Optimizing electrocardiogram analysis for efficient heart condition diagnosis

Abstract

This thesis introduces an innovative lead grouping strategy for efficient real-time Electrocardiography signal classification. This method uses a maximum of six leads instead of the traditional 12-lead approach, leading to significant reductions in sampling time (93.67%), data size at the data acquisition device (50%), and signal processing time (84.72%). Importantly, these benefits come with a minimal loss in accuracy (0.08%). The thesis presents the CardioDiverse dataset, a publicly available resource that highlights key ECG leads associated with specific cardiovascular conditions. This resource can transform ECG-based diagnoses by focusing on the most pertinent leads. The proposed lead grouping strategy has been successfully integrated with a real-time platform, demonstrating its practical robustness and applicability. This contribution brings a considerable change in the field of ECG analysis by providing an efficient and viable lead grouping method that balances accuracy and resource efficiency, marking significant advances in ECG analysis.