What: Google’s DeepMind AI can now detect bread cancer better than a human doctor.
Despite increased efforts of early detection of breast cancer, interpretation of the data medical specialists use, “suffers from suboptimal rates of false positives and false negatives”, DeepMind said.
How: The AI was trained on two sets of mammograms of women in the US and the UK collected between 2001 and 2018.
The company claims that its system application led to a reduction of 9.4% and 2.7% for the US and UK false negatives respectively. The reduction in the false positives was 5.7% for the US data set and 1.2% for the UK.
The AI’s model was then tested through an independent study by six human radiologists, in which the AI outperformed all of them by a significant margin, DeepMind said in a paper published in Nature.
To make sure that the model was applicable across different populations, DeepMind ran an experiment in which an AI trained solely on UK subject was used to evaluate US data. The results showed that the system was still able to outperform the human experts on US data, the company claims.
Why is it important: Breast cancer is the second most common cause of death from cancer among women after lung cancer.
Early detection is crucial for successful treatment but efficient screening remains challenging. The complexity is further exacerbated by a shortage of qualified human radiologists across many hospitals. Though still in its early days, the new model could prove helpful as a “second reader” opinion in the screening process, the company suggests.
You can access the full text of the paper here.