Lecture: 'A Computational Approach to Medical Image Analysis'

Researcher Gabriel Bernardino will present his work on machine learning for medical image analysis, focused on specific applications in cardiology and obstetrics. Medical imaging differs from classical computer vision due to unique challenges such as scarce, biased, noisy, and incomplete datasets, lacking ground truth. Additionally, mistakes and unfairness in this field can have real-world consequences, making safety and robustness essential.

Despite the current trend toward complex models, Gabriel advocates for the use of classical models and the incorporation of prior knowledge. He will show examples of previous and current projects where his methods have helped identify the impact of certain pathologies on the cardiovascular system, enabling quicker diagnoses and a better understanding of underlying pathophysiology.

About

Gabriel Bernardino is a Ramon y Cajal fellow at the Physense group at Universitat Pompeu Fabra, specializing in medical imaging analysis. He holds an Engineering Degree in Computer Science and a Bachelor's in Mathematics from the Universitat Politècnica de Catalunya, and a Master’s in Mathematics from the University of Bonn. He completed his PhD in Information and Communication Technologies at Universitat Pompeu Fabra, in collaboration with Philips Research, as part of the MSCA Industrial Doctorate Cardiofunxion, where he used computational methods to analyze regional cardiac shape changes. He later worked as a postdoctoral researcher at CREATIS, investigating reinforcement learning strategies for medical diagnostics.