On Visualization and Quantification of Lesion Margin in CT Liver Images

Abstract

Cancer is the one of the leading causes of death worldwide, and cancer incidence increases every year. The analysis of lesion margin is quite important to diagnose malignant and benign masses and to detect the presence and the stage of tumor invasion in case of cancer. Accordingly, the aim of the study is to visualize and quantify margin of lesions on radiological images by means of a digital computer. In this study, computed tomography (CT) images of liver have been employed for analysis because the liver has crucial tasks in our body and liver cancer-related deaths is ranked as the forth among the cancer-related deaths. The proposed method consisted of four main steps: image cropping and smoothing, specification of target lesion, the boundary detection of target lesion, and visualization and quantification of margin. First, the images were converted to gray scale. The blank regions surrounding the liver in the CT images were removed before specification of target lesion, and further were smoothed with a bilateral filter. Next, the target region was specified roughly by drawing it manually. The boundary of lesion was more precisely determined with the active contour method employing the sketched borderline as the initial curve. Next, the properties of the target region: the centroid, major axis length, and the orientation values were computed. The intensities along a line passing through the center of the tumor were obtained for eighteen different rotation angles. A pulse model was fit to each of the intensity signal corresponding to a rotation. Then, the intensity change, margin sharpness and width were acquired from the pulse approximation associated to each rotation angle. The level difference provided the intensity change, the slope of edges gave the margin sharpness, and distance between the start and end points of the pulse edge represented margin width. Besides, the inner (core) and outer diameter with respect to angle were also displayed.