A groundbreaking trial investigating the use of artificial intelligence (AI) in breast cancer screening has yielded promising results, significantly reducing the workload of radiologists by almost 50 percent. Conducted with over 80,000 women in Sweden, the study compared AI-supported screening with standard care directly.
Preliminary findings published in the Lancet Oncology journal revealed that AI screening proved as effective as the traditional double-reading approach by radiologists, without increasing false positives. Among women recalled from AI-supported screening, 28 percent were diagnosed with cancer, slightly higher than the 25 percent in the standard screening group, leading to the detection of 41 additional cancers. Importantly, the use of AI did not generate more false positives, with both groups showing a false-positive rate of 1.5 percent.
The trial resulted in an astounding 44 percent reduction in the workload of radiologists in the AI-supported group, amounting to 36,886 fewer readings. This reduction in workload could address the radiologist shortage faced by many countries. Although the results are promising, the lead author of the study, Dr. Kristina Lång, emphasized the need for further research to understand the long-term impact on patient outcomes, cost-effectiveness, and detection of interval cancers missed by traditional screening.
While experts praised the study’s potential benefits in terms of faster diagnosis, earlier cancer detection, and improved patient care, they also cautioned against potential AI-driven increases in breast cancer detection leading to the overdiagnosis of less harmful lesions. While awaiting final results in the coming years, stakeholders in breast screening programs are optimistic about AI’s potential but acknowledge the need to address current issues such as outdated IT systems to fully unlock AI’s benefits in healthcare.