Brain Tumor MRI Segmentation Using Hidden Layer Convolutional Neural Networks and Discrete Wavelet Transform Technique With KSVM

Abstract

Early and precise discovery of brain tumors in MRI data is serious for effective analysis and therapy planning. This study presents an automated methodology combining Discrete Wavelet Transform (DWT) with Convolutional Neural Networks (CNNs) for tumor localization, discovery, and cataloguing. The proposed system integrates advanced image processing techniques through a pre-processing, feature removal using a 2D - DWT combined with Grey Level Co-occurrence Matrix (GLCM), and classification with Hidden layer Convolutional Neural Networks (HL-CNNs). This hybrid approach addresses challenges such as noise in MRI scans, enabling accurate differentiation between tumorous and non-tumorous MRIs and further classifying tumors as benign or malignant. The system demonstrates robust performance, achieving high diagnostic accuracy and reliability when tested on Magnetic Resonance Images (MRI) from the Harvard Dataset. By automating the screening process, this methodology highlights the potential of DWT and HL- CNNs to advance brain tumor characterization, offering significant improvements over conventional manual methods.