GLCM and GLRLM Based Texture Analysis: Application to Brain Cancer Diagnosis Using Histopathology Images

Abstract

Objective This work focuses on design and development of an automated diagnostic system using brain cancer histopathology images. Background Detection and classification of cancerous tissues are one of the challenging tasks in histopathology image analysis. The features based on cell morphology, cell distribution, and randomness in growth and placements have been considered as important aspects of the cancer diagnosis. Results of the conventional methods are subjective and dependent on the skills of histopathologists. The computer-assisted diagnosis (CAD) offers the diagnostic results based on standard algorithms and standard test database. Material and Methods This work shows the contribution of gray-level co-occurrence matrix (GLCM) based Haralick features and gray-level run length matrix (GLRLM) features in analysis and classification of Brain histopathology images. The primary focus of this study is to analyze the complex random field of cancer images with co-occurrence matrix features, run length matrix features, and also the fusion of both features. These features are used for classifying healthy and malignant tissue. The SVM classifier with RBF kernel has been used for classification. The Confusion matrix is formed showing true positive (TP), true negative (TN), false positive (FP), false negative (FN) information. The classifier performance has been described with sensitivity, specificity, precision, accuracy, and F-score measures. Results Outcome of this work shows that GLCM-based and GLRLM-based features offer excellent discriminating features for statistical study of histopathology images and can be useful for cancer detection. Overall accuracy improvement is seen by fusing the GLCM and GLRLM based texture features. Conclusion The work describes an innovative way of using GLCM and GLRLM based textural features to extract underlying information in brain cancer imagery.