Pathological Brain Tumor Detection Using CLAHE and LS-SVM

Abstract

The segmentation, early detection and removal of infected tumor region from Magnetic resonance images is a main problem but tedious and time-consuming task conducted by radiologists and their precision depends only on their knowledge. To solve these constraints, it becomes very important to use computer-aided technology. In this study, the medical image involves improving performance and reducing complexity. This paper proposes an efficient PBDS based on MR images that significantly enhances recent results. To improve the quality of input of MR images, the proposed system uses CLAHE. Subsequently segmented using OTSU and K means segmentation methods. On the segmented image, morphological operations are performed to obtain the information about the tumor area, size and density. Using a discrete wavelet transform (DWT) strategy, the segmented image is then transformed to extract features. Subsequently, the PCA approach reduce the dimensionality of the features. The reduced features were submitted to a Least square support vector machine (LS-SVM). The strategy of 5×k-fold stratified cross validation (SCV) test has been carried out to enhance LS-SVM generalization. We performed our proposed methods with four di?erent kernels and found that the GRB kernel has the highest classi?cation accuracy of 99.38%. The LIN, HPOL, and IPOL kernel achieves 95%, 96.88%, and 98.12%, respectively. We also compared our method to those from literatures in the last decade, and the results showed our CLAHE+DWT+PCA+LS-SVM with GRB kernel still achieved the best accurate classi?cation results It could be applied to the ?eld of MR brain image classi?cation and can assist the doctors to diagnose where a patient is normal or abnormal to certain degrees at the early stage.