A Web-based Tool for Semi-interactively Karyotyping the Chromosome Images for Analyzing Chromosome Abnormalities

Abstract

Chromosome abnormalities relate to several genetic diseases. These abnormalities can be diagnosed based on the analysis of karyogram of the human chromosomes. However, the manual chromosome karyotyping process is often time consuming. This paper presents a novel web-based tool, Biochrom, to assist the cytogeneticist in producing the karyogram. Biochrom is a semi-automated tool, which provides both manual and automated functions in chromosomes segmentation and classification using image processing combined with machine learning techniques. The study is carried on 612 metaphase images with 48 of those containing abnormal chromosomes. We compare the proposed tool to a conventional public tool, Metasel, for karyotyping based on performance by two cytogeneticists using user experience metrics such as number of manual interactions and processing time. Moreover, we quantitatively evaluate the accuracy of the classification of two approaches: Support Vector Machine (SVM) and deep learning. The evaluation results show that the deep learning classification outperforms SVM classification, and our proposed tool requires fewer interactions and less time consuming to complete the karyotyping task on average.