O.N. Diagnostics Warned on AI Changes to Osteoporosis Software
FDA has warned O.N. Diagnostics that it made significant changes to its VirtuOst vertebral fracture assessment software — including a new machine-learning algorithm and conversion to a web-based platform — without obtaining required agency clearance.
A just-posted 9/1 Warning Letter following a March inspection of the Berkeley, CA, company said VirtuOst VFA 3.0.0 is adulterated and misbranded because modifications that could significantly affect the device's safety or effectiveness were made without submitting a new 510(k).
The software was originally cleared for analyzing CT images to identify and characterize existing vertebral fractures and help physicians manage patients with osteoporosis. However, FDA investigators found that version 3.0.0 introduced a new machine-learning algorithm for identifying vertebral landmarks.
“A new model trained on different data, with different architecture, or optimized for a different objective may perform differently across patient populations, imaging equipment types, or clinical settings,” the letter said. “This new feature fundamentally changes how the device performs, fails, and affects patients in ways that go beyond what was originally cleared and therefore the feature must be adequately validated. Inadequate validation of this new feature may lead to a device that is unable to perform clinically as expected, which may result in patient harm such as missed or inadequate diagnosis, absence or delay of results, leading to missing, inaccurate, or delayed patient assessment.”
FDA also cited the company's migration of VirtuOst from a desktop application to a web-based architecture, which introduced additional cybersecurity risks. The company acknowledged during the inspection that the change expanded the device's “cyberattack surface area.” FDA also raised concerns about migration of the underlying technology platform, saying the changes required comprehensive software re-verification and re-validation.
The Warning Letter separately cited quality-system deficiencies involving corrective and preventive actions and supplier controls, as well as inadequate medical device reporting procedures.