CHILDHOOD MEDULLOBLASTOMA DIAGNOSIS USING MULTISCALE FRAMEWORK

This paper proposes an efficient Shearlet Based Childhood MedulloBlastoma (SBCMB) detection system. It is a classification system that extracts prominent characteristics for childhood MedulloBlastoma diagnosis from a given collection of histopathological images. Then, those extracted features are used with specific decision-making algorithms to categorize the histopathological images. After representing the histopathological images by Shearlet, a multiscale framework, Improved Sequential Forward Selection (ISFS) algorithm, is employed to select the dominant feature subset. Finally, Multi-Layer Perceptron (MLP) with ten hidden layers is utilized for melanoma classification. Based on the results of the evaluation of the SBCM system, it seems that the classification may be accomplished exclusively by the ISFS-based features and that the accuracy of the classifier depends on these selected features. The SBCM system provides 97.62% accuracy on 10x magnified images and 98.77% on 100x magnified images for childhood MedulloBlastoma diagnosis.