Diagnosis of papilledema requires neuro-ophthalmic expertise and expensive ancillary testing. In neurology clinics and emergency departments, its ophthalmoscopic diagnosis can be easily mistaken with pseudo-papilledema due to ODD, especially when ODDs are buried within the optic nerve head. A DLS was previously shown to accurately recognize papilledema from normal optic nerves and performed as well as expert neuro-ophthalmologists. It was also able to reliably grade the severity of papilledema.

We developed and trained a DLS using a total of 3,779 ocular fundus photographs (3,014 papilledema and 765 ODD [289 buried; 476 visible]) obtained from 1810 patients in 28 centers from 20 countries. The DLS’s performance to classify papilledema and ODD was subsequently tested on a separate dataset of 631 images obtained from 3 independent centers (424 papilledema; 207 ODD [112 buried; 95 visible]) using a receiver operating characteristic analysis. Performance metrics included the area under the receiver-operating-curve (AUC), sensitivity and specificity. The reference ophthalmologic and neurological diagnoses were established by the enrolling expert neuro-ophthalmologists, using state-of-the-art diagnostic methods.