|
 
 |
| REVIEW ARTICLE |
|
| Year : 2021 | Volume
: 5
| Issue : 4 | Page : 357-365 |
|
Identification and categorization of brain tumors using ensemble classifiers with hybrid features
Royappan Savarimuthu Sabeenian, Vadivelan Vijitha
Department of Electronics and Communication Engineering, Sona College of Technology, Salem, Tamil Nadu, India
| Date of Submission | 05-Aug-2021 |
| Date of Acceptance | 15-Oct-2021 |
| Date of Web Publication | 14-Dec-2021 |
Correspondence Address:
Royappan Savarimuthu Sabeenian
Department of Electronics and Communication Engineering, Sona College of Technology, Salem - 636 005, Tamil Nadu
India
 Source of Support: None, Conflict of Interest: None
DOI: 10.4103/bbrj.bbrj_163_21
Day by day the new rate of brain tumor causes is diagnosed. To make use of technology, the standard of lifestyle for many patients was increased their survival rate. Medical imaging modalities were contributing more because the internal formation of the human brain is more complicated to diagnosis. In this field, various modalities are used to detect but the invention of magnetic resonance imaging (MRI) gives more attention, especially to detect brain tumor. The proposed work is focused on segmenting the tumor region in the MR images. The Discrete Wavelet Transform (DWT) is used to extract attributes from tumour images, which is then used with PCA to reduce the dimensionality of the attributes. Gray level co-occurrence method based different features are extracted in the MRI which is given as input to the classification learner such as ensemble, support vector machine, K-nearest neighbor, Naïve, and Fine Tree which are used to classify efficiently. The method is achieved 99% of accuracy according to that trained with 30 images and tested with 100 images. From that tumor images have been split up into benign or malignant. With this trouble-free method is used in MR images to give high accuracy.
Keywords: Brain tumor, discrete wavelet transform, ensemble, fine tree, gray level co-occurrence method, K-nearest neighbor, magnetic resonance imaging, Naïve, principle component analysis, support vector machine
How to cite this article:
Sabeenian RS, Vijitha V. Identification and categorization of brain tumors using ensemble classifiers with hybrid features. Biomed Biotechnol Res J 2021;5:357-65 |
How to cite this URL:
Sabeenian RS, Vijitha V. Identification and categorization of brain tumors using ensemble classifiers with hybrid features. Biomed Biotechnol Res J [serial online] 2021 [cited 2023 Jun 5];5:357-65. Available from: https://www.bmbtrj.org/text.asp?2021/5/4/357/332451 |
| Introduction | |  |
Tumor in the human body refers to unprecedented to growth of cells at any part. An occurrence of tumors in the human brain is referred as brain tumor. Brain tumor can be classified as primary and secondary tumors. While the former is benign and represents immature cancerous cells, the latter is malignant and represents matured cancerous cells. Malignant tumours are graded from I to IV. With Grade I refers to a smaller growth and Grades II and III representing larger cell sizes. Grade IV is a case, wherein cancerous cells begin to spread around the body.
Medical researchers have documented[1],[2] that brain tumor along with other nervous system cancer causes death to mankind. Surveys show that brain tumors are rated as the tenth most prevalent disease. Brain tumor is much common for adults below 40 and for pediatrics. Surveys by loading Indian details[3],[4] show that there has been a vital increase in the rate of cancer patients to 10% in the past 4 years. For a highly populated country like India, this vital increase would give fearful numbers in the years to come. In addition, cancer is the second leading causes of deaths in the United States (U. S), every year in the United States[5] approximately 80000 people are diagnosed with brain tumor (both primary and secondary). In this, 32% are considered as malignant (or cancerous). The cause for brain and spinal cord tumor was diagnosed more in U. S. In recent days, various types of brain tumor detection methods have been identifying the tumor location, yet researchers still focus on developing novel approach for tumors identification and classification. The availability of technological advancement has kindled the authors to carry out diagnosis using magnetic resonance imaging (MRI) scan. This work has focused on identifying abnormal growth of tumor in the human brain using MRI images.
