Automated Brain Tumor Segmentation and Detection in MRI Using ...

Available online at www.sciencedirect.com

ScienceDirect Procedia Computer Science 92 (2016) 475 – 480

2nd International Conference on Intelligent Computing, Communication & Convergence (ICCC-2016) Srikanta Patnaik, Editor in Chief Conference Organized by Interscience Institute of Management and Technology Bhubaneswar, Odisha, India

Automated Brain Tumor Segmentation and Detection in MRI using Enhanced Darwinian Particle Swarm Optimization(EDPSO) Vasupradha Vijaya ,* Dr.A.R.Kavithab,*S.Roselene Rebeccac b

a PG Student,Jerusalem College of Engineering,Chennai *Professor,c*Assistant Professor Jerusalem College of Engineering,Chennai

Abstract

Medical Image segmentation is the most challenging problems in the research field of MRI scan analysis. Automated brain tumor segmentation and detection are eminently important in medical diagnostics because it provides information related to functional structures as well as potential abnormal tissue necessary to demarcate surgical plan. But automatic tumor segmentation is still challenging because of low contrast and ill-defined boundaries and accuracy problem. Therefore Enhanced Darwinian Particle Swarm Optimization (EDPSO) is proposed for automated tumor segmentation which overcomes the drawback of existing Particle Swarm Optimization(PSO).This innovative method consists of four steps. First step is pre-processing, film artifacts and unwanted portions of MRI images are removed using tracking algorithm .Second step involves the process of removing the noises and high frequency component using Gaussian filter. Third step, segmentation is done using Darwinian Particle Swarm Optimization and Fourth step is classification, which is done by Adaptive Neuro Fuzzy Inference System. The performance of the proposed method is systematically evaluated using the MRI brain images. © by Published Elsevier B.V. This is an open access article under the CC BY-NC-ND license © 2016 2016 Published The Authors. by Elsevier B.V. (http://creativecommons.org/licenses/by-nc-nd/4.0/). Selection and peer-review under responsibility of scientific committee of Interscience Institute of Management and Technology. Peer-review under responsibility of the Organizing Committee of ICCC 2016 Keywords: Brain Tumor;Magnetic Resonance Imaging(MRI);Enhanced Darwinian Particle Swarm Optimization (EDPSO);Particle Swarm Optimization(PSO);Adaptive Neuro Fuzzy Inference System(ANFIS).

1877-0509 © 2016 Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Peer-review under responsibility of the Organizing Committee of ICCC 2016 doi:10.1016/j.procs.2016.07.370

476

Vasupradha Vijay et al. / Procedia Computer Science 92 (2016) 475 – 480

1. Introduction Tumor cells are abnormal group of cells which form lumps or growths in the body. Different types of tumors grow and behave uniquely, depending on whether they are noncancerous (benign) or cancerous (malignant). Precancerous conditions have the potential to develop into cancer.MRI is a medical imaging technique, and it is used for visualization of the internal structure of the body. MRI of head uses powerful magnetic fields, radio waves and computer to produce detail picture of the brain which are more detailed than other imaging techniques. MRI can provide ample information about anatomy of human tissues, also helps in identification of tumor cell in body. Image segmentation refers to the process of partitioning a digital image into multiple regions with similar properties such as gray level, color, texture, brightness, and contrast. There are many difficult problems in the field of pattern recognition and image processing. Local search approaches were used to solve difficult problems in the field of pattern recognition and image processing. A PSO is population-based stochastic optimization algorithm and has been successfully applied to solve a optimization problem. Due to its simplicity and efficiency in navigating large search spaces for optimal solutions, PSOs are used in this research to develop efficient, robust and flexible algorithms to solve a selective set of difficult .In general problem with the PSO and with the other optimization algorithms is that they may get trapped in local optimum points, and the algorithm may work in some problems but may fail in others. To overcome such a problem, Tillett 9 presented the Darwinian PSO (DPSO). In the DPSO, multiple swarms of test solutions performing just like an ordinary PSO may exist at any time, with rules governing the collection of swarms that are designed to simulate natural selection. The main goal of the paper is to propose a computationally efficient segmentation method, which is for detection and segmentation of brain tumor. 2. Proposed Work In this section, describe the proposed technique for effectively segment and detect the tumor from the MRI brain images consists of four phases namely pre-processing, image enhancement, segmentation and classification. 2.1 Pre-Processing The raw input image MRI is subjected to a set of pre-processing steps so that the image gets transformed to be suitable for the further processing. In this phase the raw image which contains many unwanted information (noise), actual noise which gets added due to MRI scan are removed in a sequential manner. First using tracking algorithm the artefacts (Name, Age, Sex, Date...) other such textual contents are removed and the Tracking Algorithm is as follows Step 1: Load the MRI image of size 256*256 with each element corresponds to gray value between 0 to 255 Step 2: Store the image in a two dimensional matrix. Step 3: Select the peak threshold value for removing white labels Step 4: Set flag value to 255. Step 5: Select each pixels whose intensity value is equal to 255. Step 6: If the intensity value is 255 then, the flag value is set to zero then the labels are removed from MRI Step 7: Otherwise skip to next pixel. 2.2 Image Enhancement Image Enhancement is the processing of image to improve their appearance in terms of better contrast and visibility. In Image Enhancement Gaussian filter is applied to remove noise and to improve the quality of image. Gaussian is random occurrence of white intensity value and its intensity value is drawn from Gaussian distribution, it is very much useful to reduce noise and as it linear filter so it is computationally efficient and enhances image quality with the image boundaries. Through Gaussian filter high frequency components such as impulse noise, salt and pepper noise are removed. 2.3 Segmentation The proposed method implements the EDPSO method for the detection and segmentation of tumor cells from MRI images. The Architecture diagram of proposed method is shown in the Figure 1

