Classification of Human Sperm Head in Microscopic Images Using Twin Support Vector Machine and Neural Network

Abstract

This paper describes a computer-aided diagnosis (CAD) system for the classification of healthy/unhealthy human sperm heads based on texture analysis obtained from the microscopic images. This work is based on 60 human sperm microscopic images selected from the National Research Centre in Egypt. Texture analysis is done by using intensity histogram, Haralick’s Spatial Gray-Level Co-occurrence Matrix (SGLCM), statistical feature matrix (SFM), gray-level run length matrix (GLRLM), and invariant moment’s features. A total of forty-six texture features are computed for healthy as well as unhealthy region of interest (ROI). Since many numbers of the features may effect on accuracy of classification, optimal feature selection is done to eliminate the less informative and excessive features. The feature selection is adapted using twin support vector machine (TWSVM). The accuracy of these features in distinguishing healthy/unhealthy human sperm heads has been evaluated by back-propagation neural network (BPNN) and linear support vector machine (SVM) classifiers. From the results analysis, it was found that ANN classifier gave an overall classification accuracy of 100% with 100% sensitivity. The results show that it is feasible to identify human sperm heads based on texture features extracted from microscopic images. This CAD system is shown to be useful for increased accuracy and increased speed for the classification of healthy/unhealthy sperm heads for improving the fertilization process.