Magnetic resonance imaging image
MRI is one among the approaches being used by medical practitioners for diagnosing soft tissues in the human body. When compared to other approaches, MRI imaging provides a directional imaging of the human body. Image analysis on MRI images is much easier when compared to other scanning approaches.
Discrete wavelet transform (DWT) decomposes a signal into a set of mutually orthogonal wavelets. It is built on sub-band coding. Moreover, the authors contributed more to extracting feature from MR images. For the past two decades, scientists have been utilizing machine learning algorithms for classifying images. To list out a few ensemble classifiers, logistic regression, support vector machine (SVM), k- nearest neighbor (k-nn), Naïve Bayes are widely used.
When the set of classification is called ensemble, it gives high accuracies to accomplish on various models, in this mainly bagged tree is used to get high accuracy compared to all other classification. Logistic regression can be categorized in nature. It is commonly used for binary output (0 or 1) classification. SVM is mainly used in classification purpose. The label trained based on the two groups of classification algorithm (ex: 0 or 1). k-NN is focused to calculate distance from different function. These measurements based on the nearby pixel in the tumor images. Bayes classifier algorithm helps to build quick prediction and it is very easy to predict the probability of the object.
| Literature Review | | |
Ahmad et al.[6] used radial basis function (RBF) and linear function for classifying brain tumor using MRI images. They were able to achieve 98.7% and 94.7% of accuracy, respectively. DAUB-4 wavelet method was utilized to extract features from MR image. Principle component analysis (PCA) method was evolved for getting accurate features classification. SVM kernel function was utilized for prediction.
Zikic et al.[7] presented convolutional neural network-based brain tumor tissues segmentation. Multichannel image intensity information was used for extracting from input image in a windowed manner features; the method proposed in the work has provided better accuracy when compared to other approaches. The developed process was trained with BraTs 2013 (BRATS-SICAS Medical Image Repository [smir.ch]) images.
El Abbadi[8] et al. have used singular value decomposition algorithm to detect brain tumor. The accuracy of the system performance was up to 97%. %. Varuna shree et al.[9] formulated to segment the tumor with help of K-means clustering method. Gray level co-occurrence method (GLCM) and PCA were utilized for extracting the features and then fed to a neural network-based classifier. The method was capable of discriminating, benign, and malignant tumor.
Nilesh bhaskarro et al[10] From MRI images, the scientists employed BWT and SVM to classify the amount of benign and malignant brain tumours. These experimental results show 95% accuracy.
Suriya et al.[11] proposed a couple of approaches to detect the tumor. The method proposed was based on image enhancement and image fusion. DWT techniques were used to get effective accuracy and response from MRI and positron emission tomography images.
Gurbina et al.[12] have proposed a technique to differentiate between tumor and nontumor cells. They have also proposed an approach to detect and classify the range of benign and malignant. The technique was experimented with help of DWT and SVM. The proposed work accomplished 92% of accuracy in classification for binary SVM and binary linear.
Thillakkarasi and Saravanan[13] conceived novel Deep learning algorithms to classify tumor. The M-SVM approach was also applied in this system to achieve effective outcomes. The authors have included various steps to segment the brain tumor. MRI image has been enhanced by using contrast limited adaptive histogram equalization and Laplacian of Gaussian Filtering Method to extracted feature.
Jalali and Kaur[14] investigated various techniques and algorithms were discussed for feature extraction and classification of tumor. To evaluate the experimental results, specificity, sensitivity, and accuracy were calculated.
| Dataset Description | | |
Experiments were conducted on brain MRI dataset. The dataset of brain MR images was resourced from Kaggle.[15] In this study, we have termed these brain MRI images as BT dataset. [Figure 1] shows the sample dataset of BT images. This BT dataset comprises 253 images, out of which 155 images contain malignant tumors while remaining 98 images are with benign tumors. | Figure 1: The examples of brain images showing normal and tumor with benign, malignant from the BT dataset
Click here to view |
| Proposed Method | | |
This study has proposed four steps to segment brain tumors using MRI images. The flow chart for the proposed system to detect tumors in MRI images is shown in [Figure 2].
