This project first explores how the number of electrodes used to record ECG signals from the body surface affects the ability to accurately reconstruct electrical activity on the heart. After the first investigation is finished, a second question is to investigate which part of the body has the key impact on the heart signal reconstruction accuracy.
Reconstructing heart surface electrical signals from body surface ECG measurements is a powerful noninvasive method for evaluating heart rhythm abnormalities. However, clinical and wearable applications often face limitations in how many electrodes can be used due to cost, comfort, and practicality. This project addresses the critical question: How few sensors can we use while still maintaining accurate reconstruction of heart activity? The student will work with a physics-constrained deep learning model that links body surface measurements to the heart’s electrical behavior. By systematically reducing the number of body surface electrodes in simulated datasets, the student will help investigate how reconstruction accuracy trends as sensor quantity decreases.
In the second phase of the project, the student will contribute to identifying which body surface regions provide the most valuable information. A machine learning model is to be built to highlight key electrode locations that contribute most to accurate heart signal prediction. The findings could directly inform the design of more efficient ECG systems for clinical and wearable use.
• Good programming capability (e.g., Python, PyTorch)
• Basic knowledge in deep learning and neural networks
• Interest in healthcare problems
• Curiosity to learn new techniques in AI and healthcare
1. Understand how noninvasive techniques are used to study the heart.
2. Learn how the quantity and location of sensor data affect model performance.
3. Gain experience in scientific reasoning and health analysis.
4. Explore how physics-based and deep learning models can work together in healthcare.