Graph Partitioning approach for Segmentation of Banding Pattern of G-band Metaphase Human Chromosomes

摘要

Chromosomes display a banding pattern when they are treated with some stains. Bands are simply alternating light and dark stripes that appear along the lengths of chromosomes. Banding pattern is one of the unique parameters in karyotyping chromosomes to detect chromosomal abnormalities. These banding patterns are used to identify the type of chromosomes and to diagnose chromosomal aberrations, including chromosome breakage, duplication, loss, translocation or inverted segments. There are four different segmentation algorithm is used for efficient detection of chromosome bands. These segmentation algorithms directly work on the intensity value. Firstly, watershed algorithm using distance transform is applied to the chromosome for segmentation. Over segmentation due to watershed segmentation is overcome by marker controlled watershed segmentation. Bands are not identified correctly for all the tested chromosomes using marker controlled watershed segmentation. Multi thresholding is applied where only few bands are identified. The segmentation algorithm based on watershed transform and graph cut method proves itself efficient and much better for the separation of bands.