Researchers at the University of Copenhagen have demonstrated that using a commercially available AI system can reduce the workload of radiologists by over 33% while enhancing the overall performance of breast cancer screenings.

Mammography exams are routinely used for early detection of breast cancer, significantly reducing mortality rates. However, these exams create a substantial workload for radiologists, who must read numerous mammograms, most of which do not warrant patient recall. “The reading workload is further compounded when screening programs employ double reading to improve cancer detection and decrease false-positive recalls,” said Andreas D. Lauritzen, PhD, a researcher at the University of Copenhagen and Gentofte Hospital in Denmark. Recalling women for further tests also increases the workload, a challenge exacerbated by the shortage of specialized breast radiologists.

In a study published in Radiology, Lauritzen and colleagues investigated the effects of using an AI system to assist with breast cancer screening, focusing on the workload and performance of radiologists. The AI system, called Transpara, was trained using deep learning models to identify suspicious lesions within mammograms and assign them a score out of 100 indicating the likelihood of the lesion being cancerous.

The study analyzed mammograms from women aged 50 to 69 who underwent routine screenings between October 2020 and October 2022 in the Capital Region of Denmark. A total of 60,751 women were screened without using AI, while 58,246 women were screened with AI assistance. In the AI-assisted group, mammograms were first analyzed by the AI. The exams deemed to have a low likelihood of cancer (67% of screenings) were then read by a specialized breast radiologist to confirm the results. The remaining exams (33% of screenings) were read by two radiologists with AI-assisted decision support.

In cases where radiologists identified a cancerous lesion, patients were recalled and the cancers were confirmed with a biopsy. All women in the study were followed for at least 180 days.

The results showed that AI-assisted screening detected breast cancer in 0.82% of screenings, compared to 0.70% without AI. Importantly, the use of AI reduced the rate of false positives from 2.39% to 1.63%. Additionally, a higher proportion of small cancers (1 cm or less) were detected in the AI-assisted group (44.93%) compared to the non-AI group (36.60%).

The benefits of using AI extended beyond improved screening performance. Lauritzen and colleagues found that AI-assisted screenings decreased the recall rate by 20.5% and reduced radiologists’ workload by 33.4%, potentially saving cancer specialists significant time in their practice.

Lauritzen noted that more research is needed to evaluate the long-term outcomes of AI-assisted cancer screenings. “Around November 2024, we will have full two-year follow-up data for the cohort of women screened with AI. In future work, we aim to quantify the effects of AI stratification, AI decision support, and radiologist access to prior screenings separately,” he said.

By Impact Lab