Long-term electrocardiogram signal analysis method and system based on sparse self-attention
By employing a sparse self-attention-based long-term ECG signal analysis method, the accuracy problem of existing systems in detecting complex arrhythmias and multi-lead settings has been solved. This method enables efficient multi-lead, arbitrary-duration ECG signal analysis, improving the applicability and diagnostic efficiency of the automatic analysis system.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- FUWAI HOSPITAL CHINESE ACAD OF MEDICAL SCI & PEKING UNION MEDICAL COLLEGE
- Filing Date
- 2026-06-10
- Publication Date
- 2026-07-17
AI Technical Summary
Existing ECG signal analysis systems lack sufficient accuracy in identifying complex arrhythmias, subtle abnormal events, and key feature points. They are also ill-suited to diverse lead settings and rhythm relationships within longer time windows, resulting in insufficient clinical applicability and accuracy, requiring extensive manual review by physicians.
A long-term ECG signal analysis method based on sparse self-attention is adopted. Through self-supervised pre-training and multi-task supervised fine-tuning stages, combined with the sparse attention mechanism, it achieves efficient processing of multi-lead configuration and input of arbitrary duration. The integration of self-supervised pre-training and multi-task supervised fine-tuning strategies enhances the model's generalization ability.
It significantly improves the applicability and reliability of automated analysis systems, reduces the workload of clinicians, and promotes the standardization of electrocardiogram interpretation and the improvement of the efficiency of cardiovascular disease diagnosis and treatment.
Smart Images
Figure CN122413337A_ABST