Classification of Multiple Sclerosis Patients From Geometry of White Matter Lesions

Multiple Sclerosis (MS) is a debilitating disease that affects the brain's white matter, the 'wires' that connect neurons in gray matter.  Doctors track disease progress with Magnetic Resonance Imaging (MRI) scans, which show the type and location of lesions in the white matter.  The MRI scans are scored qualitatively, and if any quantitative measure is obtained, it is simply the total "lesion load", the volume of all lesions in the brain.  MS is a heterogeneous disease, and different subtypes of the disease can have dramatically different treatment plans.  Currently, MS subtype is determined by a combination of clinical scores and subjective rating of lesion images.  This project is concerned with using the geometry of individual lesions to better explain and predict MS subtype.

In this project the student will extract geometric features on each lesion (e.g. volume, surface area, capiler distance) as well as other aspects of the brain image data to build a rich collection of features on each subject.  The project will then require the use of supervised learning methods on these features to predict the sub-type of MS each patient. There are existing Matlab tools to compute geometric features, the ideal student will have  familiarity with Matlab; experience with image is also a plus.  The end result of this project will contribute to more objective diagnosis and classification of MS patients.

Supervisors: Tom Nichols, Prof E-W Radue (Medical Image Analysis Centre, Basel University Hospital)