Intuitive ways have been found out to interpret and describe the inherent ambiguity and vagueness in the medical images. The efficiency of the use of intuitionistic fuzzy set theory in medical image processing is demonstrated in the context of contrast enhancement, segmentation.
Intuitive ways have been found out to interpret and describe the inherent ambiguity and vagueness in the medical images. The efficiency of the use of intuitionistic fuzzy set theory in medical image processing is demonstrated in the context of contrast enhancement, segmentation.
Intuitive ways have been found out to interpret and describe the inherent ambiguity and vagueness in the medical images. The efficiency of the use of intuitionistic fuzzy set theory in medical image processing is demonstrated in the context of contrast enhancement, segmentation.
NCRAMT|June 24-26, 2011|Haldia, India International Journal of Soft Computing and Engineering (IJSCE)
ISSN: 2231-2307, Volume-1, Issue-NCRAMT2011, July 2011
35
AbstractMedical imaging has been an active area of research where abnormalities are detected non invasively. But dealing with medical images is a challenging task. Due to the complexity in the images, many of the structures are hardly visible; various soft computing techniques are applied by different researchers to process medical images. In order to carry out this task, intuitive ways have been found out to interpret and describe the inherent ambiguity and vagueness in the medical images in terms of intuitionistic fuzzy set theory. The efficiency of the use of intuitionistic fuzzy set theory in medical image processing is demonstrated in the context of contrast enhancement, segmentation where the performance is observed to be much better.
I ndex Termsintuitionistic fuzzy set, hesitation degree, clustering, enhancement.
I. INTRODUCTION Motivated from Zadehs definition of fuzzy sets theory [1], where uncertainty or vagueness is considered only in the form of membership function, different notions of higher-order fuzzy sets have been presented by various researchers. Among them, intuitionistic fuzzy sets, proposed by Atanassov, has been a suitable tool for modelling the hesitancy arising from imprecise or/and imperfect information. This hesitation is due to the lack of knowledge or the personal error in defining the membership function. Intuitionistic fuzzy sets are defined by two characteristic functions, namely the membership and the non-membership, describing the belongingness or non-belongingness of an element respectively. Medical images contain uncertainties and intuitionistic fuzzy set (IFS) considers more number of uncertainties (two), so IFS will be a good tool in analyzing medical images. The initial step for medical image processing is the image enhancement. As many of the structures are not highlighted properly, so enhancement or pre processing is necessary. If the image is enhanced properly, post processing results will be better and accurate. So, after enhancement, segmentation is carried out for analyzing and diagnosing medical images. Medical image segmentation refers to the segmentation of known anatomic structures from medical images. Structures include organs such as cardiac ventricles or kidneys, lungs, abnormal lesions such as tumors, clots, and cysts, pathological cells/blood vessels, as well as other structures such as bones, vessels, brain structures etc. The overall
Manuscript received May 15, 2011. Tamalika Chaira, is with Centre for Biomedical Engg., Indian Institute of Technology Delhi, New Delhi, India (e-mail: tamalika @yahoo.com).
