Analysis of Active Contours Without Edge-Based Segmentation Technique for Brain Tumor Classification Using SVM and KNN Classifiers

Classification of brain tumors using machine learning technology in this era is very relevant for the radiologist to confirm the analysis more accurately and quickly. The challenge lies in identifying the best suitable segmentation and classification algorithm. Active contouring segmentation without edge algorithm can be preferred due to its ability to detect shapeless tumor growth. But the perfectness of segmentation is influenced by the image enhancement techniques that we apply on raw MRI image data. In this work, we analyze different pre-processing algorithms that can be applied for image enhancement before performing the active contour without edge-based segmentation. The accuracy is compared for both linear kernel SVM and KNN classifiers. High accuracy is achieved when image sharpening or contrast stretching algorithm is used for image enhancement. We also analyzed that KNN is more suitable for brain tumor classification than linear SVM when active contouring without edge method of segmentation technique is used. MATLAB R2017b is used as the simulation tool for our analysis.