Diagnostic Value of Convolutional Neural Network Algorithm and High-Sensitivity Cardiac Troponin I Detection Under Machine Learning in Myocardial Infarction

Abstract

Background: It aimed to the diagnosis and examination of acute myocardial infarction (AMI) using echocardiography under improved convolutional neural network (CNN) algorithm and high-sensitivity (Sen) cardiac troponin I (hs-cTnI) detection. The application effect was also evaluated. Methods: Ninety AMI patients were recruited as the AMI group, and ninety healthy individuals who underwent physical examinations simultaneously were chosen as control (Ctrl) group. Improved CNN algorithm-based echocardiography combined with hs-cTnI detection was applied, and its diagnostic efficiency was evaluated. Results: The optimal dataset scale (ODS), optimal image scale (OIS) and average precision (AP) of the proposed algorithm were better than those of manual labeling, Canny algorithm, and structured edge (SE) algorithm (P < 0.05). The left ventricular ejection fraction (LVEF) of the patients in the AMI group was inferior to that of Ctrl group ((55.09±2.78)%) versus (65.01±3.19)%), the left ventricular end-diastolic dimension (LVEDD) was superior to that of Ctrl group ((54.89±6.56) mm vs. (45.98±5.77) mm), and the cTnI level was also superior to that of Ctrl group ((2.90±0.31) pg/L vs. (0.73±0.42) pg/L) (P < 0.05). The diagnostic Sen (91.89%), specificity (Spe) (81.25%), accuracy (Acc) (90.00%) and consistency (0.56) of echocardiography combined with hs-cTnI were superior to those of single echocardiography or cTnI detection (P < 0.05).