A Novel Image Based Method for Detection and Measurement of Gall Stones

Accurate gall bladder segmentation is a challenging task because the surrounding tissues present in computed tomography (CT) images have densities similar to that of the gall bladder tissue and lesions at the edges of gall bladder. This work focuses on delineation of gall bladder contours on CT images whose contrast is enhanced. This gall stone detection scheme exploits a snake algorithm using Fuzzy C-means clustering (FCM) to extract the features. To improve the performance of the gall bladder contour segmentation, Sobel operator is used an edge map, followed by a template based modification using concavity removal algorithm. The unwanted edges are eliminated inside the gall bladder to obtain a modified edge map. The contour of the gall bladder is obtained using Support Vector Machine (SVM) algorithm. The segmentation of the adjacent region is done on part-by-part basis so that the result is constrained on the segmentation algorithm. Five hundred two‐dimensional gall bladder images from which 400 CT images with cholecystitis spreading through the gall bladder were delineated using SVM whose optimal parameters are inferred using chaotic whale optimisation algorithm (CWOA). This method detects sizes corresponding to small (S), medium (M) and large (L) sized gall stones in comparison with the normal gall bladder condition. An opinion from the radiologist is taken for manual evaluation. The difference ratio, defined as the ratio of percentage of mismatched detection between the algorithm and the radiologist’s. This value obtained to be 1.9%.