Tissue Segmentation in MRI as an Informative Indicator of Disease Activity in the Brain

The presented tissue segmentation technique is based on a multispectral analysis approach. The input data were derived from high resolution MR images. Usually, only two inputs, proton density (PD) and T2-weighted images, are utilized to calculate the 2D feature map. In our method, we introduced a third input, T1-weighted MR image, for segmentation based on 3D feature map. k-Nearest Neighborhood segmentation algorithm was utilized. Tissue segmentation was performed in phantoms, normal humans and those with brain tumors and MS. Our technique utilizing all three inputs provided the best segmentation (p<0.001). The inclusion of T1 based images into segmentation produced dramatic improvement in tissue identification. Using our method, we identified the two distinctly different classes of tissue within the same MS plaque. We presume that these tissues represent the different stages involved in the evolution of the MS lesions. Further, our methodology for measuring MS lesion burden was also used to obtain its regional distribution as well as to follow its changes over time. The segmentation results were in full accord with neuropsychological findings.