We were delighted to speak with Susan Shelmerdine this week for our interview series, “On Artificial Intelligence”. Among Shelmerdine’s many accolades are her roles as the Chairperson for the Artificial Intelligence Taskforce at the European Society of Paediatric Radiology as well as the prestigious Roentgen Professorship at The Royal College of Radiologists. Join us as we do a deep dive into Shelmerdine’s interest in AI, her recommendations for young radiologists interested in the topic, and what the future holds for AI in radiology.
What is your background/experience with artificial intelligence?
I am an NIHR-funded post-doctoral academic radiologist leading a multidisciplinary team of researchers, including both engineering and medical experts, on various AI projects related to children’s imaging. Together, we have curated a large national multicentric dataset of paediatric radiographs, developed in-house image classification models, and externally validated commercial AI tools. We are engaged in a range of multi-reader AI studies, collaborating with adult imaging colleagues to benefit from their experiences. As the AI taskforce chair for the European Society of Paediatric Radiology, I am also passionate about advocating for the safe and responsible use of AI for children and working with international societies to adopt streamlined, collaborative approaches on this issue.
What first attracted you to the topic?
As a deeply philosophical person, I am drawn to AI because it compels us to think creatively and reflectively about our society and the value we contribute to it as individuals. In our data-driven, digital environment, we have the ability to control many aspects of our world, prompting us to ask: What is the biggest problem we need to solve? How can we make that a reality using data insights and applications? The presence of bias in AI systems urges us to examine ourselves and societal constructs. Furthermore, as AI handles more tasks and thinking, we must consider the value we bring to the world and how we can support it to work better for us. In a nutshell, AI encourages us to think big, think creatively and think deeply.
What are the biggest challenges to AI adoption in clinical practice?
The primary challenges to AI adoption in clinical practice are funding and education. In a healthcare system constrained by financial and staffing shortages, securing additional funds for AI technology is difficult. Many existing IT infrastructures are outdated, necessitating significant upgrades to support modern AI applications. Without adequate funding, it is also challenging to hire and train knowledgeable staff to manage and implement AI systems effectively.
Education is equally critical. There is considerable misinformation and varied perceptions about AI’s capabilities and limitations, as well as misunderstandings regarding cybersecurity, data sharing, and safety. These discrepancies lead to different levels of understanding and acceptance, causing communication barriers within teams. To achieve meaningful large-scale adoption, it is essential to have well-trained and knowledgeable staff and appropriate resources to support the integration and maintenance of AI tools.
Give us an example (or an educated guess) of what you think AI will be able to do in 3 years. What about in 10 years?
In the next 3 years every hospital will have an established AI department with multidisciplinary staff overseeing AI governance, and the subspecialty of ‘AI & informatics’ within medicine will start to become established in its own right (similar to radiology, dermatology, cardiology etc.).
Within a decade AI will significantly transform organisational processes in medicine, fostering a more patient-centred ‘self-service’ approach. This will include booking appointments/tests/scans, choice of doctor, hospital and treatment options all tailored to the individual needs and background, possibly co-ordinated through smartphone applications, biometrics and with nationalised improvements in health data sharing across medical facilities. Enhanced AI tools will streamline the vetting and protocolling of appropriate requests by patients. Autonomous AI (even if not officially sanctioned) will play a role in some diagnostic processes, driven by increased staffing shortages, increased demand from an ageing population and growing backlogs.
As a radiologist, what is your advice for younger colleagues wanting to dive into the topic of Artificial Intelligence, Machine/Deep Learning, and/or Radiomics? What are your tricks for staying up to date in this fast-evolving field?
Firstly, choose a niche that interests you, whether it’s model development, regulation, deployment, surveillance, ethics etc. Traditional methods, like reading specialty journals, are often too slow to keep pace with technological advancements. Instead, supplement these by following trusted key opinion leaders on social media, subscribing to newsletters such as Doctor Penguin, Health Data Nerd etc. and attending relevant conferences. Networking is crucial, especially with industry, as it allows you to learn about novel projects early on. Finally, complement your new ‘inside’ knowledge with webinars to better grasp some of the challenging and nuanced concepts in more detail. These strategies combined will keep you informed and ahead in this rapidly evolving field.
