Standards Development

Scope

This document provides a framework for the data life cycle processes for management of data used to train, test or validate an AI model that is part of a medical device.

Data acquisition and management lifecycle the following considerations apply, amongst others: data suitability, data quality and integrity insurance, data privacy and security, data governance and documentation, data sampling and bias mitigation, data versioning and traceability, data storage and infrastructure, data access and sharing, and data labeling and annotation.

This document outlines the requirements for the data lifecycle, covering stages from planning and acquisition to usage and decommissioning. It emphasizes maintaining data quality, including aspects such as dataset classification, data annotations, traceability, metadata comprehensiveness, representativeness, and validity periods.

The scope is limited to the high-level process concepts applicable across medical specialties and device types and does not include specific requirements that might be covered by modality- or device-specific standards documents.

This document outlines the requirements for a quality management system, where an organization must demonstrate its capability to manage data in accordance with applicable medical device guidance and standards. Organizations can be involved in one or more stages of the life-cycle, including design and development, production, storage and distribution, installation, or servicing and maintenance of a medical device that incorporates AI. This document can also be used by suppliers or external parties that provide data, including quality management system-related services to such organizations.

Purpose

New work item proposal