmology","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-08-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142217435","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-08-19DOI: 10.1101/2024.08.18.24312188
Darragh Liam Harkin, Julie-Anne Little, Sara J. McCullough
Background: Visual stress is a reading disorder characterised by perceptual distortions, asthenopia and headache whilst reading, alongside increased sensitivity to repeated striped patterns (‘pattern glare’), in the absence of underlying ocular pathology. Coloured filters including tinted spectacle lenses and coloured overlays/acetates have been reported to ameliorate visual stress symptoms. However, evidence on coloured spectacle lenses efficacy at managing symptoms of visual stress, particularly in adults, is lacking, with two recent systematic reviews advocating the need for large-scale randomised control trials. Methods: This is a double-masked randomised placebo-controlled trial. University students identified with symptoms of visual stress, through use of a reading symptom questionnaire and mid-spatial frequency pattern glare test, will be recruited. Sample size for power of 90% at 5% significance, accounting for 10% dropout will be 65. Participants will be randomly assigned experimental and control coloured spectacle lenses to wear for six weeks followed by a two week washout period, prior to wear of the alternate lenses for a further six weeks with a two week washout period. Participants will compare both sets of spectacle lenses in a ‘head-to-head’ comparison after the secondary washout period, prior to choosing the preferred lenses for voluntary future wear. Long-term adherence to the preferred lenses will be assessed three months post-comparison. Researchers and participants will be masked to spectacle lenses worn throughout the duration of the trial. Reading performance will be assessed with both sets of lenses at various time points within the trial. A range of reading tests, reading symptoms and pattern glare evaluation will be used to monitor change in reading performance and visual stress symptoms during the trial. Discussion: The study will evaluate the hypothesis that coloured spectacle lenses increase reading speed and reduce severity and frequency of reading symptoms in adults with visual stress. Trial Registration: The trial is registered at clinicaltrials.gov: NCT04318106
{"title":"Reading; through the eyes of a university student: A double-masked randomised placebo-controlled cross-over protocol investigating coloured spectacle lens efficacy in adults with visual stress.","authors":"Darragh Liam Harkin, Julie-Anne Little, Sara J. McCullough","doi":"10.1101/2024.08.18.24312188","DOIUrl":"https://doi.org/10.1101/2024.08.18.24312188","url":null,"abstract":"Background:\u0000Visual stress is a reading disorder characterised by perceptual distortions, asthenopia and headache whilst reading, alongside increased sensitivity to repeated striped patterns (‘pattern glare’), in the absence of underlying ocular pathology. Coloured filters including tinted spectacle lenses and coloured overlays/acetates have been reported to ameliorate visual stress symptoms. However, evidence on coloured spectacle lenses efficacy at managing symptoms of visual stress, particularly in adults, is lacking, with two recent systematic reviews advocating the need for large-scale randomised control trials. Methods:\u0000This is a double-masked randomised placebo-controlled trial. University students identified with symptoms of visual stress, through use of a reading symptom questionnaire and mid-spatial frequency pattern glare test, will be recruited. Sample size for power of 90% at 5% significance, accounting for 10% dropout will be 65. Participants will be randomly assigned experimental and control coloured spectacle lenses to wear for six weeks followed by a two week washout period, prior to wear of the alternate lenses for a further six weeks with a two week washout period. Participants will compare both sets of spectacle lenses in a ‘head-to-head’ comparison after the secondary washout period, prior to choosing the preferred lenses for voluntary future wear. Long-term adherence to the preferred lenses will be assessed three months post-comparison. Researchers and participants will be masked to spectacle lenses worn throughout the duration of the trial. Reading performance will be assessed with both sets of lenses at various time points within the trial. A range of reading tests, reading symptoms and pattern glare evaluation will be used to monitor change in reading performance and visual stress symptoms during the trial. Discussion:\u0000The study will evaluate the hypothesis that coloured spectacle lenses increase reading speed and reduce severity and frequency of reading symptoms in adults with visual stress. Trial Registration:\u0000The trial is registered at clinicaltrials.gov: NCT04318106","PeriodicalId":501390,"journal":{"name":"medRxiv - Ophthalmology","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-08-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142217436","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-08-16DOI: 10.1101/2024.08.15.24312074
