On Tuesday 29 June, Stefano Cerri will defend his PhD thesis "Computational Imaging Biomarkers of Multiple Sclerosis".
Time: 15:00
Place: Zoom, registration is required via this link
https://dtudk.zoom.us/meeting/register/u5wtf-ChqjIiH93R-TvCVpikQfChpRoNZWeq
Supervisor: Associate Professor, Koen Van Leemput,
Co-supervisor: Head of DRCMR Hartwig Siebner,
Co-supervisor: VP Clinical Trials, Annemie Ribbens, Icometrix
Assessment Comittee:
Associate Professor Tim Dyrby, DTU Health Compute
Professor Frederik Barkhof, University College London
Professor Tal Arbel, McGill University
Abstract:
Multiple Sclerosis (MS) is a chronic disease of the central nervous system, characterized by the formation of lesions in the brain and marked atrophy. MRI scans are the primary tool used by clinicians to detect lesions and atrophy patterns. Manually labeling white matter lesions and many brain structures on MRI scans is time-consuming, prone to inter-rater variability. For such reasons, many automatic tools have been proposed in the literature. However, these methods usually work on a narrow range of MRI data and fall well below the threshold of what is required in clinical applications.
In this thesis, we first focus on the development of a method for simultaneously segmenting white matter lesions and dozens of brain structures. The method is adaptive to scanner and contrast changes, allowing its use to a large variety of MRI scans. Furthermore, the method has competitive lesion segmentation performance compared to benchmark methods while being able to segment 41 brain structures.
We then extend this method to track brain changes and lesion evolution accurately on longitudinal MRI scans. This extension leads to more temporally consistent segmentations than its cross-sectional counterpart while better detecting biological changes in the brain.
We made the methods proposed in this thesis publicly available as part of the neuroimaging package FreeSurfer, with approximately 50000 licenses worldwide.