As shown in the flow chart, the MRI image is initially preprocessed to remove the unwanted noise. Wiener filter minimizes the overall mean square error by the process of inverse filtering and noise smoothing[16] is being used for the above-mentioned step. The preprocessed MRI images are then fed for segmentation. Otsu approach[17] has been used to increase the contrast of the filtered image. DWT[18] and Gray Level Co-occurrence Matrix (GLCM)[19] have been utilized for extracting features from the segmented image. PCA[20] is been applied to the extracted DWT features, such that vital features are extracted. These vital features are fed to the classification system.
This study has used ensemble tree, logistic regression, Naive Bayes, k-NN, and SVM classifiers. The confusion matrixes of each of the classifiers have been identified and the accuracy of the extracted features is computed.
The rest of section is structured as follows:
- Image preprocessing
- Segmentation using Otsu method
- Feature extraction using DWT
- Feature selection using PCA
- Image classifier.
Preprocessing[25] includes removal of unnecessary noise, smoothening inner regions, and framing the edges. When the wiener[21] filter is used to minimise undesirable noise in an MRI picture to estimate the desired noiseless signal performance. A statistical-based approach is present, i.e. when BT image is blurred by low pass filter. With help of inverse filtering can get back to the original image. This filter is based on additive noise with sensitive. However, this approach is causing degradation at a critical period when restoration approaches are being applied.
The techniques come up with optimal trade-off between reverse filter and noise smoothing filter. From this, it is used to eliminate the additive noise signal also blurs sequentially. Wiener filtering is foremost used to minimize the errors in the process of noise smoothing and inverse filtering. The linear evaluation of original image gets back in the Wiener filter. When the orthogonal theory implicit that filter system can be expressed as,
Where,
fxx(s1,s2) fnn(s1,s2)
- Power spectra of the additive noise and the original image,
H(s1,s2)- Blurring filter.
When the Wiener filter is separated into two parts, one belongs to inverse filtering and another one belongs to noise smoothening. In this filter not only shows the high-pass filtering but also helps to remove noise in low-pass filtering during this execution process.
The proposed study has utilized different classifiers. Each of the classifiers has their own input layer with different sizes of input to be given. To reduce the computational complexity, the authors have fixed the size of the preprocessed image to 216 × 216.
From this process, it is important to resize the MRI images to eliminate the unwanted information and process only required information from the MRI images. With help of this preprocessing, the pixel strength of every input image is defined clearly for getting the enhanced results.
Segmentation
In brain tumor image, image segmentation is a big challenging process to segment the part of tumor. In this method, Otsu-based thresholding is concentred to segment the MRI brain tumor images. Otsu method is easy and effective for segmenting images.
Segmentation means dividing an image into several parts on basis of similarities or dissimilarities based on this property. This is done to extract useful information from segmented image. In the case of brain tumor, it segments the tumor part. Thus, it is an extremely tricky. The threshold-based techniques give rise to segment own pixels with indistinguishable intensities. It is handy approach for initiate boundaries in that tumor objects contain on contrasting backdrop, for complete binary algorithm is 0 or 1 (benign or malignant) in [Figure 3] shows segmented part of cluster and segmented image.
During this segmentation process, initially, the Otsu binary threshold method is used to convert the gray level image (Sushil Kumar et al.) and to measure the distance, Euclidean Metric is used. This clustering allows you to select clusters from three different cluster images. Cluster 1 is chosen in this case based on the segmented tumour area.
