A CoRIPS success story: shaping the role of AI in radiography

Earlier this year, research radiographer Avneet Gill received funding to hold an AI conference in Belfast, focusing on the bottlenecks to the technology’s integration and what more can be done to promote research into it. Synergy spoke with Avneet to discuss the findings

“I’m really looking forward to how AI is going to shape the radiographer role,” explains radiography researcher Avneet Gill. “But we just need to accept it first.” That sentiment has informed much of the work Avneet has been involved in in recent years, both clinically and academically.

Avneet, a PhD researcher at Ulster University, was one of the driving forces behind the recent AI in Healthcare Conference, held in Belfast, which brought together experts from across the globe to discuss the future of AI in health services.

The event was funded by the College of Radiographers Industry Partnership Scheme (CoRIPS), which provides financial support and funding for research projects undertaken by radiographers at all stages of their careers.

Held on 24-25 October in Belfast, the conference saw speakers discuss patient safety, clinical cardiology perspectives, LLM research at scale, the role of collaboration and more.

Synergy spoke to Avneet to hear more about her research, the conference and the role of AI in radiography.

Held on 24-25 October in Belfast, the conference saw speakers discuss patient safety, clinical cardiology perspectives, LLM research at scale, the role of collaboration and more.

Synergy spoke to Avneet to hear more about her research, the conference and the role of AI in radiography.

Supporting AI integration

Avneet has been stationed at Ulster University’s Coleridge campus since 2023. After completing her undergraduate degree in radiography in 2017, she worked clinically for a few years before deciding to return to academia. Rather than studying for a master’s degree, Avneet found herself motivated to immediately begin a doctorate.

She has also been part of the SoR’s AI Advisory Group for the past few years, though her tenure finished in July. While there, she had the opportunity to meet many people working in AI, including in clinical, research and academia. The group met frequently to discuss policy and research, and helped Avneet to meet the supervisors under whom she eventually chose to do her PhD.

With such broad swathes of experience, Avneet has found herself drawn more and more to trying to understand the best ways to support AI integration and ensure that the technology is used safely and effectively.

A common theme that has emerged throughout Avneet’s PhD project – and the AI in Healthcare Conference – has been the fear around using AI, particularly when it comes to the possibility of mistakes, and the question of who takes responsibility in such a situation.nAvneet believes, however, that people are simply scared because they haven’t used it. She says that in the focus groups the team conducted, people said they felt they would be more comfortable accepting AI if they could see their peers use it or see it in practice before it is formally deployed in their departments.

Left to right: conference facilitators Dr Clare Rainey, Avneet Gill, Dr Jordan Perchik (speaker), Dr Sonyia McFadden, Oliver Marshall (attendee) and Robby Emsley (speaker)

AI interface and design

In 2024, Avneet and her team applied for grant funding from Northern Ireland’s Department for the Economy for a project aiming to understand the differing opinions surrounding AI in imaging. This project focused on the design of user interfaces for AI in radiology – the look and feel of the computer systems that radiographers would be interacting with while using artificial intelligence. Across the UK and internationally, the user interfaces that AI vendors use for their products are nowhere near standardised. Some use salience maps, some use interest boxes or text, while others can use colour and highlight scans themselves.

Her focus on the subject of user interfaces has helped Avneet discover what people like and dislike as part of the project. That exploration included a ‘design-a-thon’, a timed event that involved graphic designers creating a prototype using the opinions of imaging professionals.

Using that prototype, the team has now started to collect opinions and ask people to evaluate it. While the prototype itself can’t be used, radiographers can rate it on a system usability scale to see how comfortable they and their peers would be while using it. Now, the team is waiting to write up and disseminate the results.

The AI in Healthcare Conference

After hearing from many interviewees that radiographers need more training and education on using AI, Avneet and her supervisory team came up with an idea. “Why don’t we have a conference?” she asked. “There’s a lot of fear. We know so many people in our network – radiologists, radiographers, healthcare professionals and even non-healthcare professionals who work in AI – so we need funding to support that.”

With the support of the CoRIPS Research Grant, the team were able to bring that vision to life. “People found it very valuable to hold an event with clinical professionals, academics and industry,” Avneet explains. “You don’t really get a chance to speak to the AI vendor that’s making the product you’re using in your hospital, so the feedback was really positive.