Matthew T McLaughlin, Caleb P Ganansky, Ayman W Taher, Melissa A Trudrung, William Van De Car, Jonathan Le, Kyle D. Peterson, Kimberly E Stepien, Melanie A Schmitt
Inherited Retinal Degenerations (IRDs) are a group of diseases where genetic variants lead to retinal photoreceptor dysfunction and subsequent visual impairment. We aimed to compare the characteristics of pediatric versus adult IRD patients at the time of presentation at a tertiary care IRD clinic. A retrospective chart review of 538 patients with IRDs was conducted. Information obtained included age, diagnosis, presenting characteristics, demographics, distance from the clinic, and referring physician. This study found that high hyperopia, congenital syndactyly, high refractive error, high astigmatism, and a history of developmental delay were most predictive of pediatric presentation. In adults, we found reduced central vision, peripheral vision loss, color vision deficits, nyctalopia, flashes/floaters, cataracts, diabetes mellitus, obesity, cardiac conditions, and a family history of cataract were most predictive of presentation. There was a greater proportion of pediatric patients presenting from 20 or more miles away. Additionally, there was no significant difference between the type of healthcare provider referring adult or pediatric patients. This study identifies characteristics predictive of pediatric and adult presentation in IRD patients thus addressing current knowledge gaps. A better understanding of these characteristics may provide for quicker recognition, education of clinicians likely to encounter IRDs, and allow for earlier treatment.
{"title":"Characteristics and Predictors of Pediatric and Adult Patients with Inherited Retinal Degenerations at Time of Presentation: Tertiary Care Ophthalmology Clinic Data","authors":"Matthew T McLaughlin, Caleb P Ganansky, Ayman W Taher, Melissa A Trudrung, William Van De Car, Jonathan Le, Kyle D. Peterson, Kimberly E Stepien, Melanie A Schmitt","doi":"10.1101/2024.08.15.24312074","DOIUrl":"https://doi.org/10.1101/2024.08.15.24312074","url":null,"abstract":"Inherited Retinal Degenerations (IRDs) are a group of diseases where genetic variants lead to retinal photoreceptor dysfunction and subsequent visual impairment. We aimed to compare the characteristics of pediatric versus adult IRD patients at the time of presentation at a tertiary care IRD clinic. A retrospective chart review of 538 patients with IRDs was conducted. Information obtained included age, diagnosis, presenting characteristics, demographics, distance from the clinic, and referring physician. This study found that high hyperopia, congenital syndactyly, high refractive error, high astigmatism, and a history of developmental delay were most predictive of pediatric presentation. In adults, we found reduced central vision, peripheral vision loss, color vision deficits, nyctalopia, flashes/floaters, cataracts, diabetes mellitus, obesity, cardiac conditions, and a family history of cataract were most predictive of presentation. There was a greater proportion of pediatric patients presenting from 20 or more miles away. Additionally, there was no significant difference between the type of healthcare provider referring adult or pediatric patients. This study identifies characteristics predictive of pediatric and adult presentation in IRD patients thus addressing current knowledge gaps. A better understanding of these characteristics may provide for quicker recognition, education of clinicians likely to encounter IRDs, and allow for earlier treatment.","PeriodicalId":501390,"journal":{"name":"medRxiv - Ophthalmology","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-08-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142217440","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-08-16DOI: 10.1101/2024.08.13.24311912
Nandi Bao, Liang Jia
Background: Color vision deficiency (CVD) is a common congenital ophthalmic disease, and no effective therapeutic measures are currently available for this patient population. This study aimed to explore the efficacy of photobiomodulation (PBM) in CVD patients. Methods: A randomized controlled trial design was applied in this study, whereby 80 patients with red-green CVD were randomly divided into PBM treatment and control groups. Color vision was assessed after 1, 2, and 4 weeks of treatment. Statistical analysis of treatment was carried out using the Kruskal-Wallis test with Dunn's test for multiple comparisons using SPSS and R software. Findings: Compared to the control group, the color vision of CVD patients exhibited a significant improvement in the PBM treatment group for different parameters (Ps＜0.05). Interpretation: PBM therapy could significantly improve color vision in patients with congenital CVD, especially in patients with green CVD.
