The FDA has issued their draft recommendations (January 2025), regarding the content of marketing submissions for medical devices that include AI-enabled software capabilities. The recommendations include guidance on the design and development of the AI-enabled device during its entire lifecycle and development of the documentation for the assessment of the device by the FDA.
What are AI-Enabled Devices?
A medical device is considered an AI-enabled device when the device includes one or more AI device software functions (AI-DSFs), that generates an interpretation or prediction based on new data as input or captured, to achieve its intended purpose. The goal of the FDA with this guidance is to provide recommended strategies to help make the biases of AI-enabled devices transparent and ensure the device’s safety and effectiveness throughout the total product lifecycle (TPLC).
Importance of the FDA Glossary
In addition, the FDA’s glossary the “FDA Digital Health and Artificial Intelligence Glossary – Educational Resource”, should be used when reading this or other FDA documents on AI and Machine Learning (ML) as terms differ in the AI and ML community as compared to common terminology used in the medical device space. A copy of this glossary can be found at FDA Digital Health and Artificial Intelligence Glossary – Educational Resource | FDA.
Key Recommendations in the Draft Guidance
The FDA has proposed specific recommendations related to the information as provided for marketing submissions (510k) and how this data should be organized. It is recommended that the information submitted be organized into the following sections:
Key details provided in the draft guidance further include:
Key Performance Validation Considerations
Description of performance data attributes used for validation activities. The validation activities are conducted to ensure the device will perform as intended safely and effectively.
Usability Evaluation Considerations
Human factors studies or another suitable approach for testing usability, is required as a part of the marketing submission process for AI-enabled devices. The usability evaluation should be used to support the control of risks.
Data Management
Data management details should be included in marketing submission for AI-enabled devices. This data is important in identifying and mitigating biases.
Example 510(k) Summary and Device Model Card Examples
An example submission summary and device model card example are provided with recommendations for the order, formatting and content of information to be provided with the marketing submission.
Model Description and Development Guidance
The details to accurately describe the device and its functionality are to be included in marketing submission for AI-enabled devices. The model description should include key details related to inputs and outputs, software architecture, description of key features and functionality, process for optimization, and model parameters.
Device Performance Monitoring Guidance
Details related to the intended monitoring of the performance of the device during its TPLC are required. Considerations regarding the description of the tools, methods and management of the monitoring activities are provided.
Cybersecurity Considerations
Due to the intended use environment of AI-enabled software devices, cybersecurity must be considered. Recommendations for the design and maintenance of cybersecurity are provided in detail including controls and security risk management information.
Public Submission Summary Guidance
The Public Submission Summary is required as part of the marketing submission process for AI-enabled devices. The summary provides key details supporting the regulatory decision-made and provides reasonable assurance of safety and effectiveness details of the device for transparency to the public.
Transparency Design Considerations
Provides important considerations for manufacturers of AI-enabled devices to ensure they understand the user characteristics and needs as these details influence the interface of the design and the output of the AI-generated data.
Risk Management Guidance
Gives details related to the importance of identifying potential risks associated with the device, data and outputs, interaction risk considerations, and software quality risks.
_____________________________________________
This FDA draft includes recommendations regarding the marketing submissions of AI-enabled devices. For help navigating these regulations, contact us today.
Learn How MedEnvoy Can Assist you:
Related Guidance
We share our expertise because we believe in removing barriers to global commercialization.
The Business Impact of EUDAMED on Market Access
The European Database on Medical Devices (EUDAMED) is often viewed by manufacturers purely as a compliance burden when…
MedEnvoy at RAPS Euro Convergence 2026
We are excited to share that MedEnvoy will be participating in RAPS Euro Convergence 2026 as an exhibitor! This year’s…
Key FDA Changes MedTech Manufacturers Need to Know
The FDA has issued an update for MedTech Manufacturers to its List of FDA-Recognized Consensus Standards, published as …