Review of AI guidance published by the SOR AI working group (Malamateniou et al., 2021)

This paper was published due to the increasing development and use of artificial intelligence within the radiography profession. It aims to provide baseline guidance for those that may encounter AI in education, research, clinical practice and stakeholder partnership. This interests me particularly as I am involved in two of those areas, and hope to be involved in AI research in the future.

The concept of AI in healthcare has been around for a while, however the actual applications are new, and there has been a global realisation of the potential AI has with regard to patient care and clinical practice. WHO established an expert group in 2019 to help address the ethics, governance and regulation of AI in health, with the purpose that a global framework will be conceived. This paper highlights the essential role of radiographers in elevating patient care with the use of AI, and builds on the WHO commitment.

This guidance is well written and clear, using a format that has been used in previous similar publications. It does note that this is a working document that will require regular updating due to the rapidly evolving nature of AI.

Supplementary material is referenced which I shall review at a later date, this indicates a good evidence base for these guidelines and increases the validity of the paper. They also state that they sought expert opinion, which is essential as this subject matter is essentially computer science, which as radiographers we may not have as good a grasp on as those that are involved in the building and writing of software.

The authors used peer professional discussion and debate with specialised subgroups to draw conclusions, increasing the robust nature of this paper further. The subgroups represented the 4 previously mentioned areas that may encounter AI: education, research, clinical practice and stakeholder partnership. Using these four areas as headings gave useful structure to the paper, along with summarising tables, and will be extremely helpful for those requiring specific guidance to be able to find what they need efficiently.

There are 6 strategic priorities for clinical practice, each has an associated recommendation, a body responsible for implem entation and the status of that implementation. Examples of these priorities include: Validation of AI tools, Ensuring equity and fairness, Ensure auditing, CPD/training, etc. One priority is the engagement of professionals, and the recommendation is that relevant research is undertaken with the input of stakeholders. Those responsible for ensuring this are; researchers, clinicians, academics, healthcare bodies, etc. The status of this priority is pending, as there are a number of initiatives and research ongoing. This format is clear and easy to understand.

A lot of importance is given to internal and external validation of AI tools, and the need for human oversight. There is a discussion to be had on how this is an addition to the radiographer’s role, and how it will be integrated as such. There is also a need for standardisation and audit of these systems, and this will also be an extension of the role.

The second table shows 11 priorities for education, and includes: collaborative learning, engagement with stakeholders, ensuring training, creating learning opportunities, considering the inclusion of AI within curriculum networking, etc. The paper does mention the new post-graduate module ‘ Introduction to AI for radiographers’, this is an interesting looking module on investigation, however it does seem to be one of a kind, and this is highlighted as an ongoing priority.

There are 8 priorities for research, including; baseline knowledge exploration, the impact of AI on patient experience, the radiographer’s role, practices and protocols, clinical decision making, promoting radiographer researchers, offer support and funding. The paper describes a lack of information on the impact of AI on the role of a radiographer and service delivery, and it provides some recommendations for research. There are two further guidelines to investigate if looking into AI for research, these are SPRIT-AI and CONSORT-AI guidelines, which may offer further recommendations.

The table for stakeholders shows 5 priorities, including: collaboration between industry and the other sectors development of ambassadors, establish terminology, and implementation. The main takeaway from this section is the need to build partnerships between industry and healthcares, educators and researchers.

The paper notes that it is still early in development, and long term cooperation with the SOR and other organisations is required to keep the guidance up to date.