Ventricular Arrhythmia Classification Using Similarity Maps and Hierarchical Multi-Stream Deep Learning.

Area of Science:

• Cardiology
• Biomedical Engineering
• Artificial Intelligence in Medicine

Background:

• Ventricular arrhythmias, including ventricular tachycardia (VT) and ventricular fibrillation (VF), are primary causes of sudden cardiac death.
• Accurate classification of these arrhythmias is critical for timely intervention and therapeutic development.

Purpose of the Study:

• To develop and evaluate a novel method for discriminating between VT, VF, and non-ventricular rhythms (NVR).
• To improve the accuracy of ventricular arrhythmia detection and classification using advanced machine learning techniques.

Main Methods:

• Development of novel 'similarity maps' to capture ECG trace regularity.
• Integration of similarity maps with features from learnable Parzen band-pass filters and derivative features.
• Implementation of a hierarchical multi-stream ResNet34 architecture for feature fusion and classification.

Main Results:

• Similarity maps significantly enhance the accuracy of distinguishing between VT and VF.
• The proposed approach achieved an overall average class sensitivity of 89.68%.
• Individual class sensitivities: 81.46% for VT, 89.29% for VF, and 98.28% for NVR.

Conclusions:

• The developed method demonstrates high accuracy in detecting and classifying ventricular arrhythmias.
• This advancement holds significant potential for improving patient outcomes and advancing translational medicine in cardiology.