• Nu S-Au Găsit Rezultate

View of Classification of Brain Tumor Using Firefly Optimisation Algorithm


Academic year: 2022

Share "View of Classification of Brain Tumor Using Firefly Optimisation Algorithm"


Text complet


Classification of Brain Tumor Using Firefly Optimisation Algorithm

1B Sundaramurthy, 2Mubarak A,3Murali R,4KavinKumar A

1Associate Professor

1Department of Computer Science and Engineering,


UG Student


Department of Computer Science and Engineering, Vinayaka Missions KirupanandaVariyar Engineering College, Vinayaka Missions Research Foundation (Deemed to be University),Salem

1[email protected]

2[email protected]

3[email protected]

4[email protected] Abstract:

Tumor present in the human body is regarded as an undesired mass that grows unconditionally in brain. Manual classification of brain tumor is regarded as a time consuming process that is carried out in diagnostic centers. Classification of brain tumors using image processing schemas on other hand is a crucial task since the classification is entirely based on the size and location.

In order to improve the task of classification, it is essential in developing a meta-learning heuristics that enables optimal classification of instances. In this paper, a meta-heuristic model using Firefly optimization algorithm is developed to classify the brain tumor regions. The simulation is conducted to validate the proposed method with existing methods in terms of various performance metrics. The results of simulation validates that the firefly algorithm obtains improved classification rate than other methods.

Keywords:Brain Tumor, Classification, Firefly Optimisation, MRI

1. Introduction

A brain tumor is an uncontrolled mass of brain tissue which prevents the human body from working properly. It can be either good or bad. It is capable of growing under brain gravity and could cause brain injury in the worst stage [1]. Varied classifications of brain cancers have different appearances in the MRI data, and the use of MRI scans to detect and classify the tumors is meaningful. Hand analysis is used for the recognition, segmentation, and sorting of brain tumors from MRIs by radiologists and various specialists [15]-[25]. Diagnosis helped by computers (CAD) can improve diagnostic accuracy [2]. The developments in machine vision and algorithms for learning [3] have made the use of automatic clinical assistance systems for brain tumors more flexible. However, due to the variety of tumor types, forms, and appearance qualities, segmentation and characterization of MRI brain tumors remains a challenge. Therefore, the development of an automatic, robust diagnostic tool to detect, segment and classify tumors is highly demanded.

In recent years, MRI scans have been employed extensively to automate tumor identification by engine learning and deep learning algorithms [4]. However, time complications have produced a balance between complexity and precision in the management of big data sets [5] [10] [12].

These problems must be efficiently mitigated to enable accurate and time-consuming tumour detection. By examining all of these facts, this article has produced and developed a computer- aided decision support system to further classify and diagnose brain tumor stages. The


combination of probability theory and neural networks in order to build a probable neuron network whose smoothing parameters are optimally defined using firefly optimization proposes an effective noise removing technique employing an ensemble genetic filter and a classifier [6]

[8]. The system proposed includes preprocessing, segmentation, extraction and classification stages. A median filter and a wiener filter are used for the preprocessing of the noise removing procedure, graph cutting approach for the removal of skulls and picture improvement. The objective of the proposed system is to classify the brain tumour images into three sub-types:

Meningioma, Glioma and Pituitary using Firefly optimization algorithm.

2. Related Works

Recent research and classification of brain tumor MRI has included soft computing techniques.

The authors in [7] have built a brain tumor clinical support system for the Lion Optimized selection of features and the BSVM-based classification model. Brain tumor clinical support system. This technique uses GMF to remove acoustic noise, followed by segmentation of the brain tumor region using a Hierarchical Fuzzy Clustering Algorithm. The GLCM functionality is extracted and selected through Lion optimization and the functionality is employed in the classification of tumor images using BSVM.

The authors in [9] have proposed a distinctive strategy for SVM classification for brain tumor detection. Different approaches were applied for segmentation of the lesion candidate and then form, texture and intensity were extracted. Finally, this feature is used by the SVM classifier to classify tumor images.

