Although one of the rarer forms, pancreatic cancer has one of the lowest survival rates. This is mainly due to how difficult pancreatic cancer is to detect and correctly diagnose. An $80,000 award through a joint seed funding program offered by the Data Institute for Societal Challenges and the Stephenson Cancer Center at the University of Oklahoma, is supporting the development of an artificial intelligence-based computer-aided diagnosis system that will help more accurately detect pancreatic cancer and predict cancer prognosis.
According to Johns Hopkins Medicine, pancreatic cancer usually shows little or no symptoms until has advanced and spread, which results in a five to ten percent five-year survival rate. However, if a tumor is found before is has metastasized, patients tend to have longer survival rates.
Lead investigator Samuel Cheng, Ph.D., associate professor of Electrical and Computer Engineering at the Gallogly College of Engineering, stated that this new technology will help radiologists make more accurate decisions regarding cancer treatment, which will help boost the rate of early detection and help save more lives.
“Image segmentation is the core component of any AI-based computer-aided diagnosis system for cancer treatment,” Cheng said. “Therefore, improving image segmentation can significantly impact cancer treatment and diagnosis.”
Image segmentation is the process in which an image is divided into regions with similar properties such as color, texture, brightness, and contrast. Medical image segmentation aims to identify abnormalities like tumors or lesions, as well as measure tissue volume to monitor tumor growth and aid in calculating the appropriate radiation dose prior to chemotherapy.
The team hopes to create a 3D image segmentation model that will provide significantly more accurate results than the existing models. This model will utilize a novel capsule network, or CapsNet, which bundles neurons into a capsule with increased representation power which results in higher segmentation power.
“Image segmentation is a fundamental problem in image processing and computer vision, and thus improving image segmentation alone can have a wide impact, but it is definitely not easy,” said Cheng. “The trick we used to improve capsule networks may also be applied to enhance other models and architectures.”
A second award was made to a research team led by Abdul Rafeh Naqash, title and affiliation, for their project, "A proof-of-concept...Team members include Chongle Pan, title and affiliation, and Kun Lu, title and affiliation.
This project was one of two projects selected to be funded by DISC and SCC during 2022. For more information regarding the seed funding program, click here.
