Mathematical approach for segmenting chromosome clusters in metaspread images.

Karyotyping is an examination that helps in detecting chromosomal abnormalities. Chromosome analysis is a very challenging task which requires various steps to obtain a karyotype. The challenges associated with chromosome analysis are overlapping and touching of chromosomes. The input considered for chromosome analysis is the metaspread G band chromosomes. The proposed work mainly focus on separation the overlapped and touching chromosomes which is considered to be the major challenge in karyotype. There are various research contribution in chromosome analysis in progress which includes both low (Machine Learning) and high level (Deep Learning) methods. This paper proposes a mathematical based approaches which is very effective in segmentation of clustered chromosomes. The accuracy of segmentation is robust compared to high level approaches.

Convolutional Neural Network for Segmentation and Measurement of Intima Media Thickness.

The measurement of Carotid Intima Media Thickness (IMT) on Common Carotid Artery (CCA) is a principle marker of risk of cardiovascular disease. This paper presents a novel method of using deep Convolutional Neural Network (CNN) for identification and measurement of IMT on the far wall of the artery. The Region of Interest (ROI) is extracted using CNN architecture with 8 layers. 110 subjects are taken for the study. Each subject is recorded with one Right Common Carotid Artery (RCCA) and Left Common Carotid Artery (LCCA) frame resulting in 220 recordings. Patch based segmentation with 2640 patches are given to the training network for ROI localization. Intima Media Complex (IMC) is the area where IMT is measured. This region is extracted after defining the ROI. Keeping in mind the end objective of measurement of IMT values binary threshold with snake algorithm is applied to extract the lumen-intima and media-adventitia boundary. IMT values are measured for 20 cases and mean difference is found to be 0.08 mm.

Overlapped chromosome segmentation and separation of touching chromosome for automated chromosome classification.

Karyotyping is the process of arranging pairs of the chromosomes in an order to examine chromosomal abnormalities which are the causes for Birth Defects. The overlapping and touching chromosomes of input metaphase spread images should be segmented and disentangled for this process. This paper proposes an algorithm to separate the touching chromosomes and the segmentation of overlapping chromosomes from G-Band metaspread images. Segmentation is performed by obtaining the interesting (concave) points on the image contour and constructing proper hypotheses for segmentation and separation.