SSO-RBNN driven brain tumor classification with Saliency-K-means segmentation technique

https://doi.org/10.1016/j.bspc.2022.104356Get rights and content

Highlights

  • Presents a new hybrid Saliency-K-mean based segmentation technique.

  • Discusses novel texture and statistical based features.

  • Presents an efficient classification approach by combining SSO with RBNN.

  • Presents experimental validation for the proposed method.

Abstract

Early-stage diagnosis of Brain Tumor leads to better chance of cure from this deadliest disease across the globe. Existing schemes on brain tumor classification use machine learning, convolutional neural networks, Generative Adversarial Networks, and deep learning schemes. However, more execution time and uncertain predictions leading additional process to cross-check the obtained results. In this paper, a classification model Saliency-K-mean-SSO-RBNN is formulated including a new hybrid salience-K-mean segmentation technique along with utilizing the advantage of social spider optimization (SSO) algorithm in Radial Basis Neural Network (RBNN). Hybrid Saliency Map with K-means cluster-based segmentation approach is formulated to segment the tumor region. As Saliency map spotlights on eye catching region within target image, segmented image is fetched to feature extraction phase by considering multiresolution wavelet transform, Principal Component, Kurtosis, Skewness, Inverse Difference Moment (IDM), Cosine transform. Feature vector is then processed for an efficient classification using RBNN by optimizing the cluster center through SSO. RBNN with Gaussian kernel depicts a low complex model for classification. Saliency-K-mean-SSO-RBNN and new hybrid Saliency-K-mean segmentation are validated on standard datasets and compared with existing schemes with regard to specificity, precision, sensitivity, F1 score, MCC, Kappa coefficient and complexity. Saliency-K-mean-SSO-RBNN, yielding a classification accuracy in three datasets as 96%, 92%, and 94%.

Introduction

Brain tumor is a well known deadly diseases where early detection of tumor is a crucial aspect of getting cure. An unusual and uncontrolled growth of cells in human brain is termed as brain tumor which leads to drastic variations in brain behavioral pattern. It causes cancer which may yield around 13% of deaths worldwide. Moreover, statistically it has been found that 36% of people with cancerous brain tumor survive for 5 years and 31% survive for 10 years [1], [2], [3]. Tumors in brain are categorized into two different types such as malignant and benign. The tumor that contains cancerous cell is considered to be harmful and termed as malignant whereas the tumor with absence of cancerous cell is less harmful and termed as benign. The tumor size, style, location, behavior, and growth rate are the factors that show the level of risk in brain. These information helps to take decision for proper treatment like radiation, chemotherapy, and surgery. Accurate detection and localization of brain tumor in early stage provides a greater chance in survival of a person [4]. There exist various imaging techniques such as magnetic resonance imaging (MRI), computed, tomography (CT), single-photon emission computed tomography (SPECT), positron emission tomography (PET), Ultrasound, and X-ray to deal with brain tumors, Corona virus, and cancerous cells [5]. MRI is the most popular non-invasive technique which offers best contrast high spatial resolution-based image with cross-sectional views of tumor cells [6]. Detection and classification of tumor is an essential task to provide good treatment. However, medically it is risky and time-consuming stage in which a patient undergoes through few conventional process such as biopsy, manual inspection, expert suggestions etc. Therefore, it requires a Computer-Aided Design for early detection and classification of brain tumor without human intervene. For this, an accurate segmentation to locate tumor and efficiently classify it as normal or abnormal (tumorous) are the challenging issues.

