Ensemble Classifications of Wavelets based GLCM Texture Feature from MR Human Head Scan Brain Slices Analysis

Abstract

This paper presents an automatic image analysis of multi-model views of MR brain using ensemble classifications of wavelets based texture feature. Primarily, an input MR image has pre-processed for an enhancement process. Then, the pre-processed image is decomposed into different frequency sub-band image using 2D stationary and discrete wavelet transform. The GLCM texture feature information is extracted from the above low-frequency sub band image of 2D discrete and stationary wavelet transform. The extracted texture features are given as an input to ensemble classifiers of Gentle Boost and Bagged Tree classifiers to recognize the appropriate image samples. Image abnormality has extracted from the recognized abnormal image samples of classifiers using multi-level Otsu thresholding. Finally, the performance of two ensemble classifiers performance has analyzed using sensitivity, specificity, accuracy, and MCC measures of two different wavelet based GLCM texture features. The resultant proposed feature extraction technique achieves the maximum level of accuracy is 90.70% with the fraction of 0.78 MCC value.