Diving brief:
- The FDA published on Tuesday a action plan To establish a regulatory approach to the rapid development field of artificial intelligence and software based on automatic learning as a medical device (SAMD).
- The action plan comes in response to substantial comments from stakeholders, including hundreds of public comments, on April 2019 discussion This proposed a framework to regulate the modifications in AI and Samd.
- The latest publication of the FDA on the subject is the next step on a path to write advice for a “predetermined changes control plan” which would include the types of covered modifications and the methodology used to implement the changes of a manner manages the risks for patients, patients, patients, patients, patients, patients, patients, patients, patients, patients, patients, patients, patients, patients, patients , patients, patients, patients, patients, patients, patients, patients, patients, patients, patients, patients, patients, patients said the agency.
Diving insight:
The action plan intervenes in the midst of Advamed’s regulatory clarity calls and other automatic learning algorithms that are constantly evolving without the need for manual updates. The FDA has so far approved or erased only devices that use locked “algorithms that do not change this way.
Ge Healthcare, Medtronic and Philips are among those who work in integration of AI capabilities in their treatments and investment in technology. Medtronic, for example, is focused on AI -helped technologies that would support robotics, navigation, imagery and preoperative planning for spine surgery. The Medtech giant in November acquired the French vertebral surgery company Medicrea, accessing an AI database of more than 5,000 surgical cases.
And last month, Philips announced an agreement of $ 2.8 billion to buy bio telemetry, which specializes in remote heart diagnostics and monitoring, including portable heart monitors and data analysis based on ‘Ia.
Since the publication of the FDA discussion document almost two years ago, the industry has been waiting for an update on its surveillance plan. Since the FDA has proposed a total approach to the life cycle of AI / machine learning products that could allow improvements to devices after use, the advice or finalized regulations would help developers better understand the Potential requirements and responsibilities, said a GAO / National Academy report in November.
In the new document of the action plan, the FDA has presented the ability to learn from real world experience to improve the performance of devices as one of the greatest advantages of the software. He suggested that “predetermined changes control plan” could allow the evaluation and monitoring of pre-market development software by the postal market market.
The approach has aroused strong interest since its described in the 2019 discussion document, said the FDA. The agency noted that health legend, which received Marketing authorization In February for the first cardiac ultrasound software with AI to guide users, used a predetermined changes control plan.
“This framework would allow the FDA to provide reasonable safety and efficiency insurance while adopting the power of iterative improvement in artificial intelligence and software based on automatic learning as a medical device,” said The regulator.
The FDA said that it will issue an advice project on this concept in response to stakeholder suggestions. The project will propose what should be included in the SAMD pre-specials and the algorithm change protocol to support the safety and efficiency of SAMD AI / ML algorithms.
The agency also responded to the concerns of stakeholders concerning algorithmic biases, saying that it will support the regulatory sciences to develop the methodology to assess and improve automatic learning algorithms. Such a methodology would target the identification and elimination of the bias, as well as the evaluation and promotion of the robustness of the algorithms.
“The provision of health care is known to vary according to factors such as race, ethnicity and socio-economic status; therefore, it is possible that the biases present in our health care system are inadvertently introduced In algorithms, “said the agency. The FDA “recognizes the crucial importance so that medical devices are well suited to a population of racial and ethnically diverse patient patients and the need to improve methodologies for the identification and improvement of automatic learning algorithms” , said the action plan.
The FDA is committed to encouraging the harmonization of good automatic learning practices and holding a public workshop on labeling to support transparency concerning AI / ml devices.
The agency also said that it will work with stakeholders on a voluntary basis of the collection and monitoring of real world performance data, in order to support the total approach to product life cycle to monitor SAMD based on the IA / ml.