CLASSIFICATION OF POSITRON EMISSION TOMOGRAPHY IMAGES USING OTSU’S AND NEUTROSOPHIC METHOD

International Journal of Pure and Applied Mathematics Volume 118 No. 18 2018, 781-789 ISSN: 1311-8080 (printed version); ISSN: 1314-3395 (on-line version) url: http://www.ijpam.eu Special Issue

ijpam.eu

CLASSIFICATION OF POSITRON EMISSION TOMOGRAPHY IMAGES USING OTSU’S AND NEUTROSOPHIC METHOD Vijayan T1, Sridhar raja 2 ,kalaiselvi B 3 Asst. Professor School of Electrical Sciences, BIST, BIHER, Bharath University tvij16@gmail.com, sridharraja.eie@bharathuniv.ac.in, kalaiselvi.eie@bharathuniv.ac.in 1,2,3

1

Abstract The segmentation is used to simplify or to change the representation of an image into something that is more meaningful and easier to analyze. Many foreground segmentation methods are available to increase the efficiency of the image. The foreground segmentation can be improved by the classification performance on weakly annotated datasets – those with no additional annotation other than class labels. A new co-segmentation algorithm called triclass that looks at all training images jointly and automatically segments out the most classdiscriminative foregrounds for each image.

Ultimately, those foreground segmentations are used to train a classification system.

Triclass solves the co-segmentation problem by minimizing losses at three different levels: the category level for foreground/background consistency across images belonging to the same category, the image level for spatial continuity within each image, and the dataset level for discrimination between classes. The segmentation is based on Otsu’s thresholding but iteratively searches for sub regions of the image for segmentation, instead of treating the full image as a whole region for processing.

And further the Neutrosophic thresholding is applied to improve the segmentation result. These methods are combined together and implemented in a new advanced scanning technique called as Positron Emission Tomography (PET). Tests on synthetic and real images showed that the new iterative method and neutrosophic method can achieve better performance than the standard Otsu’s method in many challenging cases, such as

781

Identifying weak objects and revealing fine structures of complex objects and also improves the segmentation score.

Keywords: Otsu’s thresholding, Triclass, Neutrosophic technique, Segmentation score.

1. Introduction Image segmentation is the process of partitioning a digital image into multiple segments or set of pixels called super pixels.

There is various foreground segmentation methods are available to improve classification performance on weakly annotated datasets – those with no additional annotation other than class labels. (1-2) The existing segmentation method is based on Otsu’s iterative tri-class thresholding.

Triclass was a new co-segmentation algorithm along with Otsu’s method that looks at all training images jointly and automatically segments out the most class-discriminative foregrounds for each image. It solves the cosegmentation problem by minimizing losses at three different levels.(3) This method is applied only for microscopic images and it is less efficient to noisy images and it also has less segmentation score.

The proposed segmentation method has two techniques such as iterative triclass thresholding and Neutrosophic algorithm. This algorithm considers the whole image for processing. (4-5) The objective of the proposed method is to apply segmentation techniques in the scans (Ex: PET) and to make it efficient to operate in various noisy images and also to improve the segmentation score.Matrix Laboratory (MATLAB) is the software used for the convenient extraction of the image which is faster than other computational software.(6-11)

International Journal of Pure and Applied Mathematics

Special Issue

2. Block Diagram

Fig: Overall operation and analysis of PET image.

3. Input Image Positron emission tomography (PET) is a nuclear medicine exam that produces a three dimensional image of functional processes in the body. It is a highly sensitive non-invasive technology that is ideally suited for pre-clinical and clinical imaging of cancer biology, in contrast to anatomical approaches. (12-17) By using radio labeled tracers, which are injected in non-pharmacological doses, threedimensional images can be reconstructed by a computer to show the concentration and location(s) of the tracer of interest.With the recent advances in molecular/cell biology that have led to target discovery, it is now possible to design specific tracers to image events noninvasively in small animals and humans with positron emission tomography (PET). (18-21) PET is ideally suited for monitoring cell/molecular events early in the course of a disease, as well as during pharmacological or radiation therapy. Furthermore, it can be used for prognostic information and to image for disease recurrence. PET should become increasingly important in cancer imaging in the next decade.(22-27)

782

Some common uses of PET  To detect cancer  To evaluate the heart for blood flow, signs of coronary artery disease or heart function  To evaluate the brain for memory disorders, brain tumors or seizure disorders  To provide valuable information to physicians in many neurological diseases such as Alzheimer's disease, dementias, Parkinson's disease, Huntington's disease, and Down's Syndrome.  Used in Psychiatry, because PET is very sensitive to biological brain alterations during episodes of schizophrenia, depression and other disorders.  To provide unique information— including details on both function and anatomic structure of the body that is often unattainable using other imaging procedures. For many diseases, nuclear medicine scans yield the most useful informationneeded to make a diagnosis or to determine appropriate treatment, if any.

