A Deep Learning Approach Toward Differentiating Left versus Right for Idiopathic Ventricular Arrhythmia Originated...

Area of Science:

• Cardiology
• Medical Imaging
• Artificial Intelligence

Background:

• Idiopathic ventricular arrhythmias (VA) are common cardiac conditions, often originating from the complex outflow tract (OT).
• Accurate diagnosis of VA origin is challenging due to anatomical complexity and similar ECG features between right and left OT origins.
• Misdiagnosis can complicate treatment decisions, particularly for catheter ablation procedures.

Purpose of the Study:

• To develop a method for detecting the site of origin for VAs originating from the outflow tract.
• To improve diagnostic accuracy for guiding catheter ablation treatment.
• To leverage deep learning for non-invasive VA diagnosis.

Main Methods:

• Utilized a public dataset of 334 patients with idiopathic VA.
• Processed standard 12-lead ECG data into one-dimensional (1D) streams of premature beats.
• Compared the performance of 1D deep learning models: LSTM, GRU, and 1D-CNN.

Main Results:

• The 1D-CNN model achieved the highest performance.
• Achieved an accuracy of 93.4% and an F1-score of 0.9313.
• Demonstrated superior performance compared to other 1D DL models.

Conclusions:

• Deep learning, specifically 1D-CNN, effectively detects VA origin from the outflow tract.
• This DL approach surpasses conventional computerized ECG analysis methods.
• Offers a promising, non-invasive diagnostic tool for future clinical applications in guiding treatment.