A Study Of Data Analysis And Applications Of Ultimodal Magnetic Resonance Imaging

One of the most prominent advantages using magnetic resonance imaging (MRI) is capable of conducting multiple imaging modalities (or contrasts). It has become an important probing approach in many research fields, particularly in studying neurological and psychiatric disorders. However, how to effectively employ multimodal MRI for exploring neuroimaging markers in neurological and psychiatric disorders is a field under investigated. Although MRI has been widely used, it is yet a very young technology in its early development, and the related methodologies are very limited and under developed. In this study, we made efforts to investigate these critical issues along two axes, as detailed below.In the first effort, we mainly focused on strategies of how to effectively utilize multimodal MRI for assisting investigations of neurological and psychiatric disorders. First, we introduced the fundamentals and the respective sensitivities of three MRI modalities, including functional MRI (fMRI), diffusion tensor imaging (DTI) and high-resolution Ti-weighted imaging. Second, we studied the commonly used analysis methods as well as their drawbacks for the three MRI modalities in detail. In addition, we also proposed strategies to resolve their drawbacks. Especially, we introduced an effective graph-theory analysis method for investigating the topological properties of both resting-state brain networks and anatomical networks of brain cortex; we proposed an optimal voxel-based morphometry approach to improving the accuracy in measuring the difference of diffusion derived indicies (e.g., FA-fractional anisotropy). Additionally, we also proposed a novel strategy for fiber tracking that is based on tensor-based registration for overcoming the drawbacks inherited in the traditional approach that has to be done within individual spaces. Finally, to demonstrate the effectiveness of our proposed strategy, we carried out a specific study using multimodal MRI data of stroke patients as an application instance. Comparing with previous studies, we observed many vital new findings, which may provide valuable neuroimaging markers for the pathophysiological fundamental of different outcomes in hand function after subcortical stroke.In a second effort, we studied two methodological issues on MRI data analysis.(1) We know that negative correlations constantly exist in functional connectivity analysis. Unfortunately, the physiological underpinnings of negative correlations are not exactly clear. For reconstructing the brain network, negative correlations are usually treated using either the following two strategies:(a) adopting the absolute value of a negative correlation, or (b) setting all negative correlations to zero. However, little is known concerning the effect of taking the two different strategies for dealing with negative correlations on the topological properties of brain network. We therefore for the first time examined the differences in the topological properties of the resulting brain networks reconstructed using the two strategies. The work not only provides insight into the role of negative correlations in configuring the topology of brain functional network, but also offers a new view for studying the negative correlations.(2) Using traditional single tensor model (TSTM), we often observed extremely low FA at locations with edema in the brain, which might be partly attributed to the partial volume effect induced from edema. In fact, anisotropic tissue may also exsit at the loacations of edema. To resolve this issue, we proposed a partial-volume tensor model (PVTM) to approximate and model the diffusion weighted signals, and the model was applied in the diffusion-weighted datasets collected from stroke and glioma patients, respectively. We found that our PVTM can effectively enhance the FA measurement in the ischemic stroke and glioma tumor regions, significantly greater than that calculated using TSTM, which would therefore facilitate the success of fiber tracking within such regions. Moreover, the f map generated from PVTM can well display the lesion. Our results demonstrated that PVTM can isolate the component of edema from lesion to some extent and might be more suitable than TSTM for measuring diffusion properties in the regions of edema.