Using Artificial Intelligence to Find Pancreatic Cancer Earlier
A team of Johns Hopkins researchers, led by Elliot Fishman, M.D., is harnessing the power of machine learning to detect tiny, early-stage tumors on CT scans. The FELIX program uses data from thousands of scans to teach computers to detect tumors small enough to be missed by even the most experienced radiologists.
FELIX research has demonstrated sensitivities of more than 90% and specificities of more than 85% on average-sized pancreatic cancer, meaning it is highly effective in identifying people who have the disease and those who don’t. The research is especially promising: 25% of pancreatic cancers identified on scans by FELIX imaging were not previously detected.
Researchers are now perfecting the technology to discover tumors measuring one centimeter or less, when life-saving surgery is often possible. The goal is to ensure information from the FELIX program is incorporated as standard imaging in MRI and CT scan machines.
The FELIX program is named for Harry Potter’s Felix Felicis, also known as “Liquid Luck.” In the books, the magical potion makes the drinker lucky and turns the ordinary into the extraordinary. Like the potion, the goal of the program also is to provide extraordinary results through research, technology and maybe even a bit of luck.