Pathologists at OSUCCC-James have been evaluating an AI algorithm for the past year. The results are extremely promising.
By Chris Wolski
First, it was analog. Then digital. Is artificial intelligence (AI) the next step in the evolution of pathology? Yes and no—at least according to Anil Parwani, MD, PhD, director of the division of Anatomical Pathology at the Ohio State University Comprehensive Cancer Center—Arthur G. James Cancer Hospital and Richard J. Solove Research Institute (OSUCCC-James).
For the past year, Parwani and the pathologists on his team have been evaluating an AI algorithm designed to improve their accuracy and workflow. And he’s come to a definite conclusion about AI.
“It has become a clinical tool,” he says, but that doesn’t mean it’s ready to replace the pathologist. “AI plus the pathologist is a good option.”
The OSUCCC-James pathologist and his team are no strangers to updating their tools in an effort to streamline their workflows. In 2018, they introduced digital pathology with great success.
Artificial Intelligence and Time Savings
Parwani, who specializes in urology, has been using the AI algorithm to diagnose prostate cancer. The AI algorithm is also being used for prostate, breast, gastric, and other cancers for diagnosis validation and risk assessment.
He says the AI algorithm has been critical in helping him to save time and focus on making cancer diagnoses. Parwani evaluates between 30 and 40 cases per day. The AI algorithm helps by grading samples prior to his review—helping him to prioritize cases while relieving him of the initial mundane, low-level evaluation of the digital slides. This helps to speed diagnosis and, more importantly for the patient, receive treatment sooner.
Because he is freed from the routine evaluation of the slides, Parwani estimates that the AI algorithm cuts the time needed to review a case by about 20%.
Parwani stresses that AI is just a new tool in the toolbox, and that it does not make the diagnosis. The pathologist will continue as the one making the cancer diagnosis into the foreseeable future.
“We have to trust human intelligence over [artificial intelligence],” he says. “What we don’t want is to blindly follow AI.”
What he predicts will happen is that AI will become a reliable tool for handling the routine tasks.
“We won’t question those. But at the higher level [of diagnosis] is where human intelligence comes in,” he says.
That being said, he does see AI as being useful in training new pathologists and as a quality check when necessary.
Integrating the AI Algorithm
Currently, the AI algorithm exists outside the regular pathology workflow at OSUCCC-James, but Parwani expects to be integrated it into the workflow sometime in the future.
“Ultimately we want to be in a position that no matter the disease type, we can use the AI algorithm,” he says.
While Parwani and his OSUCCC-James colleagues are forward-looking, they are also cautious in their plans to use the AI algorithm. They carefully reviewed and evaluated it for the past year, and have kept it separate from the regular workflow.
Parwani does acknowledge the potential danger of overreliance on an AI algorithm, particularly in parts of the country or world where there is a significant shortage of pathologists. There could be a temptation to use AI simply to make the diagnosis.
“You have to use the tool safely and carefully,” he says.
Chris Wolski is chief editor of CLP.
Featured Image: Anil Parwani, MD, PhD, examines a digital image of a prostate biopsy in microscopic detail at The Ohio State University Comprehensive Cancer Center—Arthur G. James Cancer Hospital and Richard J. Solove Research Institute. Digital pathology supports pathologists with artificial intelligence that flags concerning biopsies and identifies traits that help inform cancer treatment. Photo: Ohio State University