An Ensemble Machine Learning Method for Single and Clustered Cervical Cell Classification

Cervical Cancer was in recent history a major cause of death for women of childbearing age. This changed when in the 1950s the Papanicolaou (Pap smear) test was introduced to identify and diagnose cervical cancer in its infancy. The introduction of the Pap smear test dropped cervical cancer related deaths by 60% but still approximately 4,210 women die from cervical cancer in the United State annually. The goal of our research is to aid in the methods of identifying and classifying cervical cancer used in the Pap smear or Liquid-based Cytology (LBC) with cutting edge machine vision, and ensemble learning techniques. The contribution of this research is to develop an automated Pap smear screening system that identifies cells within a cervical cell slide sample and classify cells and clusters of cells as abnormal or normal as defined by the Bethesda System for reporting cervical cytology. Achieving an accuracy of 90.4% when evaluated with a five-fold cross-validation demonstrates promise in the creation of an automated Pap smear screening test.