Artificial intelligence can detect breast cancer as well as radiologists and has the potential to improve the accuracy of mammogram screenings, according to new research.
Researchers at Google's DeepMind AI unit, which merged with Google Health in September, worked with clinical researchers at Cancer Research UK Imperial Centre, Northwestern University and Royal Surrey County Hospital to study whether AI could support radiologists to spot the signs of breast cancer more accurately.
Google's algorithm spotted breast cancer in de-identified screening mammograms with greater accuracy, fewer false positives, and fewer false negatives than experts, according to the company.
This sets the stage for future applications where the model could potentially support radiologists performing breast cancer screenings, Shravya Shetty, technical lead at Google Health and Daniel Tse, M.D., product manager at Google Health wrote in a blog post published Wednesday.
The health data used in the breast-cancer project doesn’t include identifiable information, according to Google Health officials said, and the data was stripped of personal indicators before being given to Google.
Screening mammograms miss about 1 in 5 breast cancers, according to the American Cancer Society, and half of all women who get the screenings over a 10-year period have a false-positive result.
Researchers trained the AI system to identify breast cancer using de-identified mammograms from more than 76,000 women in the U.K. and more than 15,000 women in the U.S. The model's performance was then tested on different mammograms based on a de-identified data set of more than 25,000 women in the U.K. and over 3,000 women in the U.S.
Thes study results indicated Google's AI can identify breast cancers with a similar degree of accuracy to radiologists while reducing false-positive results, or incorrect positive readings, by 5.7% in the U.S.-based group and 1.2% in the U.K.-based group.
The AI system also reduced the number of missed cases by 9.4% in the U.S. and 2.7% in the U.K. compared with the original radiologist diagnoses, according to the study.
The study authors report that the AI system outperformed both the historical decisions made by the radiologists who initially assessed the mammograms and the decisions of six expert radiologists who interpreted 500 randomly selected cases in a controlled study.
While radiologists have access to patient histories and prior mammograms, the AI model only processed the most recent anonymized mammogram with no extra information. "Despite working from these X-ray images alone, the model surpassed individual experts in accurately identifying breast cancer," Shetty and Tse wrote.
"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," they wrote.
Continued research, prospective clinical studies and regulatory approval is needed to understand and prove how software systems could improve patient care, Shetty and Tse said.
While the study findings are promising for AI's potential to improve breast cancer screening, the clinical implications are uncertain, Etta Pisano, chief research officer at the American College of Radiology wrote about the study in a Nature editorial.
"The real world is more complicated and potentially more diverse than the type of controlled research environment reported in this study," Pisano wrote, noting that the study did not include all the different mammography technologies currently in use.
And Pisano also highlighted the need to protect patient privacy when using vast volumes of patient data to develop AI systems.
"If such AI systems are to be developed and used widely, attention must be paid to patient privacy, and to how data are stored and used, by whom, and with what type of oversight," she wrote.
Google has come under scrutiny for privacy concerns related to the use of patient data, including its recently announced deal with Ascension to use AI to mine personal, identifiable health information from millions of patients to improve processes and care.