Student Presenter(s): Krushang Kamleshkumar Pandya, Caroline Fernandez, Tanvi Patel, Brian Siuni, Sai Srija
Faculty Mentor: Niharika Nath
Department: Biological and Chemical Sciences
School/College: College of Arts and Sciences, New York City

Cancer is the second leading fatal disease after heart diseases. Cervical cancer can be prevented by having regular screenings to find any pre-cancerous cytopathological abnormalities of the cells such as enlarged nucleus and to start early to treat them. The Pap smear test looks for any abnormal or precancerous morphological changes in the cells on the cervix. However, the manual screening of Pap smear in the microscope is subjective, time consuming and prone to human error. Our aim was to quantify morphology data based on their nuclear features and visualize the data for classification into two classes and further into multiclass categories of cell abnormality. The nucleus region of the images were extracted using a greedy active contour model in MATLAB, and various shape features including nuclear area, nuclear perimeter, solidity, and eccentricity were examined. The ground truth data for each shape feature was gathered and this served as a base to compare the visual analysis of the cell classes using Tableau. One of our main analysis is clustering and is based on the K-means algorithm which can group the data into one class based on its relativity to the mean of that cluster. Clusters and potential two and three classes were obtained which we then used to examine the accuracy of the classified cells. The results of the study explain that, at least in the two classes obtained, more than 85% of the plot from clustering was matched with the existing ground truth.