Google Health today shared an AI model trained with more than 90,000 mammogram x-rays that achieve better results than human radiology experts while recognizing more of false negatives, the kind of images that look normal but contain breast cancer. Initial findings were published in Nature.
The model achieves lower rates of false-positives (5.7% in U.S. and 1.2% in the U.K.). It also achieves lower rates of false-negatives (9.4% in U.S. and 2.7% in U.K.). The result is a collaboration between Google’s DeepMind, Cancer Research UK, Northwestern University, and Royal Surrey County Hospital.
According to the National Health Service in the U.K. and American Cancer Society roughly 1 in 8 are diagnosed with breast cancer in their lifetime, making it the most common cancer diagnosis among women.
The model is trained on mammogram imagery from 76,000 women in the U.K. and more than 15,000 women in the United States, then evaluated using a data set of imagery from more than 25,000 U.K. women and 3,000 women in the U.S.
“Looking forward to future applications, there are some promising signs that the model could potentially increase the accuracy and efficiency of screening programs, as well as reduce wait times and stress for patients,” Google said in a blog post today. “But getting there will require continued research, prospective clinical studies and regulatory approval to understand and prove how software systems inspired by this research could improve patient care.”
The news follows news a year ago that Google developed AI for metastic breast cancer detection with 99% accuracy. IBM Research is also focusing on breast cancer. In 2019, IBM created a model that predicts when women will develop malignant breast cancer within a year, and AI to analyze breast cancer cells.