A statistical 3-D segmentation algorithm for classifying brain tissues in multiple sclerosis

The authors have previously (2001) successfully used the deterministic annealing (DA) algorithm to segment simulated magnetic resonance images (MRI) of a normal brain. This paper presents the results of applying the same algorithm to simulated and actual clinical multiple sclerosis (MS) MRI brain data with the objective of developing a computer-aided diagnostic (CAD) tool for the early detection and follow-up of MS lesions. MS lesions can be obtained by segmenting the image data in T1 simulated brain images using the DA algorithm and then performing further arithmetic manipulations on these segmented images. MS lesions in clinical T2 MRI are isolated entities in the segmented images of white matter, gray matter and cerebrospinal fluid. The achieved results demonstrate the ability of the DA algorithm to isolate MS lesions from clinical MRI data, thus providing a potential CAD tool for clinicians