Title: A tool for handling uncertainty in segmenting regions of interest in medical images

Authors: Despina Kontos, Qiang Wang, Vasileios Megalooikonomou, Alan H. Maurer, Linda C. Knight, Steve Kantor, Hrair P. Simonian, Henry P. Parkman

Addresses: Data Engineering Laboratory (DEnLab), Department of Computer and Information Sciences, Temple University, 319 Wachman Hall, 1805 N. Broad St., Philadelphia PA 19122, USA. ' Data Engineering Laboratory (DEnLab), Department of Computer and Information Sciences, Temple University, 319 Wachman Hall, 1805 N. Broad St., Philadelphia PA 19122, USA. ' Data Engineering Laboratory (DEnLab), Department of Computer and Information Sciences, Temple University, 319 Wachman Hall, 1805 N. Broad St., Philadelphia PA 19122, USA. ' Department of Radiology, Nuclear Medicine Division, Temple University School of Medicine, 3401 N. Broad St., Philadelphia, PA 19140, USA. ' Department of Radiology, Nuclear Medicine Division, Temple University School of Medicine, 3401 N. Broad St., Philadelphia, PA 19140, USA. ' Department of Medicine, Gastroenterology Section, Temple University School of Medicine, 3401 N. Broad St., Philadelphia, PA 19140, USA. ' Department of Medicine, Gastroenterology Section, Temple University School of Medicine, 3401 N. Broad St., Philadelphia, PA 19140, USA. 'Department of Medicine, Gastroenterology Section, Temple University School of Medicine, 3401 N. Broad St., Philadelphia, PA 19140, USA.

Abstract: We have developed intelligent software tools for handling the uncertainty in delineating the boundaries of complex structures when segmenting regions of interest (ROIs) in medical images. The focus is on efficiently delineating the boundary of complex 3D organ structures, enabling accurate measurement of their structural and physiologic properties. We employ intensity based thresholding algorithms for interactive and semi-automated analysis. We also explore fuzzy-connectedness concepts in order to deal with the uncertainty in identifying organ surrounding tissue and fully automate the segmentation process. We apply the proposed tools to 3D single-photon emission computed tomography (SPECT) images visualising gastric accommodation and emptying and compare their performance to that of the manual segmentation performed by a human expert. We show that the proposed tools achieve highly accurate delineation of the complex three-dimensional gastric boundaries shown in 3D SPECT images. We also demonstrate their ability to obtain accurate volume calculations based on the segmentation procedure, in order to quantitatively assess organ functional properties such as measuring the gastric mass variation.

Keywords: medical images; image analysis; regions of interest; ROIs; uncertainty; region segmentation; image processing; fuzzy-connectedness; software tools.

DOI: 10.1504/IJISTA.2006.009904

International Journal of Intelligent Systems Technologies and Applications, 2006 Vol.1 No.3/4, pp.194 - 210

Published online: 01 Jun 2006 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article