Google’s AI improved lung cancer diagnosis

Lung cancer kills more Americans every year than any other cancer, and diagnosing it through CT scans can be extraordinarily difficult. Some cancerous nodules are missed only to turn lethal, while others look dangerous but prove benign — and less damaging than the treatments to remove them. A new study raises cautious hope that artificial intelligence can improve accuracy. A Google algorithm trained on 42,000 patient CT scans outperformed six radiologists in determining if patients had cancer, detecting 5% more cancers and cutting false positives 11%. Those results need to be tested more widely before being put into practice, but they make widespread screening more feasible, especially for smokers at high risk.