|

ForNEWScience

Deliver the most talked about and important scientific findings to the world in simple language

About


2020

According to a research article,The goal of mammography screening is to identify breast cancer at an earlier stage of the disease, so that it can be treated more successfully. Although screening procedures exist throughout the world, the interpretation of mammograms is affected by the false alarm rate and the high false alarm rate. Here, this article proposes an artificial intelligence (AI) system that can surpass human experts in breast cancer prediction. In order to evaluate its performance in the clinical environment, this article selects a large representative dataset from the United Kingdom and a large rich dataset from the United States. This article shows that the false alarm rate has definitely decreased by 5.7% and 1.2% (US and UK), and the false alarm rate has definitely decreased by 9.4% and 2.7%. This article provides evidence of the system's ability to promote from the UK to the US. In an independent study of six radiologists, the artificial intelligence system performed better than all human readers: the area of ​​the artificial intelligence system under the receiver operating characteristic curve (AUC-ROC) is larger than the average radiologist’s AUC-ROC The absolute amplitude is 11.5%. This article carried out a simulation, in which the AI ​​system participated in the dual reading process used in the United Kingdom. It was found that the AI ​​system maintained inferior performance and reduced the workload of the second reader by 88%. The powerful evaluation of the AI ​​system paves the way for clinical trials to improve the accuracy and efficiency of breast cancer screening.

The research Appeared in Nature 2020,577(7788)

Search via

Altmetric Statistics

International evaluation of an AI system for breast cancer screening