Application of support vector based methods for cervical cancer cell classification

Abstract

Cervical cancer is known to be one of the deadly diseases in women. On the other hand, it is one of the most curable cancer if detected early. A well-known method in screening cervical cancer is by performing the Papanicoulaou test or Pap test. However, the screening process is laborious since it requires visual inspection of individual cells by experts. A large number of cases processed by a limited number of experts can lead to misclassification due to human errors. To solve this problem, using an automatic classification method may help improve the screening process. In this paper, we explore various support vector based classifiers, namely support vector machine (SVM), twin support vector machine (TWSVM), and twin-hypersphere support vector machine (THSVM), and test their performance on cervical cancer cell classification in 2-class and 4-class scenarios. The cervical cancer cell dataset named the LCH dataset used in this paper was collected and extracted from Lampang Cancer Hospital in Thailand. The experimental results show that TWSVM is preferable to SVM and THSVM in the cervical cancer cell classification.