Mayo Clinic Says Its AI Model Can Help Detect 'Hidden Cancers'
Researchers demonstrated the ability to automate detection of pancreatic cancer earlier than typical.
An AI model built by researchers at the Mayo Clinic can potentially flag signs of cancer in patients that might otherwise go undetected until treatment becomes ineffective.
The clinic's Comprehensive Cancer Center recently published a paper describing its work on an AI model built on "the world's most extensive imaging dataset," comprising CT scans from over 3,000 patients with a diverse range of backgrounds. The model was able to detect pancreatic cancer in patient CT scans well before they would normally be detected -- a significant breakthrough considering about four out of 10 "small pancreatic cancers" go undiagnosed until "they've advanced to an incurable stage," according to the Mayo Clinic.
"This creates a critical 'last-mile' barrier for early detection efforts where, in most patients...imaging detects the cancer at a stage when a cure is unlikely," the organization said in a press release this week describing the paper's findings. "This makes imaging the final frontier in the quest for early cancer detection."
The AI model developed by the Mayo Clinic -- a major healthcare adopter of generative AI technologies, partnering with Google and Microsoft -- effectively shortens that "last mile." According to the researchers, the model was able to identify "visually imperceptible cancer from normal-appearing pancreases" by a median of 438 days before official diagnosis.
As an added benefit, the model also demonstrated consistency across different patient populations, medical devices and imaging techniques. To protect against bias, the researchers also "deconstructed the AI's decision-making process to ensure transparency."
The Mayo Clinic model's success sets the stage for clinicians to use AI to detect other cancers earlier, improving the effectiveness of treatment and patient survival rates.
"These findings suggest that AI has the potential to detect hidden cancers in asymptomatic individuals, allowing for surgical treatment at a stage when a cure is still achievable," said Dr. Ajit H. Goenka, one of the paper's authors, in a prepraed statement. "We're only at the beginning but stand ready to address the challenges of early cancer detection, leveraging the capabilities of AI and next-generation molecular imaging in conjunction with complementary biomarkers."
The next steps for the reseachers include clinical validation and regulatory approval for the AI model, as well as screening trials.