Automatic Cervical Cell Classification Using Features Extracted by Convolutional Neural Network

Based on World Health Organization (WHO), cervical cancer is one of deadliest disease in human life. There are some researches related to this issues, but the need of improvement is still needed in order to help doctors in making a decision to the patients. This paper aims to provide a better method in cervical cell classification case indicated by giving better results compared to the previous researches. The proposed method mainly consists of three stages, features extraction, features reduction, and the last is classification. Convolutional Neural Network (CNN) algorithm which has been proven as a good algorithm in image domain was implemented in feature extraction stage. As CNN results high numbers of features, this research proposes a feature reduction stage consist of Linear Discriminant Analysis (LDA) followed by Principal Component Analysis (PCA). Those features were eventually classified by using robust kernel based classifier, Support Vector Machine (SVM) and softmax classifier. The results show that the proposed method has better performance than the previous researches.