"Medical image analysis has paved a way for research in the field of medical and biological image analysis through the applications of image processing. This study has special emphasis on nuclei segmentation from digitized histology images and pill segmentation. Cervical cancer is one of the most common malignant cancers affecting women. This can be cured if detected early. Histology image feature analysis is required to classify the squamous epithelium into Normal, CIN1, CIN2 and CIN3 grades of cervical intraepithelial neoplasia (CIN). The nuclei in the epithelium region provide the majority of information regarding the severity of the cancer. Segmentation of nuclei is therefore crucial. This paper provides two methods for nuclei segmentation. The first approach is clustering approach by quantization of the color content in the histology images uses k-means++ clustering. The second approach is deep-learning based nuclei segmentation method works by gathering localized information through the generation of superpixels and training convolutional neural network.
The other part of the study covers segmentation of consumer-quality pill images. Misidentified and unidentified pills constitute a safety hazard for both patients and health professionals. An automatic pill identification technique is essential to address this challenge. This paper concentrates on segmenting the pill image, which is crucial step to identify a pill. A color image segmentation algorithm is proposed by generating superpixels using the Simple Linear Iterative Clustering (SLIC) algorithm and merging the superpixels by thresholding the region adjacency graphs. The algorithm manages to supersede the challenges due to various backgrounds and lighting conditions of consumer-quality pill images"--Abstract, page iii.
Stanley, R. Joe
Moss, Randy Hays, 1953-
Stoecker, William V.
Electrical and Computer Engineering
M.S. in Computer Engineering
National Library of Medicine (U.S.)
Missouri University of Science and Technology
ix, 40 pages
© 2017 Sudhir Sornapudi, All rights reserved.
Thesis - Open Access
Print OCLC #
Electronic OCLC #
Link to Catalog Recordhttp://laurel.lso.missouri.edu/record=b12136226~S5
Sornapudi, Sudhir, "Nuclei segmentation of histology images based on deep learning and color quantization and analysis of real world pill images" (2017). Masters Theses. 7710.