Arab J Sci Eng (2014) 39:1017–1037
DOI 10.1007/s13369-013-0664-4
RESEARCH ARTICLE - ELECTRICAL ENGINEERING
A Fast Geodesic Active Contour Model for Medical Image
Segmentation Using Prior Analysis and Wavelets
Sharif M. S. Al Sharif · Mohamed Deriche ·
Nabil Maalej · Sami El Ferik
Received: 29 November 2011 / Accepted: 4 July 2013 / Published online: 5 September 2013
© King Fahd University of Petroleum and Minerals 2013
Abstract The deformable geodesic active contour (GAC)
method is one of the most popular techniques used in object
boundary detection in images. In this work, we improve the
automatic GAC technique by incorporating prior information
extracted from the image region of interest. In addition, we
propose a new stopping function to speed up convergence
and improve accuracy. The proposed technique was applied
to both synthetic and real medical images. The results show
both an improvement of more than 40 % in convergence
speed together with an excellent accuracy when compared
with the previous work.
Keywords Deformable models · Geometric active contour
(GAC) · Snake method · Boundary detection · Prior
information · Medical image segmentation
S. M. S. Al Sharif (B ) · S. El Ferik
Systems Engineering Department, King Fahd University
of Petroleum and Minerals, Dhahran 31261, Saudi Arabia
e-mail: sharif182@yahoo.co.uk
S. El Ferik
e-mail: selferik@kfupm.edu.sa
M. Deriche
Electrical Engineering Department, King Fahd University
of Petroleum and Minerals, Dhahran 31261, Saudi Arabia
e-mail: mderiche@kfupm.edu.sa
N. Maalej
Physics Department, King Fahd University of Petroleum and Minerals,
Dhahran 31261, Saudi Arabia
e-mail: maalej@kfupm.edu.sa
1 Introduction
One of the most important tasks in medical image analysis is
segmentation. The main objective is to be able to subdivide
a given image into a number of regions exhibiting different
properties. Image segmentation is typically used to locate
objects and boundaries (lines, curves, etc.) in images. More
precisely, image segmentation is the process of assigning a
label to every pixel in an image such that pixels with the
same label share certain visual or functional characteristics
[1]. For instance, in therapy planning for cancer patients, this
approach is widely used in extracting tumour volumes from
medical images [2]. Other medical applications for image
segmentation include but are not limited to measuring tissue
volumes [3], computer-guided surgery, diagnosis [4], local-
ization pathology [5], treatment planning and studying of
anatomical structure [6, 7].
123