Even as AI moves ahead quickly, both in radiology and in pathology, the exact applications or use cases have been under development. Do we use AI to confirm if a diagnosis is correct? To provide a "second reading" of benign slides, as a backup? Other uses?
One general use case that has emerged both in radiology and in pathology is pre-reading large numbers of tests (whether CT of the head or digital prostate slides) and flagging slides for priority review. A couple of these applications in radiology have not only been FDA-approved, but have garnered extra Medicare inpatient payments (under the 3-year New Technology Add-on Payment rules). See CMS NTAP for Avicenna.AI's CINA HEAD software for stroke (here).
Here we have the use case brought through FDA approval, and for pathology.
PAIGE-AI garners FDA clearance for a tool to flag likely positives among prostate cancer biopsies, assuming the data is available as digital files. The use case is clever because it provides clinical value (faster diagnosis for positive cases) while not directly affecting the pathologist's (or radiologist's) diagnosis, so it shouldn't create false negatives or false positives in actual sign-outs. However, collateral data for PAIGE.AI suggested it increased actual (true positive) cancer diagnoses by 7%.
See the FDA press release here. This was a De Novo clearance. Typically, FDA posts clearance paperwork and reviews within a few weeks of a device approval.
See an article by Elise Reuter at MedCityNews here.
See home page for Paige.AI here.
Paige previously had FDA clearance for its full-slide imaging system, prior to the addition of the AI software. Here; K201005, using 21 CFR 864.3700.