Performance evaluation of machine learning techniques for screening of cervical cancer

This paper presents comparative analysis of various machine learning algorithms in order to evaluate their predictive performance for screening of cervical cancer by characterization and classification of Pap smear images. Papanicolaou smear (also referred to as Pap smear) is a microscopic examination of samples of human cells scraped from the lower, narrow part of the uterus, called cervix. The sample is observed under microscope for any unusual developments indicating any precancerous and potentially precancerous changes. Examining the cell images for abnormalities in the cervix provides grounds for provision of prompt action and thus reducing incidence and deaths from cervical cancer. Pap smear test, if done with a regular screening programs and proper follow-up, can reduce cervical cancer mortality by up to 80% [1]. Authors have applied fifteen different machine learning algorithms under different platforms over two databases and evaluated their screening performances for prognosis of cervical cancer. The results indicate that among all the algorithms implemented, the Ensemble of nested dichotomies (END) is the best predictor and Naïve Bayes was the worst performer.