Automatic segmentation of white matter lesions in T1-weighted brain MR images

White matter lesions are common brain abnormalities. In this paper, an automatic method for segmentation of white matter lesions in T1-weighted brain magnetic resonance (MR) images is presented. A subject's T1-weighted MR image is first segmented into the three major tissue types, white matter (WM), gray matter (GM) and cerebral spinal fluid (CSF) solely based on each voxel's intensity. Since WM lesions are typically classified as GM based on their intensity characteristics, the GM class is then separated into normal GM and WM lesions. This is accomplished using a statistical model of tissue distribution of healthy brains in a stereotaxic space. The proposed method is tested on 10 MR images with WM lesions and the results of the method are visually compared with WM lesions manually labeled by an experienced radiologist.

1. INTRODUCTION

WM lesions are common abnormalities of the brain, which may be the result of different brain diseases, such as multiple sclero-sis (MS) [1], Alzheimer's disease [2], stroke and head injury [3]. They also appear in normal elderly subjects [4]. MR imaging is sensitive to these WM changes and thus widely used as a noninvasive method for diagnosing such diseases clinically. This generates a large number of MR images to be analyzed. Manual lesion segmentation by trained experts, a commonly used method, is extremely time consuming, labor intensive, and suffers from high intra-observer and inter-observer variability, This raises the demand for automatic lesion segmentation methods which can reduce both the observers' burden and the intra-observer and inter-observer variability.