International Journal of Pure and Applied Mathematics

Special Issue

T= arg

σw2(T)

Where σw2(T)=q0(T) σ02(T)+ q1(T) σ12(T) where the subscript 0 and 1 denote the two classes, background and foreground, respectively, and qi and σi , i = [0,1] are the estimated class probabilities and class variances, respectively. These quantities are calculated as q0=

Fig No. 1 A PET image of brain with parkinson’s disease

q1= and the individual class variances are given as Fig No. 2 A PET image of brain with tumor

σ02 σ12

4. Method 4.1 Otsu’s Thresholding The simplest method of image segmentation is called the thresholding method.

This method is based on a clip-level (or a threshold value) to turn a gray-scale image into a binary image. There is also a balanced histogram thresholding. (28-31) The key of this method is to select the threshold value (or values when multiple-levels are selected). Several popular methods are used in industry including the maximum entropy method, Otsu's method (maximum variance), and k-means clustering.

Recently, methods have been developed for thresholding computed tomography (CT) images. The key idea is that, unlike Otsu's method, the thresholds are derived from the radiographs instead of the (reconstructed) image.

Otsu’s method searches the histogram of an image to find a threshold that binarizes the image into two classes, the background with a mean of μ0 and the foreground with a mean of μ1, as shown in the top of Fig. 1. Without loss of generality, here we assume that the foreground is brighter than the background, i.e., μ1 > μ0.

The calculation of threshold T is as follows:

783

0(T)]

2

1(T)]

2

Where we assume that the pixel values of the image are from0 to K. So from the above equations we can see that T is function of the pixel values of both the foreground and the background. If the signal intensity changes, it may affect Tin such a way that the segmentation result may become less optimal.(32-36) 4.2Triclass Thresholding For an image u, at the first iteration, Otsu’s method is applied to find a threshold T [1] where the superscript denotes the number of iteration. We then find and denote the means of the two classes separated by T [1]as μ0[1] and μ1[1] for the background and foreground, respectively.

Then we classify regions whose pixel values are greater than μ1[1] as foreground F[1]and regions whose pixel values are less than μ0[1] as background B[1].For the remaining pixels u(x, y) such that μ0[1] ≤ u(x, y) ≤μ1[1] .we denote them as the TBD class Ω[1].

So our iterative process assumes that the pixels that are greater than the mean of the “tentatively” determined foreground are the true foreground. Similarly, pixels with values less than μ0 are for certain the background. But the

International Journal of Pure and Applied Mathematics

pixels in the TBD class, which are the ones that typically cause mis-classifications in the standard Otsu’s method, are not decided at once and will be further processed. By our definition, we have U = F[1]

B[1]

Ω[1

where ∪ is the logical union operation. At the second iteration, we apply Otsu’s method to find threshold T [2] on region Ω[1]only. We then calculate the two class means in Ω[1]separated by T [2]as μ0[2]and μ1[2] . Similarly, the second iteration will generate a new F[2], B[2], and Ω[2]such that Ω[1] = F[2] B[2] Ω[2] where F[2]is defined as the region in Ω[1]with pixel values greater than μ1[2], B[2]as the region in Ω[1]with pixel values less than μ0[2] and Ω[2]are the new TBD region. The iteration stops when the difference between two consecutive threshold |T [n+1] − T [n]| is less than a preset threshold. At the last iteration Ω[n +1], Ω[n+1]is separated into two instead of three classes, i.e., foreground F[n+1] is defined as the region of Ω[n] that is greater than T [n+1]instead of μ[n+1] and background B[n+1]is defined as the regions with pixel value less than T [n+1]. The innovation of the new method is to iteratively define the TBD regions to gain a high distance ratio, which will result in better segmentation by applying Otsu’s method.

