实验性AI乳腺筛查系统在研究中较传统方法多检出25%间歇性癌症(通常被漏诊)及更多浸润性癌症,总体发现更多病例,同时预计减少临床医生40%工作量。该研究由Google与帝国理工学院及NHS合作完成,发表于Nature Cancer。
Amazing to see how our experimental research AI system for mammography interpretation identified more cases of interval cancers, as well as invasive cancer, and more cases overall, than conventional methods. Plus it could reduce clinicians' screening workloads by an estimated 40%.
Big step towards a future where AI serves as a reliable collaborator for clinicians.