Association for Research in Vision and Ophthalmology
Association for Research in Vision and Ophthalmology
Source/Disclosures
Source:

Milea D, et al. Deep learning for image-based detection of sight- and life-threatening neuro-ophthalmic conditions affecting the optic nerve head. Presented at: Association for Research in Vision and Ophthalmology annual meeting; June 12, 2020 (virtual meeting)

Disclosures: The authors report no relevant financial disclosures.
June 29, 2020
1 min read
Save

Fundus photograph-based deep learning system can classify optic discs

Source/Disclosures
Source:

Milea D, et al. Deep learning for image-based detection of sight- and life-threatening neuro-ophthalmic conditions affecting the optic nerve head. Presented at: Association for Research in Vision and Ophthalmology annual meeting; June 12, 2020 (virtual meeting)

Disclosures: The authors report no relevant financial disclosures.
You've successfully added to your alerts. You will receive an email when new content is published.

Click Here to Manage Email Alerts

We were unable to process your request. Please try again later. If you continue to have this issue please contact customerservice@slackinc.com.

A newly developed deep learning system based on fundus photographs can automatically discriminate normal optic discs from discs with papilledema or other abnormalities.

“Our system could, in our data set, discriminate optic discs on photographs. ... The accuracy was quite high, especially for papilledema,” Dan Milea, MD, PhD, said at the virtual Association for Research in Vision and Ophthalmology meeting.

Deep learning software normal from abnormal optic discs 0.99, papilledema 0.98 infographic

Milea and colleagues developed a deep learning system to automatically classify optic discs as normal or abnormal. The system could also detect papilledema using ocular fundus photographs. The researchers used 14,341 ocular fundus photographs to train the deep learning system and for validation, plus 1,505 photographs for external testing. The photographs were from patients in the Brain and Optic Nerve Study with Artificial Intelligence international consortium, Milea said.

The performance of the deep learning system was evaluated by calculating the area under the receiver operating curve (AUC), specificity and sensitivity. The system’s conclusions were referenced with neuro-ophthalmologist-performed clinical examinations and ancillary investigations, he said.

The researchers used 9,156 images of normal discs, 2,148 images with papilledema and 3,037 images with other optic disc abnormalities. The deep learning system successfully discriminated normal optic discs from abnormal optic discs with an AUC of 0.99. The system successfully determined papilledema discs from other discs with an AUC of 0.98.

Using an external data set, similar results from the deep learning system were observed. The system’s AUC for the detection of normal discs was 0.98, the sensitivity was 95.3 and the specificity was 86.6. The deep learning system’s detection of papilledema discs had an AUC of 0.96, a sensitivity of 96.4 and a specificity of 84.7.

“There was a high negative predictive value, meaning that a patient who was negative had a very high chance to be normal,” Milea said.