AI-based methods show high accuracy in prediciting graft failure risks in DMEK

Machine learning techniques can effectively predict early graft failure (GF) in patients undergoing Descemet membrane endothelial keratoplasty (DMEK), according to a study.

Duration of intensive care unit (ICU) stay and death-to-preservation time (DPT) are significant predictors of GF risk.

The study found that machine learning models achieved a classification accuracy of 96%, with a precision of 0.95, recall of 0.81, and an F1-score of 0.90. The analysis indicated that longer ICU stays and extended DPT were significant predictors of GF risk (P < 0.05), while donor age, endothelial cell density, and other factors did not show a significant relationship with GF.

Reference
Karaca EE, Bulut Ustael A, Keçeli AS, et al. Predicting Success in Descemet Membrane Endothelial Keratoplasty Surgery Using Machine Learning. Cornea. 2024;doi: 10.1097/ICO.0000000000003599. Epub ahead of print. PMID: 38913970.