AI Able to Diagnose Cancer as Accurately as Humans But Much Faster

Neurosurgeons and pathologists from the Univerisity of Michigan were able to demonstrate how AI can diagnose cancer tumors just as accurately as humans. 

The work led by Todd Hollon and a team of neurosurgeons and pathologists at the University of  Michigan Medicine School showed how AI could diagnose cancer in imaged tissue. The work was published in journal Nature Biomedical Engineering.

Researchers imaged tissue from 100 neurological patients via both existing methods and using AI. Both techniques produced accurate results but the AI method came to the conclusion much faster.  

RELATED: NEW AI SYSTEM SNIFFS OUT MISSED TUMORS IN CANCER PATIENTS 

Surgeries could be sped up thanks to AI

If the AI method is applied throughout medical facilities the researchers say it could speed up operations. 

“By achieving excellent image quality in fresh tissues, we're able to make a diagnosis during surgery,” said first author Daniel A. Orringer, M.D., assistant professor of neurosurgery at the University of Michigan Medical School in a press release highlighting the results.  “This eliminates the lengthy process of sending tissues out of the OR for processing and interpretation.”

According to the researchers as it stands today, to determine a diagnosis during an operation the surgeon has to wait as long as forty minutes while the tissue is analyzed by the pathology lab. But if that was done more efficiently it could save money and reduce the time spent in the operating room. 

Computer keeps getting smarter 

“Our technique may disrupt the intraoperative diagnosis process in a great way, reducing it from a 30-minute process to about 3 minutes,” Orringer said. “Initially, we developed this technology as a means of helping surgeons detect microscopic tumor, but we found the technology was capable of much more than guiding surgery.”

The scientists were also able to teach a computer how to use the images to make diagnoses. They created and validated a machine learning process that identified a brain tumor subtype 90% of the time in 30 patient samples.

Orringer said in the press release the more information the computer was fed the more accurate the diagnosis became.