Integration of Multimodal MR Data |
| | Abstract | The brain is an extremely complex organ. Understanding the function and the anatomy of the brain that underlies normal and adaptive behaviours, are the foundation for understanding the pathology and the pathophysiology of neurological diseases. Both functional magnetic resonance imaging (fMRI) and
diffusion weighted resonance imaging (DWI) give an insight into how the brain works by imaging the functional and the structural connectivity of the brain, respectively. The stochastic blockmodel, the Infinite Relational Model (IRM), can be used to find the structure in relational data. It partitions the graph such that equivalent nodes are assigned to the same cluster. The optimal partition is found by maximizing the likelihood of the model. In this project the IRM is used to find the structure in fMRI and DWI data.
In this project fMRI data from 29 healthy subjects is used. During the acquisition the subjects perform finger tapping with alternating left and right hand. The pre-processing of the fMRI data is performed in SPM8, and the activated regions in the motor cortex and the striatum is found. The fMRI graphs are created by extracting the time series in the voxels in the regions of interest (ROIs), and by calculating the correlation between the time series. Three graphs for each subject are created, one for each of the states: Baseline, Left, and Right. DWI data from 14 of the subjects is used. The processing of the DWI data is performed in FSL. Probabilistic tractography is performed to find the streamlines between the voxels in the ROIs. The DWI graphs are obtained by assuming a link between two voxels if a streamline exists.
Two different analyses are performed. An analysis only regarding fMRI data and an analysis of combined fMRI and DWI data. In each of the analyses, two hypotheses are tested: 1) The partition of the ROIs is the same in the different states and 2) The communication pattern is the same in the different states, assuming the partition is the same.
When testing the first hypothesis in both analyses, it is found that the partition of the ROIs is significantly different between the different states, but despite of the significant difference, the differences are hard to interpret by visual inspection. The different partition between the states show that the IRM is highly sensitive, since, it is sensitive to the task performed. The structure found by the IRM is consistent across states, as the number of clusters only vary a little between the states. When testing the second hypothesis, it is found that the communication pattern between the clusters, found by the IRM, is significantly different in the different states in both analyses. This result suggest that the link probability matrix is highly sensitive to the task performed. The inspection of the cluster pairs with pair-wise most significant difference in link probability between the states, indicates that there is a coordination between cortex and striatum, as clusters in cortex and striatum are connected. Furthermore, it is found that the IRM is reliable, as it is stable across runs. | | Type | Bachelor thesis [Academic thesis] | | Year | 2013 | | Publisher | Technical University of Denmark, DTU Compute, E-mail: compute@compute.dtu.dk | | Address | Matematiktorvet, Building 303-B, DK-2800 Kgs. Lyngby, Denmark | | Series | IMM-B.Sc.-2013-04 | | Note | DTU supervisor: Professor Lars Kai Hansen, lkh@imm.dtu.dk, DTU Compute | | Electronic version(s) | [pdf] | | Publication link | http://www.compute.dtu.dk/English.aspx | | BibTeX data | [bibtex] | | IMM Group(s) | Intelligent Signal Processing |
|