Development and performance evaluation of processing techniques for automatic Karyotyping

Abstract

The main objective of this work is to develop intermediate steps for automatic Karyotyping, aimed at improving the efficiency of the process. In Karyotyping, the input is an image of metaphase Chromosomes and the output is an ordered arrangement of Chromosomes, referred to as a Karyotype, which is used to ascertain genetic factors associated with diseases. The steps in automatic Karyotyping are: preprocessing, segmentation, feature extraction and classification. The factors that limit the efficiency of automatic Karyotyping systems are (1) The presence of unwanted components like Interphase cells that interfere with the segmentation of Chromosomes. (2) The presence of occlusion which hinders the extraction of features from the partially hidden Chromosomes.(3) The closely spaced features of intra-group Chromosomes in the feature space make it difficult to accurately classify them. This work proposes methods to address these problems. To efficiently remove the Interphase cells which exhibit similar characteristic as that of the Chromosomes, a segmentation technique termed as pixel ranking through multidirectional parameter evaluation has been proposed. To resolve occlusion, a stochastic search algorithm based on Evolutionary Computation and Particle Swarm Optimization has been proposed in this work. In this technique, a set of initial random solutions are created from the occluding Chromosomes and using the operators borrowed from Genetics and swarm behavior, the initial solutions are transformed into probable solutions for the problem over certain number of generations. Some of the intra-group Chromosome classes have overlapping and closely spaced features that limits the classification accuracy. The performance of this technique has been compared with other techniques. An improvement in the efficiency of automatic Karyotyping using the proposed techniques, by comparing with the performance of an existing automatic Karyotyping package, has been established.