Article

EyeArt Screening System Detects Diabetic Retinopathy

A presentation at AAO 2019 highlighted the ability of the EyeArt screening system to detect diabetic retinopathy with 95.5% sensitivity.

Srinivas Sadda, MD

Srinivas Sadda, MD

Data on the effectiveness of the artificial intelligence screening system EyeArt was presented at the American Academy of Ophthalmology (AAO) 2019 Annual Meeting in San Francisco.

Results of the study examining EyeArt revealed the system had the ability to accurately predict diabetic retinopathy 95.5% of the time, without the input of an ophthalmologist, and in less than a minute.

With the diabetes epidemic continuing to plague patient populations in the US and across the globe—and 1 in 4 diabetics developing diabetic retinopathy—investigators sought to evaluate the effectiveness of the artificial intelligence screening system in a group of 893 patients. Led by Srinivas Sadda, MD, of the Doheny Eye Institute at UCLA, investigators conducted the study at 15 different centers across the US.

For inclusion in the study, patients needed to have a diagnosis pf diabetes mellitus, be at least 18 years of age, and provide written informed consent. Exclusion criteria included history of macular edema or retinal vascular occlusion, history of ocular injections, and persistent visual impairment in one or both eyes, among others.

All patients in the study underwent an undilated 2-field fungus imaging to for automated eye-level detection of referable diabetic retinopathy by EyeArt. Investigators defined referable diabetic retinopathy as refractory diabetic retinopathy, clinically significant diabetic macular edema. moderate nonproliferative diabetic retinopathy or higher on the International Clinical Diabetic Retinopathy scale. Patients also underwent 2 dilated 4 wide-field fungus imaging for adjudicated ETDRS reference standard.

Primary endpoints of the study included the number of eyes whose EyeArt results match the reading center grading for identifying referable diabetic eye disease and eyes whose results match the reading center grading for identifying vision threatening diabetic eye disease.

Upon analyses, EyeArt achieved sensitivity of 95.5%(95% CI, 93.0%-97.9%), specificity of 86.0%(95% CI, 83.8%-88.3%) and gradeability of 87.5%(95% CI, 85.4%-89.7%) with undiluted images. When using a dilate-if-ungradable protocol, the system achieved sensitivity of 95.5%(95% CI, 93.1%-97.8%), specificity of 86.5%(95% CI, 84.4%-88.6%) and gradeability of 97.4%(95% CI, 96.4%-98.5%).

Based on the observed results from the trial, investigators suggest the results of the study support the use of the EyeArt system as it compared favorably with ETDRS reference standards and met the study’s predetermined sensitivity and specificity endpoints.

“Diabetic patients already outnumber practicing ophthalmologists in the United States, and unfortunately, that imbalance is only expected to grow,” Sadda said. “Accurate, real-time diagnosis holds great promise for the millions of patients with diabetes.”

This study, “AI Eye Screening for Diabetic Retinopathy: Results From a Pivotal Multicenter Clinical Trial,” was presented at AAO 2019 by Srinivas Sadda, MD.

Related Videos
Ashfaq Marghoob, MD: Artificial Intelligence, Smartphone Use for Pigmented Lesion Classification
Steve Nissen, MD | Credit: Cleveland Clinic
Major Diagnostic Challenges for Pigmented Lesions, with Ashfaq Marghoob, MD
Sherona Bau, NP | Credit: UCLA Health
Jessica Crimaldi, NP | Credit: Jessica Crimaldi on LinkedIn
Benjamin Scirica, MD | Credit: Brigham and Women's Hospital
© 2024 MJH Life Sciences

All rights reserved.