A growing body of research shows computer algorithms may soon outperform human pathologists in both accuracy and speed.
But rather than hand their diagnostic keys to machines, two pathology professors from Yale argue that artificial intelligence will emerge as a “powerful tool” that can accurately identify the drug or therapy that will be most beneficial to an individual patient’s disease.
“Should we human pathologists fold our tents and pledge supplication to our computer overlords?” Balazs Acs, M.D., and David Rimm, M.D., Ph.D., wrote in JAMA Oncology. “No, just the opposite; intelligent digital pathology will make us more valuable to patients, providers, and payers."
The authors pointed to previously published research that has shown the ability of neural networks to accurately identify breast cancer or abnormal tumors, but also cautioned that there are several barriers to adaption, including validating clinical utility and regulatory pathways.
However, they acknowledged the inevitability of this digital transition, noting that AI tools will be particularly beneficial as the industry moves towards value-based payments. Providers that want to excel in a pay-for-performance environment will need to rely on the speed and predictive power of AI to assist a pathologist’s diagnosis.
RELATED: AI’s success in healthcare relies on quality data
“The pathologist will need to be armed with new tools to provide the needed diagnostic sensitivity and specificity, and it now seems clear that intelligent digital pathology will be a critical tool in that tool box,” they wrote.
Industry experts predict AI will be particularly beneficial for image-based specialties like radiology and pathology, in part because algorithms can process thousands of images far faster than a human clinician. That’s prompted some to argue that pathology and radiology should merge into a single specialty and adopt the role of “information specialists” to make room for new technology.
At the same time, researchers have called for more realistic expectations for AI in healthcare and better access to quality healthcare data to build effective algorithms.