Detection and Classification of Brain tumor tissues from Noisy MR Images using hybrid ACO-SA based LLRBFNN model and modified FLIFCM algorithm

Abstract

This paper presents a hybrid ACO-SA (Ant colony optimization- Simulated Annealing) based LLRBFNN (Local Linear Radial Basis Function Neural Network) model for detection and classification of brain tumor tissues from noisy images. Denoise of rician affected noisy images becomes a complex and difficult task. To denoise the MR (Magnetic Resonance) images the FLICM (Fuzzy Local Information C-Means) segmentation algorithm has been considered initially and modified the cost function of the algorithm for better segmentation result. The images are segmented using Modified FLICM algorithm and the features are extracted by using GLCM (Gray Level Co-occurrence Matrix) feature extraction technique which are fed as input to the proposed ACO-SA based LLRBFNN model for the purpose of classification of malignant and benign tumors from the MR (magnetic resonance) images. The proposed model has been compared with (particle swarm optimization) PSO-LLRBFNN, (Adaptive Particle Swarm Optimization) APSO-LLRBFNN models and the comparison results are presented. It is found that the proposed ACO-SA based LLRBFNN algorithm shows better results than the conventional methods.