AI Tool Assists in Predicting the Likelihood of Pancreatic Cancer

Published in Nature Medicine, new research led by investigators from Harvard Medical School and the University of Copenhagen describes an artificial intelligence (AI)-based tool that aims to enhance the detection and treatment of pancreatic cancer through early prediction up to three years before diagnosis.

Estimates from the American Cancer Society indicate that about 64,050 people will receive a pancreatic cancer diagnosis in 2023, with about 50,550 deaths due to this condition.

Research from the National Center for Biotechnology Information also noted expectations for the prevalence of this condition to increase.

Traditionally, the tools that screen patients for pancreatic cancer engage with only those suspected of developing the condition based on genetics. Resources that perform population-based screenings have yet to emerge.

However, the growth in AI capabilities shows the potential to diagnose and treat medical conditions faster.

“One of the most important decisions clinicians face day to day is who is at high risk for a disease, and who would benefit from further testing, which can also mean more invasive and more expensive procedures that carry their own risks,” said study co-senior investigator Chris Sander, PhD, a faculty member in the Department of Systems Biology in the Blavatnik Institute at Harvard Medical School, in a press release.

“An AI tool that can zero in on those at highest risk for pancreatic cancer who stand to benefit most from further tests could go a long way toward improving clinical decision-making,” he continued.

Sander also noted that improvements leading to faster detection could result in earlier treatment. Since pancreatic cancer diagnoses often occur after prime opportunities for treatment success have passed, the AI tool could provide significant value to pancreatic cancer patients and their families.

The tool leverages an AI algorithm trained on two data sets containing 9 million patient records. Researchers directed the AI tool to determine indications of the condition based on patient data.

From there, the model was able to predict the patients at the highest risk for developing pancreatic cancer. Various versions of the AI models also identified people with a high risk of developing the disease six months, one year, two years, and three years before diagnosis.

Researchers also noted that the AI training process allowed them to recognize how the patterns of other diseases impacted the risk of pancreatic cancer. Gallstones and type 2 diabetes, for example, were commonly correlated with a higher risk of pancreatic cancer within three years.

But researchers did caution that there was no definitive evidence to confirm a direct correlation between these conditions and pancreatic cancer. Rather, they serve as indicators, allowing physicians to gain a sense of those who may be at risk.

Further, researchers noted that leveraging an AI algorithm to predict pancreatic cancer improves the traditional process as it widens the pool of those screened from only those with a known family history of the disease.

Amid the emergence of healthcare-focused AI, various research efforts have aimed to use it to predict and treat different conditions.

In November 2022, researchers from the University of California Davis gained financial support from the National Cancer Institute (NCI) to support AI projects to improve breast cancer screening.

The funding allows researchers to operate three projects that use AI to predict which patients are at risk for breast cancer, mainly those with no history of the condition. They also aim to determine factors contributing to screening inequities and create a tool to identify patients at risk for cancer recurrence surveillance failure.