Figure 1 - Architecture diagram of proposed system

Enhanced Darwinian Particle Swarm Optimization, particles travel through the search space to find an optimal solution, by interacting and sharing information with neighbour particles, namely their individual best solution (local best) and computing the


477

neighbourhood best. Each step of the procedure, the global best solution obtained in the entire swarm is updated. Using all of this information, particles realize the locations of the search space where success was obtained, and are guided by these successes. In each step of the algorithm, a fitness function is used to evaluate the particle success. To model the swarm, each particle moves in a multidimensional space according to position (୲୬ ) and velocity (୲୬ ) values which are highly dependent on local best (෬୲୬ ), neighborhood best ( ෭୬୲ ) and global best (෬ ୬୲ ) information. The proposed EDPSO Algorithm is as follows Initialize swarm (Initialize‫ݔ‬௡௧ ,‫ݒ‬௡௧ ,‫ݔ‬෱ ௡௧ ,݊෱ ௡௧ and݃෱ ௡௧ ) Loop: for each particle in swarm Evaluate the fitness of each particle in-terms of velocity and position. ‫ݒ‬௡௧൅ͳ =‫ݒݓ‬௡௧ ൅ ߩ1 ‫ݎ‬1 (݃෱ ௡௧ -‫ݔ‬௡௧ )൅ߩ2 ‫ݎ‬2 (‫ݔ‬෱ ௡௧ -‫ݔ‬௡௧ ) ൅ߩ3 ‫ݎ‬3 (݊෱ ௡௧ -‫ݔ‬௡௧ ) //Velocity of next particle ‫ݔ‬௡௧൅ͳ =‫ݔ‬௡௧ + ‫ݒ‬௡௧൅ͳ // Position of next particle Update ‫ݔ‬෱ ௡௧ ,݊෱ ௡௧ and݃෱ ௡௧ Update ‫ݒ‬௡௧ and ‫ݔ‬௡௧ Check if the swarm has improved using the fitness value If swarm improved Retain the best swarm particle and extend swarm life Else Delete the inferior swarm particle and reduce swarm life end until stopping criteria. The coefficients w, ɏ1 and ɏ2 assign weights to the inertial influence, the global best and the local best when determining the new velocity, respectively 1 and 2 are constant integer values, which represent “cognitive” and “social” components. The parameters ɏͳ and ɏ2are random vectors with each component generally a uniform random number between 0 and 1. 2.4 Classification ANFIS being neuro-fuzzy system harness power of both hence it proves to be a sophisticated framework for multi-object classification. A comprehensive feature set and fuzzy rules are selected to classify an abnormal image to the corresponding tumor type. This proposed technique is fast in execution, efficient in classification and easy in implementation. Neuro-fuzzy systems use the combined power of two methods: fuzzy logic and artificial neural network (ANN).This type of hybrid system called as ANFIS ensures detection of the tumor in the input MRI image. The ANFIS architecture consists of two fuzzy if-then rules based on a first order Sugeno model: Rule 1: If (x is A1) and (y is B1) then (f1 = p1x + q1y + r1) Rule 2: If (x is A2) and (y is B2) then (f2 = p2x + q2y + r2) where x and y are the inputs, Ai and Bi are the fuzzy sets, f1 are the outputs within the fuzzy region specified by the fuzzy rule, p1, q1 and r1 are the design attributes that are calculated during the training process. 3. Results and Discussion The MRI image dataset have utilized in proposed image segmentation technique is taken from the publicly available sources. This image dataset contains 101 brain MRI images, including 87 brain images with tumor and the other 14 brain images without tumor. The brain image dataset are divided into two sets: training dataset and testing dataset. In this, the 65 images are utilized for the training purpose and the remaining 36 images are utilized for testing purpose. The dataset used for analysis consists of 256–256 pixel MRI brain images. The input and output of various stages of MRI images are shown in Figure 2(a)Input MRI Image, 2(b)Pre-processed Image,2(c)ANFIS Classification, 2(d)Tumor Detection,2(e)Tumor Classification.