BrainMRNet has been built by Toğaçar et al. [11] to detect brain tumors using the new CNN model. This model transfers pre-processed images into the focus modules with image augmentation, and then classifies the CNN.

The authors in [13] built a tumor detection algorithm based on an extremist learning machine (ELM). The triangular, fluctuating medium filter system is used to increase the picture and extract Gabor characteristics and comparable texture (ST) functions. In this approach, however, the precision of segmentation of noisy images is still limited.

The authors in [14] have implemented a 3D CNN 3D technique for microscopic brain tumor identification and classification. The 3D CNN was used to extract the characteristics and to employ the correlation-based selection method. These features are used for final classification in the feed-forward neural network. In the 10-layer resnet50 architecture for brain tumour identification, the authors in [23] constructed a hybrid CNN model.

According to literature research, the compromise between precision and time complexity is still a major issue in the classification of brain tumors. If the classifiers are pretrained with MRI images, the accuracy improves significantly, but the training period lengthens. This longer training time introduces the problem of complexity. In the majority of investigations, this challenge was tackled and high precision and less complicated linkages were ensured. The vast variants in the size, form, and position of the brain tumor generate problems with the treatment of these data sets. In this study, a hybrid classifier was designed that could significantly reduce this issue.

3.Proposed Classification

Figure 1 depicts a functional block diagram of the system used to aid in the identification and classification of brain tumors. Brain images from the MRI are gathered and processed using


techniques to remove noise and enhance the image. The characteristics are then extracted by clinical experts or doctors to analyze and classify the tumor for further diagnosis.

Figure 1: Proposed Model Pre-processing stage

Three stages are included in the pre-processing phase of the proposed diagnosis tool: noise removal, skull removal and image improvement. A median and wiener filter is used for noise removal. First of all, the employment of median and Wiener filters is proposed. The information about these features is combined and a feature vector has been created after the extraction of these features. The graph cut approach is used for skull removal of brain pictures. The chart, often used [16], was desirable for skull areas with low contrast and speckles. The skull stripping technique uses three operating groups. These are: preliminary mask thresholding, removal of narrow links by graph cutting and postprocessing. These are: The last stage is to re-establish, via inadvertency following the threshold, the partial volume of the gray matter voxels. The graph is defined by the entire picture, given the starting mask. Then, the aim is to determine seed areas and edge weights, so that all permissible cuttings, i.e. cuts separating seeding regions, have the least value required for the desired cut. These technologies provide effective pre-processing with the efficient reduction of noise and the removal of skulls.

Feature Extraction

Texture, form and intensity characteristics are retrieved from this proposed technology. GLCM and HOG descriptors are used for extraction of functions. The technology used to divide common standards at a certain offset describes GLCM for an image. If the gray-scale image values or various shading actions are taken into consideration, GLCM is used to estimate the image texture. Given that GLCMs are reliably wide and intermittent, the characteristics produced with this technique. HOG feature vectors for the gradient nature of brain images have also been recovered. To portray the gradient character of a block around a key point, a local feature comparable to the HOG descriptor is employed. The first step is to process the gradient image by

Brain image


Feature Extraction

Classification using FFA



converting the input image into a filter mask. A grid of histograms is then built to organize each histogram in a container according to its introduction. In order to conserve space, a histogram on an equally separated grid is recorded for each cell. Therefore, each cell has an identical number of tilts (based on the cell estimate) and a histogram. The cells are then grouped into rectangular blocks that can be covered. All cells in a single block have histogram values tied to a vector frame. Each block's vector is then normalized and so the connection of each block vector results in the last vector feature.

Firefly Classification

The Firefly Algorithm is a stochastic method of global optimization and the hybrid clustering is based on Firefly algorithm and utilizes distance metrics. The objective function f(x) uses a difference in central clusters between one data point and another, and then will become finally the best data point. The fireflies move in a cluster to another data point to the best location. The new location provides the ideal data points, and the best centre value in a similar group of data points can be found. Classification results can obtain without any degradation in cluster quality.

