Design Implementation and Assessment of Efficient Brain Tumor Detection and Classification System Using Improved Machine Learning Techniques

Main Article Content

Nidhi Mittall, Yashika Saini, Manish Kumar Mukhija, Satish Kumar Alaria

Abstract

A tumor is a mass of tissue formed by the accumulation of abnormal cells that continue to grow. The brain is the most important organ in the human body, responsible for controlling and regulating all critical life functions for the body, and a tumor is a mass of tissue formed by the accumulation of abnormal cells that continue to grow. A brain tumor is a tumor that has grown in the brain or has spread across the brain. To yet, no one reason has been found for the development of brain tumor. Though brain tumor are uncommon (approximately 1.8 percent of all recorded cancers in the world), the death rate of malignant brain tumor is quite high due to the tumor's location in the body's most vital organ. As a result, it is critical to accurately detect brain tumor at an early stage in order to reduce mortality. As a result, we've presented a computer-assisted radiology system that will assess brain cancers from MRI scans for brain tumor diagnosis treatment. In this paper, we developed a model that accurately separates and extracts characteristics from images using the DWT and PCA techniques, as well as SVM. For medical analysis and interpretation, automated and reliable classification of MR brain images is critical. Several techniques have previously been proposed in the recent decade. We introduced a novel method for classifying a given MR brain image as normal or abnormal in this paper. The proposed approach used wavelet transform to extract features from images, then PCA to minimize the dimensionality of the features. A kernel support vector machine was used to process the reduced features (KSVM). To improve KSVM generalization, the K-fold cross validation approach was applied. The accuracy of the LIN, HPOL, and IPOL kernels is higher. We also compared our technique to previous literatures, and the results revealed that our DWT+PCA+KSVM with specific kernel method still produced the most accurate classification results.

Article Details

Section
Articles