About My Project
Project Title
EcgNet: A Hybrid Multimodal Deep Learning Approach for Cardiovascular Disease (CVD) Diagnosis
Problem
EcgNet solves the shortage of fast, accurate, and explainable ECG diagnosis—especially in places where expert doctors aren’t available.
Approach
The project follows a five-phase pipeline:
- Clean raw ECG data using Butterworth filters and wavelet denoising to remove noise and artifacts.
- Train three separate models (1D-CNN, 2D-CNN, and LSTM) to learn different types of patterns from the ECG data.
- Combine the outputs of all three models to make a more accurate and robust diagnosis.
- Use SHAP and LIME to explain why the AI made a specific prediction, improving trust and transparency.
- Test the system on real ECG datasets (like MIT-BIH) to measure accuracy and ensure clinical relevance.
Expected Outcome
By the end of the program, the project will result in a working prototype of EcgNet: a multimodal AI system capable of diagnosing cardiovascular conditions from ECG signals. The final deliverables will include a trained model combining 1D-CNN, 2D-CNN, and LSTM architectures, along with explainable AI outputs using SHAP and LIME to visualize the reasoning behind each diagnosis. The system’s performance will be benchmarked using the MIT-BIH dataset, with metrics such as accuracy, precision, recall, and F1-score. A poster presentation will summarize the model architecture, preprocessing pipeline, fusion strategy, and evaluation results. This project will contribute toward scalable, interpretable solutions for AI-assisted cardiac diagnostics in resource-limited settings.
Final Report: View PDF