Feature extraction
The feature extraction is used to extract the valid or essential information from the MRI images. There are different approaches were used to extract the feature from MRI. In this segmenting process, the extraction of tumor image is very important. In this study of tumor detection, the extraction is deal with Gray-Level Co-Occurrence Matrix (GLCM) features, DWT based brain tumor segmentation to reduce the complexity of the system and improve the performance.
It is the process of extracting valued information from the segmented tumor image (it may be size, shape, color, and texture).
In brain tumor images, the defined various features are extracted from MRI image. The segmented part of the tumor has different in nature based on the prototype of MRI image; from these entire images, it is captured and derived.
Due to the feature extraction and analysis, the dimensionality of the segmented image is given more clear and improved. Hence, the feature extractions of the dimensionality of MRI dataset have to be decreased for increasing the overall MRI brain tumor recognition process. For this purpose, in the future, the feature selection approach is going through to optimize the BT features.
Discrete wavelet transform
In general,[26] the DWT was used to examine different level frequencies of various images using different variety of scales. DWT is more useful tool for extraction features from extraction. When the features are extracted from the coefficients, discrete wavelet is used to extract the brain tumor MRI images. In this classification, the importing wavelet-based localized frequency information is important, which is based on 2D DWT and so on.
DWT is an essential ride for extracting feature from tumor images since it allows investigation of images placed at numerous amount of resolution by cause of multiscale approximation property. MRI information is retrieved successfully with help of multiscale analytic. The 2D transform is feasible to decompose into four nonidentical resolutions of sub-bands such as low-low (LL), low-high (LH), high-low (HL), and high-high (HH) (both levels of frequencies in one and the other directions). By applying the DWT model, four sub-bands are framed based on the Region of Interest.
The sub-bands of DWT systems are classified into four ways that mentioned in above is described in detail manner that is,
[Figure 4]a describes normal and first-level decomposition, from that image easy to get back the level of composition and execution process of the 1DWT. | Figure 4: (a) Shows input image with the 1st level decomposition, (b) 2nd level decomposition. (LH1, HH1, HL1-1st level, LH2, HH2, HL2-2nd level, LH3, HH3, HL3-3rd level decomposition)
Click here to view |
In the second and third-level decomposition shows [Figure 4]b, it deals with high-frequency part of the image those are represented as the LH2, HL2, HH2, LH3, HL3, and HH3 gives the detailed explanation of all directions such as horizontal, vertical, and diagonal part the MRI image. As of now know that LL1 denotes the approximation of MRI original image and it is processed for next level decomposition. This process is frequently repeated to get desired level of image resolution.
In this, the step is based on:
- First-order statistical analysis
- Second-order statistical analysis.
Extraction based on DWT + PCA based statistical features listed as,
From this step, the first-order statistical features[27] were extracted based on the frequencies of gray level, and image intensities at different random level were extracted[28] in the MRI images.
Mean (M):
Standard deviation (S. D): the square root of difference is split by whole digit representative,
Mean of variance (V):it measures as first-order statistical feature value of variance calculated as,
Entropy (E): Used to measure the disarray or difficulty of image f (x, y), the entropy of image calculated as,
Skewness (Sk): with help of skewness can measure the probability distribution of asymmetry of real value,
Kurtosis (Ku): it is heavy-tailed data or light-tailed data based on the distribution of probabilities,
Inverse difference moment (IDM): To compute image homogeneity and local uniformity of an image, it detected between valued and nonvalued
Root mean square (RMS): It measures the predictable error of pixel strength and square root of arithmetic mean,
Smoothness (SM): it can be measured by number of values which are continuous in the image,
Contrast (Cont): it can be calculated as highest and lowest values of pixel intensities of the image,
Extraction based on gray level co-occurrence matrix
To extract the information[29], GLCM-based extraction is used. The likelihood of various levels at different frequencies are retrieved in the MRI picture in this second order feature extraction.
