3. Brain tumor image segmentation and classification using SVM, CLAHE, and ARKFCM

Abstract

With the modern lifestyle and environmental changes, many life-threatening diseases are stealing the normal livelihood of human society. One of these diseases is brain tumor, which if not detected and treated on time may even cause loss of life. For any disease to be treated on time, early detection is the key factor. A tumor is generally an abnormal or a malignant growth of tissues in any organ of human body and does not contribute toward any physiological functionality. Magnetic resonance imaging (MRI) is one of the most widely used scanning techniques where magnetic and radio waves are used to create a detailed study of the bones and tissues of the targeted area for detecting any abnormal growth. To detect the MRI images efficiently and to locate the tumor position accurately, a support vector machine (SVM) technique is used. The imaging classification totally depends on the quality of the image in terms of contrast, illumination, blurring, and so on. Thus, an equalization technique called contrast limited adaptive histogram equalization is used for improving the image contrast. Along with this pixel adjustment techniques are also used for quality enhancement of the image. Tumor image segmentations are done using adaptively regularized kernel fuzzy C-means clustering algorithm, threshold algorithm, and morphological operations. These techniques rely on the image gray level intensity. This segmentation procedure helps to extract the information about the abnormal growth of the tissues related to some physical parameters such as area, perimeter, and the overall size of the tumor. The extracted parameters are taken into account to dictate whether the tumor condition is normal or malignant. High level of accuracy is required to classify and segment the image that could give an accurate result to segregate a normal tumor and a malignant tumor. The effective proposed method decreases the variance of the prediction error and increases the flexibility to detect the nature of the tumor, which extend helping hands to the medical practitioners. The majority of patients affected from brain tumor throughout the world do not come from affluent financial background. The proposed technique for detecting and classifying brain tumor cells proves to be affordable and efficient.