Chapter 8. Artificial Intelligence-based Software
132 Regulatory Affairs Professionals Society (RAPS)
Table 8-6. Summary of Key Digital Health Activities
Year Guidance/Workshops/Other Developmental Activities
2013 i. The IMDRF formed the SaMD Working Group 14 to develop guidance for manufacturers to support innovation
and timely access to safe and effective SaMD globally. The working group developed guidance around the defini-
tions, framework for risk categorization, quality management system, and clinical evaluation 17-19
2016 i. The FDA released draft guidance on the clinical evaluation of SaMD to help emphasize the essential clinical
considerations18
ii. Section 3060(a) of the 21st Century Cures Act was amended to exclude certain software functions (e.g., adminis-
trative functions, healthy lifestyle, and electronic patient records) from the definition of ‘medical device’36
2017 i. The FDA announced a voluntary software precertification pilot program to evaluate the quality standards for
software design, validation and maintenance of software products.82,83
ii. The FDA published a ‘Digital Health Innovation Action Plan’ to protect and promote the public health84
2019 i. The FDA released a discussion paper that described the FDA’s foundation for a TPLC approach to premarket
review of AI/ML-driven software modifications20
ii. The FDA announced a partnership with the National Evaluation System for Health Technology Coordinating
Center Collaborative Community and the Ophthalmic Imaging Collaborative Community to develop solutions to
medical device innovation challenges,85
iii. The FDA provided a ‘Regulatory Framework for Conducting the Pilot Program within Agencies’ Authorities’86
2020 i. The Digital Health Center of Excellence was established within the FDA to advance digital health by facilitating
cross-collaborations to promote safe and effective digital products81
ii. The FDA down-classified a subset of medical image analyzers from Class III to Class II particularly, CADe devices
applied to mammography breast cancer, ultrasound breast lesions, radiograph lung nodules, and radiograph
dental caries detection from Class III to Class II32
iii. The FDA organized a public workshop to discuss emerging applications of AI in radiological imaging87
iv. The FDA organized a Patient Engagement Advisory Committee to ensure transparency in AI and ML medical
device software88
2021 i. The FDA published their AI/ML-Based SaMD Action Plan in response to stakeholder feedback summarizing the
Regulatory Framework and feedback on ‘Predetermined Change Control Plan Good Machine Learning Practice
Patient-Centered Approach Incorporating Transparency to Users and Real-World Performance’84,85,88
ii. The Department of Health and Human Services published the Trustworthy AI playbook89
iii. The FDA announced a virtual public workshop on transparency of AI/ML-enabled medical devices to patients,
caregivers, and providers87
iv. The FDA and other health authorities released 10 guiding principles for good machine learning practice for medi-
cal device development68
2022 i. The FDA released a draft cybersecurity guidance to ensure the need for robust cybersecurity controls in medical
devices90
ii. The FDA released a CDS software guidance to clarify the types of CDS software functions that are excluded
from the definition of device74
iii. The FDA released a report of the completed software precertification (Pre-Cert) pilot program. The summary of
this report suggested that the approach described in the ‘working model’ is not practical to implement under the
FDA’s current statutory and regulatory authorities82
iv. The White House released a blueprint for an AI Bill of Rights. The Bill discusses the principles for design, use and
deployment of automated systems91
v. The FDA released a list of AI/ML-enabled medical devices currently being marketed in the United States6
vi. The Association for the Advancement of Medical Instrumentation (AAMI) published a consensus report for iden-
tifying, evaluating, and managing risk for healthcare technology that incorporates AI or ML92
2023 i. NIST released their AI Risk Management Framework93
ii. The FDA issued a draft guidance on recommendations for a Predetermined Change Control Plan for AI software
devices. This guidance suggests an openness in Agency’s thinking to allow manufacturers developing AI/ML
algorithms to retrain their models from data in real time20
iii. The FDA released an updated software guidance on the information necessary to include in premarket submis-
sions for the FDA’s evaluation of the safety and effectiveness of device software functions94
iv. The Biden-Harris Administration announced the creation of the NIST Public Working Group on Generative
AI, which will build on the NIST AI Risk Management Framework to address AI technologies that can generate
content, including images95
AI, artificial intelligence CADe, computer-assisted detection CDS, clinical decision support FDA, Food and Drug Administration IMDRF,
International Medical Device Regulatory Forum ML, machine learning NIST, National Institute of Standards and Technology SaMD,
software as a medical device TPLC, Total Product Life Cycle