Energy (E): The amount of assessable pixel pairs of an image,
Homogeneity (H): Used to measure the nearness of elements in the distribution,
Correlation (C): is defined as contiguous feature of pixels depend on computations,
By these methods of feature extraction[30], the derived features are given for the classification model for training and testing, for efficient tumor detection in MRI scan images of the brain.
Feature selection by principal component analysis
The PCA is a conventional method; it focuses on linear data between high to low dimensional image subspace by least-square decomposition. To reduce the structure of the data[31], the dimensionality reduction method was used in the system with help of this system; it can reduce the size of the image as per out expectation. PCA is nearly new in exploratory statistical analysis for building predictive models. It is mainly focused on dimensionality reduction in the tumor dataset images. PCA is possible to reduce the size of a data extracted tumor be made up of a large quantity of interrelated adaptable variables while retain most of inequality contrast variations.
For example, PCA method trained as, when the number of useful information is retrained it based on the eigenvalue analysis for distinct criteria-based execution, eigenvalues are larger than one or less than one in this analysis procedure.[32]
Image classifier
Logistic regression
The logistic regression analysis is apt for testing hypotheses and describing the stages, it also well suited for relationship-based variables in categorical outcome. Logistic regression also deals with one or more or continuous variables predictors used (Chao-Ying Joanne Peng).
Ensemble
An ensemble classifier (Mohamed Shakeel et al.) contains a set of trained individual classifiers; it may be neural networks or decision tress, these predictions were combined based on the classifying system. The method for producing ensemble tree the system concentrates with bagging and boosting. For combining more than one classifier, this system is effectively used to get high accuracy. This bagging is more accurate in the ensemble method compare to boosting ensembles
Support vector machine
SVM Vikramaditya Jakkula et al,[22] classifier is used to analyse data that is regularly performed during the execution process. In this, the regions are mapped together with the help of points in the system. This system is contributed more on the image segmentation process. From this method, accuracy rate and performance can be measured.
Naïve Bayes
This classifier is branch of Bayes theorem (T M. Mitchell et al)[23]; it is called as a Naïve Bayes classifier with inter-strong relationship-based features. In this Naive Bayes, each feature problem is models around independent of the other system. The system is trained by various numbers of classification rules-based approach to find the statistics.
K-nearest neighbor
k-NN (Gongde Guo et al)[24] is based on learning process, in this method, it allows all the data for training in this classification system. It is one of the lazy systems for learning method. From this, the whole training data can be found out the best one for improving their classification accuracy.
| Performance Matrices | | |
To compute the level of performance of the system, the classifier of the performance matrix is calculated. These parameter measures accuracy, sensitivity, specificity are calculated using this below formula,
(TP-True Positive, TN-True Negative FP-False Positive, FN-False Negative)
| Results and Discussion | | |
The experimental performance of the proposed work deals with tumor identification method. In these 100 images are used to classify the stages, which are begin and malignant stage of the tumor (the images obtained from Kaggle dataset). In view of this segmentation process, tumor area detected with the help of DWT and PCA techniques and the performance also calculated. Extraction of various feature is listed below in [Table 1]. In consonance with feature, values of all images to be benign or malignant are displayed. | Table 1: Magnetic resonance tumor (benign and maligned) image features values
Click here to view |
In this proposed work to classify the tumor images, classification learner is used. With help of MATLAB features are extracted based on DWT (GLCM) and PCA techniques described in the kernel function which are linear, radial bias, quadratic, polygonal accuracy. In each stage of testing and training, dataset is used in the function to get high accuracy.
Feature extraction using discrete wavelet transform (GLCM) and principle component analysis
When the brain tumor image is loaded as input in the feature extraction, the following process is done to detect the tumor which is benign or malignant.
- Brain tumor image is loaded as input
- A segmented image is displayed, with 13 features extracted.
- The type of tumor is diagnosed (benign or malignant)
- Accuracy of RBF, polygonal, linear, and quadric is calculated.