objective of such methods is to provide computer-aided diagnosis that will be used for assisting doctors in evaluating medical imagery or in recognizing abnormal findings in a medical image. II. INTUITIONISTIC FUZZY SET THEORY A fuzzy set A in a finite set } ..., { , 2 1 n x x x X = may be represented mathematically as: } ) ( , {( X x x x A A e =
where, the function ( ) : [0,1] A x X is measure of degree of belongingness or membership function of an element x in the finite set X and the measure of non-belongingness is 1 ( ) A x . Attanassov [2] suggested that while defining the membership degree, there may be some hesitation, which arises due to the lack of knowledge in. So, with the introduction of hesitation degree, ) (x A t , non-membership degree is not the complement the membership degree as in fuzzy set, rather less than or equal the complement of membership degree. An intuitionistic fuzzy set A in a finite set X may be mathematically represented as:
} | )) ( ), ( , {( X x x x x A A A e = v where, ] 1 , 0 [ : ) ( ), ( X x x A A v are respectively the membership and the non-membership functions of an element x with the necessary condition 1 ) ( ) ( 0 s + s x x A A v and 1 ) ( ) ( ) ( = + + x x x A A A v t
III. PROCESSING OF MEDICAL IMAGES Image enhancement is necessary before processing. Enhancement increases the contrast of the image. The darker regions will be darker and brighter regions will be more brighter. Enhancement may be carried out using non-fuzzy/fuzzy/intuitionistic fuzzy set to highlight the important structures of the image. But enhancement of medical images using IFS gives better result as it considers more number of uncertainties that may be suitable to medical images as medical images also contain many uncertainties. An example on image enhancement is shown using fuzzy/intuitionistic fuzzy set theory. Fig. 1 is an image of a prostate. With fuzzy INT operator [4], the Intuitionistic Fuzzy Set Theory in Medical Imaging Tamalika Chaira
Intuitionistic Fuzzy Set Theory in Medical Imaging
36 regions are not visible where as using IFS [3], it is observed that all the regions are visible properly and clearly.
After pre processing, the images are segmented to separate different regions of interest such as tumor, abnormal lesions, blood vessels/cells and so on. Segmentation includes thresholding, clustering, and edge detection and it depends on the type of application. Thresholding - There are lots of methods on image thresholding using non-fuzzy/fuzzy techniques. It is a simple, fast and a robust segmentation technique But no such technique is there that accurately threshold the medical images. An example on image thresholding by Chaira [5] using IFS is shown in Fig. 2. where blood vessels are clearly segmented from a poorly illuminated image as compared to the non-fuzzy/fuzzy methods. Counting of blood vessels is a very difficult task and it is necessary for the pathologists in the detection of various diseases such as angiogenesis, infection so on. It uses a novel IFS divergence measure and a membership function. In this method, a window based thresholding is used where the image is divided into several windows and then thresholding is carried out for each windows. The IFS method is compared with other fuzzy and non fuzzy methods and is observed that using IFS method, almost the blood vessels are visible.
Clustering is also a type of segmentation where the image is divided into several regions introduced by Bezdek [9]. It groups similar pixels into one cluster and different pixels in different cluster based on some similarity measure. Thus there is similarity of pixels among the clusters and dissimilarity of pixels in different clusters. So, when any abnormal regions are required to detect, clustering algorithm is a suitable method that separates the abnormal region present in the image. Clustering can be done using crisp or fuzzy. In non-fuzzy/crisp clustering, each pixel present in a distinct cluster i.e. 0 or 1. But in fuzzy clustering, pixels may belong to several clusters and membership value is associated to each pixel. It considers the vagueness of the boundary. Intuitionistic fuzzy method considers the hesitation in the membership function in the fuzzy clustering method. An example on image clustering [6] is shown in Fig 3 using fuzzy/intuitionistic fuzzy set where blood cells are clearly clustered using IFS with almost no noise.
Edge detection is another type of segmentation, which is required to detect the boundaries of any abnormal regions/lesions. This information is widely used in computer-aided diagnosis, localization of pathological tissue, study of anatomical structures. Up till now there are many methods such as curve fitting methods, statistical methods but Cannys method [7] is an optimal edge detector. From the edges, morphological parameters such as area, perimeter, curvature can be evaluated. In medical images, there are many unwanted structures, so direct applying Cannys edge detector does not give better result. A better result is expected, if clustering is followed by histogram thresholding and then edge detection to find the exact region of interest. Clustering clusters the image into different regions, histogram thresholding selects the wanted lesions and then edge detector detects the edges of that lesion. An example [8] is shown in Fig. 4 where the tumour image is clustered, thresholded, and then edge detected to detect the exact boundary of the tumour in the brain. Simultaneously, the morphological parameters are computed which will help the doctors and physicians to come to a successful decision.