Special Issue

artificial images with different levels of noise and real images demonstrate that the proposed approach can reduce the indeterminate of the images and perform thresholding with better result especially, in noisy cases.

Let U be a universe of discourse, and W be a set included in U , which is composed by bright pixels. A neutrosophic image Ns P is characterized by three subset T , I and F . A pixel P in the image is described as P(T, I, F) and belongs to W in the following way: it is t% true in the bright pixel set, i% indeterminate, and f % false, where t varies in T , i varies in I , and f varies in F . The pixel P(i, j) in the image domain is transformed into neutrosophic domain PNS( i, j) = {T (i, j) I( i, j) F( i, j)} . T(i, j) , I (i, j) and F(i, j) are the probabilities belong to white set, indeterminate set and non-white set, respectively, which are defined as: T(i.j)= I( i, j)= F(i, j) = 1−T(i, j) Ho(i, j) = abs(e(i, j)) Where HO(i, j) is the homogeneity value of T at (i, j) ,which is described by the local gradient value e(i, j).

A λ-mean operation for PNS, PNS(λ ) is defined as:

4.3Neutrosophic Method Neutrosophy, a branch of philosophy, introduced by Florentin Smarandacheas a generalization of dialectics, which studies the origin, nature, and scope of neutralities, as well as their interactions with different ideational spectra. First, the image is transformed into neutrosophic set and a new operation, λ-mean, is employed to reduce the indetermination degree of the image, which is evaluated by the entropy of the indeterminate subset. Then, the image becomes more uniform and homogenous, and more suitable for thresholding. Finally, the image in Ns is thresholded using a histogram thresholding method. The experiments on

784

NS

= P ( (λ), (λ),

(λ)

λ



λ



λ

λ





International Journal of Pure and Applied Mathematics

Special Issue

I( i, j) = References Where (i, j) is the homogeneity value of T(λ) at (i, j) with the local window size.

After a λ-mean operation is performed on neutrosophic set PNS, the entropy of the indeterminate subset I is decreased and the distribution of the elements in I becomes more non-uniform. It leads to the decrease of the indetermination of the neutrosophic set PNS. Otsu's method reduces the interclass variance defined by the following equation: σ (t)=c1 (t)σ1 (t)+c2 (t)σ2 (t) where σ (t) is the sum of variances of the two classes as a function of threshold t. σi(t)and ci(t)are the respective variance and probability of class i , i = 1,2. Threshold t that results in the minimization of σ (t) separates the two classes as the foreground and background, respectively.

5. Segmentation Score Segmentation score is nothing but the efficiency of the segmentation algorithm used.

In our method, we use the neutrosophic method inorder to increase the efficiency of the scanning images. By using this technique, the score has been improved from 4.1885 to 49.9939 for our input PET image.

6. Conclusion The proposed segmentation algorithm is based on iterative triclass thresholding and Neutrosophic algorithm. These methods are implemented in new scanning technique called Positron Emission Tomography (PET) and it was finely segmented to obtain the high segmentation score than the existing method. In future, noisy images can be segmented using the proposed method.

785

1. [1] Hameed Hussain J., Anbazhgan G., Improve of the cop of vapour compression refrigeration system by using thermoelectric cooler, International Journal of Pure and Applied Mathematics, V-116, I-14 Special Issue, PP-585-589, 2017 2. Nakkeeran S., Hussain J.H., Innovative technique for running a petrol engine with diesel as a fuel, International Journal of Pure and Applied Mathematics, V-116, I-18 Special Issue, PP-41-44, 2017 3. Naveenchandran P., Vijayaragavan S.P., A sensor less control of SPM using fuzzy and ANFIS technique, International Journal of Pure and Applied Mathematics, V-116, I-13 Special Issue, PP-43-50, 2017 4. Hameed Hussain J., Nakkeeran S., Design and development of pepper plucking equipment, International Journal of Pure and Applied Mathematics, V-116, I-14 Special Issue, PP-573-576, 2017 5. Naveenchandran P., Vijayaragavan S.P., A high performance inverter fed energy efficient cum compact microcontroller based power conditioned distributed photo voltaic system, International Journal of Pure and Applied Mathematics, V-116, I-13 Special Issue, PP-165-169, 2017 6. Golden Renjith Nimal R.J., Hussain J.H., Stress analysis of gear tooth using metal, International Journal of Pure and Applied Mathematics, V-116, I-17 Special Issue, PP-317-322, 2017 7. Hameed Hussain J., Hariharan R., Development of temperature-time and pressure-time diagrams for diffusion bonding ti-aa7075 dissimilar materials, International Journal of Pure and

International Journal of Pure and Applied Mathematics

8.

