Summary of the application of artificial intelligence in medical diagnosis (artificial intelligence has been proved to be as good as doctors in interpreting medical images)

Benshaw found that many friends were interested in information that artificial intelligence proved to be as good as doctors in interpreting medical images. Bian Xiao has sorted out some relevant information, which you can share here.

Meta analysis of 14 studies shows that the accuracy of AI algorithm in diagnosing diseases in medical images is 87%, and the accuracy of excluding specific diseases is 93%.

These figures are very close to the efficiency of doctors in the same study.

The results are published in the Lancet Digital Health.

Dr Liviavus of Lucerne State Hospital in Switzerland and his colleagues wrote: "our review found that the diagnostic performance of deep learning models is comparable to that of medical professionals.

At the same time, the author notes some warnings.

First, the studies they reviewed rarely included the results of external validation. There is also a lack of comparison between deep learning models using the same samples and health care professionals. The authors also found that because of the poor reports of these studies, it is difficult to give a reliable explanation for the diagnostic accuracy of the reports.

"the new reporting standards address the specific challenges of deep learning and can improve future research and give people confidence in the future assessment of the technology," said Faes et al. " Write.

They concluded that there is "great potential" to use deep learning algorithms to diagnose diseases. "through this exploratory meta-analysis, we are careful to point out that the deep learning algorithm is as accurate as health care professionals and acknowledge that more research is needed to consider integrating the algorithm into the real environment."