OCT-PRO Model vs. Clinicians: Cataract Surgery Outcome Prediction

Overview

Cataract is the leading cause of blindness worldwide, yet 5-20% of patients fail to achieve satisfactory visual recovery after surgery. Current methods for predicting postoperative visual acuity lack accuracy, particularly in patients with co-morbid fundus diseases. The OCT-PRO model, developed by our team, uses artificial intelligence (AI) to integrate optical coherence tomography (OCT) images and clinical data to forecast surgical outcomes. This multi-center, randomized, single-blind trial aims to compare the predictive accuracy of OCT-PRO-assisted predictions versus standard clinician predictions. A total of 534 participants will be randomized 1:1 to either the experimental group (OCT-PRO-assisted prediction) or the control group (routine care). The primary outcome is the mean absolute error (MAE) between predicted and actual postoperative best-corrected visual acuity (BCVA). Secondary outcomes include patient satisfaction, informed decision-making scores, and clinician acceptance of the AI tool. This study will provide high-level evidence on the clinical utility of AI in optimizing cataract surgical decision-making and patient communication.

Cataract is the leading cause of blindness worldwide, yet 5-20% of patients fail to achieve satisfactory visual recovery after surgery. Accurate preoperative prediction remains challenging, particularly for eyes with co-morbid retinal pathologies, as current methods relying on clinician experience and traditional tests (e.g., laser interferometry) often lack reproducibility. Although AI models like OCT-PRO show promise, prospective RCT evidence comparing their accuracy against clinicians is lacking. This multi-center, randomized, assessor-blinded trial will enroll 534 adults scheduled for cataract surgery. Participants are allocated 1:1 to either the Experimental Group or the Control Group via centralized randomization. In the Experimental Group, clinicians use the OCT-PRO model-integrating OCT images and clinical data-to obtain a predicted postoperative BCVA. Physicians may confirm or adjust this prediction, and the final value is communicated to patients during preoperative counseling. The Control Group receives standard care, where predictions are based solely on conventional clinical assessments without AI assistance. Outcome assessors will be blinded to group allocation. The primary endpoint is the Mean Absolute Error (MAE) between predicted and actual postoperative BCVA. Secondary endpoints include patient-reported outcomes (expectations, informed choice, satisfaction), clinician acceptance of the model, and correlation analyses. Analysis will follow the Intention-to-Treat principle. This study aims to provide high-level evidence on integrating AI into clinical workflows to enhance prognostic accuracy and optimize shared decision-making in cataract surgery.