The development and adoption of artificial intelligence (AI) in the medical space, and in particular the imaging-intensive specialties, continues to grow rapidly. At the time of publication, the FDA’s website reflected approximately 91 clearances of AI-enabled medical devices in 2022, with around 79 of those being reviewed by radiology panels at the FDA.

This trend is not unexpected, as the FDA has approved or cleared roughly 392 such devices since 1995. Among this group are three clearances of 510(k) submissions made by iCAD, Inc. These “devices” are all software programs targeted at assisting healthcare professionals in assessing mammography results and, in particular, detecting cancerous lesions via tissue density analysis.

iCAD is far from the only developer working in this corner of the radiological world, though it took an interesting step forward when, in October 2022, it announced a collaboration with Solis Mammography, aimed at developing a program to evaluate cardiovascular disease risk based upon the presence and quantity of calcifications within the arteries of the breast (BACs).

From a medical perspective, this appears to be a common sense outgrowth of iCAD’s prior efforts. Just as a reading radiologist would flag incidental findings when reviewing mammography images, this collaboration is aimed at helping providers interpret these images, even when the primary reason for the imaging study was breast cancer detection. To the extent a connection between BACs and cardiovascular disease risk is supported by current data, having a program that can reliably detect and quantify such features would be helpful and may avoid disregarding such findings as benign (as had been traditionally thought).

What this means from a product liability risk perspective is far from clear. As an initial matter, it is wholly unsettled how the legal system would treat a strict liability or negligence claim made against an AI-enabled medical device manufacturer. There is some precedent suggesting that AI-driven programs do not constitute products, while other courts have concluded that healthcare software programs may be subject to product liability claims. Compare Rodgers v. Christie, 795 Fed. Appx. 878 (3rd Cir. 2020), with Lowe v. Cerner Corp., No. 20-2270, 2022 WL 17269066 (6th Cir. Nov. 29, 2022). Second, this increases the risks for the manufacturer, as it would have several programs interacting with the same patient and care provider, increasing the chances for misdiagnoses of the patient and miscommunications with the care provider.

What is clear is that the presence of AI will only continue to expand in the medical field. Eventually, courts may expand the scope of the term “product” to encompass these devices, subjecting manufacturers to product liability claims, but when and how that will be implemented remains unknown.