Khaberni - Researchers at Mayo Clinic developed an advanced artificial intelligence model that could change the future of pancreatic cancer diagnosis, having proven its ability to spot the disease through routine CT scans of the abdomen, years before clinical diagnosis, up to three years in advance.
This advancement offers a valuable opportunity to detect the disease at a stage when treatment is more effective.
Pancreatic cancer is among the most dangerous types of cancer because its symptoms often appear late, leading to its diagnosis after the disease has spread and significantly reducing survival chances. Therefore, early detection is considered the most crucial factor in improving survival odds.
Accurate Indicators
The new system relies on analyzing very precise indicators within the radiology images, which are not easily noticeable by the human eye. It identifies early changes in pancreatic tissue and structure before tumors become clearly visible. It can also analyze scans done for other reasons, especially in individuals at high risk, such as those newly diagnosed with diabetes.
During the tests, the researchers used about two thousand CT scans of patients who were later found to be suffering from pancreatic cancer, although these scans were previously classified as normal.
The model, known as REDMOD, has shown an ability to detect 73% of cases early, with an average time of 16 months before traditional diagnosis—an accomplishment nearly twice as effective as human review alone.
Positive Results
The results were encouraging in very early stages, as the model succeeded in detecting a number of cases three times more in scans conducted more than two years before the diagnosis, a stage where the disease often passes without any symptoms or clear signs.
The model operates entirely automatically, without the need for complex or time-consuming manual settings. The researchers confirmed its accuracy through testing it on CT scans coming from multiple institutions, using different devices and imaging methods, which demonstrated its ability to maintain consistent performance in diverse medical environments, not just within a limited data set.
The results also showed stability over time; the artificial intelligence consistently analyzed patients who underwent several examinations over spaced intervals, which enhances its potential in ongoing monitoring and early detection of the disease.
However, the system might produce some false positives, meaning some people might be asked to undergo additional tests despite not being sick.
The team is currently working on moving this technology to the actual clinical trial phase, aiming to evaluate its use in hospitals and measure its direct impact on saving patients.
If the results prove successful, artificial intelligence could become a crucial tool in combating one of the deadliest and hardest-to-detect cancers.
