NeuroCardioSense (NCS): a time-aware fuzzy decision framework for multi-lead ECG classification and arrhythmia...

Area of Science:

• Cardiology
• Artificial Intelligence
• Signal Processing

Background:

• Accurate electrocardiogram (ECG) classification is crucial for automated arrhythmia detection and clinical decision support.
• Existing deep learning methods face challenges in jointly analyzing morphological patterns, multi-lead interactions, and temporal dependencies in ECG signals.
• This limitation hinders comprehensive representation of waveform details, rhythm dynamics, and class boundary separability.

Purpose of the Study:

• To introduce the NeuroCardioSense (NCS) framework, a novel deep learning approach for enhanced ECG signal classification.
• To address the limitations of existing methods in capturing complex ECG signal characteristics.
• To improve the accuracy and robustness of automated arrhythmia detection systems.

Main Methods:

• Developed the NeuroCardioSenseNet (NCSN) using a convolutional neural network (CNN) backbone with Time-Aware Gated Convolution (TAG-Conv) and Time-Aware Gating Mechanism (TAGM).
• Introduced NeuroCardioSenseNet-Fusion (NCSNF), an enhanced variant incorporating a Time-Fuzzy Integration Module (TFIM) for improved class boundary ambiguity mitigation.
• Validated the framework on the MIT-BIH Arrhythmia Database.

Main Results:

• NCSN achieved 98.77% intra-patient and 87.82% inter-patient accuracy.
• NCSNF further improved performance to 99.16% intra-patient and 90.85% inter-patient accuracy.
• The NCS framework demonstrated superior performance compared to existing baseline methods.

Conclusions:

• The NCS framework effectively addresses the limitations of current deep learning models in ECG analysis.
• NCSN and NCSNF significantly enhance the joint characterization of morphological patterns, multi-lead interactions, and temporal dependencies in ECG signals.
• The proposed methods offer a promising advancement for automated arrhythmia detection and clinical decision support.