Search Results

Abstract

PURPOSE: This pilot study evaluated the diagnostic accuracy of a deep learning model for detecting pterygium in anterior segment photographs taken using smartphones in the Brazilian Amazon. The model’s performance was benchmarked against assessments made by experienced ophthalmologists, considered the clinical gold standard.
METHODS: In this cross-sectional study, 38 participants (76 eyes) from Barcelos, Brazil, were enrolled. Trained nonmedical health workers captured high-resolution anterior segment images using smartphones. These images were analyzed using a deep learning model based on the MobileNet-V2 convolutional neural network. Diagnostic metrics–including sensitivity, specificity, accuracy, positive predictive value, negative predictive value, and area under the receiver operating characteristic curve–were calculated and compared with the ophthalmologists’ evaluations.
RESULTS: The deep learning model achieved a sensitivity of 91.43%, specificity of 90.24%, positive predictive value of 88.46%, negative predictive value of 92.79%, and an area under the curve of 0.91. Logistic regression revealed no statistically significant association between pterygium and demographic variables such as age or gender.
CONCLUSIONS: The deep learning model demonstrated high diagnostic performance in identifying pterygium in a remote Amazonian population. These preliminary findings support the potential use of artificial intelligence–based tools to facilitate early detection and screening in underserved regions, thereby enhancing access to ophthalmic care.

Keywords: Pterygium/diagnostic imaging; Smartphone; Diagnostic techniques, ophthalmological; Deep learning; Telemedicine; Artificial intelligence; Cross-sectional studies; Brazil/epidemiology

Abstract

PURPOSE: To assess the performance of a contemporary large language model (ChatGPT-5) against ophthalmology residents on a standardized set of glaucoma multiple-choice questions.
METHODS: We conducted a cross-sectional comparative study with 189 text-only glaucoma multiple-choice questions from the Cybersight question bank. ChatGPT-5 was tested under standardized conditions, with each item placed in a new chat and limited to letter-only outputs. Six ophthalmology residents from a Brazilian training program (two Postgraduate Year 1, two Postgraduate Year 2, and two Postgraduate Year 3) answered the same questions under supervision. Accuracy was calculated using the official key. McNemar’s exact test was used to compare items between ChatGPT-5 and residents, and matched odds ratios and 95% confidence intervals (95% CIs) were calculated using the Haldane–Anscombe correction.
RESULTS: ChatGPT-5 received 164 of 189 correct responses (86.8%; 95% CI, 81.2–90.9). Residents’ overall accuracy was 62.9% (713/1,134; 95% CI, 60.0–65.6). The top-performing resident earned 76.7%. ChatGPT-5 outperformed all residents in head-to-head comparisons, with odds ratios ranging from 1.84 (95% CI, 1.10–3.08) to 13.15 (95% CI, 5.93–29.20), all p≤0.023. ChatGPT-5 correctly answered 17/189 items (9.0%), but fewer than half of residents were correct (“large language model-only wins”), whereas residents were more successful on items that ChatGPT-5 overlooked.
CONCLUSIONS: ChatGPT-5 outperformed ophthalmology residents on text-based glaucoma multiple-choice questions, indicating its potential as a subspecialty education and assessment tool. Generalizability is limited by the single question bank, text-only items, a small resident cohort, and the evaluation of one large language model version at a single time point. Before incorporating these findings into clinical decision-making, larger, multimodal, and longitudinal studies are required.

Keywords: Glaucoma; Artificial intelligence; Large language models; Education, medical; Medical staff, hospital