Overview
SIST-TS CEN ISO/TS 24971-2:2026 is a technical specification developed by the Slovenian Institute for Standardization (SIST), in alignment with ISO and CEN, to provide guidance on applying the ISO 14971 risk management process to medical devices enhanced by machine learning (ML), a subset of artificial intelligence (AI). This standard addresses the unique risks associated with ML-enabled medical devices (MLMD), and is intended to be used in combination with ISO 14971 and ISO/TR 24971.
Crucially, this document focuses on ML techniques in medical device software but excludes MLMD employing large language models (LLM) or generative AI. Its guidance covers the full lifecycle of MLMD, from design and risk analysis to post-production monitoring and updates.
Key Topics
- Machine Learning in Medical Devices: Outlines the concepts of ML models and algorithms, the process of training and testing with patient and synthetic data, and their application in clinical settings.
- Risk Management Alignment with ISO 14971: Details how standard ISO 14971 processes (hazard identification, risk estimation, risk evaluation, and control) are applied to the specific challenges of MLMD.
- Data Quality and Bias: Emphasizes robust data management, including the need for representative training and test data, managing and detecting unwanted bias, and maintaining data privacy.
- Explainability and Transparency: Addresses the “black-box problem” by encouraging manufacturers to improve the explainability of ML models so that clinicians and users can understand system decisions.
- Continuous Learning and Retraining: Recognizes that ML models may need periodic retraining and performance monitoring to maintain safety and effectiveness over time.
- Team Competence Requirements: Recommends multidisciplinary teams-covering ML, clinical context, software validation, data management, usability engineering, and IT security-to ensure comprehensive risk assessment and mitigation.
- Post-Market Surveillance: Stresses the importance of production and post-production monitoring, including planning for data collection, complaint response, and ML model updates.
Applications
The standard is intended for:
- Manufacturers of ML-enabled Medical Devices: To ensure risk management processes are comprehensive and appropriate for the unique characteristics of ML, providing a pathway to compliance with regulatory requirements and standards such as ISO 14971.
- Quality and Regulatory Professionals: To align internal procedures with international best practices and ensure ongoing safety and effectiveness of MLMD in the healthcare market.
- Developers of Medical Device Software: To integrate safety, data quality, and bias prevention considerations directly into the machine learning development cycle.
- Healthcare Providers and Clinical IT Teams: To understand how the safety and effectiveness of MLMD are assured, facilitating trust in AI-driven medical technologies.
By following the guidance in SIST-TS CEN ISO/TS 24971-2:2026, organizations can better identify, assess, and mitigate risks inherent to machine learning applications in medical devices, facilitating innovation while prioritizing patient safety.
Related Standards
For comprehensive risk management in ML-enabled medical devices, the following standards and technical reports should be considered alongside ISO/TS 24971-2:2026:
- ISO 14971:2019 – Medical devices - Application of risk management to medical devices
- ISO/TR 24971 – Guidance on the application of ISO 14971
- IEC 62304 – Medical device software - Software life cycle processes
- IEC 62366-1 – Medical devices - Application of usability engineering
- IEC/TR 80002-1 – Guidance on the application of ISO 14971 to medical device software
- IMDRF documents N67, N88 – Guidance on ML-enabled medical devices
- ISO/IEC 22989 and ISO/IEC 23894 – General AI concepts and risk management
Complying with SIST-TS CEN ISO/TS 24971-2:2026 helps organizations address the technical and process challenges unique to machine learning in medical device design, contributing to greater patient safety, improved clinical outcomes, and streamlined regulatory approval.