Automatic plaque characterization and vessel wall segmentation in magnetic resonance images of atherosclerotic carotid arteries

Composition and structure of atherosclerotic plaque is a primary focus of cardiovascular research. In vivo MRI provides a means to non-invasively image and assess the morphological features of atherosclerotic and normal human carotid arteries. To quantitatively assess the vulnerability and the type of plaque, the contours of the lumen, outer boundary of the vessel wall and plaque components, need to be traced. To achieve this goal, we have developed an automated contour detection technique, which consists of three consecutive steps: firstly, the outer boundary of the vessel wall is detected by means of an ellipse-fitting procedure in order to obtain smoothed shapes; secondly, the lumen is segmented using fuzzy clustering. The region to be classified is that within the outer vessel wall boundary obtained from the previous step; finally, for plaque detection we follow the same approach as for lumen segmentation: fuzzy clustering. However, plaque is more difficult to segment, as the pixel gray value can differ considerably from one region to another, even when it corresponds to the same type of tissue. That makes further processing necessary. All these three steps might be carried out combining information from different sequences (PD-, T2-, T1-weighted images, pre- and post-contrast), to improve the contour detection. The algorithm has been validated in vivo on 58 high-resolution PD and T1 weighted MR images (19 patients). The results demonstrate excellent correspondence between automatic and manual area measurements: lumen (r=0.94), outer (r=0.92), and acceptable for fibrous cap thickness (r=0.76).