Testing of Artificial Intelligence / Machine Learning-enabled Medical Devices
Abstract
In the rapidly evolving landscape of healthcare technology, artificial intelligence (AI) and machine learning (ML) are playing pivotal roles in medical device development. These technologies have the potential to enhance patient outcomes, improve diagnostics, and optimize treatment strategies. However, ensuring their safety, effectiveness, security, and transparency is critical. To this end, the verification of medical devices through testing plays a crucial role, particularly when they are programmable. When dealing with AI/ML-enabled Medical Devices (AI/ML-MD), testing becomes more intricate due to their dynamic nature and reliance on complex algorithms, necessitating consideration of additional aspects. As the main stakeholder responsible for verification and validation, this document addresses the manufacturer of medical devices. Nonetheless, the document can be used by any stakeholder that needs to perform testing of AI/ML-MD. For example, a test laboratory can apply the requirements of this document by substituting each occurrence of “the manufacturer” by “the test laboratory”. For medical devices, requirements on software testing are well established internationally by IEC 62304:2006/Amd 1:2015 . This document establishes additional aspects to specifically test AI component in AI/ML-MD. AI component considered in this document include locked AI model and AI model that learn under the control of the manufacturer. The case where the AI model is continuously learning in the field under indirect control by the manufacturer is treated in Annex A.
Begin
2024-10-02
Planned document number
DIN EN IEC 63450
Project number
02232611
