< Back to previous page

Project

SUBJECT-SPECIFIC GRAPH MEASURES BASED ON RESTING-STATE FUNCTIONAL MRI AS POTENTIAL BIOMARKERS FOR ALZHEIMER’S DISEASE

The brain is one of the most complex systems. To understand the brain we have to understand how the brain works, in terms of how remote/non-remote brain regions effectively interact and how brain regions are organized to process information. Therefore, it is important to understand the brain as a network. To quantify the performance of brain networks, graph theory is a very suitable method. It has been proposed that graph measures of functional networks based on rs-fMRI could be used as a biomarker of brain disease for future clinical use. As a biomarker for clinical use, it should satisfy several characteristics. It majorly includes good reproducibility, low test-retest variability, easy to measure, preferably non-invasive, ability to make a distinction between disease and normal, and being able to track or predict progression at individual level. Thereupon, to answer “Can-subject-specific graph measures based on resting-state fMRI be used as potential biomarkers for brain disease”,  we orderly explore these properties of subject-specific graph measures derived from rs-fMRI as a potential biomarker for Alzheimer’s disease (AD)  diagnosis through the entire thesis.

We started to exam the technical factors (preprocessing methods and choices of network construction) influencing the reproducibility of graph measures derived from rs-fMRI in healthy young adults. The reproducibility/test-retest variability of subject-specific graph measures was quantified. Then, we moved to the biologically oriented factors (psychological traits and sex) which associated with the reproducibility of graph measures.  The next step was to explore the performance of subject-specific graph measures derived from resting-state fMRI in AD and comprehensively evaluate if they could be used as a biomarker for AD diagnosis for future clinical usage. This step was combined with a support vector machine approach. Finally, a novel technique for modularity analysis of brain networks, edge-based modularity, combined with a machine learning technique was used to classify patients in the AD continuum.

To conclude, subject-specific normalized graph measures of rs-fMRI computing through a weighted network with the weights based on the absolute (partial) correlations showed good reproducibility, especially global graph measures. They showed potential as biomarkers for diagnosis of AD, but there still is a gap before this can be used in clinical practice. 

Date:6 Oct 2017 →  19 Dec 2022
Keywords:brain network, graph theory, fMRI, resting-state network
Disciplines:Neurosciences, Biological and physiological psychology, Cognitive science and intelligent systems, Developmental psychology and ageing
Project type:PhD project