A groundbreaking AI model can determine a person’s risk of developing pancreatic cancer with astonishing accuracy, research suggests.
Using medical records and information from previous scans, the AI was able to accurately flag patients at high risk of developing pancreatic cancer within the next three years.
There are currently no complete scans for pancreatic cancer, with doctors using a combination of CT scans, MRIs, and other invasive procedures to make the diagnosis. This prevents many doctors from recommending these studies.
The study gives doctors hope because pancreatic cancer is notoriously difficult to spot, making it one of the deadliest forms of the disease, with more than half of patients dying within five years of diagnosis.
They also hope that over time these AI models will help them develop a reliable way to screen for pancreatic cancer, which already exists for other types of the disease.
Researchers have developed an AI model that can highlight patients at increased risk of developing pancreatic cancer. They hope it can help catch the deadly disease before it starts to spread (file photo)
Unlike other cancers, there is no single way to screen for it and in its early stages it can cause mild symptoms that are often overlooked.
“One of the most important decisions clinicians face on a daily basis is who is at high risk for a disease and who would benefit from further testing, which can also lead to more invasive and costly procedures that come with their own risks,” Dr. . Chris Sander, a Harvard biologist who contributed to the study, said.
“An AI tool that can target those at highest risk for pancreatic cancer who would benefit most from further testing could go a long way in improving clinical decision-making.”
The National Cancer Institute estimates that 64,050 Americans will be diagnosed with pancreatic cancer this year and will be responsible for 50,550 deaths.
The American Society of Clinical Oncology estimates that 56 percent of all people diagnosed die from the disease.
If the cancer spreads to another part of the body — called metastasis — the survival rate drops to just three percent.
This makes finding a way to screen for pancreatic cancer early is crucial, as any delay in treatment significantly increases a person’s risk of death.
“Many cancers, especially those that are difficult to identify and treat early, take a disproportionate toll on patients, families and the health care system as a whole,” said Dr Soren Brunak, a Danish study author.
“AI-based screening represents an opportunity to change the trajectory of pancreatic cancer, an aggressive disease that is notoriously difficult to diagnose early and treat quickly when the chance of success is greatest.”
Harvard researchers collaborated with scientists from, among others, the Danish pharmaceutical giant Novo Nordisk from the US and Denmark for their study, which was published Monday in Nature Medicine.
They trained their AI model by using 500 CT scans of people who had experienced lung nodules.
These are abnormal growths in the lungs. They are not cancerous in 95 percent of cases.
However, they can also serve as a sign that pancreatic cancer has spread to the lungs.
They then used past medical records to see if the AI could accurately identify people who were more likely to be diagnosed with the disease.
In total, data from 6 million Danes and 3 million Americans were included. Among the study population, 24,000 people from Denmark suffered from pancreatic cancer, along with 3,900 Americans.
The 9 million data points were fed into the AI, which was tasked with predicting the likelihood of someone suffering from pancreatic cancer within the next three years.
They gauged the accuracy of their model by generating an area under the curve, or AUC, score.
These scores work by comparing the results of the model to the actual patient outcomes. It generates a score between 0 and 1.0.
A model that receives a 0 is worthless, 0.5 is as accurate as flipping a coin, and 1.0 indicates a perfect model.
In general, scientists consider a score of 0.8 or higher to be indicative of an accurate test.
The Harvard model scored 0.88 for estimating cancer risk over the next three years and 0.9 for detecting risk over the next 12 months.
It was also being tested to see if it would predict further intervention for people with scans considered to be at ‘average risk’ of developing cancer.
Of the 22 people with lung nodules who were eventually diagnosed with the cancer, the AI labeled 18 at high risk of developing the disease.
Researchers hope this model can be used by cancer care physicians. Current screening tools include an MRI, CT scans, or endoscopic ultrasounds — where a doctor inserts a camera deep into a person’s throat.
However, these can be inconvenient, labor intensive and expensive. This makes doctors reluctant to recommend them to patients.
Neither are 100 percent accurate ways to find the cancer.
By using this type of tool to flag high-risk patients, doctors can ensure that the people most in need of these screenings can get them.