From X-Rays to AI: Navigating US Regulations in Radiological Health
133
Control Plan.33 This development suggests that
the future is headed towards adaptive algorithms,
and the Agency seems to be open and encourag-
ing companies to develop adaptive algorithms.
As the federal government continues to
receive input on the trustworthiness of AI the
industry should expect to see more guiding prin-
ciples and recommendations. NIST continues
to lead efforts in developing AI/ML standards
and best practices for measuring and assessing of
AI/ML technologies. Towards this goal, NIST
launched the Trustworthy and Responsible AI
Resource Center which is intended to facilitate
implementation and international alignment
with risk management framework.93 The
Organization for Economic Co-operation and
Development (OECD) has also provided a set of
internationally agreed principles and recommen-
dations that can promote an AI/ML-powered
crisis response that is trustworthy and respects
human-centered and democratic values.96 Hence,
in coming years we expect to see various health
authorities across the globe coming together
to develop frameworks and recommendations
towards global harmonization.
As the use of AI products increases, pro-
tecting an individual’s information and privacy
becomes important. Hence, in the coming years
it is expected that there will be an increased
vigilance on the privacy and cybersecurity issues
mentioned in this chapter.
Summary and Conclusion
SaMD and SiMD incorporating AI/ML technol-
ogies have the potential to revolutionize clinical
practice in the radiological health space.1 The
applications of AI/ML are diverse, and they have
the potential to impact not only every step of the
diagnostic imaging chain – from protocol creation
to image generation and image analysis, to the
analysis of electronic health records – but also the
therapeutic space in radiation oncology where the
treatment planning process can be streamlined.
This chapter has focused on an area of
rapid evolution: regulatory considerations for
CAD devices. CAD devices are helping physi-
cians manage workflow, detect disease early and
accurately, and assess risk to improve patient
outcomes and medical care.
Major developments in the FDA regulation
of CAD in the last few years have included the
down-classification of some CAD devices from
Class III to Class II,36 the creation of a mechanism
for FDA review of predetermined change control
plans for AI/ML enabled device software func-
tions,97 the De Novo reclassification and creation
of regulations19 for different types of CAD devices
that can now serve as predicates for 510(k)’s, and
FDA marketing authorization of a wide range of
CAD devices via the De Novo pathway.31
While regulation helps ensure safety, inno-
vative technologies can also challenge regulatory
frameworks. For example, adaptive AI/ML
technologies with the potential to optimize their
performance in real-time have found appli-
cations in the financial sector, marketing and
e-commerce industries, but do not necessarily
fit well into the current frameworks of medical
device regulation. The regulation of these tools in
healthcare will require new approaches allowing
the devices to continually improve while provid-
ing effective safeguards.
With the increasing availability of patient data
and advancement in machine learning algorithms,
an increasing range of AI/ML-based software tools
will become available for clinical use however, mul-
tiple challenges such as algorithm generalizability,45
data privacy,80 and cybersecurity90 will need to be
addressed in order to ensure widespread clinical
adoption of these tools.
AI/ML algorithms are expected to continu-
ously improve over time, and regulation will need
to adapt alongside. Moreover, as the use of AI/
ML technologies in medical devices continues to
advance, all involved stakeholders including man-
ufacturers, developers, regulators, users, patients,
and payors across the globe should work closely
together to develop safe and effective methods to
improve patient outcomes in healthcare.
References
1. Gottlieb S, Silvis L. Regulators face novel challenges
as artificial intelligence tools enter medical Practice.
JAMA Health Forum. 2023 4(6):1-3e232300. Verified
13 June 2023.
2. Rajpurkar P, et al. AI in health and medicine. Nat Med.
2022 28(1):31-48. Verified 13 June 2023.
133
Control Plan.33 This development suggests that
the future is headed towards adaptive algorithms,
and the Agency seems to be open and encourag-
ing companies to develop adaptive algorithms.
As the federal government continues to
receive input on the trustworthiness of AI the
industry should expect to see more guiding prin-
ciples and recommendations. NIST continues
to lead efforts in developing AI/ML standards
and best practices for measuring and assessing of
AI/ML technologies. Towards this goal, NIST
launched the Trustworthy and Responsible AI
Resource Center which is intended to facilitate
implementation and international alignment
with risk management framework.93 The
Organization for Economic Co-operation and
Development (OECD) has also provided a set of
internationally agreed principles and recommen-
dations that can promote an AI/ML-powered
crisis response that is trustworthy and respects
human-centered and democratic values.96 Hence,
in coming years we expect to see various health
authorities across the globe coming together
to develop frameworks and recommendations
towards global harmonization.
As the use of AI products increases, pro-
tecting an individual’s information and privacy
becomes important. Hence, in the coming years
it is expected that there will be an increased
vigilance on the privacy and cybersecurity issues
mentioned in this chapter.
Summary and Conclusion
SaMD and SiMD incorporating AI/ML technol-
ogies have the potential to revolutionize clinical
practice in the radiological health space.1 The
applications of AI/ML are diverse, and they have
the potential to impact not only every step of the
diagnostic imaging chain – from protocol creation
to image generation and image analysis, to the
analysis of electronic health records – but also the
therapeutic space in radiation oncology where the
treatment planning process can be streamlined.
This chapter has focused on an area of
rapid evolution: regulatory considerations for
CAD devices. CAD devices are helping physi-
cians manage workflow, detect disease early and
accurately, and assess risk to improve patient
outcomes and medical care.
Major developments in the FDA regulation
of CAD in the last few years have included the
down-classification of some CAD devices from
Class III to Class II,36 the creation of a mechanism
for FDA review of predetermined change control
plans for AI/ML enabled device software func-
tions,97 the De Novo reclassification and creation
of regulations19 for different types of CAD devices
that can now serve as predicates for 510(k)’s, and
FDA marketing authorization of a wide range of
CAD devices via the De Novo pathway.31
While regulation helps ensure safety, inno-
vative technologies can also challenge regulatory
frameworks. For example, adaptive AI/ML
technologies with the potential to optimize their
performance in real-time have found appli-
cations in the financial sector, marketing and
e-commerce industries, but do not necessarily
fit well into the current frameworks of medical
device regulation. The regulation of these tools in
healthcare will require new approaches allowing
the devices to continually improve while provid-
ing effective safeguards.
With the increasing availability of patient data
and advancement in machine learning algorithms,
an increasing range of AI/ML-based software tools
will become available for clinical use however, mul-
tiple challenges such as algorithm generalizability,45
data privacy,80 and cybersecurity90 will need to be
addressed in order to ensure widespread clinical
adoption of these tools.
AI/ML algorithms are expected to continu-
ously improve over time, and regulation will need
to adapt alongside. Moreover, as the use of AI/
ML technologies in medical devices continues to
advance, all involved stakeholders including man-
ufacturers, developers, regulators, users, patients,
and payors across the globe should work closely
together to develop safe and effective methods to
improve patient outcomes in healthcare.
References
1. Gottlieb S, Silvis L. Regulators face novel challenges
as artificial intelligence tools enter medical Practice.
JAMA Health Forum. 2023 4(6):1-3e232300. Verified
13 June 2023.
2. Rajpurkar P, et al. AI in health and medicine. Nat Med.
2022 28(1):31-48. Verified 13 June 2023.