AAMI CR34971-2022 PDF

Full title and description

AAMI CR34971:2022 — AAMI Consensus Report: Guidance on the Application of ISO 14971 to Artificial Intelligence and Machine Learning. This consensus report provides practical guidance for applying the ISO 14971 risk‑management framework specifically to medical technologies that incorporate artificial intelligence (AI) and machine learning (ML). It emphasizes ML‑related risks (data management, model training and evaluation, feature extraction, and information/cybersecurity) and offers examples and suggested mitigation strategies to accompany ISO 14971 processes.

Abstract

CR34971:2022 is a companion guidance document intended to be used with ISO 14971:2019. It does not replace or change the ISO 14971 risk‑management process; rather, it explains how to interpret and apply ISO 14971 when the device or system includes AI/ML elements. The report concentrates on risks elevated or unique to ML‑enabled medical devices and provides annexed examples and tables to clarify hazards, hazardous situations, harms, and possible risk controls.

General information

• Status: Published 2022; later listings indicate it has been superseded by AAMI TIR34971:2023 (see related note in Related standards).
• Publication date: 2022 (document registries list the 2022 edition; some listings record April 1, 2022).
• Publisher: Association for the Advancement of Medical Instrumentation (AAMI).
• ICS / categories: Medical devices — risk management; aligned with ISO 14971 family (ICS reference for ISO 14971: 11.040.01).
• Edition / version: 2022 (AAMI Consensus Report, CR34971:2022).
• Number of pages: Approximately 35 pages (standard listings indicate a 35‑page report).

Scope

The report provides guidance for applying the ISO 14971 risk‑management process when evaluating medical technologies that utilize machine learning. It is intended to apply across the product lifecycle for ML‑enabled medical devices and to highlight risks that are elevated or unique to ML (for example, data set representativeness, training/test separation, model performance drift, explainability/interpretability issues, and cybersecurity impacts). It includes an annexed risk‑management process and tables of example hazards and risk‑control strategies keyed to ISO 14971 concepts.

Key topics and requirements

• Applying ISO 14971 risk‑management steps specifically to AI/ML components (hazard identification, risk analysis, evaluation, control, and monitoring).
• Data governance and dataset quality considerations: representativeness, labeling, independence of training/validation/test sets.
• Model development lifecycle topics: feature extraction, training procedures, validation and verification, performance metrics, and generalizability.
• Post‑market monitoring and model‑performance surveillance (detecting drift and managing updates/re‑training).
• Human‑AI interaction and use‑related risks, including transparency and user information needs.
• Information security and privacy considerations as risk contributors for AI/ML systems.
• Examples and suggested risk‑control measures provided in annexes to map common ML hazards to ISO 14971 controls.

Typical use and users

Intended users are medical device manufacturers, software developers of AI/ML‑enabled medical products, risk managers, clinical engineers, quality/regulatory professionals, notified bodies and conformity assessment organizations, and regulatory reviewers who need practical guidance to apply ISO 14971 to ML components and systems throughout development, release, and post‑market phases.

Related standards

Primary companion and reference: ISO 14971:2019 (Application of risk management to medical devices). CR34971:2022 is a guidance document to help apply ISO 14971 to ML; AAMI later published AAMI TIR34971:2023 as further guidance in this area. Other related documents and guidance include ISO/TR 24971 (guidance on ISO 14971 application) and international/regulatory GMLP guidance (Good Machine Learning Practice) published jointly by FDA, Health Canada, and the UK MHRA.

Keywords

AI, machine learning, ML, medical devices, risk management, ISO 14971, AAMI, dataset quality, model validation, post‑market surveillance, GMLP, cybersecurity.

FAQ

Q: What is this standard?

A: AAMI CR34971:2022 is an AAMI consensus report that provides guidance on applying ISO 14971 risk‑management principles to medical devices and systems that incorporate AI or machine learning.

Q: What does it cover?

A: It covers interpretation and practical application of ISO 14971 activities (hazard identification, risk analysis, risk controls, verification, and post‑market monitoring) specifically for ML‑related risks such as data quality, model training and evaluation, performance drift, human‑AI interaction, and cybersecurity. Annex material gives examples and suggested control strategies.

Q: Who typically uses it?

A: Device manufacturers, software developers, risk managers, quality/regulatory teams, clinical engineers, conformity assessment bodies, and regulators working with AI/ML medical technologies use this guidance to interpret ISO 14971 in an ML context.

Q: Is it current or superseded?

A: CR34971:2022 was published in 2022; registry listings note that AAMI subsequently issued AAMI TIR34971:2023 as updated guidance for the application of ISO 14971 to machine learning/AI, and many distributors list the CR as superseded by the TIR. Users should consult the AAMI TIR34971:2023 and the current ISO 14971 edition when planning compliance activities.

Q: Is it part of a series?

A: It is a consensus report (CR) from AAMI intended as a companion to ISO 14971 and sits alongside other AAMI guidance and technical reports addressing medical‑device software and risk management; AAMI later released TIR34971 as a related technical information report in 2023.

Q: What are the key keywords?

A: AI, machine learning, ISO 14971, risk management, dataset quality, model validation, post‑market surveillance, cybersecurity, AAMI.