FDA Acts on Radiological CDS

William HymanWilliam A. Hyman
Professor Emeritus, Biomedical Engineering
Texas A&M University, w-hyman@tamu.edu
Read other articles by this author

The FDA took two recent actions in the domain of CDS for radiological image interpretation. One was to permit the marketing under the De Novo program of software for the interpretation of fractures of the wrist. The De Novo process is applicable to new devices (those that don’t have a predicate) but whose safety profile may not require that the device be in the most demanding Class III where a new device would be automatically classified. In making the De Novo decision the FDA correspondingly defined Radiological Computer Assisted Detection and Diagnosis Software as Class II devices and established the Special Controls under which such software could be brought to market under the 510(k) program.

In a somewhat related action the FDA has proposed to down classify several other image analysis device types form Class III to Class II. These devices are those that are used for interpretation of mammography for breast cancer, ultrasound for breast lesions, radiography for lung nodules, and radiography for dental caries. These software devices are currently grouped together under the FDA product code MYN; Analyzer, Medical Image. There have been 5 device approvals (the appropriate term for Class III, but not Class II devices) under this code, not counting supplements. The basis for such a down classification is the belief that the risks associated with such devices can be adequately managed when the now Class II devices are subject to the enumerated Special Controls. Even though the new definition of the first action would seem to cover the devices in the second action, the latter devices have a separate product code and regulation number.

In both the De Novo and the reclassifications actions the FDA addressed certain general risks of radiological CDS including false positives, false negatives, use outside of validated populations, use on images not acquired using validated protocols, and device failure such that an automated analysis cannot be made.

More generally, CDS are the subject of an FDA Draft Guidance from December 2017. An open question for CDS generally is the extent to which the user can believe and act on what the CDS is saying, and the degree to which the manufacturer can find cover under a disclaimer that the results are just a suggestion but that it is always up to the user to make their own decision. In this regard the FDA offers the somewhat laughable assertion that “labeling ensures proper use of the device” which seems to suggest that the FDA believes that labeling is routinely read, understood and followed.