Cell Nuclei Classification and Immunohistochemical Scoring of Oral Cancer Tissue Images: Machine-learning Approach

Abstract

Listen

Translate article

In this paper, a novel cell nuclei feature extraction method has been proposed for immunohisto chemical (IHC) scoring of oral cancer tissue images. This method is based on maximal separation approaches, which was mainly used to find immuno-positive area of tissue sections. The performance of of the proposed method is evaluated using machine-learning classifiers such as support vector machine (SVM), k-nearest neighbor(k-NN), linear discriminant analysis(LDA), and naive bayes, and compared the results with other state-of-art methods including intensity and texture features. According to the performance outcomes, the best classification accuracy of 96.09% is achieved by the proposed method for LDA classifiers. However, an automatic IHC scoring technique has been implemented using highly correlated feature elements as input to the classifiers. The statistical analysis between manual and automatic techniques (LDA) revealed that the agreement between both techniques have high correlation coefficient (CC), low mean absolute difference(MAD), and within the range of 95% confidence interval(CI) of manual techniques (CC>0.966, MAD 0.649). Therefore, the proposed feature extraction method and automatic scoring technique have high potential in IHC stained tissue image analysis.