Research
Explainable AI-Based ECG Heartbeat Classification
An IEEE-published research project addressing the interpretability gap in AI-driven ECG classification.
Explainable AIECGSHAPLIME
Overview
Co-authored a paper on explainable ECG heartbeat classification for IEEE AISP 2024.
Framed the work around the clinical trust problem in black-box deep learning systems.
Combined performance with interpretability using a CNN-LSTM-Attention ensemble on MIT-BIH data.
Results
98.25% classification accuracy on MIT-BIH arrhythmia benchmark
Published and indexed in IEEE Xplore (AISP 2024)
3 interpretability methods integrated — SHAP, LIME, Grad-CAM
Key Outcomes
Reached 98.25% accuracy while preserving interpretable decision support.
Integrated SHAP and LIME for transparent prediction analysis.
Focused on making AI-driven ECG classification more clinically applicable.
Need the full implementation story?
Walkthrough, private repo access, or implementation details available.