ABSTRACT

Inverse problems are fundamental for inferring unknown parameters from observations but face ill posedness, high dimensionality, and noise sensitivity. This report proposes an interdisciplinary framework integrating inverse problem modelling with AI agents to address key medical engineering challenges. For neurodegenerative disease diagnosis, we develop an eye tracking fMRI data fusion model with AI agents for noise suppression and cross modal mining, enabling non invasive Alzheimer’s diagnosis and risk stratification. For cardiovascular disease, an AI driven framework infers abnormal myocardial activity from noisy ECG signals, supported by a cardiac digital twin that converts drug target interactions into a solvable inverse problem. In cardiac electrophysiology, an AI enhanced method integrates anatomical priors to improve ECG imaging accuracy. These innovations demonstrate the superiority of combining inverse problem modelling with AI agents, providing a standardised paradigm for transitioning from experience driven to data mechanism dual driven approaches.