AI-supported mammography analysis is as effective as having two breast radiologists work together to detect breast cancer, according to a groundbreaking trial published in The Lancet Oncology.
The trial involved over 80,000 women in Sweden, with the AI technology almost halving the screen-reading workload and not increasing false positives.
HOW THE TRIAL WAS CONDUCTED
The randomised control trial compared the performance of AI-supported screening with standard care. Half of the mammograms were assessed by two radiologists (standard care), while the other half were evaluated using the AI-supported screening tool, followed by at least one radiologist’s interpretation.
RESULTS OF AI-SUPPORTED MAMMOGRAPHY
The results showed that AI-supported detection could identify breast cancer in mammograms at a similar rate to two radiologists. The AI technology did not lead to an increase in false positives, maintaining a low false positive rate of 1.5% in both groups.
The use of AI technology resulted in a 44% reduction in the screen-reading workload for radiologists. With 36,886 fewer screen readings, AI has the potential to alleviate the strain on radiologists and expedite the screening process.
FUTURE IMPLEMENTATION AND IMPLICATIONS
Although the trial shows promising results, researchers believe it will still take several years before the ultimate use of AI in mammography becomes widespread. Additional trials and program evaluations are needed to address the radiologist shortage in many countries and assess AI’s impact on patient outcomes.
ADDRESSING MISSED INTERVAL CANCERS
Dr. Kristina Lang, lead author of the study from Lund University in Sweden, emphasizes that further research is required to understand the implications on patients’ outcomes. Specifically, they aim to investigate whether combining radiologists’ expertise with AI can help detect interval cancers, which are often missed by traditional screening methods.
