Robust and accurate anomaly detection in ECG artifacts using time series motif discovery

Area of Science:

• Cardiology
• Biomedical Signal Processing
• Artificial Intelligence in Healthcare

Background:

• Electrocardiogram (ECG) anomaly detection is crucial for identifying heart abnormalities.
• Current methods struggle with differentiating ECG artifacts from genuine signals, leading to high false alarm rates.
• This inaccuracy increases physician workload and risks misdiagnosis, especially for non-specialists.

Purpose of the Study:

• To develop a novel, highly robust, and accurate ECG anomaly detection technique.
• To effectively reduce the false alarm rate in ECG analysis, particularly in the presence of artifacts.
• To create an algorithm interpretable by cardiologists and applicable to both single-lead and multilead ECGs.

Main Methods:

• Integration of expert cardiologists' knowledge with motif discovery techniques.
• Development of an algorithm where each step aligns with cardiologists' interpretation standards.
• Validation using real ECG datasets with evaluation by cardiologists.

Main Results:

• The proposed algorithm achieved near-perfect performance metrics: 100% accuracy on detection (AoD), sensitivity, specificity, and positive predictive value.
• Demonstrated a 0% false alarm rate, effectively distinguishing real ECG signals from artifacts.
• Showcased high accuracy and robustness against artifacts compared to existing methods.

Conclusions:

• The novel ECG anomaly detection method offers a significant improvement over current techniques.
• Its high accuracy and robustness, particularly in artifact-prone conditions, promise to reduce misdiagnosis and physician vigilance.
• The algorithm's clinical applicability and expert-aligned design support its integration into diagnostic workflows.