{"title":"Novel non-invasive physical photobiomodulation can treat congenital colour vision deficiency and enhance color vision recognition ability: a randomized, single blind, controlled clinical trial","authors":"Nandi Bao, Liang Jia","doi":"10.1101/2024.08.13.24311912","DOIUrl":"https://doi.org/10.1101/2024.08.13.24311912","url":null,"abstract":"Background: Color vision deficiency (CVD) is a common congenital ophthalmic disease, and no effective therapeutic measures are currently available for this patient population. This study aimed to explore the efficacy of photobiomodulation (PBM) in CVD patients. Methods: A randomized controlled trial design was applied in this study, whereby 80 patients with red-green CVD were randomly divided into PBM treatment and control groups. Color vision was assessed after 1, 2, and 4 weeks of treatment. Statistical analysis of treatment was carried out using the Kruskal-Wallis test with Dunn's test for multiple comparisons using SPSS and R software. Findings: Compared to the control group, the color vision of CVD patients exhibited a significant improvement in the PBM treatment group for different parameters (Ps＜0.05). Interpretation: PBM therapy could significantly improve color vision in patients with congenital CVD, especially in patients with green CVD.","PeriodicalId":501390,"journal":{"name":"medRxiv - Ophthalmology","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-08-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142217437","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Purpose: To determine the prevalence and causes of blindness and vision impairment among people 50 years and older in Nepal. Methods: We conducted seven provincial-level Rapid Assessment of Avoidable Blindness (RAAB) cross-sectional, population-based surveys between 2018-2021. Provincial prevalence estimates were weighted to give nationally representative estimates. Sampling, enumeration, and examination of the population 50 years and older were done at the province level following standard RAAB protocol. Results: Across seven surveys, we enrolled 33,228 individuals, of whom 32,565 were examined (response rate 98%). Females (n=17,935) made up 55% of the sample. The age-sex-province weighted national prevalence of blindness (better eye presenting visual acuity <3/60) was 1.1% (95% confidence interval [CI] 1.0-1.2%), and any vision impairment <6/12 was 20.7% (95% CI 19.9-21.5%). The prevalence of blindness and any vision impairment were both higher in women than men (1.3% [95% CI 1.1-1.5%] vs 0.9% [95% CI 0.7-1.0%]). Age-sex weighted blindness prevalence was highest in Lumbini Province (1.8% [95% CI 1.3-2.2%]) and lowest in Bagmati Province (0.7% [95% CI 0.4-0.9%]) and Sudurpaschim Province (0.7% [95% CI 0.4-0.9%]). Cataract (65.2%) was the leading cause of blindness in our sample, followed by corneal opacity (6.4%), glaucoma (5.8%) and age-related macular degeneration (5.3%). Other posterior segment diseases accounted for 8.4% of cases. Conclusion: The prevalence of blindness was higher among women than men and varied by province. The Lumbini and Madesh Provinces in the Terai (plains) region had higher prevalence of blindness than elsewhere. Cataract was the leading cause of blindness, severe vision impairment and moderate vision impairment while refractive error was the leading cause of mild vision impairment.
