Meeting News

AI system shows high sensitivity, specificity for detection of diabetic retinopathy

Jennifer I. Lim

CHICAGO — The EyeArt artificial intelligence system for diabetic retinopathy screening demonstrated nearly 96% sensitivity for the detection of moderate or higher nonproliferative diabetic retinopathy, according to a speaker here.

“EyeArt AI system compared favorably with clinical reference standard four-field stereoscopic images. ... It is therefore useful for the detection of referable DR in patients with diabetes. This can enable point-of-care screening,” Jennifer I. Lim, MD, said at the American Society of Retina Specialists annual meeting.

In the prospective study across 15 centers, 911 subjects underwent undilated two-field 45° fundus photography and dilated four-wide-field stereoscopic fundus photography. The EyeArt AI system (Eyenuk) evaluated results about referable diabetic retinopathy, defined as moderate or higher nonproliferative diabetic retinopathy based on the International Clinical Diabetic Retinopathy Severity Scale. If the images could not be graded, two-field photos were repeated after dilation, she said.

The dilated wide-field photography was the reference standard and graded by Wisconsin Fundus Photograph Reading Center graders using the ETDRS severity scale, Lim said.

Of the 911 participants, 1,674 eyes had gradable two-field and four-field images. The EyeArt system evaluated 310 eyes as positive for referable diabetic retinopathy and 1,364 as negative. Using only undilated images, the EyeArt system showed 95.5% sensitivity, 86% specificity and 87.5% gradability. Lim said 214 eyes required dilated two-field photography, of which 170 eyes were gradable and 44 remained ungradable.

With the dilate-if-ungradable photography protocol, EyeArt gradability rate improved to 97.4%, sensitivity remained at 95.5% and specificity improved slightly to 86.5%. – by Robert Linnehan

Reference:

Lim JL. Artificial intelligence screening for diabetic retinopathy: analysis from a pivotal multi-center prospective clinical trial. Presented at: American Society of Retina Specialists annual meeting; July 27-30, 2019; Chicago.

Disclosure: Lim reports no relevant financial disclosures.

Jennifer I. Lim

CHICAGO — The EyeArt artificial intelligence system for diabetic retinopathy screening demonstrated nearly 96% sensitivity for the detection of moderate or higher nonproliferative diabetic retinopathy, according to a speaker here.

“EyeArt AI system compared favorably with clinical reference standard four-field stereoscopic images. ... It is therefore useful for the detection of referable DR in patients with diabetes. This can enable point-of-care screening,” Jennifer I. Lim, MD, said at the American Society of Retina Specialists annual meeting.

In the prospective study across 15 centers, 911 subjects underwent undilated two-field 45° fundus photography and dilated four-wide-field stereoscopic fundus photography. The EyeArt AI system (Eyenuk) evaluated results about referable diabetic retinopathy, defined as moderate or higher nonproliferative diabetic retinopathy based on the International Clinical Diabetic Retinopathy Severity Scale. If the images could not be graded, two-field photos were repeated after dilation, she said.

The dilated wide-field photography was the reference standard and graded by Wisconsin Fundus Photograph Reading Center graders using the ETDRS severity scale, Lim said.

Of the 911 participants, 1,674 eyes had gradable two-field and four-field images. The EyeArt system evaluated 310 eyes as positive for referable diabetic retinopathy and 1,364 as negative. Using only undilated images, the EyeArt system showed 95.5% sensitivity, 86% specificity and 87.5% gradability. Lim said 214 eyes required dilated two-field photography, of which 170 eyes were gradable and 44 remained ungradable.

With the dilate-if-ungradable photography protocol, EyeArt gradability rate improved to 97.4%, sensitivity remained at 95.5% and specificity improved slightly to 86.5%. – by Robert Linnehan

Reference:

Lim JL. Artificial intelligence screening for diabetic retinopathy: analysis from a pivotal multi-center prospective clinical trial. Presented at: American Society of Retina Specialists annual meeting; July 27-30, 2019; Chicago.

Disclosure: Lim reports no relevant financial disclosures.

    See more from American Society of Retina Specialists Meeting