Convolutional Neural Networks Accurately Identify Ungradable Images in a Diabetic Retinopathy Telemedicine Screening Program.

Purpose: Diabetic retinopathy (DR) is a microvascular complication of diabetes mellitus (DM). Standard of care for patients with DM is an annual eye examination or retinal imaging to assess for DR, the latter of which may be completed through telemedicine approaches. One significant issue is poor-quality images that prevent adequate screening and are thus ungradable. We used artificial intelligence to enable point-of-care (at time of imaging) identification of ungradable images in a DR screening program. Methods: Nonmydriatic retinal images were gathered from patients with DM imaged during a primary care or endocrinology visit from September 1, 2017, to June 1, 2021. The Topcon TRC-NW400 retinal camera (Topcon Corp., Tokyo, Japan) was used. Images were interpreted by 5 ophthalmologists for gradeability, presence and stage of DR, and presence of non-DR pathologies. A convolutional neural network with Inception V3 network architecture was trained to assess image gradeability. Images were divided into training and test sets, and 10-fold cross-validation was performed. Results: A total of 1,377 images from 537 patients (56.1% female, median age 58) were analyzed. Ophthalmologists classified 25.9% of images as ungradable. Of gradable images, 18.6% had DR of varying degrees and 26.5% had non-DR pathology. 10 fold cross-validation produced an average area under receiver operating characteristic curve (AUC) of 0.922 (standard deviation: 0.027, range: 0.882 to 0.961). The final model exhibited similar test set performance with an AUC of 0.924. Conclusions: This model accurately assesses gradeability of nonmydriatic retinal images. It could be used for increasing the efficiency of DR screening programs by enabling point-of-care identification of poor-quality images.

Identifying core biological processes distinguishing human eye tissues with precise systems-level gene expression analyses and weighted correlation networks.

The human eye is built from several specialized tissues which direct, capture and pre-process information to provide vision. The gene expression of the different eye tissues has been extensively profiled with RNA-seq across numerous studies. Large consortium projects have also used RNA-seq to study gene expression patterning across many different human tissues, minus the eye. There has not been an integrated study of expression patterns from multiple eye tissues compared with other human body tissues. We have collated all publicly available healthy human eye RNA-seq datasets as well as dozens of other tissues. We use this fully integrated dataset to probe the biological processes and pan expression relationships between the cornea, retina, retinal pigment epithelium (RPE)-choroid complex, and the rest of the human tissues with differential expression, clustering and gene ontology term enrichment tools. We also leverage our large collection of retina and RPE-choroid tissues to build the first human weighted gene correlation networks and use them to highlight known biological pathways and eye gene disease enrichment. We also have integrated publicly available single-cell RNA-seq data from mouse retina into our framework for validation and discovery. Finally, we make all these data, analyses and visualizations available via a powerful interactive web application (https://eyeintegration.nei.nih.gov/).

Exploring Diagnostic Precision and Triage Proficiency: A Comparative Study of GPT-4 and Bard in Addressing Common Ophthalmic Complaints.

In the modern era, patients often resort to the internet for answers to their health-related concerns, and clinics face challenges to providing timely response to patient concerns. This has led to a need to investigate the capabilities of AI chatbots for ophthalmic diagnosis and triage. In this in silico study, 80 simulated patient complaints in ophthalmology with varying urgency levels and clinical descriptors were entered into both ChatGPT and Bard in a systematic 3-step submission process asking chatbots to triage, diagnose, and evaluate urgency. Three ophthalmologists graded chatbot responses. Chatbots were significantly better at ophthalmic triage than diagnosis (90.0% appropriate triage vs. 48.8% correct leading diagnosis; p < 0.001), and GPT-4 performed better than Bard for appropriate triage recommendations (96.3% vs. 83.8%; p = 0.008), grader satisfaction for patient use (81.3% vs. 55.0%; p < 0.001), and lower potential harm rates (6.3% vs. 20.0%; p = 0.010). More descriptors improved the accuracy of diagnosis for both GPT-4 and Bard. These results indicate that chatbots may not need to recognize the correct diagnosis to provide appropriate ophthalmic triage, and there is a potential utility of these tools in aiding patients or triage staff; however, they are not a replacement for professional ophthalmic evaluation or advice.

Predicting Malignant Transformation of Choroidal Nevi Using Machine Learning.

Objective: This study aims to assess a machine learning (ML) algorithm using multimodal imaging to accurately identify risk factors for uveal melanoma (UM) and aid in the diagnosis of melanocytic choroidal tumors. Subjects and Methods: This study included 223 eyes from 221 patients with melanocytic choroidal lesions seen at the eye clinic of the University of Illinois at Chicago between 01/2010 and 07/2022. An ML algorithm was developed and trained on ultra-widefield fundus imaging and B-scan ultrasonography to detect risk factors of malignant transformation of choroidal lesions into UM. The risk factors were verified using all multimodal imaging available from the time of diagnosis. We also explore classification of lesions into UM and choroidal nevi using the ML algorithm. Results: The ML algorithm assessed features of ultra-widefield fundus imaging and B-scan ultrasonography to determine the presence of the following risk factors for malignant transformation: lesion thickness, subretinal fluid, orange pigment, proximity to optic nerve, ultrasound hollowness, and drusen. The algorithm also provided classification of lesions into UM and choroidal nevi. A total of 115 patients with choroidal nevi and 108 patients with UM were included. The mean lesion thickness for choroidal nevi was 1.6 mm and for UM was 5.9 mm. Eleven ML models were implemented and achieved high accuracy, with an area under the curve of 0.982 for thickness prediction and 0.964 for subretinal fluid prediction. Sensitivity/specificity values ranged from 0.900/0.818 to 1.000/0.727 for different features. The ML algorithm demonstrated high accuracy in identifying risk factors and differentiating lesions based on the analyzed imaging data. Conclusions: This study provides proof of concept that ML can accurately identify risk factors for malignant transformation in melanocytic choroidal tumors based on a single ultra-widefield fundus image or B-scan ultrasound at the time of initial presentation. By leveraging the efficiency and availability of ML, this study has the potential to provide a non-invasive tool that helps to prevent unnecessary treatment, improve our ability to predict malignant transformation, reduce the risk of metastasis, and potentially save patient lives.

