Feature enhancement framework for brain tumor segmentation and classification

• DOI: 10.1002/jemt.23224
• Publication Date: 2019-02-15T16:02:19Z
• Authors: Bilal Tahir; Sajid Iqbal; M. Usman Ghani Khan; Tanzila Saba; Zahid Mehmood; Adeel Anjum; Toqeer Mahmood
• Keywords: Feature (linguistics)

Abstract Automatic medical image analysis is one of the key tasks being used by community for disease diagnosis and treatment planning. Statistical methods are major algorithms consist few steps including preprocessing, feature extraction, segmentation, classification. Performance such statistical an important factor their successful adaptation. The results these depend on quality images fed to processing pipeline: better images, higher results. Preprocessing pipeline phase that attempts improve before applying chosen method. In this work, popular preprocessing techniques investigated from different perspectives where grouped into three main categories: noise removal, contrast enhancement, edge detection. All possible combinations formed applied sets which then passed a predefined Classification calculated using measures: accuracy, sensitivity, specificity while segmentation dice similarity score. Statistics five high scoring reported each data set. Experimental show application proper could classification greater extent. However, characteristics type set used.