Deep learning models for multilabel ECG abnormalities classification: A comparative study using TPE optimization

Abstract The problem addressed in this study is the limitations of previous works that considered electrocardiogram (ECG) classification as a multiclass problem, despite many abnormalities being diagnosed simultaneously real life, making it multilabel problem. aim to test effectiveness deep learning (DL)-based methods (Inception, MobileNet, LeNet, AlexNet, VGG16, and ResNet50) using three large 12-lead ECG datasets overcome limitation. define-by-run technique used build most efficient DL model tree-structured Parzen estimator (TPE) algorithm. Results show proposed achieve high accuracy precision classifying for datasets, with best results 97.89% 90.83% Ningbo dataset, 42 classes Inception model; 96.53% 85.67% PTB-XL 24 Alex net 95.02% 70.71% Georgia 23 model. achieved optimum was by were 97.33% 97.71% classes; 96.60% 83.66% 94.32% 66.97% classes. DL-based TPE algorithm provide accurate abnormalities, improving diagnostic heart conditions.