Automated chromosomes counting systems using deep neural network

Karyotyping analysis is an important clinical process used in the process of diagnosing several genetic diseases by analyzing the number or shape of chromosomes. If the number of chromosomes is out of the normal range, it can cause hereditary diseases such as Down syndrome and Edward syndrome. In this study, we employed Faster R-CNN-based model for automated chromosome counting systems and showed a result of 0.914 in average precision
$\boldsymbol{[\text{IoU}=0.75]}$
. As a result, it was possible to count chromosomes by detecting each object with a high probability for independent and overlapping chromosomes.