Correlated Imaging Series: Deep learning for medical image analysis: self-supervision, graphs and priors


Published September 12, 2022
Technology
Scientific application
Category

On Friday, September 16th at 13:00 CEST, Diana Mateus, LS2N laboratory / Ecole Centrale Nantes, delivers a lecture on “Deep learning for medical image analysis: self-supervision, graphs and priors,” as part of the Correlated Imaging Series by COMULIS and Euro-BioImaging.

Abstract:

This talk will address fundamental medical image analysis tasks such as classification, segmentation, registration, and ranking with deep learning methods. The focus will be on techniques that help handle issues raised by the small medical datasets typically having few or noisy annotations. In the past years, our team has proposed different advances to counter these issues, for instance, relying on self-supervised techniques, incorporating prior knowledge, or considering the image regions' topology (graph modeling) to guide the predictions. We will also discuss a new formulation of the image registration problem following the Deep Image Prior (DIP) paradigm. We will illustrate these methodological contributions in the context of several medical applications:

  • Breast cancer diagnosis (mammography and tomosynthesis)

  • Muscle volume estimation (ultrasound)

  • Survival analysis on PET images

About Diana Mateus

Diana Mateus is a Full-Professor at “Centrale Nantes,” one of the top engineering schools in France. She holds the MILCOM Chair (Multimodal Image analysis and Learning for COmputational based Medicine), awarded by the local Connect Talent program to attract excellent foreign researchers. Her current work focuses on the design of machine learning methods to improve the acquisition and analysis of medical images in the context of nuclear medicine and ultrasound imaging. Previously, Diana Mateus was a research scientist at the Technical University of Munich (Germany, 2009-2016) and the Helmholtz Zentrum (2011-2015), where she received the Laura-Bassi award. She holds a PhD in computer vision from INRIA Rhone-Alpes, (Grenoble, France, 2004-2008), funded by a European Marie Curie project. Her previous background includes an MSc. (MsC./DEA) in Automation Systems and Robotics at the LAAS (Toulouse, France, 2003-2004) and a BSc in Electronics Engineering in the Javeriana University (Bogota, Colombia 1996-2002).

Biography originally published on: https://loop.frontiersin.org/people/1136527/bio 


More news from Euro-BioImaging

Dr. Kvåle Løvmo

September 8, 2025

Ultrasound-induced sample reorientation for enhancing penetration depth in Optical Coherence Tomography

Optical Coherence Tomography (OCT) is a label-free technique that has emerged as an excellent tool for monitoring the structure and function of organoids. Nevertheless,…

T1 maps measured in rats bearing C6 brain tumor

September 8, 2025

Spotlight - Cyclodextrin-Based Organic Radical Contrast Agents for in vivo Magnetic Resonance Imaging

Contrast agents (CA) for magnetic resonance imaging (MRI) are widely used in clinical practice to improve diagnostic accuracy. Despite the success of gadolinium (III)-chelate-based…

Robert Lees is an expert in multiphoton and lightsheet microscopy based at the Octopus Facility in Harwell, part of Euro-BioImaging’s UK Node.

September 8, 2025

Meet Robert Lees, Link Scientist at Octopus Facility, Central Laser Facility, Harwell Campus

As part of our ongoing series spotlighting the talented scientists working at the Euro-BioImaging Nodes, we sat down with Robert Lees, an expert in…