If your cardiac AI has no predicate, the FDA De Novo Classification route is your path to market, and it’s the process that sets the specific claims you can make. HeartFlow’s De Novo clearance is the perfect anchor for understanding how this works in practice, showing exactly how the claims you want to make must line up with the device’s final classification.
De Novo Classification for Novel Device Claims
The De Novo path is for new, low-to-moderate-risk medical devices that don’t have an existing predicate to compare against. When you make a De Novo submission, you’re not just describing your tech. You’re defining its intended use and establishing the specific regulatory controls, the guardrails, required to prove it’s safe and effective. It’s the classification itself that dictates what a cardiac AI can claim, a point people often miss when they get fixated on the model card. Take HeartFlow’s FFRct Analysis software, which received a De Novo because there was no other device doing what it did: creating a 3D model of coronary arteries from CT scans and using computational fluid dynamics to calculate fractional flow reserve (FFR) non-invasively. No predicate meant De Novo was the only option. FDA De Novo summary for HeartFlow FFRct
First-of-Kind Device Review Bar
The FDA’s Center for Devices and Radiological Health (CDRH) is where first-of-a-kind devices get scrutinized, and as Bakul Patel has explained, the evidence bar is high. The review digs deep into the safety and real-world effectiveness of the new technology. You can’t just show up with a cool algorithm. You need solid clinical data showing the device performs against a meaningful clinical endpoint, with a standard of evidence that’s tough enough to give a reasonable assurance it won’t hurt people and actually works. For a cardiac AI, that usually means proving diagnostic accuracy or prognostic value through proper clinical studies designed to back up every single claim you’re making. The reviewers will pour over everything, the algorithms, the training data, the validation methods, but in the end, the question is always: does this device have real clinical utility and a positive impact on patients?
Peer Submissions and Consistent Sequence
You see the same story play out with other cardiac AI players like Anumana and iRhythm Technologies. They all had to align their claims with a specific classification. For instance, Anumana got a 510(k) clearance for its ECG-AI that spots low ejection fraction by analyzing standard 12-lead ECGs, a totally new way to use AI for that specific clinical need. They’ve since followed that up with more 510(k)s for AIs that detect pulmonary hypertension and cardiac amyloidosis. Meanwhile, iRhythm’s Zio XT System (not a De Novo itself) has been building AI into its monitoring platform for years, and each new AI-based feature or indication forced them to think hard about the right regulatory path, sometimes even eyeing a De Novo for something completely new. Their history makes it clear that AI development and regulatory approvals are an ongoing cycle, not a one-and-done deal. Looking at all these companies together, HeartFlow, Anumana, iRhythm, a clear sequence emerges. Define the claim. Classify the device. Get the clearance. That’s the playbook. The De Novo process isn’t just about the paperwork. It’s the actual regulatory act that defines what your product is allowed to claim it does.
The Underrated Classification Question
Too many regulatory affairs leads blow past the classification question, but it’s a huge mistake. Your device’s classification route determines what you can claim, not the fancy model card your data scientists are so proud of. It all starts with the intended use. Nail that down, and the right regulatory pathway becomes obvious. Teams that don’t articulate the device’s intended use with absolute clarity early on are setting themselves up for a fall, because that clarity is the foundation of the entire regulatory strategy. Get the alignment between your intended use and your chosen classification wrong, and you’re looking at months or years of delays, or worse, having to completely re-scope the device from the ground up. The hard work you do upfront, defining precise claims and really grappling with what they mean for classification, is what separates a smooth clearance from a regulatory nightmare. FDA guidance on intended use and device classification
Frequently Asked Questions
What is the primary purpose of the De Novo classification pathway for a first-of-kind cardiac AI device?
The De Novo Classification pathway defines the regulatory pathway for a cardiac AI device without a predicate. This pathway establishes the specific claims a novel device may make and determines what the device is allowed to claim. It also defines the intended use and technological characteristics, and establishes regulatory controls for safety and effectiveness.
What level of risk is typically associated with devices that qualify for the De Novo pathway?
The De Novo Classification pathway applies to low-to-moderate-risk devices. This route is specifically for medical devices presenting novel technology with no established predicate device.
What kind of evidence is required for the FDA’s review of first-of-kind cardiac AI device submissions?
The review process focuses on the safety and effectiveness of the novel technology. Clinical data demonstrating the device’s performance against a relevant clinical endpoint is required. For cardiac AI, this often means demonstrating diagnostic accuracy or prognostic utility through well-designed clinical studies that address the specific claims.
How important is defining the device’s intended use early in the development process for a De Novo submission?
Teams must clearly articulate the device’s intended use early in development, as this clarity informs the regulatory strategy. Misalignment between intended use and classification pathway can lead to significant delays and necessitate extensive re-scoping of a device. The upfront investment in defining precise claims and understanding classification implications is critical.
