Google's mammography AI system slashes interpretation times by one-third
Google’s mammography artificial intelligence software can perform as well or better than human radiologists at detecting breast cancer, according to findings from the largest study on AI in breast cancer settings to date.
Published in Nature Cancer, results suggest using Google’s mammography AI system (version 1.2) may be a more efficient second reader than radiologists. The research compared the performance of two real-life readers versus an AI-human combo, revealing the latter has potential to increase detection rates and reduce interpretation times.
“Breast cancer screening offers a compelling clinical use case for artificial intelligence in healthcare,” Deborah Cunningham, MD, a consultant radiologist at London's Imperial College Healthcare NHS (National Health Services) Trust, and colleagues noted. “It is hoped that AI can improve the quality and consistency of screening while improving cost effectiveness and addressing global radiologist workforce shortages.”
The analysis incorporated more than 115,000 women who underwent breast cancer screening mammography through the NHS.
AI as a second reader outperformed the first human reader at detection, identifying 9.33 cancers per 1,000 women compared to 7.54 per 1,000 for the human reader. It also achieved superior sensitivity and noninferior specificity compared to radiologists, while simultaneously detecting more invasive and interval cancers; the tool identified 25% more interval cancers than the human readers. When the system was tasked with assessing first-time screening mammograms and did not have any prior imaging to analyze for comparison, it performed particularly well, producing nearly 40% fewer recalls while also spotting nearly 9% more malignancies for that subset of patients.
Outside of detection rates, use of the AI system as a second reader resulted in significant time savings; reading times dropped by almost 33%—something experts highlighted as big benefit for readers.
“This study demonstrates AI’s potential to substantially improve breast cancer screening efficiency and accuracy, particularly for first-time participants,” the authors noted. “However, successful implementation will require adaptive threshold management, continuous performance monitoring and careful workflow integration to ensure equitable benefit across all populations.”
Read more here.
In addition to her background in journalism, Hannah also has patient-facing experience in clinical settings, having spent more than 12 years working as a registered rad tech. She began covering the medical imaging industry for Innovate Healthcare in 2021.