132 Regulatory Affairs Professionals Society (RAPS)
Table 8-6. Summary of Key Digital Health Activities
Year Guidance/Workshops/Other Developmental Activities
2013 i. The IMDRF formed the SaMD Working Group 14 to develop guidance for manufacturers to support innovation
and timely access to safe and effective SaMD globally. The working group developed guidance around the defini-
tions, framework for risk categorization, quality management system, and clinical evaluation 17-19
2016 i. The FDA released draft guidance on the clinical evaluation of SaMD to help emphasize the essential clinical
considerations18
ii. Section 3060(a) of the 21st Century Cures Act was amended to exclude certain software functions (e.g., adminis-
trative functions, healthy lifestyle, and electronic patient records) from the definition of ‘medical device’36
2017 i. The FDA announced a voluntary software precertification pilot program to evaluate the quality standards for
software design, validation and maintenance of software products.82,83
ii. The FDA published a ‘Digital Health Innovation Action Plan’ to protect and promote the public health84
2019 i. The FDA released a discussion paper that described the FDA’s foundation for a TPLC approach to premarket
review of AI/ML-driven software modifications20
ii. The FDA announced a partnership with the National Evaluation System for Health Technology Coordinating
Center Collaborative Community and the Ophthalmic Imaging Collaborative Community to develop solutions to
medical device innovation challenges,85
iii. The FDA provided a ‘Regulatory Framework for Conducting the Pilot Program within Agencies’ Authorities’86
2020 i. The Digital Health Center of Excellence was established within the FDA to advance digital health by facilitating
cross-collaborations to promote safe and effective digital products81
ii. The FDA down-classified a subset of medical image analyzers from Class III to Class II particularly, CADe devices
applied to mammography breast cancer, ultrasound breast lesions, radiograph lung nodules, and radiograph
dental caries detection from Class III to Class II32
iii. The FDA organized a public workshop to discuss emerging applications of AI in radiological imaging87
iv. The FDA organized a Patient Engagement Advisory Committee to ensure transparency in AI and ML medical
device software88
2021 i. The FDA published their AI/ML-Based SaMD Action Plan in response to stakeholder feedback summarizing the
Regulatory Framework and feedback on ‘Predetermined Change Control Plan Good Machine Learning Practice
Patient-Centered Approach Incorporating Transparency to Users and Real-World Performance’84,85,88
ii. The Department of Health and Human Services published the Trustworthy AI playbook89
iii. The FDA announced a virtual public workshop on transparency of AI/ML-enabled medical devices to patients,
caregivers, and providers87
iv. The FDA and other health authorities released 10 guiding principles for good machine learning practice for medi-
cal device development68
2022 i. The FDA released a draft cybersecurity guidance to ensure the need for robust cybersecurity controls in medical
devices90
ii. The FDA released a CDS software guidance to clarify the types of CDS software functions that are excluded
from the definition of device74
iii. The FDA released a report of the completed software precertification (Pre-Cert) pilot program. The summary of
this report suggested that the approach described in the ‘working model’ is not practical to implement under the
FDA’s current statutory and regulatory authorities82
iv. The White House released a blueprint for an AI Bill of Rights. The Bill discusses the principles for design, use and
deployment of automated systems91
v. The FDA released a list of AI/ML-enabled medical devices currently being marketed in the United States6
vi. The Association for the Advancement of Medical Instrumentation (AAMI) published a consensus report for iden-
tifying, evaluating, and managing risk for healthcare technology that incorporates AI or ML92
2023 i. NIST released their AI Risk Management Framework93
ii. The FDA issued a draft guidance on recommendations for a Predetermined Change Control Plan for AI software
devices. This guidance suggests an openness in Agency’s thinking to allow manufacturers developing AI/ML
algorithms to retrain their models from data in real time20
iii. The FDA released an updated software guidance on the information necessary to include in premarket submis-
sions for the FDA’s evaluation of the safety and effectiveness of device software functions94
iv. The Biden-Harris Administration announced the creation of the NIST Public Working Group on Generative
AI, which will build on the NIST AI Risk Management Framework to address AI technologies that can generate
content, including images95
AI, artificial intelligence CADe, computer-assisted detection CDS, clinical decision support FDA, Food and Drug Administration IMDRF,
International Medical Device Regulatory Forum ML, machine learning NIST, National Institute of Standards and Technology SaMD,
software as a medical device TPLC, Total Product Life Cycle