RESUMO
Although magnetic resonance imaging (MRI) is a valuable aid in the initial diagnosis of multiple sclerosis (MS), quantitatively MRI has been disappointing in staging and evaluating therapy protocols by means of serial examinations. In this study, image processing algorithms were developed for the global analysis of MR images of the cerebrum. Limited three-dimensional segmentation was achieved through histogram analysis by these algorithms, which are essentially operator independent. The effects of coil response and tip angles, patient positioning, and interslice gap thicknesses were examined for 10 MS patients with repeated examinations for a total of 72 images. Effects of technique and instrumentation errors were approximately 6%, and agreement between two independent operators for measuring the total MR pixel sum from periventricular effusions and intense MS plaques was better than 97% with a standard deviation of 2.9%.
