From building the consensus on the definition of artificial intelligence responsible for talking about cases of almost independent artificial intelligence in the field of health, the industry and the government seem to be directed towards common values, even in these polarization times, and we are heading towards 2025, says the CEO of the Intelligence Health Care Alliance. Artificial, Brian. Anderson.
Anderson explained: “We need to understand policymakers and organizational officials with these types of frameworks that are developed in the private sector about the responsible form of artificial intelligence in the field of health, and then developing regulatory frameworks on this,” Anderson said. Health technology news news.
Contemporary so far
There has been a lot of thinking by the technology industry and federal organizational organizational organizations in artificial intelligence model cards, or “artificial intelligence feeding stickers” – an easy -to -digestive form of communication used to determine the main aspects of developing the artificial intelligence model for users.
On Thursday, when Chai issued its report An open source version of the artificial intelligence model draft cardWe talked with Anderson about the coalition’s recent experience and his vision about what the future future might have to develop a working framework between the public and private sectors to spread artificial intelligence in health care safely.
He said: “It is a wonderful thing to see the consensus between the direction that the innovation community is heading in the private sector and the destination that the organizational community is heading to in the public sector.”
While Chai is looking for comments about the open source of the form of the open source card this month – which “flows from” the National Coordinator Office for Health and Technology and the inter -employment of health care technology: updates of the certificates program, the transparency of the algorithm and the information base – and with a plan to put up an update at some time during the months The next six, Anderson said that he hopes that the various regulatory authorities will continue to move in the same direction.
Specifically, he pointed to a degree of consensus on the requirements of artificial intelligence with the regulatory authorities of the medical devices.
Earlier this week, the US Food and Drug Administration included an example of the voluntary artificial intelligence model card in its report Dawn of the total product life cycle recommendations Design, development, maintenance and documentation – for devices that support artificial intelligence.
“One of the exciting things is to look at the American Food and Drug Administration Card (FDA), and then look at it for sure ONC HTI-1 base Anderson said: “- It is that the CHI model card is strongly corresponding to both,” Anderson said.
According to the US Food and Drug Administration, the model card of medical devices that support artificial intelligence can address communication challenges in providing important information to healthcare users – patients, doctors, organizers, researchers – and the public.
“Research has shown that the use of the typical card can increase the user’s confidence and understanding,” said the Food and Drug Administration in a draft directive. “It is a way to summarize the main aspects of devices that support artificial intelligence constantly and can be used to describe their properties, performance and restrictions briefly.”
As one of the organizational examples of artificial intelligence in the field of health care, it is an indication of the direction that the Food and Drug Administration organizers tend to work to frame confidence about the use of artificial intelligence.
Anderson said: “However, the stakeholders groups in the public and private sectors need to work cooperatively to inform each other,” Anderson said.
When asked, he referred to the upcoming administration, “And every leader spoke to him in the House of Sheikhs and House, is very interested in understanding how they can partnership in a partnership between the public and private sectors with organizations such as Chai.”
And with the opening of the door to continue the tasks of the most stringent artificial intelligence in the field of health care – such as the compatibility between the government and the industry on artificial intelligence guarantee laboratories and how they work – “Anderson said.
He added: “We need time to do so, and the coming administration appreciation for their willingness to work with us and work alongside us – and we hope that they will save that time – I think this is very fresh and very exciting.”
Update annual label and use IRL
Anderson said that the CHI model card aims to be “a living document that breathes with the emergence of new capabilities, especially in the field of obstetric intelligence.”
He said: “It is very likely that you will need the standards and methodologies that we use to assess these emerging capabilities to change or to create.”
Even before the Food and Drug Administration issued a draft of the total life cycle guidelines for medical devices, It has completed the pre -defined change control plan For artificial intelligence transmission and machine learning – without provoking new marketing transmitters.
Anderson said: “While we think about the different sections of the model card, different data should be included – different evaluation results, different standards … and different types of use cases, and so on.
He added: “This type of flexibility will be really important,” noting that the artificial intelligence model or the system’s feeding poster will require regular updates, “certainly, at least at the annual level.”
For service providers, there is a great deal of complexity that must be taken into account when using clinical decision -making tools supported by artificial intelligence to reduce errors or omission.
He stressed that “the incomplete transparency will be something we will fight with and we need to work for it again.”
Whether the model is trained on a specific set of features that may relate to a specific patient or not, it may not be included in the easy -to -use forms cards.
He said: “You can include all the information under the sun in these model cards, but the sellers community will be at risk Useson in front of them. Tdecisionaboutwitherornottheyshoulusthatmodelwithepatienttheyhaveinfrontfthem “Andersonsaid[الملكيةالفكرية””لذا،فهوتوازنبينكيفيةحمايةالملكيةالفكريةللبائع،ولكنإعطاءالعميل-الطبيبفيهذهالحالة-المعلوماتاللازمةلاتخاذالقرارالصحيححولماإذاكانيجبعليهماستخدامهذاالنموذجمعالمريضالذيلديهمأملاوقالأندرسونأمامهم[intellectualproperty”hesaid”Soit’sabalanceofhowdoyouprotecttheIPofthevendorbutgivethecustomer–thedoctorinthiscase–thenecessaryinformationtomaketherightdecisionaboutwhetherornottheyshouldusethatmodelwiththepatienttheyhaveinfrontofthem”Andersonsaid
He admitted that “the causal relationship is really deep in terms of how this affects a specific result of the patient in front of you.”
Bring others to the AI table
While the 31st fields of HTI -1 – which target electronic health records and other approved health information technology – “is a really great starting point”, it is not enough for different cases of use of artificial intelligence – especially in the direct field of the consumer, as He said. Anderson.
“The typical cards that we are working on developing aim to use them widely through various cases of use, and in the field of consumer, especially with obstetric artificial intelligence, there will be a full set of new use cases next year and clearly.
However, over the two years to the next five years, the evaluation of artificial intelligence models for health care will become more complicated, raising questions about how to define “human prosperity”.
Anderson said he believed that the use of use will be closely related to artificial intelligence agents in the field of health, and the development of confidence frameworks about these will require support from “ethics, philosophers, sociologists and spiritual leaders” to help provide advice to technicians and artificial intelligence experts in the field of artificial intelligence. Think through an evaluation framework for these tools.
He said: “The development of this type of framework for evaluation in the future of artificial intelligence will be a real challenge.” “It is a very intimate personal space, so how do we build that confidence with these models? How can we evaluate these models?”
Starting next year, Anderson said that Chai will lead “a very intended effort to bring together members of society together and stakeholders who will not necessarily think first of the types of stakeholders who will be included in such an effort.”
“We really need to make sure that these models are in line with our values, and we do not have a title on how to assess the model. I don’t know how to do this yet. I don’t think anyone knows this, yet.”
Andrea Fox is a first editor in Healthcare IT News.
Email: Afox@himss.org
Health Care Information Technology News is one of the media publications of HIMSS.
