Performance Analysis of Gabor-Wavelet based Features in Classification of High Grade Malignant Brain Tumors

Incorporating machine learning based concepts and knowledge in medical science domain elaborates the interdisciplinary research areas. Using such the identification of the disease is accomplished at initial state of its generation with better accuracy as compared to naked eye. In past, lot of feature extraction approaches were proposed to extract the features in spectral domain. Among all approaches, two major approaches was Gabor filter based feature extraction and Wavelet based feature extraction mechanism. But it leads to the kiosk that which feature extraction mechanism is better for classification perspective. This paper focuses on machine learning based approach for performance analysis for the grade of malignant tumor types using a diverse feature set based on Gabor and Wavelet transformation. Further, various feature selection (FS) algorithms are taken into consideration for selection of the best feature set among the feature vector. For experimental purpose, several state-of-art classifiers are used for analysis of the performance for the classification of malignant brain tumors in magnetic resonance (MR) images. In experimentation several high grade malignant brain tumors like Central Neuro Cytoma (CNC), Glioblastoma Multiforme (GBM), Gliomas, Intra Ventricular Malignant Mass, and Metastasis are taken into consideration having 30 images of each tumor type. The classification accuracies achieved using various combinations of FS-Classifier's are presented in detail.