Biomedical Engineering Reference
In-Depth Information
3
DEFORMABLE MODELS AND THEIR
APPLICATION IN SEGMENTATION OF IMAGED
PATHOLOGY SPECIMENS
Lin Yang
Department of Electrical and Computer Engineering
Rutgers University, Piscataway, New Jersey, USA
David J. Foran
Center for Biomedical Imaging,
UMDNJ-Robert Woods Johnson Medical School,
Piscataway, New Jersey, USA
Microsopic evaluation of peripheral blood smears and stained tissue analysis are performed
routinely in pathology departments worldwide for cancer diagnosis and/or early detection.
Recently, there has been an increase in the number of institutions using digital imaging and
analysis to assist in assesment before a dignosis is rendered. Before the computer can be
used to index, achive, analyze, or classify an imaged specimen, it must first be delineated
into “homogeneous” regions based on the similarity of pixel attributes. Deformable models,
or snakes, have gained significant attention and have become popular image segmentation
methods since their first introduction by Kass, Witkin, and Terzopoulus in 1989. In this
chapter, we will review recent advances and improvements on deformable models. We will
focus primarily on the application and performance of different types of deformable models
for analyzing microscopic pathology specimens.
1.
INTRODUCTION
In many applications where computer-assisted analysis of pathology spec-
imens is performed ([1, 2, 3]), one of the most difficult challenges is how to
