A new Swedish study has shown that artificial intelligence could enhance the detection of breast cancer at screening by as much as 20%. Preliminary results of the large trial show that the mammography assisted by artificial intelligence detected 20% more cancers compared with double reading of mammograms by two radiologists. This study was published in the journal The Lancet Oncology with an enrollment of over 80,000 women; in fact, it is the first randomized control clinical trial involving the use of AI in breast cancer screening.
Mammograms are a critical tool that can be applied in the early detection and treatment of breast cancers since they rely on the use of X-ray images to detect early signs of disease. For the trial, the AI system analyzed mammogram images and predicted the level of risk for cancer between 1-the lowest risk and 10-the highest risk. If the AI determined that there was significant risk, two radiologists evaluated the images separately. In contrast, for lower-risk cases, only one radiologist was involved. The AI-assisted group detected 244 cancers, while the traditional double-reading method found 203. Additionally, the AI approach led to a 44.3% reduction in the radiologist’s workload without increasing the incidence of false positives.
Lead author Dr. Kristina Lång from Lund University was taken aback by the promising results, saying that AI-assisted screening enhances the detection of cancer and reduces the workload of radiologists. She added that AI is meant to supplement rather than supplant radiologists, as that will prove to be more precise and economical. Radiologists are invariably necessary for explaining AI results to minimize false positives and ensure accuracy in diagnoses.
The research is still under way but will go up to 100,000 participants. Future goals include determining whether AI can help reduce missed cancers in annual screenings. The European Commission currently recommends double reading mammograms, but the shortage of breast radiologists in many regions makes this approach increasingly difficult to maintain.
Despite the promising results, some experts have raised concerns about overdiagnosis and the potential limitations of AI in detecting specific biological features of tumors. Further analysis will examine the types and stages of cancers detected through AI and compare them with standard screening methods.
The study’s authors are hopeful that AI can significantly improve breast cancer detection, while future research will determine the long-term benefits and potential drawbacks of integrating AI into routine screenings.
