Pancreatic cancer: Researchers program AI to identify people at high risk

Artificial intelligence algorithms could be key to identifying who is most at risk of developing pancreatic cancer, one of the most difficult cancers to diagnose early, according to a new study. I have.

In a study published Monday in the peer-reviewed journal Nature Medicine, researchers, with the help of AI, found that: Can identify people at highest risk of developing pancreatic cancer It was done approximately 3 years before diagnosis by purely using patient medical records.

Researchers say this could be a game changer in combating this fast-growing and difficult-to-detect form of cancer.

“One of the most important decisions clinicians face on a daily basis is who is at higher risk of disease and who might benefit from further testing. It can also mean that the procedure is risky,” said Chris Sander. Harvard He Medical He is a faculty member in the Department of Systems Biology at the Bravatnik Institute of the Harvard School of Medicine and Research Co-Principal Investigator, said in a press release.

“AI tools that can target those at highest risk of pancreatic cancer and who stand to benefit most from further testing could greatly help improve clinical decision-making.”

A family history of pancreatic cancer and the presence of certain genetic mutations flag individual patients for targeted screening and early testing, but many patients still find themselves at higher risk. There is no way to know, and it slips through the cracks.

Early identification of cancer is important for recovery, but it is especially urgent for pancreatic cancer. Pancreatic cancer is one of the most difficult cancers to detect in its early stages, when it is most curable.

It begins in the pancreas, an organ behind the stomach that produces enzymes and hormones that help digestion and regulate blood sugar, respectively. However, pancreatic cancer does not cause symptoms until it spreads beyond the pancreas to other organs. It is much less likely to beat the cancer once it has metastasized.

Clinicians refer to the pancreas as the “angry organ” because of how difficult it is to perform a biopsy of it, the release explained.

According to the Canadian Cancer Society, only about 10% of people diagnosed with pancreatic cancer in Canada live longer than five years.

In this new study, researchers trained an artificial intelligence model based on Danish clinical data spanning 6.2 million patients over 41 years. Of this sample, approximately 24,000 will be diagnosed with pancreatic cancer at some point.

By sorting through this vast amount of data, an AI algorithm was able to compile warning signs of pancreatic cancer within a specific time frame. Once the AI ​​learns these “disease codes,” the researchers found that the AI ​​algorithm was significantly more accurate in predicting who would develop pancreatic cancer compared to population-wide estimates based on disease incidence levels. I discovered something.

To test it further, they took their most successful algorithm and used it on a new cohort: 3 million patients from the US Veterans Health Administration’s dataset spanning 21 years.

This dataset included 3,900 patients diagnosed with pancreatic cancer. The researchers found their AI algorithms to be slightly less predictive than their Danish cohort. The Danish cohort was a truly national sample, but retraining on US data improved accuracy.

Pancreatic cancer is much more difficult to screen than other cancers. Mammograms, Pap smears, and blood tests make it easier for doctors to detect breast, cervical, and prostate cancer, respectively, but pancreatic cancer screening methods are more expensive. Physicians are unlikely to order a CT scan or MRI to screen for pancreatic cancer in a patient without the family history typically used to evaluate it, researchers note.

Being able to identify those who really need more testing and those who don’t not only saves time and resources, but may also catch more cases of this aggressive cancer earlier.

If patients are diagnosed at an early stage, the five-year survival rate jumps to 44%, the release said, but currently only about 12% of cases are diagnosed early. If the tumor spreads beyond the pancreas, his 5-year survival rate may drop to 2%.

“This low survival rate is despite significant advances in surgical techniques, chemotherapy and immunotherapy,” Sander said. “Therefore, in addition to advanced treatments, there is a clear need for better screening, more targeted testing and earlier diagnosis, with AI-based approaches emerging as the first critical step in this continuum. ”

The researchers found that changes in the accuracy of AI algorithms when introduced to new country data meant that for this method to be successful, the AI ​​model needed to be trained on a large dataset or tested for specific demographics at risk. You pointed out that it suggests that you should train locally as much as possible to capture patterns.

“Many types of cancer, especially those that are difficult to detect and treat early, wreak havoc on patients, families and the entire healthcare system,” said research director of the university’s Novo Nordisk Foundation Protein Research Center. says Søren Brunak, He is a co-principal investigator in Copenhagen, the release said. “AI-based screening represents an opportunity to change the trajectory of pancreatic cancer, a notoriously difficult and progressive disease that is difficult to diagnose early and treat promptly when it has the highest chance of success.”