Corpus Callosum as a biomarker for Alzheimer and Multiple Sclerosis

Corpus Callosum as a biomarker for Alzheimer and Multiple Sclerosis

Sotiropoulos, G.

Lelieveldt, B. (mentor)
Van de Giessen, M. (mentor)
Smeets, D. (mentor)

Mechanical, Maritime and Materials Engineering

BioMechanical Engineering

BME

2015-01-14

With the increase of the life expectancy and the deterioration of the quality of life that neurodegenerative diseases can cause to the patients, they have been receiving more and more attention the last few decades in the developed countries. Two of the leading neurodegenerative diseases are Alzheimer’s Disease and Multiple Sclerosis. In diagnosis and assessment progression Magnetic Resonance Imaging (MRI) is playing an important role. MRI is a non-invasive tool and can achieve a great contrast between the different tissues that exist inside the brain. In the search of biomarkers for the faster diagnosis of these diseases, the Corpus Callosum appears to be an interesting brain structure, as it facilitates inter-hemispheric communication and it is not sensitive to brain hydration and dehydration effects. This thesis investigates if the regional and global shape and area changes of the Cor- pus Callosum could set the Corpus Callosum as a biomarker for Alzheimer’s Disease and Multiple Sclerosis. In order to achieve that a segmentation pipeline is proposed, so as to extract the Corpus Callosum from Magnetic Resonance brain images. After such a pipeline is constructed and validated, the quantification of the Corpus Callosum area is investigated for scans acquired in scanners of different vendors and in scanners of different magnetic field strengths, in order to gain better insights of the differences that can exist between such scans. Finally, the potential of the Corpus Callosum to be used as a biomarker is investigated, by attempting to correlate the global and regional shape and area changes of the Corpus Callosum with neurodegenerative diseases. During the thesis the segmentation algorithm showed a high segmentation accuracy performance with a mean dice of 93%, a segmentation reproducibility error of 1.93% and the clas- sification accuracy between AD and MS patients with NC was above 95% for both groups.

http://resolver.tudelft.nl/uuid:fa452ea4-3f2c-4fac-8935-c20ea12a990f

Student theses

master thesis

(c) 2015 Sotiropoulos, G.