Improving active shape models performance in low-contrast images using a KNN-based search algorithm - with applications in liver segmentation

Active Shape Model (ASM) is considered as a high level image processing algorithm. Typical applications include image segmentation and interpretation. A major challenge in ASMs is to repeatedly move model points towards true boundaries. It is a crucial step in the algorithm which fails in cases of low contrast images. In this paper, we present a new search algorithm for ASM to tackle segmentation of tissues with nearby organs of similar intensities. We train a KNN classifier and employ it to label the region surroundings each mesh point and move the point towards the boundary. Thus, evolution of the initial surface is performed faster in a single step. Evaluation of the proposed method was carried out by Dice and Jaccard similarity measure and accuracy index. The results of segmentation were compared with the results of Active Contour Model and conventional ASM. The Dice (Jaccard) indices are 0.93 (0.87), 0.85 (0.73) and 0.9 (0.76) for our method, conventional ASM and ACM, respectively. Moreover, the accuracy is increased in the proposed method compared to the two other methods.