An image processing algorithm for accurate extraction of the centerline from human metaphase chromosomes

Abstract

The study of human metaphase chromosomes is an important aspect in clinical diagnosis of genetic disorders. Although many image processing techniques have been developed for chromosomal karyotyping to assist in laboratory diagnosis, they fail to provide reliable results in segmenting and extracting the centerline of chromosomes due to their shape variability when placed on microscope slides. In this paper we propose a hybrid algorithm that uses Gradient Vector Flow active contours, Discrete Curve Evolution based skeleton pruning and morphological thinning to provide a reliable centerline that is robust to shape variations of the chromosomes. Effective identification of the chromosome outline with its centerline provides a basis for further operations such as automated chromosome classification and centromere identification.