Most Women Reject AI As the Only Interpreter of Their Mammograms

October 14, 2020

More than three-quarters of women prefer a radiologist to be involved with reading their screening studies.

Artificial intelligence (AI) algorithms designed to independently interpret mammograms might be gaining ground when it comes to radiologist confidence, but the public-at-large does not share this sentiment. In fact, nearly 80 percent of women say they oppose letting AI interpret their studies without a radiologist being directly involved somehow.

In a study published Oct. 12 in the Journal of the American College of Radiology, a team of investigators from the University of Groningen, in The Netherlands, revealed that, when asked whether they were comfortable with having AI alone analyze their mammograms for signs of breast cancer, 78 percent of women surveyed said no.

“Despite recent breakthroughs in the diagnostic performance of AI algorithms for the interpretation of screening mammograms, the general population currently does not support a fully independent use of such systems without involving a radiologist,” said the team led by Yfke Ongena, Ph.D. assistant professor with the university’s Center of Language and Cognition.

Related Content: Artificial Intelligence & Mammography: Where It’s Been & Where It’s Going

Given those advancements and study results that show AI can outperform radiologists, this response is surprising, they said. But, it shows that radiology as an industry should pump the brakes on implementing any AI algorithms as the sole interpreter of mammograms, as well as actively work to educate women about the capabilities and efficacy of these algorithms. Ultimately, improving the public’s perceptions about AI is critical because the industry currently has a shortage of radiologists in many regions who are qualified to interpret mammograms, the team pointed out.

The team’s survey results also showed when it comes to using AI as a second read on select studies, 42 percent disagreed, 31 percent agreed, and 27 percent were undecided.

Ongena’s team reached their conclusion after conducting a survey of 922 women, ages 16 to 75, in two waves. They included women ages 16 to 40 because they represent the next generation of women who will undergo routine screening mammography. First, they collected responses from December 2018 and, then, from April 2020. The team asked women for their feeling about the necessity of a human check, AI as a second reader, AI as a selector for second reading, and whether developers or radiologists were responsible for errors.

Their goal was to determine how women actually feel about having AI interpret their mammograms in light of how the mainstream press has portrayed these tools in a largely positive manner. They initially postulated that women’s responses would be in alignment.

Based on the survey responses, they were wrong. Still, the team concluded, there is room for AI to gain acceptance as a second reader that supports the radiologist. Only 17 percent of women categorically opposed using the tool in collaboration with her radiologist.

“The combination of a radiologist as a first read and an AI system as a second reader seems to be the most acceptable approach to the population at present, although still not fully embraced by the entire population,” they said. “Improved information supply and education about the development, possibilities, and limitations of AI algorithms in screening mammography may potentially overcome some of the perceived obstacles and increase acceptance of this new technique in clinical practice.”

The most important step to improving the acceptance of AI in the general population, they said, is bolstering patient education about the possibilities of the technology. On the whole, women who opposed the use of AI also reported having lower level of education.

The study did have a significant limitation, however, the team reported. Survey responses were collected in a country where mammography is routinely offered to women between ages 50 and 75, and those scans are interpreted by at least two radiologists. Other countries might not have the resources available to process mammograms at this level, and in the case of the United States, breast cancer screening is only routinely covered by Medicare.

Overall, the team said, it is critically important for women’s opinions of these tools to be considered.

“The voice of the population who will undergo AI-based diagnostic tests is crucial in this context, because it is a determining factor for the boundaries within which an AI system is allowed to operate,” they said. “The success of a breast cancer screening program depends on the willingness of subjects to participate, and the willingness may be affected if AI systems are used without taking into account the population’s wishes, concerns, and objections.”

For more coverage based on industry expert insights and research, subscribe to the Diagnostic Imaging e-Newsletter here.