IMDRF N88 Good Machine Learning Practice for Medical...

What does P6 require?

The choice and design of the model are evaluated and shown to fit the available data and to support active mitigation of known risks (overfitting, performance degradation, security); the product's clinical benefits and risks are well understood and translated into clinically meaningful performance goals for testing, considering the whole intended population and its subgroups and the uncertainty and variability in inputs, outputs and clinical conditions of use.

Other controls in Section 3: Guiding principles – IMDRF N88 Good Machine Learning Practice for Medical Device Development: Guiding Principles (2025)

P1 Understand the intended use deeply and bring multi-disciplinary expertise across the life cycle
P10 Monitor deployed models and manage re-training risks
P2 Apply good software engineering, device design and security practice throughout
P3 Evaluate clinically with datasets representative of the intended patient population
P4 Keep training and test datasets independent
P5 Use reference standards that are fit for purpose
P7 Assess the human-AI team in the intended use environment, not the device alone
P8 Test performance under clinically relevant conditions, independent of training data

Query this from an agent

The graph holds this control, the 0 it maps to, and the evidence behind each claim, over MCP and REST.

MCP endpoint All IMDRF N88 Good Machine Learning Practice for Medical Device Development: Guiding Principles (2025) controls