How InferTrust™ helps medical device manufacturers meet FDA post-market surveillance requirements for AI and machine learning-based Software as a Medical Device
The FDA's framework for AI/ML-based Software as a Medical Device (SaMD) includes requirements for post-market surveillance: ongoing monitoring of device performance after deployment. For AI/ML-based SaMD, this means demonstrating that the model continues to perform within its intended specifications across real-world clinical conditions.
Every InferTrust™ (Patent Pending) decision record captures the exact model version (MID), the confidence score produced, the decision action taken, and the policy that governed the decision. Over time, this creates a comprehensive, tamper-evident dataset of model performance across the full range of clinical scenarios the device encounters in production.
Because every decision record includes the confidence score and model version, manufacturers can analyze trends in model confidence over time, across patient populations, clinical settings, and input characteristics. A decline in average confidence scores or an increase in escalation rates may indicate model drift that warrants investigation, retraining, or a predetermined change control process.
When the FDA requests evidence of ongoing device performance, manufacturers with InferTrust™ can produce verifiable records demonstrating model behavior across the surveillance period. These records are cryptographically sealed and independently verifiable, providing a level of evidence that goes beyond what traditional application logging can offer.