Should radiographers fear the rise of AI automation?

Many of us have heard the term artificial intelligence (AI) and are perhaps aware of some real-life examples. For years we have been fed a steady stream of headlines touting that AI will revolutionise the world around us but have we truly witnessed this yet? 

Such predictions of an AI-powered world date back as far as the advent of the term itself, which was coined in 1956 at the Dartmouth Conference by John McCarthy, the man considered to be the "father of AI". 

I am a new PhD researcher, in my infancy within the radiography profession. The insurgence of AI, by happenstance, is occurring at the same time. My research aims to look at the workforce’s understanding of AI and at ways that we can build on the profession’s understanding and knowledge, both at curriculum level and with ad-hoc training provision. 

Utopian and dystopian visions

The dream of a true AI has been an obsession and goal of mankind since the inception of the concept. We are served a banquet of utopian and dystopian visions of AI by the media – if life were like the movies, by 21 October 2015 we would all have been cruising to our clinical departments on Back to the Future-style hoverboards¹. Or for those who prefer the darker side of science fiction, we could soon be revealed to be just organic battery packs used by a sentient AI overlord, according to The Matrix². 

If I had to choose, my personal favourite depiction of AI would be Baymax, the bumbling "personal healthcare companion" from Disney’s Big Hero 6³. Less outlandish portrayals, such as Hank Scorpio in The Simpsons⁴ with his home of the future, are more reminiscent of mid-20th century cartoon strip "Closer Than We Think!"⁵ ideals that demonstrated technology as the key to a better way of life, with automation taking care of burdensome chores and enhancing crop yields, improving travel and general wellbeing. This reverie is closer to what we are experiencing today. But what is the ubiquitous AI?

We know it makes for good sci-fi entertainment. In his 2004 paper, McCarthy defined AI as "… the science and engineering of making intelligent machines, especially intelligent computer programs. It is related to the similar task of using computers to understand human intelligence, but AI does not have to confine itself to methods that are biologically observable." ⁶

Practical applications

Unless you are tech-savvy (which I understand many of you might be!) that still doesn’t really explain very much. Rather than get bogged down in definitions or carried away with Hollywood flights of fancy, let’s instead think of practical applications of AI. 

Do you know what AI technologies are present in your home? What about in your local grocery store? In your clinical environment? For a long time, science fiction was just that, fiction. Increasingly though, AI is being embedded in all aspects of life. But to what end? Towards the end of 2022, the creative arts world was rocked by two new AI systems: an open-source AI chatbot named chatGPT⁷, which can write stories and essays comparable to those created by humans, and an image generator named Lensa⁸, which can create digital art from uploaded selfies. 

Dissenters are citing ethical quandaries as their main issues with these technologies, along with worries that AI-generated material will devalue "true" art. Another concern is a lack of understanding of how the AI generates its output. These issues are not confined to the creative world – they have been relevant in our world of healthcare since AI technologies were first implemented.

AI in healthcare

The reality is that AI in healthcare is being rolled out at a startling pace. Reams of studies are being published, funded by manufacturers and the industry, lauding the capabilities of their software and comparing its performance with that of clinicians – often matching and, in some cases, outdoing its human counterparts.

This and the fact that machines don’t suffer fatigue and can process unfathomable amounts of data at mind-blowing speeds mean it's no wonder that AI has been referred to in some publications as "the enemy"⁹, threatening to automate processes, replace some roles and make others completely obsolete^10,11. 

This is not the case. What is happening is that AI is being used to assist healthcare providers. One example is MASTER-TBI¹², a collaborative project between the Pawsey Supercomputing Centre in Perth, Western Australia, and three regional trauma centres. This technology has recently been able to improve patient outcomes in the ICU through the collation and analysis of a large dataset of historical patient outcomes and has helped predict adverse events in patients with moderate to severe traumatic brain injuries (TBI). 

This is a prime example of how AI can enhance and supplement healthcare. Being able to anticipate such events enables ICU staff to intervene before the event occurs, demonstrating symbiosis between human skill and AI capabilities.

As with all great innovations, hype occurs around the release of a new AI technology. But this is not the first outing for AI and radiology. This is a relationship that began back in the 1960s, when computer-aided diagnosis (CAD) systems were introduced in mammography. It ultimately failed when it became apparent that the gains in sensitivity were at the sacrifice of specificity¹³, and that workload was actually increasing due to radiologists having to review images that had been erroneously flagged as suspicious by the software. 

This episode led to reluctance within the medical community to adopt AI technology in relation to patient care. More recently, however, it has become accepted that AI has myriad potential applications in the world of medical imaging, from task automation to image interpretation. A total of 240 AI or machine-learning tools gained CE Mark approval for use in medical imaging between 2015-2020¹⁴. Have any of these tools crept into your departments? Perhaps they are already there, expensive and underused because they failed to live up to expectations or are not yet trusted?

Are we then destined for a future with Star Trek-style healthcare and artificially intelligent systems that can test, diagnose and treat? You, my colleagues on the frontline, know better than anyone that such a future is not yet on the horizon. 

The human touch

The technology will continue to advance rapidly and offer more complex and nuanced healthcare support, but AI will never be able to truly emulate human empathy or sincerity, to know when to whisper words of reassurance to the patient who, despite claiming to be fine, is betrayed by the fear on their face. An algorithm is not able to give a short squeeze of a patient’s hand before sending them through the scanner. 

AI is a tool that has the potential to change the way we work. It is to be embraced and harnessed, made to work for us and patients, enhancing how we do the job not taking it from us. But to embrace the unknown is difficult.

A wise general once said: “Know thy enemy”¹⁵. This sentiment is echoed (albeit in a more academic fashion) in publications from the Society of Radiographers¹⁶, NHS¹⁷ and government and other regulatory bodies¹⁸, which all outline education on AI as a future priority. Education is the key to being able to understand AI and implement it safely and effectively so, as a profession on the cutting edge of healthcare and technology, we must ensure we are educated about any technology that will be used in clinical practice. We can achieve symbiosis with AI but to do so we must understand it.

Conclusion

Although light-hearted in essence, my point is that to fulfil our roles as healthcare professionals we must move with the tide of change. To allow AI to permeate into clinical practice without educating ourselves about its capabilities and limitations would be ethically and morally reckless.

Radiography is, arguably, the most technologically driven field in healthcare, so it stands to reason that we are likely to be the practitioners who will embrace AI and ensure it is implemented safely, cost-effectively and with the best interests of patients at the centre of everything we do. To achieve this, we must understand the core concepts of AI so that we can appropriately utilise it in our individual roles and the profession as a whole.

I invite you to help me begin a process of evaluating the workforce’s current knowledge of AI, your expectations of how it could/should be implemented and your ideas about the future of radiographic practice and education in this new era. 

Recent studies have aimed to gauge your opinions and perceptions of AI. I am interested in finding out what you know (or think you know) about AI, and what you would like to know, and to use this to facilitate a discourse on how we can modify the radiography curriculum to effectively deliver education on AI at appropriate stages of a radiographer’s career.

You can do your bit to assist in this endeavour by completing the short, anonymous survey via the link/QR code below. Data collected will be used to inform the basis of a study looking at ways to improve educational provision for all imaging department staff to ensure that we, as a profession, remain at the forefront of diagnostic support.