9.

10.

11.

12.

13.

14.

15.

Applied Mathematics, V-116, I-14 Special Issue, PP-493-499, 2017 Udayakumar R., Khanaa V., Saravanan T., Saritha G., Cross layer optimization for wireless network (WIMAX), Middle - East Journal of Scientific Research, V16, I-12, PP-1786-1789, 2013 Khanaa V., Thooyamani K.P., Udayakumar R., Cognitive radio based network for ISM band real time embedded system, Middle - East Journal of Scientific Research, V-16, I-12, PP1798-1800, 2013 Khanaa V., Mohanta K., Saravanan T., Comparative study of uwb communications over fiber using direct and external modulations, Indian Journal of Science and Technology, V-6, ISUPPL.6, PP-4845-4847, 2013 Kumarave A., Udayakumar R., Web portal visits patterns predicted by intuitionistic fuzzy approach, Indian Journal of Science and Technology, V6, I-SUPPL5, PP-4549-4553, 2013 Thooyamani K.P., Khanaa V., Udayakumar R., An integrated agent system for e-mail coordination using jade, Indian Journal of Science and Technology, V-6, I-SUPPL.6, PP-47584761, 2013 Sengottuvel P., Satishkumar S., Dinakaran D., Optimization of multiple characteristics of EDM parameters based on desirability approach and fuzzy modeling, Procedia Engineering, V-64, PP-1069-1078, 2013 Udayakumar R., Khanaa V., Saravanan T., Synthesis and structural characterization of thin films of sno2 prepared by spray pyrolysis technique, Indian Journal of Science and Technology, V-6, I-SUPPL.6, PP-47544757, 2013 Udayakumar R., Kumarave A., Rangarajan K., Introducing an efficient programming paradigm for objectoriented distributed systems, Indian

786

Special Issue

16.

17.

18.

19.

20.

21.

22.

23.

Journal of Science and Technology, V6, I-SUPPL5, PP-4596-4603, 2013 Kerana Hanirex D., Kaliyamurthie K.P., Multi-classification approach for detecting thyroid attacks, International Journal of Pharma and Bio Sciences, V4, I-3, PP-B1246-B1251, 2013 Udayakumar R., Khanaa V., Kaliyamurthie K.P., Performance analysis of resilient ftth architecture with protection mechanism, Indian Journal of Science and Technology, V6, I-SUPPL.6, PP-4737-4741, 2013 Udayakumar R., Khanaa V., Kaliyamurthie K.P., Optical ring architecture performance evaluation using ordinary receiver, Indian Journal of Science and Technology, V-6, ISUPPL.6, PP-4742-4747, 2013 Udayakumar R., Khanaa V., Saravanan T., Chromatic dispersion compensation in optical fiber communication system and its simulation, Indian Journal of Science and Technology, V-6, ISUPPL.6, PP-4762-4766, 2013 Sundarraj M., Study of compact ventilator, Middle - East Journal of Scientific Research, V-16, I-12, PP1741-1743, 2013 Udayakumar R., Khanaa V., Saravanan T., Analysis of polarization mode dispersion in fibers and its mitigation using an optical compensation technique, Indian Journal of Science and Technology, V-6, I-SUPPL.6, PP-47674771, 2013 Gopalakrishnan K., Prem Jeya Kumar M., Sundeep Aanand J., Udayakumar R., Thermal properties of doped azopolyester and its application, Indian Journal of Science and Technology, V6, I-SUPPL.6, PP-4722-4725, 2013 Udayakumar R., Khanaa V., Kaliyamurthie K.P., High data rate for coherent optical wired communication using DSP, Indian Journal of Science

International Journal of Pure and Applied Mathematics

24.

25.

26.

27.

28.

29.

30.

