SPS22-110UL

Applying Machine Learning Segmentation Algorithms to Biomedical Imaging for More Accurate Tumor Detection

By: Gineton Alencar, Joel Villalpando, and Elizabeth Mathiasen       

Department: Computer Science

Faculty Advisors: Dr. Ilmi Yoon and Faye Orcales 

In this paper we attempt to demonstrate the application of a segmentation machine learning algorithm, based on U-Net, to measure the development of tumor growth. Using radiological medical imaging data, we train a model to detect the area of a tumor within a cross-sectional image and distinguish it from healthy tissue. Segmentation is a machine learning algorithm which is able to delineate objects of interest in an image. Using a computer to execute this task tackles the issue of human error in the reading of biomedical images, as well as reducing the cost of imaging analysis in the diagnostic process. Segmentation is also consequently useful for tracking the progress of tumor growth and remission, enabling more expedient clinical trials and better patient care. We hope that these techniques will add to the body of knowledge that will lead to improved outcomes for cancer patients through early and precise detection of tumor masses.