.
RESEARCH PAPER
.
Special Focus
SCIENCE CHINA
Information Sciences
May 2012 Vol. 55 No. 5: 1052–1061
doi: 10.1007/s11432-012-4556-0
c
Science China Press and Springer-Verlag Berlin Heidelberg 2012 info.scichina.com www.springerlink.com
Medical image segmentation using improved FCM
ZHANG XiaoFeng
1,2
, ZHANG CaiMing
1,3,4 ∗
, TANG WenJing
1,2
&WEIZhenWen
1
1
School of Computer Scienc e and Te chnology, Shandong University, Jinan 250101,China;
2
School of Information and Ele ctrical Engineering, Ludong University, Yantai 264025,China;
3
School of Computer Science and Technology, Shandong University of Finance and Economics, Jinan 250014,China;
4
Shandong Province Key Lab of Digital Media Te chnology, Jinan 250014,China
Received October 21, 2011; accepted December 6, 2011; published online March 19, 2012
Abstract Image segmentation is one of the most important problems in medical image processing, and the
existence of partial volume effect and other phenomena makes the problem much more complex. Fuzzy C-
means, as an effective tool to deal with PVE, however, is faced with great challenges in efficiency. Aiming at
this, this paper proposes one improved FCM algorithm based on the histogram of the given image, which will
be denoted as HisFCM and divided into two phases. The first phase will retrieve several intervals on which
to compute cluster centroids, and the second one will perform image segmentation based on improved FCM
algorithm. Compared with FCM and other improved algorithms, HisFCM is of much higher efficiency with
satisfying results. Experiments on medical images show that HisFCM can achieve good segmentation results in
less than 0.1 second, and can satisfy real-time requirements of medical image processing.
Keywords FCM, histogram, image segmentation, medical image processing
Citation Zhang X F, Zhang C M, Tang W J, et al. Medical image segmentation using improved FCM. Sci China
Inf Sci, 2012, 55: 1052–1061, doi: 10.1007/s11432-012-4556-0
1 Introduction
Image segmentation has become one common scientific issue and core technology in digital image pro-
cessing [1,2]. In medical image processing, image segmentation plays a vital role in biomedical imaging
applications such as the quantification of tissue volumes, diagnosis, localization of pathology, study of
anatomical structure, treatment planning, partial volume correction of functional imaging data, and
computer integrated surgery [3]. Currently, medical images suffer from three main problems: intensity
inhomogeneity(IIH), noise and partial volume effect(PVE) [4,5]. Specifically, IIH appears as tissue inten-
sity variation with locations, which may arise from radio frequency coils or acquisition sequences. PVE
occurs where received signals contain a mixture of several tissues; thus it is difficult to assign one single
class to the affected pixels. Therefore, conventional “hard” segmentation method cannot be applied to
this phenomenon, because it restricts each pixel exclusively to one class. As a result, fuzzy classification
has been extensively applied, since it can assign one pixel to several classes concurrently and can retain
information as much as possible. Currently, there are two popular PVE models: fuzzy C-means(FCM)
[6,7] and Gaussian distribution models [8,9]. In this paper, we will investigate FCM-based algorithms for
medical image segmentation.
∗
Corresponding author (email: czhang@sdu.edu.cn)