MED-XAI · Design and development of explainable DL/ML model for medical image analysis with high accuracy and reliability

Despite significant advances in artificial intelligence (AI) for medical imaging, healthcare providers face a critical challenge: current deep learning models function as “black boxes” producing results without offering insight into how these conclusions are reached1. This lack of transparency limits clinical trust, hinders adoption in real-world practice, and ultimately restricts the potential of AI to improve patient care. MED-XAI project focuses on solving a major problem (Explainability and Interpretability) with the analysis of medical images in an AI-based healthcare system . The purpose of MED-XAI is to create an Explainable AI (XAI) model applicable to the diagnosis of Pulmonary Hypertension (PH) based on medical imaging data, making AI results easier for doctors to understand and trust. It combines advanced interpretation methods (SHAP, LIME, Grad-CAM, Bayesian inference) to provide clearer, more reliable insights that can be applied in real clinical practice. The innovation in the multi-method XAI integration is combined with the relatively new condition, PH, at a risky stage and is not widely studied. This research aligns with EU missions in digital health and responsible AI, which provides not only scientific progress but also practical impact. I will be based at the NOVA Medical School (NMS)’s Comprehensive Health Research Centre (CHRC), a multidisciplinary and multi-institutional research unit committed to innovation across medicine, public health, elderly care, nutrition, mental health, and technology. The project will be supervised by Professor Jorge MENDES, an expert in medical imaging, large data processing and with privileged access to healthcare providers' data.