Detection of tumor region from MR image depends upon proper segmentation whereas classification depends upon extraction of relevant features and an efficient classifier. There exist numerous techniques which are broadly categorized as brain tumor object segmentation and classification techniques. Tumor segment and feature extraction technique includes mostly different clustering-based segmentation algorithm, wavelet, PCA, intensity, texture, shape, LBP etc. [7], [8], [9], [10], [11]. Similarly, brain tumor classification technique used machine learning algorithm such as linear regression, K-nearest neighbor (KNN), Neural Network, support vector machine (SVM), random forest etc. are employed for detecting and classifying the tumor into normal or abnormal [12], [13], [14], [15], [16], [17]. Recently various convolutional neural network techniques are utilized such as CNN  [18], [19], [20], [21], [22], [23], Generative Adversarial Network [24], [25], Deep Learning [26], [27], [28], and Transfer learning [29], [30] have been developed for segmentation with automatic feature extraction. However, these techniques take more time for execution and had no assurance of the accuracy which leads to an additional manual detection process to cross check the obtained results of classification models. Moreover, it leads to erroneous while dealing with bulk image data. To mitigate this issue, we formulate a brain tumor detection model including a new hybrid salience-K-mean based segmentation technique along with utilizing the advantage of social spider optimization (SSO) algorithm in Radial Basis Neural Network (RBNN) to yield significant performance.

The objective of the paper is to perform classification on the brain MR image by using proposed Saliency-K-mean-SSO-RBNN method which comprises preprocessing stage followed by segmentation and feature extraction, and finally classification. Initially, the MRI image is pre-processed to remove noise and convert it into a high contrast image. Then hybrid Saliency-K-mean is used for segmentation to determine the shape and position of tumor. Next, novel features are extracted by using Wavelet Transform (WT), PCA, kurtosis, skewness, IDM and DCT to yield a feature vector representation. Then the feature vector is applied into the RBNN to classify it by using an optimal cluster center which is further determined through SSO algorithm.

In this paper, the major contributions are listed as follows:

  • We propose a new hybrid Saliency-K-mean based segmentation technique which detects the tumor region irrespective of its size with better accuracy.

  • Novel features are extracted by using Wavelet Transform (WT), PCA coefficients for statistical characteristics, kurtosis for quantifying probability distribution of image texture, skewness for detecting the edges of a tumor by measuring the lack of symmetry, IDM for measuring local homogeneity, and DCT for zonal masking.

  • We propose an efficient classification approach by combining SSO with RBNN where SSO discovers the optimal cluster center for RBNN with a high convergence speed and able to classify the brain tumors (Normal or Tumorous).

  • Experimental results validate that the of proposed method Saliency-K-mean-SSO-RBNN achieves a better accuracy and computationally executes in a shorter time span (in seconds) as compared to existing methods.

The rest of this paper is structured as follows: Section 2 presents related works, Proposed approach is detailed in Section 3. Section 4 reports experimental result analysis and conclusion is in Section 5.

Section snippets

Related work

In present years, the detection of brain tumor from MR images are getting more challenging. An accurate detection of tumor region and prediction play vital roles in diagnosis. There exist a number of machine learning models for detecting the brain tumor which is further utilized for the development of automated system in medical field. The brain tumor detection process can broadly be categorized into brain tumor region segmentation and classification techniques as described in following

Proposed Saliency-K-mean-SSO-RBNN method

The proposed method comprises four basic steps such as pre-processing, segmentation, feature extraction, and classification. Fig. 1 depicts the overview of the proposed methodology.

SSO-based optimal cluster center

SSO algorithm follows the principles of swarm intelligent algorithm which is rooted with the cooperative aspects of social spiders  [39], [40]. In multidimensional feature space, to find the sub-optimal center point, SSO is far better as compared to existing optimization algorithms such as Genetic Algorithm (GA), Particle swarm Optimization (PSO), Ant Colony Optimization (ACO). These existing algorithms may have premature convergence whereas SSO is capable enough to provide a balanced solution

Dataset

The proposed Saliency-K-mean-SSO-RBNN is validated on three standard datasets through experimentation. Table 1 details the description of the datasets. Dataset 1 [41] contains 296 images, out of which 152 images represent normal and rest 144 represent abnormal (tumor). Dataset 2 [42] comprises 250 images, out of which 100 images are normal and 150 images are abnormal. Similarly, Dataset 3 [43] has 369 images, where 155 image is abnormal and 214 is normal. A few classified normal and abnormal

