Machine learning approach for homolog chromosome classification

Automated analysis of human chromosomes is a necessary procedure to attain karyotyping and it is highly effective in cytology analysis to detect birth defects in metaspread chromosomes. In this, chromosomes are partitioned into “abnormal” and “normal” categories. However, the success of most traditional classification methods relies on the presence of accurate chromosome segmentation. Despite many years of research in this field, accurate segmentation and classification remains a challenge in the presence of cell clusters and pathologies. Many classification methods focused on hand crafted features, such as length, centromere positions. In this manuscript, proposed method focused on chromosome classification based on deep features using convolutional neural network. It is subsequently trained on various chromosome datasets consisting of adaptively resampled image patches. In the testing phase, average the prediction scores of a similar set of image patches is performed. The proposed method is evaluated on different overlapped, nonoverlapped chromosomes and normal, abnormal datasets. Proposed method better performs than previous algorithms in classification accuracy with 98.7%, area under the curve AUC is 0.97 values, and abnormality detection accuracy is 98.4%.