Centromere and Length Detection in Artificially Straightened Highly Curved Human Chromosomes

Abstract

Analyzing the morphological characteristics of the human chromosomes is a general task for diagnosing many genetic disorders. For this purpose, it is necessary to identify each of the 23 pairs of the chromosomes within the microscopic images and to place them in a table like format known as a Karyotype. This is usually carried out manually by a skilled operator considering the features of each chromosome the most important of which are the location of the Centromere and the length of the chromosome. Automation of this procedure is a difficult image processing task due to the non-rigid nature of the chromosomes making them to have unpredictable shapes and curvatures within the image. Various automatic algorithms were developed in the past but success of most of these algorithms is limited to only straight or slightly curved chromosomes. In this paper, using our previously reported method(17) we present a novel combined algorithm for Centromer and length detection in any given (straight, curved or highly curved) chromosome. The proposed Centromere locating algorithm uses the fact that the Centromere is by definition the narrowest part of the chromosome. By generating a linearly varying Gray Level Mask (GLM) individually for any given chromosome and multiplying it to the binary version of the chromosome's image, the global minimum in the histogram of the resulted image indicates the location of the Centromere. For evaluating the performance of the proposed algorithm a data set of 54 highly curved human chromosomes provided by the Cytogenetic Laboratory of the Cancer Institute, Imam Hospital, Tehran, Iran was used. Comparing the results of the proposed Centromere and length detection algorithm to those manually identified by a skilled operator, an average absolute error of 4.2 and 5.8 pixels were obtained respectively which is acceptable according to the expert.