UCSF

Multiple sclerosis is an idiopathic, autoimmune disease that affects the central nervous system (CNS). Imaging studies have shown that gray matter volume, rather than whole brain or white matter volume, acts as the best imaging biomarker for MS progression. Previous studies were performed via cross-sectional analysis of each time point and then interrogating the difference between values. Because of variability inherent in software tools, the population of cross-sectional analysis studies is dependent on the segmentation program being utilized with smaller standard deviations allowing for smaller subject populations, particularly when the tissue volume difference being studied is small in comparison those standard deviations. Longitudinal analysis aims to minimize that variability and give more accurate segmentation results. Segmentation in MS is also plagued by the presence of white matter lesions, whose T1 hypointensities can result in the tissue being misclassified as gray matter. Two longitudinal programs that have been validated for healthy controls and patients with Alzheimer's disease - aBEAT and FreeSurfer - were explored by retrospectively analyzing 7 sets of longitudinal data both cross- sectionally and longitudinally. A comparison between programs revealed that FreeSurfer produced more accurate both segmentation and anatomical parcellation results. Quantitative analysis of gray matter volumes also showed FreeSurfer to be superior to aBEAT with FreeSurfer's cross-sectional processing yielding the smoothest transition from time point to time point. The investigation into cortical thicknesses obtained by FreeSurfer, on the other hand, yielded slightly conflicting results between R2 values and observed longitudinal trends. Further analysis of both longitudinal processing and lesion segmentation is required to evaluate the usefulness of currently available longitudinal processing programs and to avoid the need for manual segmentation respectively. Currently, cross-sectional segmentation is the optimal method for longitudinal brain volume analysis as longitudinal segmentation programs have proven inferior rather than superior to their cross-sectional counterparts.