Towards Automated Karyotyping of Curved Chromosomes

Abstract

Since the birth of the automated karyotyping systems by the aid of computers, building a fully automated chromosome analysis system has been an ultimate goal. Along with many other challenges, automating chromosome segmentation and classification has been a major challenge especially due to touching, overlapping and highly curved chromosomes in the metaphase images. In this paper, the proposed technique addresses one of these challenges of straightening highly curved chromosomes and karyotyping them. The earlier reported methods for straightening the highly arched chromosomes have limited success as they are able to straighten the chromosomes with only one bending centre and that too if the bent is at the centromere. Contribution and novelty of this work lies in the ability of the algorithm to successfully straighten the chromosomes with multiple bends and classify it in its respective class. The proposed algorithm uses watershed transform to segment the chromosomes from a metaphase image and then the curved chromosomes are artificially straightened using the first and the second minima, which are actually the bending points along the curvature of the chromosomes. The extracted features are used to classify the chromosomes using a neural network classifier. To assess the effectiveness, the algorithm was tested and analyzed using a variety of metaphase images exhibiting various levels of bends giving promising results and showing close correlation to the manually classified chromosomes.