Retina (Philadelphia, Pa.) · 2026
PREDICTING POSTOPERATIVE OUTCOMES IN FULL-THICKNESS MACULAR HOLE REPAIR SURGERY: ChatGPT Versus Clinical Decision.
СтатьяСравнительное исследованиеНазвание и аннотация: Перевод пока не подготовлен
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Wattad Aya, Saadi Razan, Bez Maxim, Loewenstein Anat, Goldstein Michaella
Аннотация
PURPOSE: To evaluate the ability of ChatGPT-5 to predict long-term anatomical and functional outcomes after full-thickness macular hole (FTMH) surgery, and to compare its performance with retinal specialists' predictions and real-world results.
METHODS: This retrospective study included 50 eyes of 50 patients undergoing pars plana vitrectomy for FTMH (2021-2024). Deidentified clinical summaries with preoperative demographics, ocular history, best-corrected visual acuity (BCVA), optical coherence tomography parameters, foveal B-scan optical coherence tomography images, and surgical details were entered into ChatGPT-5 using a standardized prompt to predict 12-month BCVA and anatomical closure. Predictions were compared with actual results and assessments from two senior retina specialists.
RESULTS: At 12 months, closure occurred in 44/50 eyes (88%), and mean BCVA improved from 20/100 (0.7 ± 0.4 logMAR) to 20/63 (0.5 ± 0.5 logMAR) ( P = 0.03). The anatomical prediction accuracy was 72% to 86% (specialists) and 90% (ChatGPT-5). ChatGPT achieved perfect accuracy in closure cases but failed to identify nonclosure, reflecting optimism bias. For functional outcomes, the accuracy was 42% to 44% (specialists) and 66% (ChatGPT-5). ChatGPT-5 performed well when vision improved (60%) but poorly for stable (≤13%) or worsened (0%) cases. The mean BCVA prediction error was 11.4 ± 10.8 letters, with ∼60% within two lines of the true outcome.
CONCLUSION: ChatGPT-5 demonstrated apparent accuracy in predicting FTMH surgery outcomes; however, this was largely driven by an optimism bias that overestimated closure and visual recovery. This model still lack clinical judgment. Larger prospective studies are needed before clinical use.
METHODS: This retrospective study included 50 eyes of 50 patients undergoing pars plana vitrectomy for FTMH (2021-2024). Deidentified clinical summaries with preoperative demographics, ocular history, best-corrected visual acuity (BCVA), optical coherence tomography parameters, foveal B-scan optical coherence tomography images, and surgical details were entered into ChatGPT-5 using a standardized prompt to predict 12-month BCVA and anatomical closure. Predictions were compared with actual results and assessments from two senior retina specialists.
RESULTS: At 12 months, closure occurred in 44/50 eyes (88%), and mean BCVA improved from 20/100 (0.7 ± 0.4 logMAR) to 20/63 (0.5 ± 0.5 logMAR) ( P = 0.03). The anatomical prediction accuracy was 72% to 86% (specialists) and 90% (ChatGPT-5). ChatGPT achieved perfect accuracy in closure cases but failed to identify nonclosure, reflecting optimism bias. For functional outcomes, the accuracy was 42% to 44% (specialists) and 66% (ChatGPT-5). ChatGPT-5 performed well when vision improved (60%) but poorly for stable (≤13%) or worsened (0%) cases. The mean BCVA prediction error was 11.4 ± 10.8 letters, with ∼60% within two lines of the true outcome.
CONCLUSION: ChatGPT-5 demonstrated apparent accuracy in predicting FTMH surgery outcomes; however, this was largely driven by an optimism bias that overestimated closure and visual recovery. This model still lack clinical judgment. Larger prospective studies are needed before clinical use.