Google has developed an artificial intelligence system that can match or even surpass doctors in detecting breast cancer. Although there continue to be cases where white coats beat the car. These, in a nutshell, the conclusions of a study conducted by British and American researchers and published in the journal Nature. Thousands of mammograms of women in the United Kingdom and the United States have been used to train artificial intelligence. Eventually, a model was obtained that could detect tumours that had been initially ignored by doctors and reduce false positives in patients who did not have cancer. In particular, the artificial intelligence system has been tested on mammograms of over 25 thousand women in the United Kingdom and 3 thousand women in the United States.
Results showed that the Google model reduced false negatives (i.e. cancer cases not identified by doctors) by 9.4 percent in the United States and 2.7 percent in the United Kingdom compared to original radiological diagnoses. It also reduced “false positives” by 5.7 percent and 1.2 percent respectively. In the UK, where typically two radiologists read a mammogram, the study found that the Google model did not perform worse than the second “reader” and could potentially reduce the workload by 88 percent.
However, scientists specified that the algorithm is not yet ready for clinical use. Although radiologists and AI specialists consider this new system to be promising and Google Health officials said it could possibly support radiologists in detecting breast cancer, improving efficiency in reading mammography, and could help doctors establish the prognosis for each case. “Using this type of technology is a huge opportunity to make screening fairer and more accurate, not just in breast cancer,” says Dominic King, head of Google Health in the UK. “It looks like this is another step towards this technology that could really make a difference in the real world,” he adds.