Patients want a say in AI: 5 ways to help radiology get there

Radiology has not shied away from offering its viewpoint on artificial intelligence, but there’s been less attention paid to how patients see the technology altering their own imaging care.

That’s according to a trio of researchers from the University of Saskatchewan in Canada, who helped conduct roundtable discussions asking what patients thought about AI in radiology, their concerns related to the tech and how they believe it should evolve.

In total, 17 people participated in the discussions; and many expressed anxieties related to a fear of the unknown, lack of trust, human connection issues and cultural acceptability.

On top of their fears, patients also had a number of priorities they believe AI should address. Those have been summarized and are listed below:

1. Improve access and decrease wait times: Like many imaging experts, patients see AI as a tool to reduce image acquisition and interpretation times. This, in turn, can boost the number of scans an institution would perform in a given time period, while also reducing wait times.

Respondents also pointed to AI's potential to bring imaging to more rural areas and make screening widely available.

2. Shorten time to diagnosis: Participants said when an incidental or indeterminate finding requires follow-up imaging, it often leaves them frustrated and concerned.

“I don’t want to know and have to worry for six months before they say, ‘Oh it’s nothing,’” one patient responded.

AI, they noted, could help radiologists stratify patients according to risk and improve how they communicate the significance of a finding upon initial examination.

3. Improve accuracy: Patients want their imaging results as accurate as possible. And although they don’t completely trust AI, they want their provider to have the “best tools available” and make sure such resources are validated. Participants believe a human-AI hybrid is the best option for this.

4. Better communication: Respondents don’t want to be left in the dark during their care journey. And a large part of that includes understanding their exam results.

This has been a long-standing problem for patients. There have been promising studies on this topic, however. One included computer-human based systems that improved communication and another made radiology reports easier to digest.

5. Empowerment: When an artificial intelligence tool is created and evaluated, it should be done with input from patients, respondents said. And at the very least, developing such technologies must be completed with them in mind.

All told, radiology may want to consider educational outreach campaigns to better inform patients about AI, Scott J. Adams, MD, with the University of Saskatchewan, Saskatoon, Canada, and colleagues wrote. Taken together with the results of their discussions, a patient-centered AI future should be top of mind for the specialty.

“Findings from this workshop—and from future patient engagement efforts—may guide radiologists, healthcare leaders, and AI developers in developing and implementing AI tools that meet patients’ needs and priorities,” they wrote, Feb. 14 in the Journal of the American College of Radiology.