Key Steps To Harmonize Medical Device Software Machine Learning Best Practices

The International Medical Device Regulators Forum (IMDRF) has published a draft guideline on good machine learning practices for medical device software development. This guideline provides principles for manufacturers to consider and apply to their software development and post-market processes. The principles cover aspects such as intended use, software engineering, data representation, testing, and user information. They closely align with the FDA's guiding principles, with some minor differences.

Both the IMDRF and FDA guidelines emphasize the importance of multidisciplinary expertise, good software engineering practices, and representative clinical study participants and datasets. The IMDRF guidelines also address usability, quality management systems, suitable datasets for clinical evaluations, human factors, and clear communication of benefits and risks to users. Both guidelines stress the importance of testing device performance under clinically relevant conditions and monitoring deployed models for performance and re-training risks. The IMDRF guidelines also mention the use of real-world monitoring for model maintenance and improvement.

The article also discusses the importance of targeted medicines in enhancing efficacy, safety, and regulatory success in the pharmaceutical industry. It highlights the shift to the European Medicines Agency's Post-Marketing Surveillance and explores the balance between customization and standardization in the development of targeted medicines.