When screening for breast cancer, the most effective strategy to utilize artificial intelligence (AI) may involve collaboration with human radiologists, according to a recent study published by Ahsen et al in Nature Communications. The findings could help shape how hospitals and clinics integrate AI into their diagnostic workflows amid a growing demand for early breast cancer detection and a shortage of radiologists.
Background
Breast cancer screening is a critical public health tool, with nearly 40 million mammograms performed annually in the United States. Nonetheless, the process is time-intensive and costly in both labor and follow-up procedures necessitated by false-positive results. When cancers are not detected, the resulting false negative findings can lead to significant risks among patients and health-care providers.
“One of the issues in mammography is, because of the sheer number of screenings performed, that it generates so many false positives and false negatives,” explained lead study author Mehmet Eren Ahsen, PhD, Professor of Business Administration and the Deloitte Scholar at the University of Illinois Urbana-Champaign as well as the Health Innovation Professor at the Carle Illinois College of Medicine. “If you have a 10% false-positive rate out of 40 million mammograms per year, that’s 4 million women who are being recalled to the hospital for more appointments, screenings and tests, and potentially biopsies,” he added.
The resulting follow-up appointments can create challenges for patients, including increasing their anxiety.
“Follow-up appointments often take weeks, leaving patients with [uncertainty and distress]. It’s a very stressful time for them,” Dr. Ahsen stressed. “We often hear the question: can AI replace this or that profession? In this case, our research shows that … the real value of AI comes not from replacing humans, but from helping them via strategic task-sharing,” he continued.
Study Methods and Results
In the study, researchers developed a decision model to compare three decision-making strategies in breast cancer screening: an expert-alone strategy—the current clinical standard—in which radiologists read every mammogram; an automation strategy in which AI assessed all mammograms without human oversight; and a delegation strategy in which AI performed an initial screening, triaged low-risk mammograms, and flagged higher-risk cases for closer inspection by human radiologists.
The model accounted for a wide range of costs, including implementation, radiologist time, follow-up procedures and potential litigation. It evaluated outcomes using real-world data from a global AI crowdsourcing challenge for mammograms—which was sponsored as part of the White House Office of Science and Technology Policy’s Cancer Moonshot initiative from 2016 to 2017.
The researchers found that the delegation strategy outperformed both the full automation and the expert-alone approaches, yielding up to 30.1% in cost savings without compromising patient safety.
Although fully automating radiologic tasks was appealing from an efficiency standpoint, the researchers cautioned that current AI systems still fall short of replacing human judgment in complex or borderline cases.
“AI is excellent at identifying low-risk mammograms that are relatively straightforward and easy to interpret,” noted Dr. Ahsen, “[b]ut for high-risk or ambiguous cases, radiologists still outperform AI. The delegation strategy leverages this strength: AI streamlines the workload, and humans focus on the toughest cases,” he indicated.
Discussion
With AI and the delegation model, the researchers suggested that health-care providers could streamline the breast cancer screening process. The findings also raised broader questions about how AI should be implemented and regulated in medicine.
“It has the potential to be that much more efficient of a workflow,” Dr. Ahsen argued. “The delegation strategy works best when breast cancer prevalence is either low or moderate. In high-prevalence populations, a greater reliance on human experts may still be warranted, [b]ut an AI-heavy strategy also might work well in situations where there aren’t a lot of radiologists—in developing countries, for example,” he highlighted.
However, the researchers pointed to potential issues regarding legal liability. For instance, if AI systems are held to stricter liability standards compared with human physicians, then health-care organizations may be less compelled to automate strategies involving AI, regardless of their cost effectiveness.
In addition, the results of the study are potentially applicable to other fields of medicine such as pathology and dermatology, where diagnostic accuracy is critical, but AI is potentially able to improve workflow efficiency.
“AI is only going to continue to make inroads into health care, and our framework can guide hospitals, insurers, policymakers and health-care practitioners in making evidence-based decisions about AI integration,” Dr. Ahsen underscored. “We’re not just interrogating what AI can do—we’re asking if it should do it [as well as] when, how, and under what conditions it should be deployed as a tool to help humans,” he concluded.
Mehmet Eren Ahsen, PhD, is the corresponding author. Dr. Ahsen and Mehmet US Ayvaci, PhD, contributed equally to this study.
Disclosure: For full disclosures of the study authors, visit nature.com.
