An Improved Canny Edge Detection for SVM-Based Brain Tumor Image Classification

Abstract

This paper proposes an efficient method for automatic detection and classification of brain tumor. The proposed system incorporates a multi-stage processing that includes preprocessing, segmentation, feature extraction, feature selection, and classification. Preprocessing techniques including median filtering and high pass filtering are employed to enhance image quality; the segmentation process involves Canny edge detection and moving average filtering. Feature ex-traction is performed using Convolutional Neural Network (CNN), and feature selection is conducted using Minimum Redundancy Maximum Relevance (mRMR) to ensure optimal feature representation. Support Vector Machine (SVM) trained on datasets of varying sizes is deployed as a classifier. Evaluation results revealed that the proposed method achieved detection and classification accuracy of 95.83%. The proposed method is expected to facilitate early-detection, improve the accuracy in diagnosis of brain tumors, and ultimately make efficient the work of medical practitioners.