Computerised volumetric analysis of lesions in multiple sclerosis using new semi-automatic segmentation software

The paper describes the application of new semi-automatic segmentation soft ware to the task of detection of anatomical structures and lesion and their three-dimensional (3D) visualisation in 23 patients with secondary progres sive multiple sclerosis (MS). The purpose is to study the correlation betwe en magnetic resonance imaging (MRI) parameters (volumes of plaques and cere brospinal fluid spaces) and clinical deficits (neurological deficits in the form of EDSS and RFSS scores, and neuropsychological deficits). The softwa re operates in PC/Windows and PC/NeXTstep environments and utilises graphic al user interfaces. Quantitative accuracy is measured by performing segment ation of fluid-filled syringes (relative error of 1.5%), and reproducibilit y is measured by intra- and inter-observer studies (3% and 7% variability, respectively). The mean volumes of MS plaques show significant correlations with the total RFSS scores (p=0.04). Relative intracranial cerebrospinal f luid (CSF) space volumes show statistically significant correlation with ED SS scores (p=0.01). The mean volume of MS plaques shows a significant corre lation with the overall neuropsychological deficits (p=0.03). 3D visualisat ion helps to understand the relationship of lesions to the surrounding brai n structures. The use of semiautomatic segmentation techniques is recommend ed in the clinical diagnosis of MS patients.