Developing disease-state biomarkers using multivariate pattern analysis of neuroimaging data

The goal of this dissertation is to demonstrate the incredible potential of multivariate pattern analysis (MVPA) techniques to examine extremely rich neuroimaging datasets. The first chapter is a comprehensive review of the use of MVPA analyses with neuroimaging datasets. Surveying this literature offers perspective on the potential of this approach in the realm of neuroimaging research and in real-world applications of the resultant work. The second chapter recounts our attempt to use these methods to find suitable neuroimaging-based biomarkers of disease-state in the preclinical phase of Huntington's disease (HD). Our results corroborate previous work demonstrating that MVPA methods are capable of detecting preclinical disease states in HD, and further extend these findings to multiple neuroimaging modalities. Additionally, MVPA-based regression methods were used to predict established estimates of years-to-disease-onset in preclinical HD individuals using a number of structural and functional neuroimaging-based measures, supporting the idea that neuroimaging-based measures could potentially detect HD disease-state progression in a quantitative fashion. The final chapter addresses the next step in the development of these disease-state biomarkers, namely to determine if this data can be used to predict longitudinal changes related to disease progression in the brain. We found that voxel-based structural MRI grey matter data can be used to create MVPA regression models capable of predicting longitudinal measures of the percent change in total brain volume, as well as the percent changes in the volumes of ROIs implicated in the disease process. Results from these studies indicate that MVPA analysis methods using information-rich neuroimaging datasets may soon provide valuable biomarkers that may be used in the preclinical phase of HD. These biomarkers may allow clinicians to make predictions regarding progressive disease states, which may be useful to help monitor the effectiveness of potential preventative treatments and therapies.