Experts believe in the potential of breast AI but say more real-world data is needed
Despite the known benefits that artificial intelligence-enabled tools have for breast cancer detection, significant questions remain about how these capabilities should be incorporated into clinical practice, as prospective, real-world data on AI remains relatively limited.
A new review led by researchers at UCLA Health examines the utility of commercially available AI tools for screening mammography and the emerging evidence surrounding their ability to predict interval cancers. Such cases can arise because a cancer develops rapidly after a negative examination, but some leave subtle abnormalities on the prior mammogram that were not recognized at the time.
The latter group may represent an opportunity for AI.
“The ultimate goal of screening mammography is to eliminate these interval cancers and catch as many of them as we can earlier, at the point of screening,” Tiffany Yu, MD, assistant professor of radiology at the David Geffen School of Medicine at UCLA and senior author of the paper, wrote in the review. “As AI tools become increasingly commercially available, we wanted to provide readers with a timely overview and foundational understanding of interval cancers and potential ways AI can help detect them, because ensuring these tools are safe and clinically effective is paramount.”
UCLA researchers examined two potential applications of AI in interval cancer detection. The first is retrospective identification of cancers that were technically visible on an earlier mammogram but were missed during interpretation. Across the studies reviewed, AI identified varying proportions of these cancers, with estimates ranging from 5% to 78%, highlighting problematic variations in sensitivity.
The second evaluation focused on risk prediction. Rather than identifying an obvious lesion, some AI systems analyze mammograms that appear normal and assign a future cancer risk score based on patterns that may not be apparent to the radiologist. In one study included in the review, AI assigned its highest risk scores to 23% of women who subsequently developed interval cancer three screening rounds before diagnosis. On the mammogram immediately preceding diagnosis, the proportion increased to 39%.
These findings suggest that mammography could provide information about breast cancer risk beyond the visible abnormalities traditionally evaluated by radiologists. But they also raise a questions pertaining to how clinicians should respond when AI identifies a woman as high-risk even though their mammogram is technically negative.
In these cases, additional imaging could potentially identify cancer earlier in some women, but it could also increase recalls, false positives, unnecessary biopsies and overdiagnoses.
“Finding a cancer retrospectively is very different from demonstrating that using AI during routine screening would have led to an earlier diagnosis,” said Hannah Milch, MD, associate professor of radiology at the David Geffen School of Medicine at UCLA and co-author of the study. “We need prospective evidence showing that AI actually changes patient outcomes.”
One of the biggest hurdles preventing AI from being widely integrated is that research on these tools is hard to compare, as the methods and testing cohorts are variable. Research has used different definitions of interval cancer, screening intervals, mammography technologies, radiologist workflows and AI algorithms. Many of the studies evaluating whether AI can identify interval cancers are also retrospective.
The authors suggest that future research should focus on the prospective evaluation of whether AI actually reduces interval cancer rates while also monitoring its impact on recalls, false-positive findings, downstream testing and radiologist workload; they added that long-term patient outcomes and post-market surveillance will be important as well. Though the team believes that AI’s role for identifying interval cancers earlier is promising, they do not believe the technology is ready to be viewed as a proven solution.
Read more here.
In addition to her background in journalism, Hannah also has patient-facing experience in clinical settings, having spent more than 12 years working as a registered rad tech. She began covering the medical imaging industry for Innovate Healthcare in 2021.