CAD Scheme for differential diagnosis of lacunar infarcts and normal Virchow-Robin spaces on brain MR images

Abstract

The detection of asymptomatic lacunar infarcts on magnetic resonance (MR) images is important because their presence indicates an increased risk of severe cerebral infarction. However, accurate identification of lacunar infarcts on MR images is often hard for radiologists because of the difficulty in distinguishing lacunar infarcts and enlarged Virchow-Robin (VR) spaces. The purpose of this study was to develop a computer-aided diagnosis (CAD) scheme for the detection of lacunar infarcts on MR images. Our database consisted of T1- and T2- weighted images acquired using 1.5 T MR scanner. These images were obtained from 109 patients. Another database consisted of T1- and T2-weighted images and MRA images acquired using 3.0 T MR scanner. These images were obtained from 8 patients. We first developed a method for classification of lacunar infarcts and enlarged VR spaces by using the former database. The lesion was segmented using white top-hat transform and thresholding techniques. Image features, such as size, shape, and signal intensity, were determined from the segmented lesion. A neural network with image features was employed for distinguishing between lacunar infarcts and enlarged VR spaces. The result indicated that the area under the ROC curve was 0.945. We also developed a method for making a fusion image of T2-weighted image and MRA by using the latter database. Image registration was used to achieve a matching between T2-weighted image and MRA. Thresholding and region growing techniques were used for segmenting vessel regions in MRA. The blood flow obtained from MRA was superimposed on the lesion in T2-weighted image. Blood flow on the lesion was a crucial piece of information for the diagnosis of VR spaces. Our CAD schemes would be useful in assisting radiologists for the detection of lacunar infarcts in MR images.