Results from a groundbreaking study in Sweden have revealed the potential of artificial intelligence (AI) in mammography, showing it to be an effective tool in detecting breast cancer. Early findings from the first randomized controlled trial exploring the use of AI-supported breast cancer detection demonstrate that AI-supported analysis of mammograms, in combination with radiologists, accurately detects 20% more breast cancers compared to traditional screening methods. Published in the Lancet Oncology, the study also reports a significant reduction in radiologists’ workload, highlighting the benefits of incorporating AI into breast cancer screening programs.
The study and its impact
The Swedish trial, which is still ongoing, involved more than 80,000 women. Half of the participants had their mammograms examined by two radiologists without AI, while the other half had their mammograms analyzed by AI and radiologists. Two radiologists assessed the screening in cases where the AI generated the highest risk score. The results demonstrated that using AI alongside one or two radiologists was as effective as using two radiologists without AI, leading to the detection of 20% more breast cancers.
Reduced workload for radiologists
Notably, the study also revealed a significant reduction in the workload for radiologists. By incorporating AI into the screening process, radiologists spent 44% less time reading mammograms. Reducing the burden on radiologists’ time is important in many breast screening programs, as it allows them to allocate more time to attending to patients and providing improved care.
AI’s potential in medicine
The study’s findings come at a time of growing interest in the potential opportunities and risks of AI in the medical field. While AI is increasingly deployed in medical settings, concerns persist regarding algorithm training, validation, potential biases, and over-diagnosis. The European Union is planning stringent regulations around the use of AI, and the European Medicines Agency is actively assessing the risks and benefits of AI in drug development.
Not ready for full implementation
Lead author Kristina Lång from Lund University, Sweden, urges caution despite the positive results. She emphasizes that the interim safety results alone are insufficient to confirm that AI can be fully implemented in mammography screening. The research team is waiting for further results from the ongoing trial to assess whether AI reduces the number of cancers detected between screenings, which will determine its potential for broader implementation.
Reducing the Burden on Radiologists
Kristina Lång highlights the primary potential of AI in reducing the burden on radiologists. AI can free up radiologists’ time by eliminating the need for a second radiologist to review the mammogram, enabling them to serve more patients efficiently. This advancement promises to improve breast screening programs and enhance patient care.
While reducing radiologists’ workload is an encouraging development, experts like Professor Stephen Duffy from Queen Mary University of London express concerns that AI might over-detect harmless lesions. As AI systems can produce unpredictable and not explainable errors by human logic, it is crucial to monitor their safety through randomized controlled trials, as highlighted in a January editorial in European Radiology.
The Sweden study demonstrates AI’s remarkable potential in breast cancer detection, showing a 20% improvement in cancer detection rates compared to traditional methods. Reducing radiologists’ workload is an essential benefit, as it allows healthcare professionals to focus more on patient care. Nevertheless, further research is needed to assess AI’s long-term impact and safety in mammography screening. As AI continues to reshape medicine, stringent regulation and comprehensive studies are necessary to ensure its safe and effective integration into healthcare practices.