Prediction of atrial fibrillation in men with coronary artery disease by artificial neural network

Abstract

\n \n \n Type of funding sources: None.\n \n \n \n Atrial fibrillation (AF) is quite common in patients with coronary artery disease (CАD). However, there is no data whether AF is associated with coronary lesion of specific localization.\n \n \n \n To select the most sensitive parameters for creating artificial neural networks (ANN) model to predict AF in men with CАD using ANN.\n \n \n \n We selected the data of men of comparable age with angiographically proven CAD from the Register of coronary angiography: the main group - 180 men with AF, the comparison group - 713 men without AF. To assess the risk of AF the ANN mathematical model was applied - a multilayer perceptron with one hidden layer. The initial number of patients was divided into 3 samples: training, test and control.\n \n \n \n Patients with AF were significantly less frequently employed in budget organizations (55.0% vs 63.7%, p\u2009=\u20090.040), more often demonstrated higher (III-IV) heart failure NYHA classes (49.2% vs 21.1% , p\u2009<\u20090.001), had higher body mass index (BMI) (30.2 [27.4; 33.2] kg/m² vs 29.0 [26.1; 32.3] kg/m², p\u2009=\u20090.002), echocardiographic indices of increased left ventricular (LV) myocardial mass (163.7 [144.5; 192.4] g/m² vs 143.9 [126.1; 169.0] g/m², p\u2009<\u20090.001), LV (25.8 [24.1; 29.1] mm/m² vs 25.6 [23.9; 27.5] mm/m², p\u2009=\u20090.020) and right ventricular dimensions, enlarged left atrium (23.6 [21.7; 25.7] mm/m² vs 21.1 [19.7; 22.7] mm/m², p\u2009<\u20090.001). Rate of hemodynamically significant mitral regurgitation in AF patients was higher (48.2% vs 14.1%, p\u2009<\u20090.001), index of aortic root dimensions (7.7 [16.4; 19.0] mm/m² vs 18.3 [17.8; 20.0] mm/m², р=0.002) and LV ejection fraction (EF) in this group was lower (49 [42; 56]% vs 56 [47; 60]%, p\u2009<\u20090.001), coronary calcification (23.2% vs 15.7%, p\u2009=\u20090.024 ) and proximal lesions of the right coronary artery (RCA) (28.3% vs 22.7%, p\u2009=\u20090.025) were detected more often. Sensitivity of the final model which included 10 parameters was 85%, specificity 86%.\n \n \n \n In CAD men AF can be predicted by ANN model that takes into account the presence of significant mitral regurgitation, non-budgetary employment, severity of heart failure, coronary calcification and proximal lesion of RCA, BMI, echocardiographic indices of the left heart and aortic root dimensions, LVEF.\n