Doctor of Philosophy (PhD)
Major Depression Disorder (MDD) is a common disease throughout the world that negatively influences people’s lives. Early diagnosis of MDD is beneficial, so detecting practical biomarkers would aid clinicians in the diagnosis of MDD. Having an automated method to find biomarkers for MDD is helpful even though it is difficult. The main aim of this research is to generate a method for detecting discriminative features for MDD diagnosis based on Magnetic Resonance Imaging (MRI) data.
In this research, representational similarity analysis provides a framework to compare distributed patterns and obtain the similarity/dissimilarity of brain regions. Regions are obtained by either data-driven or model-driven methods such as cubes and atlases respectively. For structural MRI (sMRI) similarity of voxels of spatial cubes (data-driven) are explored. For resting-state fMRI (rs-fMRI) images, the similarity of the time series of both cubes (data-driven) and atlases (model-driven) are examined. Moreover, the similarity method of the inverse of Minimum Covariant Determinant is applied that excludes outliers from patterns and finds conditionally independent regions given the rest of regions. Next, a statistical test that is robust to outliers, identifies discriminative similarity features between two groups of MDDs and controls. Therefore, the key contribution is the way to get discriminative features that include obtaining similarity of voxel’s cubes/time series using the inverse of robust covariance along with the statistical test. The experimental results show that obtaining these features along with the Bernoulli Naïve Bayes classifier achieves superior performance compared with other methods. The performance of our method is verified by applying it to three imbalanced datasets. Moreover, the similarity-based methods are compared with deep learning and regional-based approaches for detecting MDD using either sMRI or rs-fMRI.
Given that depression is famous to be a connectivity disorder problem, investigating the similarity of the brain’s regions is valuable to understand the behavior of the brain. The combinations of structural and functional brain similarities are explored to investigate the brain’s structural and functional properties together. Moreover, the combination of data-driven (cube) and model-driven (atlas) similarities of rs-fMRI are looked over to evaluate how they affect the performance of the classifier. Besides, discriminative similarities are visualized for both sMRI and rs-fMRI. Also, to measure the informativeness of a cube, the relationship of atlas regions with overlapping cubes and vise versa (cubes with overlapping regions) are explored and visualized. Furthermore, the relationship between brain structure and function has been probed through common similarities between structural and resting-state functional networks.
Mousavian, Marzieh Sadat, "Machine Learning Methods for Depression Detection Using SMRI and RS-FMRI Images" (2021). LSU Doctoral Dissertations. 5546.