查看文章

Background

Leaflet thrombosis (LT) is an adverse outcome after aortic valve replacement (AVR), of unknown etiology and uncertain predictive factors. This study aims to generate a data-driven predictive model of LT to assist clinicians during preprocedural planning in selecting a valve that leads to minimal LT chance.

Methods

The pre-AVR CT scans and echocardiographic data of 58 patients at Mayo Clinic of whom 29 developed LT and 29 did not (control) were obtained. Patients were matched by age and sex. Geometric, hemodynamic and baseline clinical data were extracted. Geometric data included sinuses size, STJ height and area, coronary heights, annulus size, etc. Hemodynamic data included stroke volume, pressure gradient, etc. Baseline data included smoking, diabetes, hypertension, etc. Logistic regression trained on anatomical and hemodynamic data followed by Bayesian inference (to include …