Institutional Repository [SANDBOX]
Technical University of Crete
EN  |  EL

Search

Browse

My Space

Multi-subject fMRI processing via clustering techniques

Psychountaki Margarita-Antonia

Full record


URI: http://purl.tuc.gr/dl/dias/475183F5-56D0-4B0F-A949-B3A560A760AA
Year 2021
Type of Item Diploma Work
License
Details
Bibliographic Citation Margarita-Antonia Psychountaki, "Multi-subject fMRI processing via clustering techniques", Diploma Work, School of Electrical and Computer Engineering, Technical University of Crete, Chania, Greece, 2021 https://doi.org/10.26233/heallink.tuc.89769
Appears in Collections

Summary

Functional magnetic resonance imaging (fMRI) is a noninvasive method which provides significant insight into brain functionality. It measures brain activity by detecting changes associated with blood flow. The most common form of fMRI measures the blood-oxygen-level dependent (BOLD) signal, which is created by changes in blood flow in the brain. Since brain activity is intrinsic, any brain region will have spontaneous fluctuations in BOLD signal. Thus, even at rest, the BOLD signal reflects systematic fluctuations in the regional brain activity. It is widely believed that resting-state networks are the cause of these systematic fluctuations. Many studies acclaim that these networks are common, spatially, across subjects, and do not overlap each other. That is, there is a common brain parcellation across subjects, based on functionality.In this Diploma thesis, we study a method where the authors compute a common (over many subjects) brain parcellation, with no prior knowledge on the properties of the parcels. At first, the proposed method uses the subjects' resting state fMRI data and computes an orthonormal basis for the subjects' common spatial subspace via Generalized Canonical Correlation Analysis (gCCA). Then, an estimate of the common whole-brain parcellation is obtained as the solution of a semi-orthogonal nonnegative matrix factorization problem. We test this approach on synthetic and real data. Further, we study the preprocessing steps which are applied to raw fMRI data, in order to ensure that they do not violate the basic assumptions of the model.

Available Files

Services

Statistics