Conclusion

In this paper, proposed Saliency-K-mean-SSO-RBNN is designed for classification of brain MR image into tumor and non-tumor. Initially, the modality of the given input image is pre-processed to enhance quality of images by removing the additional noises and convert it into gray image. Then in the next stage, these images are used for segmentation to separate tumor region using new hybrid Saliency-K-mean. Feature extraction is performed after segmentation to extract novel features such as Wavelet

Abbreviations

ANN:Artificial Neural Networks
CT:Computed Tomography
CNN:Convolutional Neural Network
DCT:Discrete Cosine Transform
DWT:Discrete Wavelet Transform
GAN:Generative Adversarial Network
IDM:Inverse Difference Moment
KNN:K-nearest neighbor
LBP:Local Binary Pattern
MCC:Matthews Correlation Coefficient
MRI:Magnetic Resonance Imaging
PCA:Principal Component Analysis
PET:Positron Emission Tomography
RBNN:Radial Basis Neural Network
SPECT:Single-Photon Emission Computed Tomography
SSO:Social Spider Optimization
SVM:

CRediT authorship contribution statement

Aparajita Nanda: Conceptualization, Validation, Investigation, Resources, Writing – original draft. Ram Chandra Barik: Software, Formal analysis, Investigation, Data curation, Writing – review & editing, Visualization. Sambit Bakshi: Methodology, Formal analysis, Investigation, Data curation, Writing – review & editing, Supervision, Project administration, Funding acquisition.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgment

This research is supported by the following projects:

  • a.

    Project titled “Deep learning applications for computer vision task” funded by NITROAA, India with support of Lenovo P920 and Dell Inception 7820 workstation and NVIDIA Corporation with support of NVIDIA Titan V and Quadro RTX 8000 GPU.

  • b.

    Project titled “Establishment of Bioinformatics and Computational Biology Centre: Animal Bioinformatics - BIC at National Institute of Technology Rourkela” by Department of Biotechnology, Ministry of Science

References (45)

  • CuevasE. et al.

    A swarm optimization algorithm inspired in the behavior of the social-spider

    Expert Syst. Appl.

    (2013)
  • BahadureN.B. et al.

    Image analysis for MRI based brain tumor detection and feature extraction using biologically inspired BWT and SVM

    Int. J. Biomed. Imaging

    (2017)
  • Cancer. net

    Brain Tumour: Stat.

    (2020)
  • BoroleV.Y. et al.

    Image processing techniques for brain tumor detection: A review

    Int. J. Emerg. Trends Technol. Comput. Sci. (IJETTCS)

    (2015)
  • CoatrieuxG. et al.

    A watermarking-based medical image integrity control system and an image moment signature for tampering characterization

    IEEE J. Biomed. Health Inf.

    (2013)
  • AlamM.S. et al.

    Automatic human brain tumor detection in MRI image using template-based K means and improved fuzzy C means clustering algorithm

    Big Data Cogn. Comput.

    (2019)
  • KharratA. et al.

    A hybrid approach for automatic classification of brain MRI using genetic algorithm and support vector machine

    Leonardo J. Sci.

    (2010)
  • SharmaK. et al.

    Brain tumor detection based on machine learning algorithms

    Int. J. Comput. Appl.

    (2014)
  • AggarwalN. et al.

    First and second order statistics features for classification of magnetic resonance brain images

    (2012)
  • V.P. Rathi, S. Palani, Brain tumor MRI image classification with feature selection and extraction using linear...
  • ZacharakiE.I. et al.

    Classification of brain tumor type and grade using MRI texture and shape in a machine learning scheme

    Magn. Reson. Med.: Off. J. Int. Soc. Magn. Reson. Med.

    (2009)
  • ChenX. et al.

    Automated brain tumor segmentation using kernel dictionary learning and superpixel-level features

  • Cited by (0)

    View full text