Research Questions:
1. Can a multimodal deep learning model integrating electrocardiogram (ECG), clinical history, and cardiac biomarkers (e.g., troponin) improve the early diagnosis of myocardial infarction (MI)?
2. How does the diagnostic performance of the multimodal model compare to single-modality models (e.g., ECG-only or troponin-only)?
3. What is the contribution of each modality to the final diagnosis, and can the model provide interpretable insights for clinicians?
Research Aims:
This project aims to develop and validate a multimodal deep learning framework for the early and accurate diagnosis of myocardial infarction by integrating heterogeneous clinical data sources. The model will combine raw ECG signals, structured clinical history (including risk factors and symptoms), and laboratory biomarker levels to enhance diagnostic precision. The project will also assess the relative contribution of each modality, evaluate model performance using real-world datasets, and explore explainable AI techniques to ensure clinical interpretability. Ultimately, this work seeks to demonstrate that multimodal learning can provide a more robust and timely decision support tool for emergency and cardiology settings.