Study Uses AI To Detect Pancreatic Cancer Before Diagnosis
Key Takeaways:
- A new Mayo Clinic study found that an AI model can help detect pancreatic cancer up to three years before clinical diagnosis
- Pancreatic cancer is the third-leading cause of cancer deaths in the U.S. and is often diagnosed late because many patients do not show symptoms early on
- Experts believe AI and radiomics could improve early detection and treatment outcomes
A new study spearheaded by the Mayo Clinic is using an artificial intelligence (AI) model to detect pancreatic cancer up to three years before clinical diagnosis on routine CT scans.
Pancreatic cancer is currently the third-leading cause of cancer deaths in the United States, with an overall five-year survival rate of 13%. The disease is often diagnosed at a late stage due to many cases being asymptomatic.
-
13%
Overall five-year survival rate with pancreatic cancer
According to a Mayo Clinic release, the AI model uses information and workflows that mirror clinical practice, including CT scans from multiple centers, imaging systems and protocols.
AI Finding More Cancers?
In the study, researchers using the AI model evaluated around 2,000 CT scans and found twice the amount of pre-diagnostic cancers compared to those who evaluated the same scans without AI assistance.
Daniel Jeong, MD, a diagnostic radiologist at Moffitt Cancer Center, says typically, pancreatic cancer is diagnosed using CT, MRI or endoscopic ultrasound.
“Pancreatic cancer is unique because once the mass reaches a certain size where it’s visible on a CT scan, oftentimes the patient’s life expectancy has already decreased,” he explained. “Although early detection does not always mean the disease is 100% curable, it does often lead to more treatment options and better outcomes.”
Using Radiomics To Predict Disease
At Moffitt, Jeong is part of a team using radiomics, the extraction of quantitative data from medical images, to study precursor lesions, abnormal tissue areas that are not cancerous but have a high likelihood of developing into cancer.
“One of the precursor lesions we are evaluating is called intraductal papillary mucinous neoplasm, which is the most common noninvasive precursor lesion,” Jeong said. “A number of these types of benign lesions transform into pancreatic adenocarcinoma over time. Our goal is to use advanced imaging analyses to predict which of these precursor lesions will become cancerous and to treat these patients as early as possible.”
As treatment options continue to evolve, many experts, including Jeong, are hopeful that tools such as AI and radiomics can improve early detection rates for pancreatic cancer.