In the Firefly algorithm, the major problems are the differentiation in light intensity and absorption. The study estimates that absorbent luminosity is mainly dependent on the objective function. As absorption is directly linked with the light intensity of the adjacent Firefly in Firefly algorithm. The absorption rate (β) is given in Eqn.(1):

 

0 2

r e r

  (1) where,

0is regarded as absorbency factor that tends to provide a total absorption rate obtained the light source during the condition r = 0.

γ is regarded as a Parameter that tends to find the speed of convergence and the behaviour of Firefly. This is determined entirely by absorbance changes that lie within the limits of [0, ∞).

The total distance exists between the fireflies can be defined in terms of relations using Cartesian distance is given in Eqn.(2).

 


1 d

ij i j ik jk


r x x x x

  



k is defined as a spatial coordinates part of a firefly that lies in a two-dimensional state and hence the above equation is further refined as in Eqn.(3),

  



ij i j i j

rxxyy (3)

The firefly movement and the rate of absorption behaviour of fireflies is thus determined by the Eqn.(4),

   


i i i j

x  xr xx  rand (4)

From the above equation, the rate of absorbency can be given in Eqn.(5)

 


xi xj

 (5)

Where, α is regarded as the chance maker parameter, and rand [0,1] is regarded as the random number, which is obtained through a uniform distribution.

The chance maker is considered to be the third statement, and hence concerning the random number, it is defined as in Eqn.(6).


rand 0.5

  (6)

Fitness function, which is defined in the following equation, to assess the problem of clustering via measurement based on distance in Eqn.(7).


1 M

i i


f D C



where, Ci is regarded as the cluster centre, Di is regarded as the total number of available data and M is regarded as the maximum distances between the cluster centres.

4. Performance Evaluation

In this section, we validation the FFA model with existing deep learning models that includes genetic algorithm, ant colony optimisation, particle swarm optimisation, bees swarm optimisation and artificial bee colony. The FFA is evaluated on BRATS 2013 with synthetic images for training and real-images for testing. The evaluation are measured against various performance metrics that includes the following:

Fig.2. Accuracy

Fig.3. F-measure

0 10 20 30 40 50 60 70 80 90




0 10 20 30 40 50 60 70 80 90





Fig.4. Sensitivity

Fig.5. Specificity

Fig.6. Geometric mean

0 20 40 60 80 100 120




72 74 76 78 80 82




65 70 75 80 85 90


Geometric mean



Fig.7. Percentage error

From the results of Fig.2 – Fig.7, proposed FFA obtains improved rate of accuracy, f-measure, sensitivity, specificity, geometric mean and reduced percentage error than existing classification models.

5. Conclusion

In this paper, FFA is used to classify the brain tumour images into three sub-types: Meningioma, Glioma and Pituitary using Firefly optimization algorithm. The results of simulation shows that the proposed method achieves improved classification rate than other existing algorithms. The overall results shows that the proposed method obtains improved accuracy rate of 81.234 than existing methods with reduced percentage error of 16.45%.


[1] Bacanin, N., Bezdan, T., Venkatachalam, K., & Al-Turjman, F. (2021). Optimized convolutional neural network by firefly algorithm for magnetic resonance image classification of glioma brain tumor grade. Journal of Real-Time Image Processing, 1-14.

[2] Bezdan, T., Zivkovic, M., Tuba, E., Strumberger, I., Bacanin, N., & Tuba, M. (2020, July).

Glioma brain tumor grade classification from mri using convolutional neural networks designed by modified FA. In International Conference on Intelligent and Fuzzy Systems (pp. 955-963). Springer, Cham.

[3] Kapoor, A., &Agarwal, R. (2021). Enhanced Brain Tumour MRI Segmentation using K- means with machine learning based PSO and Firefly Algorithm. EAI Endorsed Transactions on Pervasive Health and Technology, 7(26), e2.

[4] Sharma, A., Chaturvedi, R., Kumar, S., &Dwivedi, U. K. (2020). Multi-level image thresholding based on Kapur and Tsallis entropy using firefly algorithm. Journal of Interdisciplinary Mathematics, 23(2), 563-571.