31.

and Technology, V-6, I-SUPPL.6, PP4772-4776, 2013 Kerana Hanirex D., Kaliyamurthie K.P., Kumaravel A., Analysis of improved tdtr algorithm for mining frequent itemsets using dengue virus type 1 dataset: A combined approach, International Journal of Pharma and Bio Sciences, V-6, I-2, PP-B288-B295, 2015 Thooyamani K.P., Khanaa V., Udayakumar R., Using integrated circuits with low power multi bit flipflops in different approch, Middle - East Journal of Scientific Research, V-20, I12, PP-2586-2593, 2014 Gopalakrishnan K., Sundeep Aanand J., Udayakumar R., Electrical properties of doped azopolyester, Middle - East Journal of Scientific Research, V-20, I11, PP-1402-1412, 2014 Thooyamani K.P., Khanaa V., Udayakumar R., Partial encryption and partial inference control based disclosure in effective cost cloud, Middle - East Journal of Scientific Research, V-20, I-12, PP-2456-2459, 2014 Sundar Raj M., Saravanan T., Srinivasan V., Design of silicon-carbide based cascaded multilevel inverter, Middle East Journal of Scientific Research, V20, I-12, PP-1785-1791, 2014 Thooyamani K.P., Khanaa V., Udayakumar R., Wide area wireless networks-IETF, Middle - East Journal of Scientific Research, V-20, I-12, PP2042-2046, 2014 Kanniga E., Srikanth S.M.K., Sundhararajan M., Optimization solution of equal dimension boxes in container loading problem using a permutation block algorithm, Indian Journal of Science and Technology, V7, PP-22-26, 2014 Arulselvi S., Sundararajan M., Smart control system in traffic analysis using RTK-GPS standards, International

787

Special Issue

32.

33.

34.

35.

36.

37.

38.

Journal of Pure and Applied Mathematics, V-116, I-15 Special Issue, PP-349-352, 2017 Arulselvi S., Karthik B., Sundararajan M., A frame work for road network extraction from remotely sensed high resolution images, International Journal of Pure and Applied Mathematics, V116, I-15 Special Issue, PP-355-360, 2017 Arulselvi S., Karthik B., Sundararajan M., Super resolution method for Thumbnail Web image, International Journal of Pure and Applied Mathematics, V-116, I-15 Special Issue, PP-369-373, 2017 Arulselvi S., Karthik B., Sundararajan M., Linear framework free rewriting systems, International Journal of Pure and Applied Mathematics, V-116, I-15 Special Issue, PP-363-367, 2017 Kanniga E., Selvaramarathnam K., Sundararajan M., Kandigital bike operating system, Middle - East Journal of Scientific Research, V-20, I-6, PP685-688, 2014 Lakshmi C., Ponnavaikko M., Sundararajan M., Improved kernel common vector method for face recognition varying in background conditions, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), V6026 LNCS, PP-175-186, 2010 M. Rajesh, K.Sathesh Kumar, K.Shankar, M. Ilayaraja., SENSITIVE DATA SECURITY IN CLOUD COMPUTING AID OF DIFFERENT ENCRYPTION TECHNIQUES, Journal of Advanced Research in Dynamical and Control Systems, Volume 18, PP2888-2899, 2017 M. Rajesh, A signature based information security system for vitality proficient information accumulation in wireless sensor systems, International

International Journal of Pure and Applied Mathematics

39.

40.

41.

42.

Journal of Pure and Applied Mathematics, Volume 118, Issue 9, Pages 367-387 M. Rajesh, A Review on Excellence Analysis of Relationship Spur Advance in Wireless Ad Hoc Networks, International Journal of Pure and Applied Mathematics, Volume 118, Issue 9, Pages 407-412 Rajesh, M., and J. M. Gnanasekar. "Path Observation Based Physical Routing Protocol for Wireless Ad Hoc Networks."

Wireless Personal Communications 97.1 (2017): 12671289.

Rajesh, M., and J. M. Gnanasekar.

"Congestion Control Scheme for Heterogeneous Wireless Ad Hoc Networks Using Self-Adjust Hybrid Model." International Journal of Pure and Applied Mathematics 116 (2017): 537-547.

Rajesh, M., and J. M. Gnanasekar. "GetUp-And-Go Efficientmemetic Algorithm Based Amalgam Routing Protocol." International Journal of Pure and Applied Mathematics 116 (2017): 537-547.

788

Special Issue