Fig. 1 (a) Image, b) enhancement using INT operator, c) intuitionistic fuzzy enhancement (c) (b) (a) Fig. 3: a) blood cell image, b) FCM cluster, c) intuitionistic fuzzy cluster (b) (a) (c)
Fig. 2: a) Original image, b) Robert method, c) fuzzy method, d) Intuitionistic method
(c) (d) (a) (b)
NCRAMT|June 24-26, 2011|Haldia, India International Journal of Soft Computing and Engineering (IJSCE) ISSN: 2231-2307, Volume-1, Issue-NCRAMT2011, July 2011 37
IV. CONCLUSION This paper provides the usefulness of IFS theory in medical image processing. It is observed that the results using IFS is much better than the fuzzy/non fuzzy set theory. The reason for obtaining better results is that in fuzzy set theory, the only uncertainty is the membership function. IFS considers more number of uncertainties (membership function and the hesitation degree), and as the medical images are low contrasted, with vague region/boundaries, IFS gives better result. Thus, when membership function is not always accurately defined due to the lack of personal error, an intuitionistic fuzzy set may help in solving the problem conclusion section is not required.
ACKNOWLEDGMENT Author would like to acknowledge the Department of Science and Technology for carrying out the work under the scheme DST-SERC Fast Track for young scientist. REFERENCES [1] L.A. Zadeh, Fuzzy Sets : Information and Control, 8, pp.338-353, 1965. [2] K.T. Atanassov, Intuitionistic fuzzy set, Fuzzy Sets and Systems 20 (1986) 87-97. [3] I.K.Vlachos, G.D.Sergiadis, Role of entropy in intuitionistic fuzzy contrast enhancement, Lecture notes in artificial intelligence- 4529, pp. 104-113, Springer, 2007. [4] T. Acharya, A.K. Ray, Image processing: Principles and application. Wiley, 2005. [5] T. Chaira, Intuitionistic fuzzy segmentation of medical images, IEEE Transaction of Biomedical Engineering, vol. 57, No. 6 (2010) 1430-1436. [6] T. Chaira, Intuitionistic fuzzy color clustering of medical images, Proc. of IEEE, World Congress on Nature and Biologically inspired soft computing (Coimbatore), (2009). [7] J.Canny, Computational approach to edge detection, IEEE Transaction on PAMI 8(6), (1986), 679-698. [8] T. Chaira, , S. Anand, A novel intuitionistic fuzzy approach for tumour/hemorrhage detection in medical images. Journal of scientific and industrial research (2011) (accepted)., [9] J.C. Bezdek, , L.O.. Hall, and L.P. Clark.,Review of MR segmentation technique in pattern recognition, Med Phy,10:20, pp.33-48, 1993 .
Tamalika Chaira rerceived her Bachelar degree from Bihar Institute of Technology, Sindri, Jharkhand, Masters from Bengal Engineering and Science university from Electronics and Communication Department. She obtained her PhD from Indian Institute of Technology Kharagpur in image processing (Electronics and Communication) in the year 2003. Her research interest is medical imaging that includes image enhancement, segmentation, clustering using fuzzy/intuitionistic fuzzy set theory. Currently she is a young Scientist of DST, Govt. of India working with the Centre for Biomedical Engg., Indian Institute of Technology, Delhi. She is an author of a book on Fuzzy image processing and application with MATLAB with CRC press, 2009. She has 1 US patent and published several papers in international IEEE and Elsevier journals. She is also a reviewer of several international journal. She has received an young scientist award from Department of Science and Technology 2007 (Fast Track scheme for young scientist). She has also received a prestigious National award (IYBA 2010), Innovative young Biotechnologist award from Department of Biotechnology, Govt. of India.
(b) Fig.4 a) CT scan tumor image, b) edge of the tumour using IFS method, c) edge detected image using IFS method