Artificial Intelligence-Enabled 8-Channel ECG Diagnosing of Abnormalities with Wide QRS Complexes

Area of Science:

• Cardiology
• Artificial Intelligence
• Medical Diagnostics

Background:

• Existing research on AI for electrocardiogram (ECG) abnormalities primarily uses the 12-channel format.
• There's a gap in studies focusing on differentiating ECG abnormalities with wide QRS complexes using simplified channel formats.

Purpose of the Study:

• To develop and evaluate an AI model for classifying ECG abnormalities with wide QRS complexes.
• To compare the performance of AI models using 4, 8, and 12-channel ECG formats.
• To assess the AI model's generalizability across different datasets.

Main Methods:

• A convolutional neural network (CNN) was developed using an 11,808-ECG dataset from Tongji Hospital.
• The CNN was trained and tested on 4, 8, and 12-channel ECG formats.
• Model performance was evaluated using F1 score, AUROC, and accuracy, and compared against 6 cardiologists' diagnoses.

Main Results:

• The 8-channel format achieved the highest performance, with 95.0% accuracy, a mean F1 score of 0.969, and a mean AUROC of 0.997.
• The AI model outperformed cardiologists, achieving higher accuracy and AUROC scores.
• The model demonstrated good generalizability on external datasets (JX-Test set and public validation data).

Conclusions:

• The AI model accurately distinguishes various ECG abnormalities with wide QRS complexes in normal heartbeats.
• The 8-channel format provides a robust and efficient input for AI-driven ECG analysis.
• This AI model serves as a foundation for AI-aided clinical decision-support systems in ECG differential diagnosis.