Beat-Level Interpretation of Intra-Patient Paradigm Based on Object Detection

Area of Science:

• Cardiology
• Biomedical Engineering
• Artificial Intelligence in Medicine

Background:

• Electrocardiogram (ECG) is crucial for diagnosing heart conditions.
• Manual analysis of extensive ECG data is time-consuming and burdensome for clinicians.
• Automated ECG analysis is essential for efficient cardiac abnormality detection.

Purpose of the Study:

• To develop an automated, beat-level ECG interpretation method.
• To improve the efficiency and accuracy of arrhythmia diagnosis.
• To reduce the reliance on manual R-peak detection and separate heartbeat classification.

Main Methods:

• Utilized an object detection model for unified QRS complex localization and heartbeat classification.
• Employed an automatic annotation algorithm for ground truth generation.
• Trained the model on the MIT-BIH dataset for interpretation of 12 heartbeat types.

Main Results:

• Achieved superior performance compared to existing advanced methods.
• Attained an average accuracy of 99.60% for heartbeat type interpretation.
• Demonstrated high average sensitivity (97.56%) and specificity (99.78%).

Conclusions:

• The proposed method effectively automates ECG interpretation at the beat level.
• This approach offers a robust clinical auxiliary tool for arrhythmia diagnosis.
• The unified object detection framework enhances diagnostic accuracy and efficiency.