Posted in | Medical Robotics

A.I. System Trained to Detect Often-Missed Cancer Growths

Thanks to the University of Central Florida's Computer Vision Research Center, doctors may soon have help in the fight against cancer.

Assistant Professor Ulas Bagci leads the group of engineers at the University of Central Florida that have taught a computer how to detect tiny specks of lung cancer in CT scans, which radiologists often have a difficult time identifying. The artificial intelligence system is about 95% accurate, compared to 65% when done by human eyes, the team said. (Image credit: University of Central Florida, Karen Norum)

Engineers at the center have programmed a computer to detect minute specks of lung cancer in CT scans, which radiologists repeatedly have a difficult time identifying. The artificial intelligence (AI) system is about 95% accurate, compared to 65% when performed by human eyes, the team said.

We used the brain as a model to create our system. You know how connections between neurons in the brain strengthen during development and learn? We used that blueprint, if you will, to help our system understand how to look for patterns in the CT scans and teach itself how to find these tiny tumors.

Rodney LaLonde, a doctoral candidate and captain of UCF's hockey team

The method is like the algorithms that facial-recognition software employs. It scans numerous faces looking for a specific pattern to locate its match.

Engineering Assistant Professor Ulas Bagci leads the team of scientists in the center that concentrates on AI with promising medical applications.

The team fed over 1,000 CT scans - provided by the National Institutes of Health through a partnership with the Mayo Clinic - into the software they designed to help the computer learn to hunt for the tumors.

Graduate students involved in the project had to teach the computer various things to help it learn correctly. Naji Khosravan, who is pursuing his doctorate degree, created the framework of the system of learning. His proficiency at innovative machine learning and computer vision algorithms led to his summer as an intern at Netflix assisting the company with different projects.

LaLonde taught the computer how to disregard nerves, other tissue, and other masses it came across in the CT scans and examine lung tissues. Sarfaraz Hussein who earned his doctorate degree this past summer, is tweaking the AI's ability to detect cancerous versus benign tumors, while graduate student Harish Ravi Parkash is borrowing lessons learned from this project and using them see if another AI system can be created to help detect or predict brain disorders.

I believe this will have a very big impact. Lung cancer is the number one cancer killer in the United States and if detected in late stages, the survival rate is only 17 percent. By finding ways to help identify earlier, I think we can help increase survival rates.

Professor Ulas Bagci

The researchers will present their finding in September at the largest premier conference for medical imaging research - the MICCAI 2018 conference in Spain. The team's efforts have been published in advance of the conference.

The following step is to transfer the research project into a hospital venue; Bagci is seeking partners to accomplish that. After that, the technology could be a year or two away from commercialization, Bagci said.

"I think we all came here because we wanted to use our passion for engineering to make a difference and saving lives is a big impact," LaLonde said.

Ravi Prakash agrees. He was studying engineering and its applications to agriculture prior to hearing about Bagci and his research at UCF. Bagci's research is in the domain of biomedical imaging and machine learning and their applications in clinical imaging. Earlier, Bagci was a staff scientist and the lab manager at the NIH's Center for Infectious Disease Imaging lab, in the Department of Radiology and Imaging Sciences.

Tell Us What You Think

Do you have a review, update or anything you would like to add to this news story?

Leave your feedback
Submit