Macular findings expedite accurate diagnosis of MIDD in a young female patient with newly diagnosed diabetes.

Purpose: To report a case of a 34-year-old female patient with newly diagnosed Maternally Inherited Diabetes and Deafness (MIDD) in the setting of undifferentiated macular dystrophy and newly discovered diabetes. Observations: A 34-year-old woman presented to the retina service with new-onset diabetes and a history of hydroxychloroquine use. Ophthalmologic examination showed findings early in the patient's presentation that within the context of her recent diabetes diagnosis and family history pointed to MIDD as the specific cause of the patient's many different symptoms. This diagnosis was further supported through obtaining previous ophthalmic images of the patient's mother demonstrating circular areas of geographic atrophy seen in advanced MIDD, and the diagnosis was confirmed through genetic testing. Conclusions and importance: As was observed in the patient discussed in this manuscript, recognition of macular dystrophy findings suggestive of MIDD can hasten a timely diagnosis for a patient with diabetes of unspecified etiology. Additionally, knowledge of the underlying cause being MIDD can optimize care for patients in terms of treatment, understanding their risk for various diabetes complications, screening for additional systemic manifestations, and initiating valuable genetic counseling for patients and their families. Given these factors and the surprisingly high prevalence of MIDD among diabetes patients, increased awareness of MIDD and its manifestations can help to optimize diagnosis and management for these patients.

Timeout? The Epidemiology of Pediatric Sports Injuries During the COVID-19 Pandemic.

BACKGROUND: The COVID-19 pandemic resulted in closure of schools and playgrounds while requiring social distancing, changes that likely affected youth sports participation. The purpose of this study was to identify changes in the epidemiology of pediatric sports injuries during the COVID-19 pandemic. METHODS: This retrospective cohort study included patients between the ages of 4 and 18 years who presented to orthopaedic clinics within a single children's hospital network with an acute injury sustained during athletic activity between March 20, 2020, and June 3, 2020 (the strictest period of state-level shelter-in-place orders). These patients were compared with those within the same dates in 2018 and 2019. Chi square and Mann-Whitney U tests were used, as appropriate. RESULTS: Significantly less sports injuries were seen during the pandemic (n = 257) compared with the same dates in 2018 (n = 483) and 2019 (n = 444) despite more providers available in 2020 (P < 0.001). During the pandemic, patients with sports injuries were younger (median age 11 versus 13 years, P < 0.001) and had less delay in presentation (median 5 versus 11 days, P < 0.001). A higher proportion were White (66.9% versus 47.7%, P < 0.001), privately insured (63.4% versus 48.3%, P < 0.001), and seen at a nonurban location (63.4% versus 50.2%, P < 0.001). Most sports injuries during the pandemic were fractures (83.7%). Although 71.4% of all injuries in the prepandemic period occurred in the context of formal sports, only 15.2% were sustained in a formal athletic context in 2020 (P < 0.001). The frequency of surgical treatment was higher during the pandemic (14.8% versus 7.8%, P = 0.001), mainly because most of these injuries were fractures requiring surgical intervention. CONCLUSIONS: Fewer sports injuries were seen in the outpatient setting during the COVID-19 pandemic, and most of these injuries were fractures and occurred outside of organized sports settings. Patients were more likely to be White, privately insured, and seen at a nonurban location.

Retinal Vessel Density and Treatment Intensity among Adults with Retinal Vein Occlusion: A Swept-Source Optical Coherence Tomography Angiography Study.

Previous studies have shown retinal vein occlusion (RVO) is associated with changes in vessel density visible on swept-source optical coherence tomography angiography (ss-OCTA). This study aimed to characterize retinal changes on ss-OCTA among RVO patients stratified by the need for continuous anti-VEGF therapy. This cross-sectional study of 24 RVO patients ≥ 18 years were imaged with SS-OCT-A. Patients were categorized into continuous vs. limited therapy (≥1 vs. no injections in previous 12 months) based on recurrence of intraretinal fluid (IRF) on OCT. Images were analyzed using ImageJ. T-tests were used to compare vessel density of the macula and peripheral retina. Overall, RVO patients undergoing continuous therapy (n = 14) had higher diabetes prevalence, worse baseline visual acuity, and higher baseline macular thickness compared to the limited (n = 10) therapy group. Continuous therapy was associated with lower macular VD in the combined retina layer and the superficial capillary plexus (SCP), but not in the deep capillary plexus (DCP). Further, the continuous therapy group exhibited lower peripheral VD in the combined retina layer, and no difference in the SCP and DCP layers when analyzed separately. In conclusion, RVO patients requiring continuous anti-VEGF injections demonstrate reduced VD of the macula and in the periphery on SS-OCTA imaging. SS-OCTA may be valuable for monitoring and prognosticating treatment for RVO patients.