Image registration and fusion based Brain tumor detection methods using Machine learning algorithms

Abstract

The uncontrollable developments of cells in human brain are called as tumors which are generally classified into benign and malignant based on the cells in the human brain. In case of benign tumor regions, the cells are inactive and it does not spread to the nearby regions in the brain. In case of malignant tumor regions, the cells are active and it spreads to the other regions of the brain. The benign tumor cells can be controlled and cured by proper medication and it does not lead to death. The malignant tumor cells cannot be controlled and cured by medication. It can be controlled by only surgical operations in the affected brain newlineregions. Generally, the brain regions are scanned by Computer Tomography (CT) and Magnetic Resonance Imaging (MRI) techniques. In this work, MRI scanning approach is used to detect the abnormal tumor regions in brain due to the high evel of accuracy. This research work proposes an efficient and automated computer aided methodology for brain tumor detection and segmentation using image registration technique and classification approaches. This proposed work consists of the following modules as image registration, Contourlet transform, and feature extraction with feature normalization, classifications and segmentation. The extracted features are optimized using Genetic Algorithm (GA) and then Adaptive Neuro Fuzzy Inference System (ANFIS) classification approach is used to classify newlinethe features for the detection and segmentation of tumor regions in brain Magnetic Resonance Imaging (MRI). The quantitative analysis are performed to evaluate the proposed methodology for brain tumor detection using sensitivity, specificity, segmentation accuracy, precision and Dice similarity coefficient. This work research work also proposes an efficient approach for developing the brain tumor detection framework using fusion based classification approach. The brain MRI images from open access dataset are fused with each other to enhance the internal low resolution border pixels. newline newline