A groundbreaking new test that has the ability to detect the deadliest form of cancer years before a diagnosis could save thousands of lives.
Researchers at Minnesota‘s Mayo Clinic announced they have developed an AI-assisted test to detect pancreatic cancer up to three years before a patient receives a diagnosis.
The AI model, called REDMOD, was able to pick up even the most subtle tissue changes of pancreatic ductal adenocarcinoma, the most common form of pancreatic cancer.
Conventional imaging and the human eye alone often find these subtle changes very difficult to detect and they can go unnoticed.
Pancreatic cancer has earned that reputation not just for how many lives it claims, but for how swiftly it can advance before patients realize anything is seriously wrong.
In its early stages, symptoms are vague and easily dismissed: a dull back ache, intermittent indigestion, unexplained fatigue, subtle yellowing of the eyes or skin that comes and goes.
Doctors often describe it as a cancer that ‘whispers’ rather than shouts – and by the time it finally makes itself heard, it is frequently a death sentence. Its stealth is what makes pancreatic cancer uniquely dangerous.
Around 80 percent of cases are diagnosed only after the disease has spread beyond the pancreas, at which point surgery – currently the only potential cure – is no longer an option.
Holly Shawyer of North Carolina was diagnosed with pancreatic cancer in her 30s despite being a marathon runner. Her main symptom was a stomach ache. ‘I was in great health before this,’ she said
Overall, just 12 percent of patients survive for five years after diagnosis, and the majority do not live more than a year.
Each year, pancreatic cancer is diagnosed in around 67,000 Americans and kills more than 52,000.
Now, researchers believe the AI-assisted technology could detect the cancer at stage 0, making it more treatable and increasing chances of survival.
Dr Ajit Goenka, the study’s senior author, and a Mayo Clinic radiologist and nuclear medicine specialist, said: ‘The greatest barrier to saving lives from pancreatic cancer has been our inability to see the disease when it is still curable.
‘This AI can now identify the signature of cancer from a normal-appearing pancreas, and it can do so reliably over time and across diverse clinical settings.’
In the study, published in the journal Gut, REDMOD, which stands for Radiomics-based Early Detection MODel, was used on hundreds of CT scans from 219 patients’ abdomens that were deemed by a radiologist to show no evidence of disease.
However, the patients were later diagnosed with pancreatic cancer.
REDMOD was able to detect the ‘invisible’ signature of pre-clinical pancreatic cancer an average of 475 days before diagnosis.
Additionally, it performed better than radiologists and was twice as sensitive – meaning it had a superior ability to pick up true positive cancer results.
Ryan Dwars of Iowa with his family. He was diagnosed with stage four pancreatic cancer at 36
Panel A shows a CT scan of a 63-year-old man, which was interpreted as normal, with the pancreas outlined (yellow dashes). Panel B shows a CT scan from the same patient 2.4 years later with the red arrow pointing to a large pancreatic ductal adenocarcinoma (cancer). Lastly, panel C shows texturized maps generated by the REDMOD AI tool. The color map indicates that areas of high feature expression (red/yellow) are concentrated in the region of the pancreas where the tumor subsequently developed
It correctly detected cancer in 73 percent of cases, compared to 39 percent of cases among radiologists.
REDMOD was also nearly three times as accurate as radiologists when it came to detecting cases more than two years before diagnosis – accurate in 68 percent of cases compared to 23 percent.
The researchers acknowledged that their patient set was not diverse and would like to expand its test subjects.
They still concluded, however: ‘This study validates REDMOD as a fully automated AI framework capable of identifying the imaging signatures of stage 0 pancreatic ductal adenocarcinoma in normal pancreas, achieving this with substantial lead times and performance superior to expert radiologists.
‘While prospective validation is paramount to confirm clinical utility, the REDMOD framework represents a significant advance towards shifting the paradigm for sporadic [pancreatic ductal adenocarcinoma] from a late-stage symptomatic diagnosis to proactive pre-clinical interception, offering tangible hope for improving outcomes in this challenging disease.’
