Google claims its pattern-matching systems beats humans in detecting breast cancer

Early detection of breast cancer currently remains a big challenge for specialists

Google Health claims to have created a machine learning system that can detect breast cancer with greater accuracy - and more quickly - than specialists.

According to Google, the system outperformed radiologists in a study by correctly spotting cancerous nodes, which had been missed by specialists in mammograms. The system also reduced false positives, in which specialists wrongly believe patients have tumours.

Breast cancer is the most common cancer in the UK, according to the NHS. But early detection and diagnosis of breast cancer remains a major challenge. Detection involves reading digital mammography images, which often result in both false negatives and false positives, causing unnecessary stress for patients, delays in proper treatment, and a higher workload for radiologists.

Google's system follows on from a 2017 research on using deep learning algorithms to spot metastatic breast cancer from lymph node specimens. In this study, Google Health partnered with DeepMind, Royal Surrey County Hospital, Cancer Research UK Imperial Centre, and Northwestern University to see if machine learning could help experts in more accurately detecting the signs of breast cancer.

To evaluate the performance of their machine learning model in the clinical setting, the researchers trained the model on a data set comprising of de-identified mammograms obtained from over 76,000 women in the UK and nearly 15,000 women in the US.
It was then assessed on a separate data set of over 28,000 British and US women.

The model indicated an absolute reduction of 5.7 per cent in false positives for US women and a 1.2 per cent reduction in the sample of British woman. It also showed a 9.4 per cent and 2.7 per cent reduction in false negatives for the US and British women, respectively.

In a separate study of six radiology experts, Google's system demonstrated its effectiveness by outperforming all of the human readers.

"There's enormous opportunity, not just in breast cancer but more widely, to use this type of technology to make screening more equitable and more accurate," Dominic King, the UK lead at Google Health, told the WSJ.

"It feels like this is another step towards this technology actually making a difference in the real world."

The details of the system are presented in a paper published in the scientific journal Nature.