Is AI just what the doctor ordered for the radiologist shortage? Yes and no
AI shines in radiology when tasked with interpreting images of patients whose odds of illness are either very high or very low. Used for such cases, the technology could help mitigate the worldwide shortage of radiologists, according to a new review of the literature.
However, ethical, regulatory and legal issues will probably continue to challenge radiology operations seeking to deploy AI for workload relief. Particular tripping points are likely to include nagging doubts over the trustworthiness of the systems and, by extension, radiologists’ concerns about liability when mistakes are made.
Researchers in the U.K. make these predictions after analyzing 22 studies from various countries, including the U.S. Their report posted April 25 in Health and Technology, a Springer WHO-affiliated journal of the International Union for Physical and Engineering Sciences in Medicine (IUPESM).
Nebil Achour, PhD, of Anglia Ruskin University and colleagues note AI’s now-established prowess for reading scans with high sensitivity. Given this capability, setting AI to “recognize and eliminate the high and low likelihood threshold will reduce the number of exams to be conducted by the radiologists which can reduce workload to approximately 53%,” they write.
On the other hand, they add, leaning on the technology “can increase fragmentation in the workflow, which might increase workload.”
Achour and co-authors comment in some detail on three key things to watch for when tapping AI to ease the impact of radiologist shortages.
1. The literature suggests researcher consensus on AI’s potential to reduce radiologists’ workloads.
However, individual researchers disagree on how to view AI assistance.
“AI systems have developed very high sensitivity, to the extent that they can replace radiologists in specific decisions where the probability of being ill is very little or when it is almost certain that the patient is diagnosed with a particular illness, such as cancer,” the authors write. “Despite the high sensitivity, there is a risk that [any given] diagnosis is not accurate and thus someone needs to take the responsibility when there is an error. This is one of the debates of this opinion.”
2. Many hospital practices, management and staff are hesitant to accept and adopt the new technology.
Common concerns include ethical considerations such as respect for autonomy and confidentiality, Achour and colleagues report. Also in the mix are worries over legal exposure should harm come to patients with AI use.
Further, some fear for the future of the radiology profession while others, including healthcare consumers, distrust the technology across the board.
Meanwhile:
“It is crucial that trainees are not taught to rely on AI but to interpret the images they see with any access to AI as they need to learn to develop their interpretive skills,” the authors maintain. “Organizations need to ensure that these biases are minimized through ensuring that radiologists have the final decision at least in a transition period, by the end of which more evidence will be developed to support the way forward.”
3. Studies of AI in radiology skew toward reviews.
Most of the articles Achour and colleagues deemed suitable for the present literature review, 55%, were themselves reviews. This suggests a need for more primary work to “help generate new knowledge and identify aspects related to the impact of AI on radiologists and radiographers,” the authors comment.
The findings show little diversity and “too much” consistency, they add, suggesting the extent of repetition they noted “gives the impression of data saturation, but could be due to the influence of researchers on each other.”
Younger physicians are reluctant to choose careers in radiology, Achour and co-authors remark, suggesting the nature of the reticence deserves deeper investigation than it has so far received. One specific inquiry they call for is a look at how radiological AI can influence staff recruitment and retention.
“The shortage of radiologists is a global concern that can have significant impact on the provision of health service” and could “threaten the lives of many people specifically with increasingly complex and serious illnesses such as cancer,” Achour et al. write. “Problems such as [heavy] workloads and burnout are associated with errors and misdiagnoses, which reduce the quality of care, increase unnecessary work or overlook the severity of serious health conditions for some patients.”
Against this backdrop, they add, AI can improve care quality and resilience due to its ability to read and interpret images, videos and audios—and to “generate evidence that can enable quicker and potentially more accurate decisions.”
There is more, and the study is posted in full for free.
Dave P. has worked in journalism, marketing and public relations for more than 30 years, frequently concentrating on hospitals, healthcare technology and Catholic communications. He has also specialized in fundraising communications, ghostwriting for CEOs of local, national and global charities, nonprofits and foundations.