An artificial intelligence (AI) system capable of surpassing expert radiologists in the ability to detect breast cancer is reported in a paper published in Nature this week. This deep-learning model may contribute to prospective clinical trials to improve the accuracy and efficiency of breast cancer screening.
Many developed nations have implemented large-scale mammography screening programmes to detect breast cancer as early as possible. Despite the widespread adoption of mammography, the interpretation of these images remains challenging. There is high variability in accuracy achieved by experts in cancer detection and the performance of even the best clinicians leaves room for improvement. False positives can lead to patient anxiety, unnecessary follow-up appointments and invasive diagnostic procedures.
Shravya Shetty and colleagues developed a deep-learning AI model that could identify breast cancer by screening mammograms and evaluated this AI system using two large datasets from the UK (25,856 mammograms) and USA (3,097 mammograms). Using the model, the authors show an absolute reduction of 5.7% and 1.2% (USA and UK, respectively) in false positives and 9.4% and 2.7% (USA and UK, respectively) in false negatives. The AI system outperformed all six radiologists in an independently conducted study. They also found that using the AI system in the double-reading process — in the UK, the mammogram is read by two radiologists during screening — could reduce the second reader’s workload by 88%.
Planetary science: Modelling electrolyte transport in water-rich exoplanetsNature Communications
Robotics: Taking millimetre-scale origami robots for a spinNature Communications