The New Jersey Institute of Technology's
Electronic Theses & Dissertations Project

Title:	Automatic registration and segmentation of abdominal images and detection of pancreatic cancer
Author:	Maniprasad, Girish Kumar
View Online:	njit-etd2007-037 (xii, 84 pages ~ 8.0 MB pdf)
Department:	Department of Biomedical Engineering
Degree:	Master of Science
Program:	Biomedical Engineering
Document Type:	Thesis
Advisory Committee:	Biswal, Bharat (Committee co-chair) Foulds, Richard A. (Committee co-chair) Alvarez, Tara L. (Committee member)
Date:	2007-05
Keywords:	Pancreatic cancer Abdominal imaging
Availability:	Unrestricted
Abstract:	Localized and detailed analyses of 3D abdominal images obtained through different imaging modalities help greatly in determining the progression of a disease or for postoperative treatment / evaluation. However, such analyses become difficult and sometimes unfeasible due to the effects of patient motion and breathing. This is particularly evident during analysis of the pancreas for cancer, due to its proximity to other intra-abdominal organs. Within subject registration thus becomes imperative for pathological analysis of pancreatic cancer. An intensity-based, global image registration algorithm was developed in the present work, for registration of pancreatic abdominal images. The registration algorithm was automatic and could register three dimensional MR and CT images of the abdomen. Once registered, localization and analysis of the pancreas was facilitated by a semi-automatic k-means clustering based segmentation procedure. Such a registration and segmentation based method could be used as a valuable tool for pancreas cancer screening and analysis.