AI Can Predict Pancreatic Cancer Three Years Ahead of Humans

Researchers have discovered that an AI tool can identify individuals at high risk for pancreatic cancer up to three years before a diagnosis.

According to the research published in the journal Nature Medicine, the diagnosis is based solely on the patient’s medical records.

Pancreatic cancer is one of the most lethal forms of cancer, and this discovery offers hope of earlier diagnosis, and therefore, better chances of survival.

Currently, pancreatic cancer is difficult to diagnose early on, leading to a low survival rate, and a need for better screening tools.

Without any population-based tools to screen for pancreatic cancer, only individuals with a family history of the disease or certain genetic mutations are screened. This approach leaves out cases that fall outside these categories.

Chris Sander, faculty member in the Department of Systems Biology at Harvard Medical School and co-senior investigator of the study, noted that determining who is at high risk for a disease is one of the most important decisions clinicians face.

Sander stated that an AI tool that can identify those at high risk of pancreatic cancer and would benefit most from further tests, could improve clinical decision-making, leading to earlier diagnosis, treatment, and better outcomes.

The research team trained the AI algorithm on two data sets of nine million patient records from Denmark and the United States.

The model used disease codes and the timing of their occurrence to predict which patients were 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.

The AI model was tested on different versions for its ability to detect people at elevated risk of disease development within different timescales, ranging from six months to three years.

The researchers found that 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.

They found the model to be at least as accurate as current genetic sequencing tests available for only a small subset of patients in data sets.

One significant advantage of the AI tool is that it can be used on all patients with available health records and medical history, not just those with a known family history or genetic predisposition for the disease.

This is particularly important because many patients at high risk may not be aware of their genetic predisposition or family history. Without clear indications that someone is at high risk of pancreatic cancer, clinicians may be cautious to recommend expensive testing.

An AI tool that identifies those at high risk of pancreatic cancer would ensure that clinicians test the right population, while sparing others unnecessary testing and additional procedures, the researchers said.

Pancreatic cancer has a low survival rate, and early diagnosis is essential for better outcomes.

Only 12% of pancreatic cancer cases are diagnosed early, which has led to a call for better screening tools. The AI tool developed by the researchers offers hope for earlier diagnosis and treatment, as well as improved survival rates.

According to Sander, there is a clear need for better screening, more targeted testing, and earlier diagnosis, and this is where the AI-based approach comes in.