The FDA, Health Canada and the United Kingdom's Medicines and Healthcare products Regulatory Agency have partnered to release 10 principles to be included in developing artificial intelligence practices.
The 10 principles identify areas where regulation and research are needed to create worldwide standardization in the International Medical Device Regulators Forum, according to an Oct. 27 news release.
Here are three of the guidelines:
- Study participants and data sets represent the patient population.
Protocols for data collection should mandate that relevant characteristics of the intended patient population are represented in clinical studies and datasets. Doing so helps to ensure that the results from studies are useful to the population of interest, such as age group, ethnicity or gender. - Training data sets and test sets are independent.
Training and test sets need to be independent of one another. All potential sources of dependence, such as patient, site factors and data acquisition, need to be considered and addressed to ensure independence. - Place focus on human-AI team performance.
When an AI model requires human interaction, human factors and human interpretability need to be addressed on the performance of the human and AI team, rather than how the model performs in isolation.
To read the full list of guidelines, click here.