Contour based segmentation of chromosomes in G-band metaphase images

Cytogenetics deals with the study of human chromosomes. This helps in detecting birth defects. Segmentation of overlapped chromosomes in metaphase images is very important in chromosome analysis. This becomes a complex task when more than one overlap occurs in the chromosomes mainly due to the non rigid nature of the chromosomes. This paper proposes an algorithm for identification of overlap zone which is the important step in segmentation of overlapping chromosomes without human intervention. Outer contour of the chromosomes is obtained including the overlapping zones. Then the curvature function is used to identify the concave and convex points. These concave points are used in finding the overlap zones. More than one overlap is identified correctly. G-band metaphase chromosome images are taken for analysis. The algorithm is robust. It is tested on more than 350 images with various number and degree of overlaps. The overall accuracy of the algorithm is 96%.