目的：确定尼泊尔 50 岁及以上人群中失明和视力损伤的患病率和原因：我们在 2018-2021 年间进行了七次省级可避免盲症快速评估（RAAB）横断面人口调查。对省级患病率估计值进行了加权，以得出具有全国代表性的估计值。对 50 岁及以上人口的抽样、点名和检查均按照 RAAB 标准协议在省一级进行：在七次调查中，我们共登记了 33,228 人，其中 32,565 人接受了检查（应答率为 98%）。女性（人数=17,935）占样本的 55%。经年龄-性别-省份加权后，全国失明（较佳眼睛呈现视力<3/60）患病率为 1.1%（95% 置信区间 [CI] 1.0-1.2%），任何视力损伤<6/12 的患病率为 20.7%（95% CI 19.9-21.5%）。女性的失明率和视力受损率均高于男性（1.3% [95% CI 1.1-1.5%] vs 0.9% [95% CI 0.7-1.0%]）。年龄-性别加权失明率在蓝毗尼省最高（1.8% [95% CI 1.3-2.2%]），在巴格马蒂省（0.7% [95% CI 0.4-0.9%]）和苏杜帕辛省（0.7% [95% CI 0.4-0.9%]）最低。在我们的样本中，白内障（65.2%）是导致失明的主要原因，其次是角膜混浊（6.4%）、青光眼（5.8%）和老年性黄斑变性（5.3%）。其他后节疾病占 8.4%：结论：女性失明率高于男性，各省的情况也不尽相同。特莱（平原）地区的蓝毗尼省和马德什省的失明率高于其他地区。白内障是导致失明、严重视力损伤和中度视力损伤的主要原因，而屈光不正则是导致轻度视力损伤的主要原因。
{"title":"Prevalence of blindness and vision impairment among people 50 years and older in Nepal: a national Rapid Assessment of Avoidable Blindness survey","authors":"Sailesh Mishra, Ranjan Shah, Parikshit Gogate, Yuddha Dhoj Sapkota, Reeta Gurung, Mohan Shrestha, Islay Mactaggart, Ian McCormick, Brish Bahadur Shahi, Rajiv Khandekar, Matthew Burton","doi":"10.1101/2024.08.06.24311588","DOIUrl":"https://doi.org/10.1101/2024.08.06.24311588","url":null,"abstract":"Purpose: To determine the prevalence and causes of blindness and vision impairment among people 50 years and older in Nepal.\u0000Methods: We conducted seven provincial-level Rapid Assessment of Avoidable Blindness (RAAB) cross-sectional, population-based surveys between 2018-2021. Provincial prevalence estimates were weighted to give nationally representative estimates. Sampling, enumeration, and examination of the population 50 years and older were done at the province level following standard RAAB protocol.\u0000Results: Across seven surveys, we enrolled 33,228 individuals, of whom 32,565 were examined (response rate 98%). Females (n=17,935) made up 55% of the sample. The age-sex-province weighted national prevalence of blindness (better eye presenting visual acuity <3/60) was 1.1% (95% confidence interval [CI] 1.0-1.2%), and any vision impairment <6/12 was 20.7% (95% CI 19.9-21.5%). The prevalence of blindness and any vision impairment were both higher in women than men (1.3% [95% CI 1.1-1.5%] vs 0.9% [95% CI 0.7-1.0%]). Age-sex weighted blindness prevalence was highest in Lumbini Province (1.8% [95% CI 1.3-2.2%]) and lowest in Bagmati Province (0.7% [95% CI 0.4-0.9%]) and Sudurpaschim Province (0.7% [95% CI 0.4-0.9%]). Cataract (65.2%) was the leading cause of blindness in our sample, followed by corneal opacity (6.4%), glaucoma (5.8%) and age-related macular degeneration (5.3%). Other posterior segment diseases accounted for 8.4% of cases.\u0000Conclusion: The prevalence of blindness was higher among women than men and varied by province. The Lumbini and Madesh Provinces in the Terai (plains) region had higher prevalence of blindness than elsewhere. Cataract was the leading cause of blindness, severe vision impairment and moderate vision impairment while refractive error was the leading cause of mild vision impairment.","PeriodicalId":501390,"journal":{"name":"medRxiv - Ophthalmology","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-08-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141942029","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-08-07DOI: 10.1101/2024.08.07.24311607
Gunjan Naik, Saoud Al-Khuzaei, Ismail Moghul, Thales A. C. de Guimaraes, Sagnik Sen, Malena Daich Varela, Yichen Liu, Pallavi Bagga, Dun Jack Fu, Mariya Moosajee, Savita Madhusudhan, Andrew Webster, Samantha De Silva, Praveen J. Patel, Omar Mahroo, Susan M Downes, Michel Michaelides, Konstantinos Balaskas, Nikolas Pontikos, William Woof
Purpose: To develop an automated system for assessing the quality of Fundus Autofluorescence (FAF) images in patients with inherited retinal diseases (IRD). Methods: We annotated a dataset of 2445 FAF images from patients with Inherited Retinal Dystrophies which were assessed by three different expert graders. Graders marked images as either gradable (acceptable quality) or ungradable (poor quality), following a strict grading protocol. This dataset was used to train a Convolutional Neural Network (CNN) classification model to predict the gradability label of FAF images. Results: Retinograd-AI achieves a performance of 91% accuracy on our held-out dataset of 133 images with an Area Under the Receiver Operator Characteristic (AUROC) of 0.94, indicating high performance in distinguishing between gradable and ungradable images. Applying Retinograd-AI to our full internal dataset, the highest proportion of gradable images was found in the 30-50 years age group, where 84.3% of images were rated as gradable, while the lowest was in 0-15 year olds, where only 45.2% of images were rated as gradable. 