Key points:
- A new study finds AI can predict pancreatic cancer 3 years prior to diagnosis using only medical records.
- The AI algorithm was trained on two separate data sets from 9 million patient records obtained from Denmark and the United States.
- The solution offers a population-based screening tool that is much needed given the late diagnosis and quick acceleration of pancreatic cancer.
Pancreatic cancer is one of the deadliest cancers in the world, and its toll projected to increase. Currently, health officials screen those with a family history and certain genetic mutations that predispose them to the disease. However, this specific targeted screening can miss other cases.
Researchers at Harvard Medical School and the University of Copenhagen developed an artificial intelligence tool that can identify people at the highest risk for pancreatic cancer up to three years prior to diagnosis using the patients’ medical records.
The study findings, published in Nature Medicine, suggest an AI-based population screening could be valuable in finding those at more risk for the disease, and could expedite the diagnosis—currently, 85% of patients already have metastatic disease at diagnosis.
In the new study, the AI algorithm was trained on two separate data sets from 9 million patient records obtained from Denmark and the United States. The researchers had the AI model look for telltale signs based on the data contained in the records. Based on combinations of disease codes and the timing of their occurrence, the model was able to predict which patients are likely to develop pancreatic cancer in the future.
Notably, many of the symptoms and disease codes were not directly related to or stemming from the pancreas. For example, diagnoses such as gallstones, anemia, type 2 diabetes, and other GI-related problems portended greater risk for pancreatic cancer within 3 years of evaluation.
The researchers then tested different versions of the AI models for their ability to detect people at elevated risk for various timetables—6 months, one year, two years, and three years. Overall, each version of the AI algorithm was substantially more accurate at predicting who would develop pancreatic cancer than current population-wide estimates of disease incidence.
The scientists said they believe the model is at least as accurate in predicting disease occurrence as are current genetic sequencing tests that are usually available only for a small subset of patients.