“I hope this helps get the ball rolling and get people talking about it a bit more. That’s the first step to integration and adoption, because radiographers and imaging professionals are a bit resistant to the use of AI at the moment. The deployment process has slowed down a bit. Hopefully we can start a dialogue where people aren’t scared of it.”

The team held its conference in Belfast at the Ulster Reform Club. The funding helped them to cover accommodation for attendees, fly in expert speakers and put together a post-conference digital flipbook summarising the event’s discoveries. “We wouldn’t have been able to do that without funding,” says Avneet. “We also offered it free to everyone who attended.”

Taking place over two days, the conference was open to healthcare and non-healthcare professionals alike. On the first day clinical AI usage was the main focus. While many sessions did explore AI for imaging, there was just as much attention paid to its use in cardiology, or more generally the large language models that use AI, such as ChatGPT. Other experts spoke about regulatory approvals in AI.

On the second day, attention turned to research – speakers discussed the literature that was available, industry attitudes and various research methodologies that were being used to investigate AI’s potential. Throughout this day, attendees were treated to focus group workshops where they could provide feedback, as well as panel discussions where they were able to ask questions of the experts. “It was really cool to get the audience involved,” says Avneet. “We had online attendees giving their input, and all these people who you don’t normally see together gathered in one place – an academic, a clinician and an industry professional, all speaking about the same thing.”

The conference also confirmed the team’s previous research findings about the importance of including the end user (in this case, imaging professionals) in the design of AI. This was a theme that emerged from the focus groups relating to clinical professionals wanting to be involved in guiding appropriate AI design and manufacturing.

The focus groups also raised some issues that had not been discussed as much as the more popular AI topics; for example, the infrastructural limitations that come with implementing AI. That could relate to cybersecurity, funding or regulatory differences between different parts of the country. “I hadn’t considered that geographical differences would affect the implementation of AI,” Avneet admits. “Funding disparities do exist.”

Starting small

The team is expecting to continue gathering responses to their AI user interface prototype until mid-late November. The results from the AI conference will be disseminated in October. Now, Avneet has moved on to writing her thesis paper.

During her time working clinically, she was interested in getting involved in research, but found it difficult to seek out opportunities. “You can start small,” she explains. “You can start with an audit, examine the way things are done, how many patients come in, how many listen to the changing instructions, how many exams within a week have metal on a scan – just small things that would definitely affect the patient pathway positively. You can break down what is happening that is leading to inefficiency in the process.

“You can contribute to research at any stage of your career. When conducting research formally, you don’t have to be afraid. It’s not as scary as it seems when you’re looking at it from the other side, as long as you have someone who can help guide you. Going to conferences and study days is really helpful for meeting and networking with people who can provide mentorship.”

When it comes to the future of AI in radiography, Avneet is particularly excited. While the role of the radiographer will undoubtedly be transformed, she doesn’t think it will be in a negative way. As it’s integrated further, she expects AI to leave room for radiographers to become advanced practitioners, and perform aspects of the job that require more skill and time. AI, meanwhile, will alleviate the pressure on mundane tasks that need to be done daily in the clinical environment. “It’s not something to be afraid of, and I think we need to take a mindful and meaningful approach when we move forward with AI,” she says.

“Even though it will always require clinical oversight, we need to be engaged with the process to avoid automation bias, or other kinds of cognitive biases that can result from its usage.

“We’re in charge. We’re always going to be involved with the patient. It’s exciting. I’m really looking forward to how AI is going to shape the radiographer role – but we just need to accept it first.”

Starting small

“Even though it will always require clinical oversight, we need to be engaged with the process to avoid automation bias, or other kinds of cognitive biases that can result from its usage.

More about AI in radiography

The SoR’s AI policy sets out its objectives to inform and direct the advancement of quality services when considering the development and implementation of AI, machine learning and deep learning. Find out more information here.

The SoR Artificial Intelligence Advisory Group offers members of the SoR and AI researchers the opportunity to discuss AI policy, research and clinical implementation with peers and experts in the field. Find out how you can join here.

Image credit: Getty Images

Read more