83.4% of images from male patients were rated as gradable, and 90.6% of images from female patients. By genotype, from the 30 most common genetic diagnoses, the highest proportion of gradable images was in patients with disease causing variants in PRPH2 (93.9%), while the lowest was RDH12 (28.6%). Eye2Gene single-image gene classification top-5 accuracy on images rated by Retinograd-AI was 69.2%, while top-5 accuracy on images rated as ungradable was 39.0%. Conclusions: Retinograd-AI is the first open-source AI model for automated retinal image quality assessment of FAF images in IRDs. Automated gradability assessment through Retinograd AI enables large scale analysis of retinal images, which is an essential part of developing good analysis pipelines, and real-time quality assessment, which is essential for deployment of AI algorithms, such as Eye2Gene, into clinical settings. Due to the diverse nature of IRD pathologies, Retinograd-AI may also be applicable to FAF imaging for other conditions, either in its current form or through transfer learning and fine-tuning. Retinograd-AI is open-sourced, and the source code and network weights are available under an MIT licence on GitHub at https://github.com/eye2gene/retinograd-ai.
{"title":"Retinograd-AI: An Open-source Automated Fundus Autofluorescence Retinal Image Gradability Assessment for Inherited Retinal Dystrophies","authors":"Gunjan Naik, Saoud Al-Khuzaei, Ismail Moghul, Thales A. C. de Guimaraes, Sagnik Sen, Malena Daich Varela, Yichen Liu, Pallavi Bagga, Dun Jack Fu, Mariya Moosajee, Savita Madhusudhan, Andrew Webster, Samantha De Silva, Praveen J. Patel, Omar Mahroo, Susan M Downes, Michel Michaelides, Konstantinos Balaskas, Nikolas Pontikos, William Woof","doi":"10.1101/2024.08.07.24311607","DOIUrl":"https://doi.org/10.1101/2024.08.07.24311607","url":null,"abstract":"Purpose:\u0000To develop an automated system for assessing the quality of Fundus Autofluorescence (FAF) images in patients with inherited retinal diseases (IRD). Methods: We annotated a dataset of 2445 FAF images from patients with Inherited Retinal Dystrophies which were assessed by three different expert graders. Graders marked images as either gradable (acceptable quality) or ungradable (poor quality), following a strict grading protocol. This dataset was used to train a Convolutional Neural Network (CNN) classification model to predict the gradability label of FAF images. Results:\u0000Retinograd-AI achieves a performance of 91% accuracy on our held-out dataset of 133 images with an Area Under the Receiver Operator Characteristic (AUROC) of 0.94, indicating high performance in distinguishing between gradable and ungradable images. Applying Retinograd-AI to our full internal dataset, the highest proportion of gradable images was found in the 30-50 years age group, where 84.3% of images were rated as gradable, while the lowest was in 0-15 year olds, where only 45.2% of images were rated as gradable. 83.4% of images from male patients were rated as gradable, and 90.6% of images from female patients. By genotype, from the 30 most common genetic diagnoses, the highest proportion of gradable images was in patients with disease causing variants in PRPH2 (93.9%), while the lowest was RDH12 (28.6%). Eye2Gene single-image gene classification top-5 accuracy on images rated by Retinograd-AI was 69.2%, while top-5 accuracy on images rated as ungradable was 39.0%. Conclusions:\u0000Retinograd-AI is the first open-source AI model for automated retinal image quality assessment of FAF images in IRDs. Automated gradability assessment through Retinograd AI enables large scale analysis of retinal images, which is an essential part of developing good analysis pipelines, and real-time quality assessment, which is essential for deployment of AI algorithms, such as Eye2Gene, into clinical settings. Due to the diverse nature of IRD pathologies, Retinograd-AI may also be applicable to FAF imaging for other conditions, either in its current form or through transfer learning and fine-tuning. Retinograd-AI is open-sourced, and the source code and network weights are available under an MIT licence on GitHub at https://github.com/eye2gene/retinograd-ai.","PeriodicalId":501390,"journal":{"name":"medRxiv - Ophthalmology","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-08-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141942032","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-08-06DOI: 10.1101/2024.08.05.24311503
Alex S Baldwin, Marie-Celine Lorenzini, Annabel Wing-Yan Fan, Robert F Hess, Alexandre Reynaud
