Abstract
Machine learning (ML) has a remarkable ability to solve many key tasks in medical image analysis from restoration, reconstruction, segmentation to image synthesis or classification. While DL results reported so far are impressive, serious reservations have been raised regarding their robustness to domain shifts and to which extent we can trust their output. I will first overview the different aspects of “responsible” ML which aims at reinforcing the trustworthy behavior of models, a key need in the adoption of deep learning for healthcare applications. I will then briefly present our contributions in that context with focus on the tasks of automated quality control and uncertainty estimation in neuroimage analysis and show its translation to two different applications: early brain development and support evaluation of multiple sclerosis patients.

Biosketch
Meritxell Bach Cuadra received the MSc in Electrical Engineering from the Universitat Politècnica de Catalunya (UPC) in 1998, the PhD degree from the Ecole Polytechnique Fédérale de Lausanne (EPFL) in 2003 and was postdoctoral fellow in the Signal Processing Laboratory till 2005. She then joined the CIBM Center for Biomedical Imaging as research staff scientist in the Signal Processing Core at the Lausanne University Hospital (CHUV) and was a lecturer at the School of Biology and Medicine of the University of Lausanne (UNIL). In March 2011 she became a Senior Lecturer & has been a Privat Docent since 2018 and Since 2020 Meritxell Bach Cuadra is head of the CIBM Signal Processing CHUV-UNIL Computational Neuroanatomy & Fetal Imaging Section, link. In October 2025 she was promoted to Associate Professor at Faculty of Biology and Medicine of UNIL and she teaches and leads research activities in the Medical Image Analysis Laboratory (MIAL), UNIL hosted in the CHUV Radiology Department. Her research interests are focused on novel image processing and safe machine learning-based medical image analysis. Her research aims are to ensure the trustworthy behaviour of machine learning to support diagnosis and prognosis, lead social conscious machine learning methods to tackle biases in healthcare, together with efforts in translational research, dissemination, and access of advanced medical technology through domain-shift robust and reproducible large-scale/longitudinal validation of the developed image analysis methods. Her major research projects are applied to paediatric brain MRI analysis, lesion segmentation and classification in Multiple Sclerosis and eye MR image analysis.