Guideline for AI for medical products

A) Meta information

1. Objective of the guideline

The objective of this guideline is to provide medical device manufacturers and notified bodies instructions and to provide them with a concrete checklist to

• understand what the expectations of the notified bodies are,
• to promote step-by-step implementation of safety of medical devices, that implement artificial intelligence methods, in particular machine learning,
• to compensate for the lack of a harmonized standard (in the interim) to the greatest extent possible.

The guideline is not meant to serves as a training manual or guideline to achieve the safety of AI based medical devices. It is to be a guideline for its review.

The annex lists the recitals which led to the development of this guideline.

2. Scope of applicability and target group.

This guideline is only applicable to medical devices that use AI methods, in particular machine learning. The guideline applies in particular to

• Manufacturers of these products
• Their service providers (such as engineering providers)
• People and organizations that must assess the safety of these products, such as auditors, authorities and notified bodies.

3. Instructions for use

a) Structure of the guideline

This guideline follows the thought that the safety of AI based medical products can only be achieved through a process-oriented approach, whereby all relevant processes and phases of the life cycle must be considered such as:

1. Research and development
2. Data management
3. Post-market surveillance

Accordingly, the guideline does not set forth specific requirements for the products, but for the processes. It contains the following chapters:

1. General requirements
2. Requirements for product development
   1. Intended use
   2. Software requirement specification
   3. Data management
   4. Model development
   5. Product development
   6. Product release
3. Requirements for phases following development

b) Binding character of the guideline

This guideline is neither a legal requirement nor a harmonized standard. Accordingly, there is no differentiation between normative and informative elements.

Much more, the guideline brings together best practices to best describe the required “state-of-the-art”.

Some of these best practices are not applicable in all situations, for all products or for all methods of machine learning. The manufacturers should at least justify non-obvious exclusions.

c) Use of the guideline

Creating and reviewing specifications

The manufacturers should first use the guideline to review the completeness of the specifications (process and work instructions, checklists, etc.). These tasks are normally assumed by the following roles:

• Process-related persons, in particular head of development
• Quality manager and quality management deputy
• Regulatory affairs manager

Assessing products and QM system

Then the people responsible for the following tasks should use the guideline:

• Reviewing the conformity of products with the underlying safety and performance requirements
• Assessing the conformity of the technical documentation with the regulatory requirements
• Assessing the efficacy of the internal quality management system (e.g. for design reviews or audits)

The following roles are normally responsible for these tasks:

• Quality managers
• External and internal auditors (including notified bodies)
• Internal and external reviewers of technical documentation (including notified bodies and authorities)
• Testers
• Data scientists
• Clinical affairs specialists
• Regulatory affairs manager
• Risk manager