Methods for Detecting Cervical Cancer via Machine Learning with k-NN and Artificial Neural Networks

Abstract

Among women worldwide, cervical cancer and micro classification are two of the most common types of cancer. When cervical
cancer develops, the cells in the cervix either lose their nucleus or undergo morphological changes. Some of the characteristics
of these cells include an abnormal number of nucleuses, damaged or absent cytoplasm, or dissolving cytoplasm, among others.
Cervical cancer cells do not differ much in texture or color from normal cells, making it very difficult to detect the disease
in a microscopic smear test (cervix fluid studied under a microscope).Hence, in order to detect anomalies in cancer detection
systems involving human cells, advanced digital image processing techniques are necessary. Thus, this study proposes an automated,
all-encompassing machine learning method. The suggested method reveals the cervix cell’s nucleus and cytoplasm’s
form and color. The cell’s nucleus and cytoplasm are isolated using a state-of-the-art fuzzy-based method. This method involves
training a KNN or Neural Network using the form and color attributes of the segmented cell units. Then, unknown cervical
cell samples may be categorized using these networks. While ANN achieved a classification accuracy of 54%, KNN achieved an
impressive 88.04%. Including more classifiers may improve the performance of the suggested system even further.