Preoperative Assessment For Intraoperative Floppy Iris Syndrome (Ifis), Using Ai
Published 2025 - 43rd Congress of the ESCRS
Reference: PP11.15 | Type: Free paper | DOI: 10.82333/3ew1-5h04
Authors: Young Joon Choi* 1
1Department of Ophthalmology,Ajou University School of Medicine,Suwon,Korea, Republic Of
Purpose
To assess the use of an AI image algorithm, to identify preoperatively patients with risk for Intraoperative floppy iris syndrome (IFIS).
Setting
A single tetherary medical center
Methods
Preoperative infra-red anterior segment images were analyzed and labeled according to intra-operative IFIS or no-IFIS.
The data was allocated by a ratio of 8:1:1 for training, testing, and validation.
Several convolutional-neural-network-based models were proposed to classify the probability for IFIS based on the binary image label.
The model's performance was internally validated, and the evaluation indicators included accuracy, specificity, and negative predictive value.
Results
Overall, 5030 patients were identified, with 5,013 eyes of 3458 patients included for analysis.
The median age of the study participants was 72.0 (interquartile range [IQR], 60.0-82.0) years, and they included 143 men (58.6%).
The AI test algorithm achieved an AUC of 65% compared to two cataract surgeons’ graders, who achieved 45% and 55% AUC.
Conclusions
A novel AI image algorithm was developed that can help ophthalmologists identify patients at low risk for developing IFIS. This insight may contribute to the safe planning of cataract surgery and minimize the risk of intraoperative complications.