Modelling of resting state fMRI
|Abstract||Multiple sclerosis is a complex disease that can affect all parts of the brain. The symptoms are varying from patient to patient and the disease is difficult to diagnose. Changes in the default mode network has been observed in patients with multiple sclerosis.|
In this thesis resting state networks are detected by using the infinite relation model. SVM and KNN is used to classify subjects into two groups: A healthy group and a group of patients with multiple sclerosis. The elements in the (link density) matrix are used as feature vector for classification. In the end the correlation between link densities and the progression in the disease is evaluated. The participants in the resting state fMRI study were 30 healthy subjects and 42 patients with multiple sclerosis. The two groups were matching in sex and gender.
The highest mean classification rate was 0.65 when using SVM and 0.61 for KNN. A large variation between the runs were found. For one run the classification rate was 0.73 when using SVM and 0.66 when using KNN. These results are comparable to classification results represented in the literature. 32 communities were detected by using the infinite relation model, and some of the communities were comparable with the default mode network, primary motor network, and the frontal network presented in the. One community seems to be comparable with both the primary visual and the extra-striate visual network. No significant correlation between EDSS and link density was found.
The results show that commonly represented resting state networks in the literature can be detected when using the infinite relation model. The best classification rates for using SVM and KNN were comparable with previous results, but the large variability in the classification rate is not optimal.
|Type||Master's thesis [Academic thesis]|
|Publisher||Technical University of Denmark, DTU Compute, E-mail: email@example.com|
|Address||Matematiktorvet, Building 303-B, DK-2800 Kgs. Lyngby, Denmark|
|Note||DTU Supervisor: Lars Kai Hansen, firstname.lastname@example.org, DTU Compute, and Kasper Winther Andersen|
|BibTeX data|| [bibtex]|
|IMM Group(s)||Intelligent Signal Processing|