478


2(a)

2(b)

2(c)

Figure 2(a)Input MRI Image, 2(b)Pre-processed Image,2(c)ANFIS Classification.

2(d)

2(e)

Figure 2(d)Tumor Detection,2(e)Tumor Classification The step size will be decreased if the error measure encounters two consecutive combinations of an increase followed by a decrease. After the conclusion of 65 training epochs, the network error (mean square error) which is represented as error convergence of each ANFIS shown in Figure -3

Figure 3-Curve of error convergences of ANFIS The performance is evaluated by comparing the existing PSO algorithm with DPSO algorithm in terms of the quality rate is shown in the Figure 4.The Blue color curve indicate EDPSO and red color indicate PSO algorithm. The Quality rate of EDPSO


algorithm in tumor segmentation is compared against the PSO algorithm based on execution time and accuracy is shown in Table 1. Table 1 - Comparison of PSO and DPSO based on Execution time and Accuracy Performance Evaluation

PSO

EDPSO

Iteration Time Accuracy

68 iteration 62.345 sec 92%

56iteration 53.456 sec 95%

Figure 4 –Performance Evaluation of Quality Rate in PSO and EDPSO based on accuracy 4. Conclusion and Future Enhancement In this paper, EDPSO algorithm is comparatively studied against PSO algorithm. The proposed technique consists of pre-processing, Image Enhancement, Segmentation and final classification. The MRI image dataset that have been utilized in project. This image dataset contains 101 brain MRI images, including 87 brain images with tumor and the other 14 brain images without tumor. Comparative analysis is based on execution time and accuracy of proposed EDPSO algorithm with PSO Algorithm. From the results obtained EDPSO Algorithm received a better quantity rate for all the input images. Future Enhancement will be a combination of several optimization algorithms which shall be studied further in order to improve the performance of the system and also be more robust and exhaustive. Acknowledgement I must thank, first and foremost my guide Dr.A.R.Kavitha and, Mrs.S.Roselene Rebecca for his valuable suggestions and kind co-operation to the complete the work. I wish to avail myself of this opportunity, express a sense of gratitude and my Husband Mr.G.Vijay for manual support, strength and help for everything, without whose guidance and patience, this dissertation would not be possible References 1.Ahmed kharrat, Karim Gasmi, “A Hybrid Approach for Automatic Classification of Brain MRI Using Genetic Algorithm and Support Vector Machine”, Leonardo Journal of Sciences, pp.71-82, 2010. 2.Bandhyopadhyay, S.K.Paul,T.U,“Automatic segmentation of brain tumour from multiple images of brain MRI”,Int J Appl Innovat Eng Manage (IJAIEM), Volume 2, Issue 1, 2013, pp. 240–248. 3.GlavanC.C,Holban.S, “Segmentation of bone structure in X-ray images using convolutional neural network”,Adv Electr Comput Eng, Volume 13, Issue 1, pp. 1–8,2013. 4.Jaya.J,Thanushkodi.K,Karnan.M,“Tracking Algorithm for DeNoising of MR Brain Images”, UCSNS International Journal of Computer Science and Network Security, Vol. 9, No.ll,November 2009. 5. Kennedy.J, Eberhart.R, “A new optimizer using particle swarm theory,” in Proc. IEEE 6th Int. Symp. Micro Mach. Human Sci., 1995,pp.39–43. 6. Marcello.J,Marques.F,Eugenio.F,“Evaluation of thresholding techniques applied to oceanographic remote sensing imagery” , SPIE, 5573, pp. 96-103, 2004. 7. Mharib.A, Ramli.A,Mashohor.S, Mahmood.R, “Survey on liver CT image segmentation methods” , Artif. Intell. Rev. 37(2),

479

480


pp.83–95 ,2012. 8.Muthukrishnan.R,M.Radha,“Edge Detection Techniques For Image Segmentation”, International Journal of Computer Science & Information Technology (IJCSIT) Vol 3, No 6, 2011. 9.Tillett.J,Rao, T. M, Sahin,F, Rao.R, Brockport.S, “Darwinian particle swarm optimization”, in Proc. 2nd Indian Int. Conf. Artif. Intell., pp. 1474–1487,2005. 10.Zafer Iscan,Ziimray Dokur,Tamer Olmez,“Tumor detection by using Zernike moments on segmented magnetic resonance brain images”, Expert Systems with Applications, Vol: 37, No: 3, pp.2540-2549,2010.