Brain atrophy is a proposed MRI marker of irreversible pathologic damage in multiple sclerosis (MS). The brain parenchymal fraction (BPF) is the ratio of brain parenchymal volume to the total volume within the surface contour. We developed a semiautomated measure of BPF using commercially available edge-finding and thresholding software (30-min analysis time per patient). We measured BPF in 78 patients with MS and 17 healthy controls. BPF was lower in a cohort of patients with MS (n=50) (0.843+/-0.042, range 0.743-0.906) age-matched to controls (0.877+/-0.020, range 0.835-0.901) (p<0.001). BPF correlated inversely with third ventricular width (r=-0.785, p<0.001), and total T1 hypointense lesion volume (r=-0.347, p=0.011), but not with total T2 hyperintense lesion volume (r=-0.213, p=0.13). BPF correlated negatively with expanded disability status scale (EDSS) score (r=-0.391, p=0.0006) and disease duration (r=-0.281, p=0.01). Stepwise regression compared the relative abilities of MRI variables to predict clinical data. By regression of age, BPF, third ventricular width, T2 lesions, and T1 lesions, BPF was the best predictor of disability score (R(2)=0.204, p<0.001). Third ventricular width was the best predictor of disease duration (R(2)=0.316, p<0.001). None of the MRI variables differed between relapsing-remitting (RR) (n=60) and secondary progressive (SP) (n=18) disease course (p>0.05). The intrarater, interrater, and scan-rescan BPF variability (COV) was 0.31%, 0.34%, and 0.41% and the accuracy against a phantom was 99.1%. We conclude that whole-brain atrophy in MS can be reliably and readily quantified by a semiautomated approach. Longitudinal studies are warranted to determine if this method provides a sensitive biologic marker of the MS disease process.