Use of Artificial Intelligence to Assess the Severity of Mitral Regurgitation

Objective : To develop high-accuracy neural network model for the diagnosis and prediction of the severity of mitral regurgitation as assessed by echocardiography. Materials and methods : A total of 80 patients were divided into two groups: Group 1 included 42 patients with an eccentric mitral regurgitation jet, and Group 2 included 37 patients with a central mitral regurgitation jet. All patients underwent transthoracic echocardiography with assessment of mitral regurgitation severity based on the percentage ratio of the regurgitation jet area to the left atrial area. The vena contracta, proximal isovelocity surface area radius, and the magnitude of the horizontal color Doppler expansion artifact were measured. The effective regurgitant orifice area and mitral regurgitant volume were calculated. Results . A neural network model was developed that predicts “moderate” and “severe” forms of mitral regurgitation based on echocardiography data with highest possible accuracy (100%). The accuracy for predicting the “mild” form was slightly lower (83.33%).