Identification of Arrhythmia by Using a Decision Tree and Gated Network Fusion Model.

Area of Science:

• Cardiology
• Artificial Intelligence
• Signal Processing

Background:

• Deep learning models (DNN) show promise in cardiac arrhythmia detection.
• Clinician confidence is limited by the lack of interpretability in current DNN methods.
• There is a need for reliable and transparent deep learning solutions in arrhythmia diagnosis.

Purpose of the Study:

• To develop a novel deep learning model for arrhythmia recognition that enhances interpretability.
• To improve clinician confidence in automated arrhythmia detection systems.
• To fuse Gate Recurrent Unit (GRU) and decision tree models for robust cardiac rhythm analysis.

Main Methods:

• A hybrid model, T-GRU, was designed, integrating GRU for time-frequency analysis and decision trees for probability analysis.
• The model incorporates multipathway processing of time-frequency domain features.
• GRU model parameters were regularized, and weights were controlled to enhance decision tree output.

Main Results:

• The T-GRU model achieved high performance on the MIT-BIH arrhythmia database.
• Low-frequency band features were identified as dominant predictors in the model.
• Achieved accuracy of 98.31%, sensitivity of 96.85%, specificity of 98.81%, and precision of 96.73%.

Conclusions:

• The T-GRU fusion model demonstrates high reliability and clinical significance for arrhythmia detection.
• The model's interpretability component addresses limitations of traditional deep learning methods.
• This approach offers a promising direction for trustworthy AI in clinical cardiology.