Requirement for utilization of AI in patient diagnosis......Third, allow clinicians to retrain AI models with local data if the needs of their patients and hospital require it.
The above extract from is from an article well above my pay grade in my ability to assess its relevance to BRN.
So I look to others to assess it, in the context of the following FEBRUARY 12, 2022 Lancet article
Bridging the chasm between AI and clinical implementationMany advances in artificial intelligence (AI) for health care using deep neural networks have been commercialised. But few AI tools have been implemented in health systems. Why has this chasm occurred? Transparency, suitability, and adaptability are key reasons. The deployment of any new technology is usually managed centrally in hospitals and health systems. For the information technology (IT) teams, there is the concern that input data are drawn from outside the health setting and the algorithm performance, source code, and input data are unavailable to review. Many commercial AI applications are in radiology, but few are supported by evidence from published studies. And there are concerns that the algorithms were tested and validated using retrospective, in-silico data that may not reflect real-world clinical practice. Regulators reviewing a company's AI data are privy to considerable data, but these data are usually unavailable to health system IT teams or clinicians.
Equity, safety, and regulation are also crucial. When AI is imported from a commercial setting into a hospital or health system little is known about which data have been used to train the AI and there is uncertainty about owner-ship of the subsequent, real-patient data that AI draws on post implementation. Additionally, AI implementations can be hindered by inadequate information about the data used to make decisions and recommendations. Clinicians receiving AI-generated decisions rarely have oversight of the datapoints used to reach a specific decision, contributing to so-called algorithmic aversion. Such algorithmic aversion can also arise when it is unclear who (eg, developer companies, researchers, clinicians, or hospitals) should accept responsibility for the algorithm's decisions. Any consideration by hospital leadership to implement AI-based diagnostic and decision-making tools should closely consider these concerns and barriers.
How can we bridge this chasm? Algorithmic aversion can be alleviated when an algorithm can be modified in the health setting or can learn from local data. Three steps will help optimise clinical use of AI. First, provide transparency about the datasets used for initial training of the AI tools. Second, enable the deconstruction of neural networks to make the features that drive the AI performance understandable for clinicians.
Third, allow clinicians to retrain AI models with local data if the needs of their patients and hospital require it.( My emphasis)
Source is the link below:
Bridging the chasm between AI and clinical implementation - The Lancet
(20min delay)