Predictive Value of Machine Learning–Based Echocardiographic Myocardial Texture After Transcatheter Aortic Valve Replacement

OBJECTIVES: Transcatheter aortic valve replacement (TAVR) is a minimally invasive procedure for treating severe aortic valve stenosis. This study aimed to develop an interpretable machine learning (ML) model based on echocardiographic myocardial texture for predicting 1-year clinical endpoints following TAVR via the femoral artery. METHODS: We retrospectively studied 121 patients who underwent transfemoral TAVR between January 2019 and October 2023, and data from external validation centers were collected between January 2019 and October 2023 to validate and evaluate the model externally. The composite outcome of all-cause mortality and heart failure events was used as the clinical endpoint. Based on echocardiographic imaging of myocardial texture analysis, nine widely used ML algorithms were used to construct models. Three predictive models were then constructed based on the best machine mode, with Shapley additive explanations for the interpretation of the model and the assessment of the contributions of the different characteristics. RESULTS: Among the nine ML models, the Extra Tree (ET) model had the best discriminative ability. An explainable final ET model was established with two features. The final model accurately predicted the clinical endpoint 1 year after transfemoral TAVR in both internal and external validations, and the results in the COX multifactorial regression analysis showed that LVEF <50% was an independent predictor of the clinical endpoint 1 year after transfemoral TAVR. In addition, Kaplan-Meier survival curves showed that patients with high Rad-Score values had lower survival rates than those with low Rad-Score values. CONCLUSIONS: This novel method based on echocardiographic myocardial texture with ML may have the potential to be used for predicting clinical endpoints 1 year after transfemoral TAVR and may be used as a convenient tool to facilitate its utility in clinical settings.