Classification learner
The [Table 2] shows, comparison of Confusion Matrix in this classification learner. The listed confusion matrix TP (True Positive), FN (False Positive), PP (Positive Predictive Values), FDR (False Discovery Rates) can identify the classification rate of the classifier as Specificity, Sensitivity, and Accuracy. The output show in [Figure 5], [Figure 6], and [Figure 7].
The accuracy of categorization is compared in the table below; five different classifications are tested for high accuracy. Compare to the entire classification method ensemble bagged tree gives more accuracy because set of classification is already in build to reach maximum accuracy. With help of bagged tree, 99% accuracy is achieved.
| Conclusion | | |
To a great extent, lot of investigation method is accessible in brain tumor research. From this, everyone is trying to get good accuracy to extract location of the tumor. In such a manner, the proposed work given a 99% accuracy shown in [Figure 8] to detect which is benign or malignant. The whole approach is carried out DWT and PCA (GLCM) techniques along with classification learner algorithm. Since the classification learner strategy aids multiple classifiers in detecting tumours in MR images with great accuracy. In the future, the outcome may be improved to identify the age and various grade levels by adopting more incomparable classifier.
Financial support and sponsorship
Nil.
Conflicts of interest
There are no conflicts of interest.
| References | | |
| 1. | |
| 2. | American Society of Clinical Oncology (ASCO). Available from: http://www.asco.org. [Last accessed on 2020 Aug 14]  |
| 3. | |
| 4. | |
| 5. | |
| 6. | Ahmad M, Ul-Hassan M, Shafi I, Osman A. Classification of tumors in human brain MRI using wavelet and support vector machine. IOSR J Comput Eng 2021;8:25-31. |
| 7. | Darko Zikic, Yani loannou,Matthew brown and Antonio Criminisi,“Segmentation of brian tumor tissues with Convolutional Neural networks”, MICCA 2014, BRATS challange, 36-39. |
| 8. | El Abbadi NK, Kadhim NE. Brain tumor classification based on singular value decomposition. Int J Adv Res Comput Commun Eng 2016;5:553-7. |
| 9. | Varuna Shree N, Kumar TN. Identification and classification of brain tumor MRI images with feature extraction using DWT and probabilistic neural network. Brain Inf 2018;5:23-30. |
| 10. | Nilesh Bhaskarrao Bahadure, Arun Kumar Ray, and Har Pal Thethi,” Image Analysis for MRI Based Brain Tumor Detection and Feature Extraction Using Biologically Inspired BWT and SVM”, Hindawi International Journal of Biomedical Imaging Volume 2017, Article ID 9749108, 12 pages https://doi.org/10.1155/2017/9749108. |
| 11. | Udhaya Suriya TS, Rangarajan P,” Brain tumour detection using discrete wavelet transforms based medical image fusion” Biomedical Research 2017; 28 (2): 684-688, ISSN 0970-938X, www.biomedres.info. |
| 12. | Gurbina, M., Lascu, M., & Lascu, D. (2019). Tumor Detection and Classification of MRI Brain Image using Different Wavelet Transforms and Support Vector Machines. 2019 42nd International Conference on Telecommunications and Signal Processing (TSP). doi:10.1109/tsp.2019.8769040. |
| 13. | Thillaikkarasi, R., & Saravanan, S. (2019). “An Enhancement of Deep Learning Algorithm for Brain Tumor Segmentation Using Kernel Based CNN with M-SVM”, Journal of Medical Systems, 43(4).Article number : 84, doi:10.1007/s10916-019-1223-7. |
| 14. | Jalali V, Kaur D. A study of classification and feature extraction techniques for brain tumor detection. Int J Multimed Inf 2020;9:271-90. |
