Image processing tools for quantitative analysis of magnetic resonance neuroimaging and application to amyotrophic lateral sclerosis.

摘要

The diagnosis and evaluation of neurodegenerative disease using MRI is qualitative, subjective, and experience-based. Such conventional approaches are low in information to data ratio and do not provide quantitative markers for disease evaluation. Currently, the tools needed for quantitative MRI (qMRI) processing are not adequate for routine clinical usage. Thus, there is a need for image analysis tools for clinical applications and trials using qMRI. This work consists of two main parts. (a) Software engineering of image processing tools to enhance and integrate existing advanced registration tools for processing multi-modality MRI in parallel on a supercomputer and development of an efficient method to automatically estimate thickness distribution of an anatomical structure in a medical image. (b) Application of the developed tools to a cross-sectional and longitudinal multi-modality qMRI study of amyotrophic lateral sclerosis (ALS). The thickness estimation tool was validated using basic shapes with known geometric dimensions, sample knee and brain MRI. Validation of the tools for their effectiveness for clinical research applications was performed via a group comparison of mild cognitive impaired (MCI) subjects with known pathology and normal control subjects. When applied to ALS cross-sectionally, differences were observed between ALS and normal controls for cortical thickness, measures of brain volume, T2, and diffusion tensor imaging (DTI). Longitudinally, the results were mixed. In general, changes due to disease were subtle, which strongly support the need for an integrated multimodality approach for detection of neurodegenerative diseases, such as ALS.