Autoregressive modelling: application to mitosis detection in brain cancer histopathology
D. Vaishali, R. Ramesh, J. Anita Christaline · International Journal of Biomedical Engineering and Technology · 2016
In traditional methods, cancer is diagnosed using clinical pathology, where pathologists inspect biopsy samples and make inferences. These inferences are based on cell morphology and tissue distribution which represent randomness in growth and/or in placement. Computer-Assisted Diagnostics (CAD) aids objective judgement that is based on very large database. This work emphasises the contribution of 2D autoregressive model for analysis and classification of histopathological images. Autoregressive model parameters represent a feature set of histopathological image made from biopsy samples taken from patients. These features are further used for analysis, synthesis and classification. The Yule-Walker Least-Squares (LS) method has been used for parameter estimation. AR parameters provide features for classification of sample in two classes: healthy tissue and malignant tissue. Feature data sets have been classified with Support Vector Machine (SVM) classifier. In view of cancer diagnosis, this work also explains the concepts of sensitivity, specificity and classification accuracy.