The team used anonymized data from mammograms – low energy X-ray to detect breast cancer – of more than 76,000 women from the UK and more than 15,000 women from the US to train the AI model. Later, they tested it using a de-identified dataset of more than 25,000 women from the UK and 3,000 women from the US. The AI helped reduce false negatives by 5.7 and 1.2 percent in the US and UK respectively. Plus, it also decreased the number of false positives by 9.4 and 2.7 percent in the US and UK respectively. Google noted in its blog post the AI bested doctors in breast cancer detection despite having access to less amount of data. In the coming years, the company hopes to improve the accuracy of the model so it can aid hospitals in real-life detection. Last year, the search giant released a deep learning algorithm that helps doctors detect breast cancer from a pathological perspective by analyzing lymph nodes. Using images to identify specific medical problems isn’t really new for Google’s AI team. In 2018, it released a model that can classify potential heart diseases by scanning your eyes. Last year, it unveiled an AI model that can detect lung cancer more accurately than radiologists using chest scans.
