The Potential Role of an Artificial Intelligence-Driven Tool in Decision-Making for Mitral Valve Repair Surgery.

Background

Annuloplasty ring sizing is critical for durable outcomes in surgical mitral valve repair (MVr). However, there is no clear consensus on optimal sizing strategies. Artificial intelligence (AI)-based imaging tools may help to reduce uncertainty in preoperative decision-making by providing objective, reproducible and reliable measurements. This study evaluated the predictive capability of a fully automated, computed tomography (CT)-based AI-driven tool for annuloplasty ring sizing in patients undergoing minimally invasive MVr (MI-MVr).

Methods

A total of 71 consecutive patients undergoing MI-MVr for Carpentier type II mitral valve insufficiency during the study period were included. Preoperative CT scans were analyzed using a cloud-based, fully automated AI tool to quantify mitral valve geometric parameters. Correlations between AI-derived measurements and implanted ring sizes were assessed using the Pearson correlation test. Univariable and multivariable linear regression analyses were performed to identify independent predictors of ring size selection.

Results

Several AI-derived parameters correlated significantly with implanted ring size, with the strongest correlations observed for commissural width (R = 0.693, p < 0.001) and mitral annular area (R = 0.693, p < 0.001). In multivariable regression analysis, these parameters were the strongest predictors of annuloplasty ring size (R2 = 0.504, p < 0.001). Using this model, accurate annuloplasty ring sizing could be predicted in 78.8% of patients. There were no in-hospital mortality and residual mitral regurgitation at discharge.

Conclusions

A fully automated, CT-based AI-driven tool demonstrated good accuracy for preoperative annuloplasty ring size prediction in MI-MVr and may have the potential to support surgical decision-making, reduce operator dependence, and improve reproducibility.