Machine learning leads to pediatric eye cancer detection

Photo illustration by Penelope Shirey | Design Editor

By Samantha Amaro | Reporter

The human brain is infinitely complex. The brain has over 60 billion nerves that are constantly reacting to one another and controls the entire body. Thoughts and ideas are held in the minds of people, making human beings self-aware and knowledgeable. Teaching something that is not alive human tasks is a reality in the world of computer science, according to Hamerly.

The School of Engineering and Computer Science offers several research programs, including the Competitive Learning Infrastructure Initiative, bioinformatics and machine learning. This field of computer science supplies computer systems with the capability to ‘learn’ data without manual programming The research is based on finding specific algorithms that use past examples to make decisions.

Using unsupervised learning methods, computers are actually able to become a bit more sentient, even without being able to feel. An unsupervised learning method is a kind of algorithm that makes inferences from a compilation of data sets without needing to direct the responses of the computer.

Dr. Greg Hamerly, an associate professor of computer science in the School of Engineering and Computer Science, has several ongoing research projects using machine learning.

“The main thing that I am working on right now is trying to improve this application we have for detecting symptoms of cancer and other eye diseases,” Hamerly said.

Hamerly began his projects, like the eye cancer detection research, out of curiosity. Cancer is known to be extremely difficult to treat, but to be able to identify the development of the disease could save months of recovery or even entire lives, according to Hamerly. Through the use of his research, computers are expected to see the signs before the disease advances too far for treatment to be of any help.

Projects involving machine learning research include various subjects: efficient architecture simulation, literature genre discovery and even pediatric eye cancer detection. These kinds of research all find machine learning more of a necessity than a bonus. Being able to look at a patient’s eyes and to detect the presence of cancer as the computer uses previous collections of data to make an inference would be extremely helpful. Treating a disease like cancer after seeing the beginning signs of the symptoms would save months of idling where the disease has the chance to get a firm hold within the human body.

These machine learning projects are not locked down for faculty members only — undergraduate and graduate students also have opportunities for hands-on research experience.

“My students help me a great deal,” Hamerly said. “I mostly have Masters students who help me, and they’ve done a bulk of the work in the eye cancer detection project.”

Layton, Utah graduate student Matt Griffin works in the lab with Hamerly. He said he is so invested in the research being done that he wants to continue to work on projects like those Hamerly is conducting when he graduates. The research on eye cancer detection will hopefully tell the group what to improve and how to do so.

“The ultimate goal is to figure out where we can improve the kind of model that we are using,” Griffin said. “If we can find out where we should improve, we can know what we need to work on.”