The healthcare industry is primed to push diagnostic AI beyond pilot-stage experimentation, with new studies showing gains in detection of cancer, heart disease and even substance use disorders.
As some diagnostic models approach near-perfect accuracy, broader clinical reasoning may still prove a major limitation. A recent study out of Somerville, Mass.-based Mass General Brigham found AI chatbots missed initial diagnoses more than 80% of the time when working from limited patient information, underscoring why many experts still view AI as a physician-support tool rather than a replacement.
When analysts warned AI might trigger widespread job disruption, healthcare CIOs told Becker’s the technology is more likely to reshape workflows and administrative responsibilities than replace clinicians outright.
Here are 10 ways the healthcare industry is deploying AI to refine — rather than replace — diagnostics:
- Researchers at Cleveland Clinic and Pittsburgh-based Carnegie Mellon University developed an AI system that identified certain heart conditions on cardiac MRIs with accuracy as high as 99%.
- OpenAI developed an AI model that, when evaluating 76 emergency room cases from Boston-based Beth Israel Deaconess Medical Center, identified exact or nearly exact patient diagnoses more often than two human physicians.
- Researchers at Worcester-based UMass Chan Medical School developed and tested an AI tool to diagnose cholangiocarcinoma. The tool demonstrated higher diagnostic accuracy than standard sampling methods, with 87.8% accuracy compared to 67.4%. It also outperformed visual assessments by experienced endoscopists, which had 63.1% accuracy.
- An AI model developed by researchers at Rochester, Minn.-based Mayo Clinic detected pancreatic cancer on abdominal CT scans taken up to three years before clinical diagnosis.
The model, called the Radiomics-based Early Detection Model, identified 73% of prediagnostic cancers on about 2,000 CT scans at a median of about 16 months before diagnosis. - A molecular test designed by researchers at Pittsburgh-based UPMC Hillman Cancer Center and the University of Pittsburgh School of Medicine detected nearly twice as many bile duct cancers as standard pathology.
The test, called BiliSeq, detected about 83% of bile duct cancers while pathology alone detected 44%. When BiliSeq and standard pathology were combined, cancer detection increased to nearly 90%. - Researchers at Ann Arbor, Mich.-based University of Michigan Health developed an AI model, dubbed Prima, that scanned brain MRIs in seconds and detected neurological conditions with up to 97.5% accuracy.
- Orange, Calif.-based Providence St. Joseph Hospital radiologists identified 20% more cancers up to two to three years earlier than traditional screening and reduced false positives and unnecessary callbacks by about 7% after integrating an AI algorithm with human interpretation of mammograms.
- University of Cincinnati researchers created an AI model that predicted substance use disorder-defining behaviors and addiction severity with up to 83% and 84% accuracy, respectively.
- Researchers at Rochester, Minn.-based Mayo Clinic developed an AI model that identified twice as many cases of advanced chronic liver disease in asymptomatic patients undergoing routine electrocardiograms when compared with standard diagnostic methods.
- Researchers at the University of California San Francisco developed an AI model that diagnosed pneumonia in critically ill patients with 96% accuracy, outperforming ICU clinicians. If implemented at admission, the model could have reduced inappropriate antibiotic use by more than 80%.
At the Becker's Perioperative Summit, taking place September 14–15 in Chicago, perioperative leaders and healthcare executives will focus on improving operating room efficiency, enhancing patient safety, optimizing staffing and driving innovation across surgical services. Apply for complimentary registration now.
