Optimized Solutions of Electrocardiogram Lead and Segment Selection for Cardiovascular Disease Diagnostics

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Medicine
• Cardiology

Background:

• Existing multi-lead electrocardiogram (ECG) detection methods often utilize all 12 leads, leading to high computational demands unsuitable for portable systems.
• The impact of varying ECG lead configurations and heartbeat segment lengths on diagnostic accuracy remains unclear.
• There is a need for efficient and accurate ECG analysis methods for early cardiovascular disease detection in resource-limited settings.

Purpose of the Study:

• To develop a novel framework, GA-LSLO, for automatically optimizing the selection of ECG leads and segment lengths for cardiovascular disease detection.
• To enhance the accuracy of ECG analysis by incorporating a lead attention module (LAM) for feature weighting.
• To validate the proposed method's performance and hardware feasibility for portable ECG detection devices.

Main Methods:

• A Genetic Algorithm-based ECG Leads and Segment Length Optimization (GA-LSLO) framework was developed.
• Convolutional neural networks (CNNs) were employed for feature extraction across different leads and segment lengths.
• A lead attention module (LAM) was introduced to dynamically weight selected lead features, improving diagnostic accuracy.

Main Results:

• The GA-LSLO framework achieved high accuracy in detecting arrhythmia (99.65%) and myocardial infarction (97.62%) using optimized ECG lead combinations and segment lengths.
• The method demonstrated suitability for portable devices, with successful hardware implementation verified using Raspberry Pi.
• The algorithm complexity was minimized while maintaining high classification accuracy, making it ideal for real-time applications.

Conclusions:

• The proposed GA-LSLO method offers an efficient and accurate approach for cardiovascular disease detection, suitable for portable ECG devices.
• Automatic optimization of ECG leads and segment length significantly reduces computational load without compromising diagnostic performance.
• This research paves the way for more accessible and effective remote cardiovascular health monitoring.