Histopathological image classification scheme for breast tissues to detect mitosis

Abstract

Histopathological images are widely utilized by the pathologists to determine the severity of the cancer. Usually, the pathologists analyse histopathological images manually, which demands more time and effort. The quality of image interpretation can be improvised with the help of a computerized assisting system for image analysis. This idea enhances the accuracy of the system as the computerized system helps in making decisions, which the pathologist may recheck. Understanding the benefits, this article presents a histopathological image classification scheme meant for breast tissues in detecting the mitotic cells. The goal is achieved by pre-processing the image by performing top and bottom hat transformation followed by stain normalization. The nuclei are then located with the help of Localized Active Contour Model (LACM) combined with Lion Optimization (LO) algorithm. The shape, statistical and texture features are extorted from the areas of interest and the differentiation of mitotic cells from the non-mitotic ones is performed by Support Vector Machine (SVM) classifier. The work efficiency is tested with respect to the standard performance measures such as accuracy, sensitivity, specificity and time consumption.