[5] Dey, N., Chaki, J., Moraru, L., Fong, S., & Yang, X. S. (2020). Firefly algorithm and its variants in digital image processing: A comprehensive review. Applications of firefly algorithm and its variants, 1-28.

[6] Nayak, J., Naik, B., Dinesh, P., Vakula, K., & Dash, P. B. (2020). Firefly Algorithm in Biomedical and Health Care: Advances, Issues and Challenges. SN Computer Science, 1(6), 1-36.

0 5 10 15 20 25 30 35


Percentage Error



[7] Gowrishankar, J., Narmadha, T., Ramkumar, M., &Yuvaraj, N. (2020). Convolutional Neural Network Classification On 2d Craniofacial Images. International Journal of Grid and Distributed Computing, 13(1), 1026-1032.

[8] Nayak, J., Naik, B., Dinesh, P., Vakula, K., & Dash, P. B. (2020). Firefly Algorithm in Biomedical and Health Care: Advances, Issues and Challenges. SN Computer Science, 1(6), 1-36.

[9] Natarajan, Y., Kannan, S., &Mohanty, S. N. (2021). Survey of Various Statistical Numerical and Machine Learning Ontological Models on Infectious Disease Ontology.

Data Analytics in Bioinformatics: A Machine Learning Perspective, 431-442.

[10] Rajinikanth, V., Kadry, S., & Nam, Y. (2021). Convolutional-Neural-Network Assisted Segmentation and SVM Classification of Brain Tumor in Clinical MRI Slices. Information Technology and Control, 50(2), 342-356.

[11] Mariappan, L. T., &Yuvaraj, N. (2020). Analysis On Cardiovascular Disease Classification Using Machine Learning Framework. Solid State Technology, 63(6), 10374-10383.

[12] Siva Raja, P. M., & Rani, A. V. (2020). Brain tumor classification using a hybrid deep autoencoder with Bayesian fuzzy clustering-based segmentation approach. Biocybernetics and Biomedical Engineering, 40(1).

[13] Khadidos, A., Khadidos, A. O., Kannan, S., Natarajan, Y., Mohanty, S. N., &Tsaramirsis, G. (2020). Analysis of COVID-19 Infections on a CT Image Using DeepSense Model.

Frontiers in Public Health, 8.

[14] Sara, S. B. V., Anand, M., Priscila, S. S., Yuvaraj, N., Manikandan, R., &Ramkumar, M.

(2021). Design of autonomous production using deep neural network for complex job.

Materials Today: Proceedings.

[15] Yang, X. S. (2020). Firefly algorithm and its variants in digital image processing.

In Applications of Firefly Algorithm and its Variants: Case Studies and New Developments.


[16] Balaji, M. S. P., Saravanan, S., Chandrasekar, M., Rajkumar, G., &Kamalraj, S. (2020).

Analysis of basic neural network types for automated skin cancer classification using Firefly optimization method. Journal of Ambient Intelligence and Humanized Computing, 1-14.

[17] He, H., Tan, Y., Ying, J., & Zhang, W. (2020). Strengthen EEG-based emotion recognition using firefly integrated optimization algorithm. Applied Soft Computing, 94, 106426.

[18] Yin, B., Wang, C., &Abza, F. (2020). New brain tumor classification method based on an improved version of whale optimization algorithm. Biomedical Signal Processing and Control, 56, 101728.

[19] Nayak, J., Vakula, K., Dinesh, P., & Naik, B. (2020). Applications and advancements of firefly algorithm in classification: An analytical perspective. In Computational Intelligence in Pattern Recognition (pp. 1011-1028). Springer, Singapore.

[20] Tiwari, A., Srivastava, S., & Pant, M. (2020). Brain tumor segmentation and classification from magnetic resonance images: Review of selected methods from 2014 to 2019. Pattern Recognition Letters, 131, 244-260.

[21] Raja, P. S. (2020). Brain tumor classification using a hybrid deep autoencoder with Bayesian fuzzy clustering-based segmentation approach. Biocybernetics and Biomedical Engineering, 40(1), 440-453.


