IMDRF Good Machine Learning Practice: Introduction

Artificial intelligence (AI) technologies, including machine learning (ML), have the potential to transform health care by deriving new and important insights from the vast amount of data generated every day. Algorithms can learn from real-world use and potentially improve product performance over time. For Hong Kong healthcare professionals — whether in public hospitals, private clinics, radiology, pathology, or allied health — understanding how regulators expect AI-enabled medical devices to be developed and maintained is essential for responsible use in daily practice.

Why AI-enabled devices need distinct principles

Unlike traditional software, ML models are iterative and data-driven. Their performance can change as they encounter new data, and development often continues after deployment. This creates unique considerations that standard software guidance alone may not address.

The International Medical Device Regulators Forum (IMDRF) published ten guiding principles for Good Machine Learning Practice (GMLP) in January 2025. These principles promote the development of safe, effective, and high-quality medical devices that incorporate AI across the total product lifecycle — from design and clinical evaluation through deployment, monitoring, and updates.

A call to harmonise global practice

GMLP is intended to inform international standards organisations, regulators, and collaborative bodies. Areas of collaboration include research, educational tools, international harmonisation, and consensus standards that shape regulatory policies and guidelines.

For clinicians and other healthcare professionals in Hong Kong, GMLP provides a shared vocabulary for evaluating vendor claims, participating in procurement decisions, and recognising when an AI tool's limitations may affect patient care.

Generative AI and foundation models

Recent advances — especially generative AI — highlight the importance of clearly describing a product's intended use / intended purpose and identifying its regulatory status. Technologies that incorporate generative AI may use foundation models not controlled by the device manufacturer, introducing unique risks.

Generative AI may also make it harder to demonstrate device performance. The regulatory science of measuring performance and detecting errors in these models is still maturing. As the AI medical device field evolves, so too must GMLP and consensus standards.

What this learning path covers

The following articles break down each of the ten IMDRF guiding principles in plain language, with practical implications for Hong Kong healthcare settings. Each article includes a short quiz to reinforce key concepts.

Source: IMDRF — Good Machine Learning Practice for Medical Device Development: Guiding Principles (January 2025)