In binocular vision, the relative strength of the input from the two eyes can have significant functional impact. These inputs are typically balanced, however in some conditions (e.g. amblyopia) one eye will dominate over the other. To quantify imbalances in binocular vision, we have developed the Dichoptic Contrast Ordering Test (DiCOT). Implemented on a tablet device, the program uses rankings of perceived contrast (of dichoptically-presented stimuli) to find a scaling factor that balances the two eyes. We measured how physical interventions (applied to one eye) affect the DiCOT measurement. These were: i) neutral density filters, ii) Bangerter filters, and iii) optical blur introduced by a +3 D lens. The DiCOT results were compared to those from the Dichoptic Letter Test (DLT). Both the DiCOT and the DLT showed excellent test-retest reliability, however the magnitude of the imbalances introduced by the interventions was greater in the DLT. To find consistency between the methods, rescaling the DiCOT results from individual conditions gave good results. However, the adjustments needed for the +3 D lens condition were quite different from those for the ND and Bangerter filter. Our results indicate that the DiCOT and DLT measure partially separate aspects of binocular imbalance. This supports the simultaneous use of both measures in future studies.
{"title":"The Dichoptic Contrast Ordering Test: A method for measuring the depth of binocular imbalance","authors":"Alex S Baldwin, Marie-Celine Lorenzini, Annabel Wing-Yan Fan, Robert F Hess, Alexandre Reynaud","doi":"10.1101/2024.08.05.24311503","DOIUrl":"https://doi.org/10.1101/2024.08.05.24311503","url":null,"abstract":"In binocular vision, the relative strength of the input from the two eyes can have significant functional impact. These inputs are typically balanced, however in some conditions (e.g. amblyopia) one eye will dominate over the other. To quantify imbalances in binocular vision, we have developed the Dichoptic Contrast Ordering Test (DiCOT). Implemented on a tablet device, the program uses rankings of perceived contrast (of dichoptically-presented stimuli) to find a scaling factor that balances the two eyes. We measured how physical interventions (applied to one eye) affect the DiCOT measurement. These were: i) neutral density filters, ii) Bangerter filters, and iii) optical blur introduced by a +3 D lens. The DiCOT results were compared to those from the Dichoptic Letter Test (DLT). Both the DiCOT and the DLT showed excellent test-retest reliability, however the magnitude of the imbalances introduced by the interventions was greater in the DLT. To find consistency between the methods, rescaling the DiCOT results from individual conditions gave good results. However, the adjustments needed for the +3 D lens condition were quite different from those for the ND and Bangerter filter. Our results indicate that the DiCOT and DLT measure partially separate aspects of binocular imbalance. This supports the simultaneous use of both measures in future studies.","PeriodicalId":501390,"journal":{"name":"medRxiv - Ophthalmology","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-08-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141942168","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-08-05DOI: 10.1101/2024.08.03.24311451
Benton Chuter, Alex C Lieu, Justin Huynh, Jennifer J Bu, Linda M Zangwill
Purpose: To investigate the relationship between public demand for pediatric ophthalmology services and the accessibility of such services on a statewide basis in the United States, focusing on strabismus care. Methods: Using Google Trends data, search volumes for "strabismus" were analyzed from January 2014 to December 2023. Pediatric ophthalmologist availability was assessed via the American Academy of Ophthalmology directory, normalized for state population size to create a relative demand index. Additional metrics from the United States Census Bureau and Centers for Disease Control and Prevention provided socioeconomic and health data. Correlation coefficients were used to explore relationships between search volumes, specialist availability, vision screening, socioeconomics, and demographic factors. Results: The data revealed variability in public interest in pediatric ophthalmology across states, with Alaska showing the highest relative search volume for strabismus. The data also indicated notable disparities in pediatric ophthalmologist density, with North Dakota, Vermont, and Wyoming having no pediatric ophthalmologists. A significant correlation was identified between pediatric ophthalmologist availability and vision screening (R = 0.486, p < 0.001). Negative correlations were observed between the relative demand index and urbanization (R = -0.388, p = 0.007), as well as the strabismus prevalence (R = -0.455, p = 0.001), indicating that urbanization and specialist distribution influence eyecare demand and access. Conclusion: The findings highlight disparities in accessibility and demand for pediatric ophthalmology services across the US, influenced by urbanization and distribution of healthcare professionals. The study underscores the need for targeted interventions to improve access to care and bridge gaps in pediatric ophthalmology service provision.
