Full definition
FDA SaMD (Software as a Medical Device) is the regulatory framework the FDA applies to software that performs medical functions independently of hardware medical devices. Examples: AI radiology image analysis, clinical-decision-support that influences diagnosis or treatment, sepsis-prediction algorithms, mobile-app symptom checkers that recommend care actions.
The SaMD risk classification (per IMDRF, the International Medical Device Regulators Forum) ranges from Class I (lowest risk — informational support) to Class IV (highest risk — autonomous diagnosis driving critical treatment). FDA clearance pathways depend on classification: 510(k) for predicate-equivalent, De Novo for novel low-risk, PMA for high-risk. Some SaMD falls outside FDA jurisdiction (administrative software, wellness apps).
For MOVO-X procurement: most clinical-decision-support features that augment clinician judgment fall below SaMD thresholds. Features that operate as autonomous classification (no human review) typically require FDA clearance. We document regulatory classification per AI feature and work with customer regulatory affairs teams when SaMD pathways apply.
Where fda samd (software as a medical device) is used
- AI radiology image analysis
- Sepsis prediction in clinical decision-support
- Skin-cancer detection apps
- Diabetic retinopathy screening
- Cardiac arrhythmia detection (consumer wearables that diagnose)
Types of fda samd (software as a medical device)
IMDRF Class I
Lowest risk — informational support.
IMDRF Class II
Drives diagnosis/treatment for non-serious conditions or supports decision-making.
IMDRF Class III
Drives diagnosis/treatment for serious conditions.
IMDRF Class IV
Drives diagnosis/treatment for critical care; highest oversight.
510(k)
FDA pathway for substantially-equivalent predicate device.
De Novo
FDA pathway for novel low-to-moderate risk devices without predicate.
PMA (Premarket Approval)
FDA pathway for highest-risk devices.
Quantified benefits
- ▸Patient safety
- ▸Regulatory clarity for AI in clinical use
- ▸Audit-grade governance pathway
- ▸International recognition via IMDRF alignment
Frequently asked
Does AI clinical decision-support require FDA clearance?+
Some of it does — depending on autonomy and clinical impact. Most decision-support that augments clinician judgment falls below SaMD thresholds. Autonomous classification typically requires FDA clearance.
What's the FDA pathway for new AI?+
For novel AI without predicate: De Novo or PMA depending on risk class. For substantially-equivalent: 510(k). The FDA also published a "Predetermined Change Control Plan" guidance for AI/ML SaMD that can adapt post-deployment.
How does this affect MOVO-X procurement?+
We classify each AI feature against SaMD criteria and document the regulatory pathway. For features below threshold: model card + governance + audit trail. For features that would be SaMD: we work with customer regulatory affairs.
EU equivalent?+
EU Medical Device Regulation (MDR) classifies SaMD similarly. CE marking required. EU AI Act adds additional requirements for high-risk AI in healthcare.