Improving deep-learning electrocardiogram classification with an effective coloring method.

Area of Science:

• Cardiology
• Medical Informatics
• Data Science

Background:

• Cardiovascular diseases, especially arrhythmias, are a major global cause of death.
• Electrocardiogram (ECG) analysis is crucial for diagnosing cardiovascular diseases.
• Integrating demographic data into ECG analysis presents a significant challenge.

Purpose of the Study:

• To develop an innovative ECG classification method incorporating patient demographic information.
• To enhance ECG analytics and support precision medicine through a novel colorization technique.

Main Methods:

• Demographic features are mapped to the (R, G, B) color space using normalized scaling.
• Each demographic feature is assigned a distinct color, applied to different ECG leads.
• Color correlations in statistical features are preserved to maintain data relationships.

Main Results:

• Achieved 1%-6% improvement in area under the ROC curve performance on the PTB-XL dataset.
• Demonstrated superior performance in multiclass and challenging ECG classification tasks.
• Combined color and waveform features improved prediction accuracy for deep learning models.

Conclusions:

• Colorization is a promising technique for advancing ECG classification and diagnosis.
• The method enhances the prediction and diagnosis of cardiovascular diseases.
• This approach contributes to improved clinical outcomes in cardiovascular care.