FDA Creates Regulatory Category for Machine Learning-Based Cardiovascular Alert Software

The Class II classification, which came into force on September 11, codifies the special controls resulting from the De Novo application filed by Viz.ai for its Viz HCM module at 21 CFR 870.2380.

The decision, effective September 11, 2026, does not concern an isolated product but a type of device: software that analyzes non-invasive physiological signals, identifies a suspected cardiovascular pathology and alerts the care team. The starting point is the De Novo classification request filed on January 10, 2023 by Viz.ai for Viz HCM, a module that reviews a facility's routine electrocardiograms and reports suspected cases of hypertrophic cardiomyopathy.

By placing this software in Class II rather than Class III, the FDA is reducing the regulatory burden on future equivalent products, but with specific special controls. Clinical performance should be based on independent datasets from real-world conditions, collected from at least three geographically distinct sites, with sensitivity, specificity, predictive values, and analyses by demographic subgroups.

In addition, there is comprehensive software verification—model description, inputs and outputs, integration tests, fallback strategies in the event of failure—a human factors assessment focused on the risks of misinterpretation, and labeling detailing limitations, warnings against over-dependence, and clinical interpretation framework. The coding occurs at 21 CFR 870.2380.