Google's AI is better at breast cancer screening than human experts, study claims

You may remember Google's DeepMind AI being really good at Go, but the company's also been putting it to work on a potentially life-saving mission — breast cancer screening.

According to Google's new study, published in Nature, the company's AI model was able to spot breast cancer in mammograms with "greater accuracy, fewer false positives, and fewer false negatives than experts."

This doesn't mean the AI will completely replace experts, though; instead it will support the work of radiologists who perform breast cancer screenings.

According to Google, digital mammography — x-ray imaging of the breast — is the most common method of breast cancer screening today, but actually finding and diagnosing breast cancer in the imagery isn't easy. It often results in both false positives and false negatives, which can delay treatment and increase workload for radiologists.

Google's study is based on a data set of de-identified mammograms from more than 76,000 women in the UK and more than 15,000 women in the U.S. Compared to the work of experts, Google's AI managed to significantly reduce the number of false positives (by 5.7 percent in the U.S. and 1.2 percent in the UK) as well as false negatives (by 9.4 percent in the U.S. and 2.7 percent in the UK).

The AI can also generalize to different healthcare systems, meaning the model can be trained on the data from the UK and still be effective on the data from the U.S.

Google notes that its AI had access to less information than human experts, which were able to access patient histories and prior mammograms, making this achievement even more impressive.

"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," Shravya Shetty, technical lead at Google Health, wrote in a blog post. "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."

Google's work in this area was subject of some controversy, after it was revealed that DeepMind has been granted access to the healthcare data of 1.6 million patients in UK's National Health System (NHS) without explicit permission. The NHS later decided to protect user privacy by de-identifying the data, which is reflected in this new study.