Automated Screening of Cervical Cancer Cell Images

Abstract

Cervical cancer is the first-most common type of female cancer and the second leading cause of death in Thailand. The number of cervical cancer is increasing in every year, even though it is preventable by the screening in early detection. The most popular method for the screening is so-called Pap smear test via examining morphology change in cervix cells. The aim of this research is to implement an image processing algorithm for classifying Pap smear cell images by calculating nucleus-to-cytoplasm area ratio. The algorithm used to classify the nucleus was mathematically calculated through k-mean clustering. The cytoplasm area was calculated from its edge profile relating to geometrical rotation method. Finally, the abnormal cells can be segmented using the area of nucleus-to-cytoplasm ratio with the accuracy of detection at 79%.