What is the clinical evidence from this study?

Study design: Other. Population: Arrhythmia. Intervention: Explainable deep learning approach (modified CNN). Primary outcome: Accuracy of arrhythmia detection.

What does this research mean for the field?

An explainable deep learning approach utilizing a modified convolutional neural network for ECG-based arrhythmia detection achieves 99.43% accuracy and near-perfect F1-scores on the MIT-BIH database while providing clinically understandable interpretations. Novelty: ClaimNovelty.METHODOLOGICAL. Consensus alignment: ConsensusAlignment.NEUTRAL.

January 14, 2025Technologies11 citationsOpen Access

Towards Transparent AI in Medicine: ECG-Based Arrhythmia Detection with Explainable Deep Learning

OKOleksii KovalchukKhmelnytskyi National University OBOlexander BarmakKhmelnytskyi National University PRPavlo RadiukKhmelnytskyi National University

Key Result

An explainable deep learning approach using a modified CNN for ECG-based arrhythmia detection achieved an accuracy of 99.43% and F1-scores approaching 100% on the MIT-BIH database.

Structured PICO

Population

ECG records from the MIT-BIH database

Intervention

Explainable deep learning approach for arrhythmia detection using enhanced R peak detection, a modified CNN architecture processing a triad of cardio cycles, and an interpretation method

Outcome

Accuracy and F1-scores for arrhythmia classificationsurrogate

An explainable deep learning model utilizing a modified CNN and enhanced R peak detection achieved over 99% accuracy in ECG-based arrhythmia detection, offering high performance with clinical transparency.

Abstract

Cardiovascular diseases are the leading cause of death globally, highlighting the need for accurate diagnostic tools. To address this issue, we introduce a novel approach for arrhythmia detection based on electrocardiogram (ECG) that incorporates explainable artificial intelligence through three key methods. First, we developed an enhanced R peak detection method that integrates domain-specific knowledge into the ECG, improving peak identification accuracy by accounting for the characteristic features of R peaks. Second, we proposed an arrhythmia classification method utilizing a modified convolutional neural network (CNN) architecture with additional convolutional and batch normalization layers. This model processes a triad of cardio cycles—the preceding, current, and following cycles—to capture temporal dependencies and hidden features related to arrhythmias. Third, we implemented an interpretation method that explains CNN’s decisions using clinically relevant features, making the results understandable to clinicians. Using the MIT-BIH database, our approach achieved an accuracy of 99.43%, with F1-scores approaching 100% for major arrhythmia classes. The integration of these methods enhances both the performance and transparency of arrhythmia detection systems.

Ask AI

Helpful

Bookmark

View Full Paper