| 15. | Shanthi, T., Sabeenian, R.S., Manju, K., Paramasivam, M.E., Dinesh, P.M. and Anand, R., 2021. Fundus Image Classification using Hybridized GLCM Features and Wavelet Features. ICTACT Journal on Image and Video Processing, 11(3), pp.2372-2375, year of publication 2021. |
| 16. | S.Jayaraman, S.Esakkirajan, T.Veera Kumar, “Digital Image Processing”, Mc Graw Hill Education (India) private limited, second edition. print-book edition ISBN(13) : 978-93-89811-92-6, ISBN (10) : 93-89811-92-9,2020. |
| 17. | Martin-Fernandez M, Alberola-Lopez C, Ruiz-Alzola J, Westin CF. Sequential anisotropic Wiener filtering applied to 3D MRI data. Magn Reson Imaging 2007;25:278-92. |
| 18. | S.Jayaraman, S.Esakkirajan, T.Veera Kumar, “Digital Image Processing”, Mc Graw Hill Education (India) private limited, second edition. print-book edition ISBN(13) : 978-93-89811-92-6, ISBN (10) : 93-89811-92-9,2020. |
| 19. | P. Mohanaiah , P. Sathyanarayana, L. GuruKumar,” Image Texture Feature Extraction Using GLCM Approach” International Journal of Scientific and Research Publications, May 2013 Volume 3, Issue 5, page number: 290 -294, ISSN 2250-3153. |
| 20. | Peng, Chao-Ying Joanne, Kuk Lida Lee, and Gary M. Ingersoll. “An introduction to logistic regression analysis and reporting.” The journal of educational research 96, no. 1 (2002): 3-14. |
| 21. | Shakeel, P.M., Burhanuddin, M.A. and Desa, M.I., 2020. Automatic lung cancer detection from CT image using improved deep neural network and ensemble classifier. Neural Computing and Applications, pp.1-14. |
| 22. | Jakkula, V., 2006. Tutorial on support vector machine (svm). School of EECS, Washington State University, 37. |
| 23. | |
| 24. | Guo, Gongde, Hui Wang, David Bell, Yaxin Bi, and Kieran Greer. “KNN model-based approach in classification.” In OTM Confederated International Conferences” On the Move to Meaningful Internet Systems”, pp. 986-996. Springer, Berlin, Heidelberg, 2003. |
| 25. | Rajeshwari, S., and T. Sree Sharmila. “Efficient quality analysis of MRI image using preprocessing techniques.” In 2013 IEEE Conference on Information & Communication Technologies, pp. 391-396. IEEE, 2013. |
| 26. | Khalili M, David A. Colour spaces effects on improved discrete wavelet transform-based digital image watermarking using Arnold transform map. IET Signal Process 2013;7:177-87. |
| 27. | Mustaf M, Sali AF, Illzam EM, Sharifa AM, Nang MK. Brain cancer: Current concepts, diagnosis and prognosis. IOSR J Dent Med Sci 2018;17:41-6. |
| 28. | Bahadure, N.B., Ray, A.K. and Thethi, H.P., 2017. Image analysis for MRI based brain tumor detection and feature extraction using biologically inspired BWT and SVM. International journal of biomedical imaging, 2017. |
| 29. | Ketkar D. Classification of Brain Tumors Using Principal Component Analysis and PNN Classifier, Proceedings of 12 th IRF International Conference. Vol. 17. 2015. |
| 30. | Shree NV, Nandpuru, Babu H, Salankar SS, Bora VR. MRI Brain Cancer Classification using Support Vector Machine. Electrical Electronics and Computer Science (SCEECS), 2014 IEEE Students' Conference on IEEE; 2014. |
| 31. | Haidekker M. Advanced biomedical image analysis. John Wiley & Sons; 2010 Dec 28. |
| 32. | Usman Akram M, Usman A. Computer Aided System for Brain Tumor Detection and Segmentation. International Conference on Computer Networks and Information Technology. Pakistan: IEEE; 2011. [doi: 10.1109/iccnit.2011.6020885]. |
[Figure 1], [Figure 2], [Figure 3], [Figure 4], [Figure 5], [Figure 6], [Figure 7], [Figure 8]
[Table 1], [Table 2]
|
Search Pubmed for