[22] Hadi, A. M., Al-Abass, G. J. A., &Hussain, F. F. K. (2020). Hybrid-Approach-Based Feature Selection for Magnetic Reverberation Imaging Brain Tumor Classification. Journal of Southwest Jiaotong University, 55(2).

[23] Tandel, G. S., Balestrieri, A., Jujaray, T., Khanna, N. N., Saba, L., &Suri, J. S. (2020).

Multiclass magnetic resonance imaging brain tumor classification using artificial intelligence paradigm. Computers in Biology and Medicine, 122, 103804.

[24] Gadekallu, T. R., Khare, N., Bhattacharya, S., Singh, S., Maddikunta, P. K. R., Ra, I. H.,

&Alazab, M. (2020). Early detection of diabetic retinopathy using PCA-firefly based deep learning model. Electronics, 9(2), 274.

[25] Vijh, S., Sharma, S., &Gaurav, P. (2020). Brain tumor segmentation using OTSU embedded adaptive particle swarm optimization method and convolutional neural network.

In Data visualization and knowledge engineering (pp. 171-194). springer, Cham.



The aim of this paper is to perform an automated kidney stone classification using image and data processing a back propagation neural network.. The use of a neural network for

Using a UNET-based deep neural network, Dong et al., (2017) developed a completely automated brain tumor identification and segmentation approach.. They demonstrate that

A hybrid feature extraction method with a regularized extreme learning machine (RELM) for developing an accurate Knee tumor classification approach is done,

The wavelet autoencoder-based DNN technique is used for compressing the brain image dataset [3] , integrates the autoencoder feature reduction with the decomposed wavelet

The process starts with removing of skull region from the brain image using image processing methods and later include pseudo color transformation of the axial T1-weighted

Classification and Detection of Brain Tumor Through MRI Images Using Various Transfer Learning Techniques.. Ullas kumar Agrawal1 1 , Pankaj Kumar

An optimized image pre-processing technique is proposed to enhance the quality of the facial image for improving the classification of the human emotion using

Efficacy of deep convolutional neural network algorithm for the identification and classification of dental implant systems, using panoramic and periapical

A proper confirmation of the cerebrum strangeidentification which is by basically involving some imaging techniques to obtain the result as apositive

Keywords: Particle Swarm Optimization (PSO), Convolutional Neural Networks (CNN), Image classification, Pooling

Classification and Detection of Brain Tumor through MRI Images Using Various Transfer Learning Techniques.. Amit Thakur 1 , Pawan Kumar

[6] developed the exudates detection model by employing image processing steps for extracting the features and neural network is adopted for performing the

In the proposed system we are using federated averaging algorithm which would just take the weights from distributed model and send it to the server which would take average of

Sudhakar et al.(2017) They planned Curvelet Decomposition as well as Grey Level Co- occurrence Matrix (GLCM)for spatial Feature Extraction as well as Probabilistic

genautonomousdetectionofcancerusingMachineLearningandBigDataoronlyimagerecognitiontechniques.Thetraditiona landpre-contemporaryapproachesfollowedinthedetectionofcancer using various

It is used to detect the tumor of brain[18].They proposed a Statistical Evaluation of Brain Tumor in Different MRI Protocols using support vector machine and classifies

To find the image classification of weather reports using a convolution neural network and suggest a deep learning algorithm using TensorFlow or Keras.. KEYWORDS: Computer

Fast and better results are obtained for tongue image segmentation having minor abnormal area using firefly algorithm along with k-mean algorithm along with

Through pre- processing, image fusion and initial tumor strip classification, tumor segmentation based on the final hybrid intelligent fuzzy Hopfield neural network algorithm,

In this paper, programmed tumor division approach utilizing convolutional neural systems.. A fix shrewd division method has been

Brain tumor segmentation using convolutional neural networks in MRI images. The multimodal brain tumor image segmentation benchmark

Brain Tumor Segmentation and Classification based on Deep Learning-Based Inception Networks..

Poonam, Review of image processing techniques for automatic detection of tumor in human brain, International Journal of Computer Science and Mobile Computing, 2(11),