Machine-learning algorithms have been developed that are able to diagnose disease by analysis of standard medical images.
A worldwide group of researchers sought to determine the generalizability of the effectiveness of these machine-learning algorithms at diagnosis. The researchers performed a systematic review and meta-analysis of 68 such studies, the results of which they validated in 24 other studies.
They concluded that the performance of these algorithms is on par with that of medical professionals, on average, but there are gaps in scientists’ design and reporting of deep-learning studies, which excluded several studies from the analysis.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Stower, H. Machines match humans in image-based diagnoses. Nat Med 25, 1798 (2019). https://doi.org/10.1038/s41591-019-0679-9
Published:
Issue Date:
DOI: https://doi.org/10.1038/s41591-019-0679-9
This article is cited by
-
The Crohn’s-like lymphoid reaction density: a new artificial intelligence quantified prognostic immune index in colon cancer
Cancer Immunology, Immunotherapy (2022)