Artificial intelligence (AI) tools have shown promise in reading breast cancer screening images, according to a preliminary study. The research suggests that computer-aided detection can identify cancer in mammograms at a similar rate to radiologists. The National Health Service (NHS) is already exploring the implementation of this technology in its breast screening program. However, the study’s authors emphasized that these results are not enough on their own to confirm that AI is ready for widespread use in mammography screening.
Previous studies on the accuracy of AI in diagnosing breast cancer in mammograms have been conducted retrospectively, with the technology analyzing scans that have already been reviewed by doctors. In contrast, this interim study compared AI-supported screening directly with standard care. The randomized controlled trial involved over 80,000 women from Sweden, with an average age of 54. Half of the scans were assessed by two radiologists, representing standard care, while the other half were assessed by the AI-supported screening tool followed by interpretation by one or two radiologists.
The study found that the AI-supported screening identified 244 cases of cancer, compared to 203 cases identified through standard screening. Importantly, the use of AI did not lead to more false positives, with both groups having a false-positive rate of 1.5%. The researchers also observed that AI could potentially reduce the screening workload by almost 50%. The AI-supported group required 36,886 fewer screen readings by radiologists, resulting in a 44% reduction in their workload.
The study is ongoing, with the goal of assessing whether AI tools can detect cancers that occur between screenings. The authors’ interim analysis concluded that AI-supported mammography screening has a similar cancer detection rate compared to standard double reading and significantly reduces the screen-reading workload. However, the implications on patient outcomes, such as detecting interval cancers missed by traditional screening, and the cost-effectiveness of the technology still need to be understood.
Lead author Dr. Kristina Lang from Lund University in Sweden highlighted that one of the greatest potentials of AI is to alleviate radiologists’ excessive reading workload. While AI-supported screening still requires at least one radiologist, it could eliminate the need for double reading in the majority of mammograms, thus reducing waiting times for patients and allowing radiologists to focus on more advanced diagnostics.
The NHS expressed its interest in how AI could aid breast screening by enabling rapid and scalable image analysis, potentially accelerating diagnosis and saving lives. Dr. Katharine Halliday, president of the Royal College of Radiologists, commented on the potential of AI to support clinical decision-making and improve efficiency. However, she emphasized that while AI can enhance radiologists’ work, it cannot replace them entirely. The final results of the trial are eagerly awaited to determine if AI can indeed improve breast cancer screening and increase capacity.
In conclusion, the preliminary study suggests that AI tools can safely read breast cancer screening images and identify cancer at a similar rate to radiologists. The findings indicate that AI-supported screening could significantly reduce radiologists’ workload without increasing false positives. However, further research is needed to assess patient outcomes, cost-effectiveness, and the technology’s ability to identify interval cancers. The potential of AI in mammography screening is promising, and it could lead to improved efficiency, faster diagnosis, and ultimately, better patient care.