{"title":"Assessing The Relationship Between Demand And Accessibility For Pediatric Ophthalmology Services By State In The United States","authors":"Benton Chuter, Alex C Lieu, Justin Huynh, Jennifer J Bu, Linda M Zangwill","doi":"10.1101/2024.08.03.24311451","DOIUrl":"https://doi.org/10.1101/2024.08.03.24311451","url":null,"abstract":"Purpose: To investigate the relationship between public demand for pediatric ophthalmology services and the accessibility of such services on a statewide basis in the United States, focusing on strabismus care.\u0000Methods: Using Google Trends data, search volumes for \"strabismus\" were analyzed from January 2014 to December 2023. Pediatric ophthalmologist availability was assessed via the American Academy of Ophthalmology directory, normalized for state population size to create a relative demand index. Additional metrics from the United States Census Bureau and Centers for Disease Control and Prevention provided socioeconomic and health data. Correlation coefficients were used to explore relationships between search volumes, specialist availability, vision screening, socioeconomics, and demographic factors.\u0000Results: The data revealed variability in public interest in pediatric ophthalmology across states, with Alaska showing the highest relative search volume for strabismus. The data also indicated notable disparities in pediatric ophthalmologist density, with North Dakota, Vermont, and Wyoming having no pediatric ophthalmologists. A significant correlation was identified between pediatric ophthalmologist availability and vision screening (R = 0.486, p < 0.001). Negative correlations were observed between the relative demand index and urbanization (R = -0.388, p = 0.007), as well as the strabismus prevalence (R = -0.455, p = 0.001), indicating that urbanization and specialist distribution influence eyecare demand and access.\u0000Conclusion: The findings highlight disparities in accessibility and demand for pediatric ophthalmology services across the US, influenced by urbanization and distribution of healthcare professionals. The study underscores the need for targeted interventions to improve access to care and bridge gaps in pediatric ophthalmology service provision.","PeriodicalId":501390,"journal":{"name":"medRxiv - Ophthalmology","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-08-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141942124","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-08-05DOI: 10.1101/2024.08.04.24311475
Benton Chuter, Justin Huynh, Evan Walker, Shahin Hallaj, Jalil Jalili, Jeffrey Liebmann, Massimo A Fazio, Christopher A Girkin, Robert N Weinreb, Mark Christopher, Linda M Zangwill
Purpose: To fine tune and evaluate the performance of the retinal foundation model (RETFound) on a diverse longitudinal clinical research dataset in glaucoma detection from optical coherence tomography (OCT) RNFL scans. Subanalyses of the model performance were evaluated across different subgroups, various dataset sample sizes and training cycles (epochs). Design: Evaluation of a diagnostic technology Subjects, Participants, and Controls: 15,216 Spectralis OCT RNFL circle scans of 747 individuals of diverse race (56.9% White, 37.8% Black / African American, and 5.3% Other / Not reported (5.3%), glaucoma severity (30.8% mild, 18.4% moderate-to-severe, and 50.9% no glaucoma), and age (44.8% <60 years, 55.2% >60 years) from the Diagnostic Innovations in Glaucoma Study (DIGS) and the African Descent and Glaucoma Evaluation Study (ADAGES). All OCT b scans were labeled as "Non-glaucomatous" or "Glaucomatous." Methods: RETFound was employed to perform binary glaucoma classification. The diagnostic accuracy of RETFound was iteratively tested across different combinations of dataset sample sizes (50 to 2000 OCT RNFL circle scans), epochs (5 to 50), and study subpopulations stratified by severity of glaucoma, age, and race). Main Outcome Measures: Area under receiver operating characteristic curve (AUC) for classifying RNFL scans as "Non-glaucomatous" or "Glaucomatous." Results: Performance metrics improved with larger training datasets and more training cycles, rising from an AUC of 0.61 (50 training images and 5 epochs) to AUC 0.91 (2,000 training images and 50 epochs). Gains in performance were marginal as training size increased beyond 500 scans. Performance was similar across race for all training size and cycle number combinations: African American (AUC=0.90) vs other (AUC=0.93). RNFL scans from older patients (>60 years) led to worse performance (AUC=0.85) compared to younger patients (<60 years, AUC=0.95), Performance was significantly higher for RNFL scans from patients with moderate-to-severe glaucoma vs mild glaucoma (AUC=0.99 vs 0.88, respectively). Conclusions: Good RETFound performance was observed with a relatively small sample size of images used for fine tuning and across differences in race and age. The ability of RETFound to adapt across a range of OCT training conditions and populations suggests it is a promising tool to automate glaucoma detection in a variety of use cases.
