I often get asked about the future of radiologists. AI is starting to appear in clinics, with around 5-10% of imaging doctors reporting the use of some sort of AI (unpublished data). In the eyes of many, although this is more often true of journalists rather than doctors, automation is a threat to the profession.
So far, the evidence for the death of radiology is non-existent. Not one job has been lost. It even seems likely that we will need a new workforce made up of a new sort of sub-specialist: the clinical AI informaticist.
Standards of care
RANZCR, the College of Radiology and Radiation Oncology in Australia and New Zealand, have recently posted a draft “Standards of Practice for Artificial Intelligence“, and in doing so have become the first professional organisation in the world that has recommended that all departments and practices that deploy AI employ a “chief medical information officer” to oversee the clinical use of these systems. An excerpt:
3.1
The practice shall appoint a designated clinician, a Chief Medical Information Officer (CMIO) to oversee the selection, deployment and monitoring of ML and AI tools and associated staff training within the practice.
This standard is not that unexpected, particularly in radiology where the use of information technology is so widespread. It is often highly useful to have imaging informatics (IT) specialists available to make decisions about software.
But this is not a garden variety CMIO role. Again from the Standards:
3.1 (iii)
The designated medical lead or CMIO must be trained and current in appropriate skills, knowledge and competencies for their role.
What this is essentially saying is that each practice should have a medical officer who understands enough about AI to weigh risks and benefits, oversee deployments, create contingency plans, and so on.
Guess what? This workforce flat out does not exist.
Who wants to be indispensable?
There are tentative steps being made all over the world to build this workforce. I am personally involved in several projects, and I’m announcing the first of them today:
Applications are open for the first Clinical Radiology AI Fellowship in Australia!
Well, the first clinical AI fellowship in any specialty, as far as I know.
For the next two weeks we will be accepting applications from any RANZCR* radiologist or trainee who has completed their part two exams, to undertake a mixed advanced clinical and AI research role at the Royal Adelaide Hospital (in South Australia), supervised by me and supported by our medical imaging research group at the University of Adelaide and the Australian Institute for Machine Learning. We have world leading expertise, data, and resources available, making this the ideal place to learn the skills needed to become part of the new AI workforce of the future.
In brief, the program will aim to develop an applicant’s skills, to the point they can appraise, build, test, and even deploy AI models. If all goes well, the models an applicant builds during this fellowship will be put into actual clinical practice in the department!
While I understand that this is a very strange time of year to advertise for fellowship positions**, I’m hoping that the unique role will catch the eye of anyone who wants to enter this exciting field of study and practice.
Honestly, if I was running a radiology practice I wouldn’t be waiting for new grads to come out with these skills already, I would be asking the fully qualified radiologists at the practice if any of them were willing to do a fellowship like this.
I really want this initiative to work and spread to other departments across the nation, so having a successful start now is a big deal. If you know anyone who might be interested, please let them know about this. Applications will close on the 21st of this month.
If anyone wants more information, please get in contact here, on Twitter (@drlukeor), or email me (you can find my address here).
Lauren Oakden-Rayner 
“Each practice should have a medical officer who understands enough about AI to weigh risks and benefits, oversee deployments, create contingency plans, and so on.”
I think this is such a cool and important idea. As an American medical student with an interest in clinical applications of AI who’s going into pathology, I’m curious what your thoughts are of the existing clinical informatics fellowship programs in the U.S., and how well their curriculums are set up to train fellow to do the type of work that you are describing?
LikeLike
Hi azimiv. I think the clinical informatics fellowships are moving towards this, but it is early days. You would have to reach out the the providers and ask how their curriculum suits.
Here in Australia we are offering an AI fellowship this year, as a first attempt to solve this challenge. If it goes well, hopefully more will be offered around the world.
LikeLiked by 1 person
Hi Luke, I had a heated debate with my partner yesterday about the point of having a clinician in AI team for AI-driven products in healthcare. From your experience, is it true that every AI healthcare company has clinicians who will give clinical expertise to their AI products before trialling or launching it to the marketplace? If so, asking a radiologist to ‘understand enough about AI’ would be burdensome to him/her?
I am Vinh from Melbourne, currently doing a PhD in health economics. Really glad to e-meet you here.
Thanks Luke.
LikeLike