Deep-learning System Accurately Detects Papilledema from Photographs

Using fundus photographs, an artificial intelligence (AI) system was able to detect papilledema and other optic-disk abnormalities from fundus photographs, according to an article published in the New England Journal of Medicine.

Researcher trained and validated a deep-learning system using data from 6779 patients. A total of 14,341 ocular fundus photographs obtained with pharmacologic pupillary dilation and included 9156 of normal disks, 2148 of disks with papilledema, and 3037 of disks with other abnormalities.

The system was able to detect disks with papilledema from normal disks and disks with nonpapilledema abnormalities with high accuracy (area under the curve of 0.99)

An external-testing data set of 1505 photos resulted in an AUC of 0.96% for recognizing papilledema.

The authors concluded that, “a deep-learning system using fundus photographs with pharmacologically dilated pupils differentiated among optic disks with papilledema, normal disks, and disks with nonpapilledema abnormalities.”

Reference

Milea D, Najjar RP, Zhubo J, et al. Artificial Intelligence to Detect Papilledema from Ocular Fundus Photographs [published online ahead of print, 2020 Apr 14]. N Engl J Med. 2020;10.1056/NEJMoa1917130. doi:10.1056/NEJMoa1917130