{"title":"Accuracy of a New Foundation Model in Glaucoma Detection using Ocular Coherence Tomography Images","authors":"Benton Chuter, Justin Huynh, Evan Walker, Shahin Hallaj, Jalil Jalili, Jeffrey Liebmann, Massimo A Fazio, Christopher A Girkin, Robert N Weinreb, Mark Christopher, Linda M Zangwill","doi":"10.1101/2024.08.04.24311475","DOIUrl":"https://doi.org/10.1101/2024.08.04.24311475","url":null,"abstract":"Purpose: To fine tune and evaluate the performance of the retinal foundation model (RETFound) on a diverse longitudinal clinical research dataset in glaucoma detection from optical coherence tomography (OCT) RNFL scans. Subanalyses of the model performance were evaluated across different subgroups, various dataset sample sizes and training cycles (epochs). Design: Evaluation of a diagnostic technology Subjects, Participants, and Controls: 15,216 Spectralis OCT RNFL circle scans of 747 individuals of diverse race (56.9% White, 37.8% Black / African American, and 5.3% Other / Not reported (5.3%), glaucoma severity (30.8% mild, 18.4% moderate-to-severe, and 50.9% no glaucoma), and age (44.8% <60 years, 55.2% >60 years) from the Diagnostic Innovations in Glaucoma Study (DIGS) and the African Descent and Glaucoma Evaluation Study (ADAGES). All OCT b scans were labeled as \"Non-glaucomatous\" or \"Glaucomatous.\" Methods: RETFound was employed to perform binary glaucoma classification. The diagnostic accuracy of RETFound was iteratively tested across different combinations of dataset sample sizes (50 to 2000 OCT RNFL circle scans), epochs (5 to 50), and study subpopulations stratified by severity of glaucoma, age, and race). Main Outcome Measures: Area under receiver operating characteristic curve (AUC) for classifying RNFL scans as \"Non-glaucomatous\" or \"Glaucomatous.\" Results: Performance metrics improved with larger training datasets and more training cycles, rising from an AUC of 0.61 (50 training images and 5 epochs) to AUC 0.91 (2,000 training images and 50 epochs). Gains in performance were marginal as training size increased beyond 500 scans. Performance was similar across race for all training size and cycle number combinations: African American (AUC=0.90) vs other (AUC=0.93). RNFL scans from older patients (>60 years) led to worse performance (AUC=0.85) compared to younger patients (<60 years, AUC=0.95), Performance was significantly higher for RNFL scans from patients with moderate-to-severe glaucoma vs mild glaucoma (AUC=0.99 vs 0.88, respectively). Conclusions: Good RETFound performance was observed with a relatively small sample size of images used for fine tuning and across differences in race and age. The ability of RETFound to adapt across a range of OCT training conditions and populations suggests it is a promising tool to automate glaucoma detection in a variety of use cases.","PeriodicalId":501390,"journal":{"name":"medRxiv - Ophthalmology","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2024-08-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141942030","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：