Clinical Validation Study of an automated DR Screening System against 7-field ETDRS Stereoscopic Reference Standard
PURPOSE AND BACKGROUND:
Comprehensive clinical validation of an automated diabetic retinopathy (DR) screening system, EyeArt v2.0 for detecting referable diabetic eye disease (DED) (moderate nonproliferative DR (NPDR) or higher on the ICDR scale and/or surrogate markers for clinically significant macular edema (CSME)). To the best of our knowledge, this is the first time an automated DED screening system has been comprehensively validated against ETDRS 7-field stereoscopic reference standard.
STUDY DESIGN AND METHODS:
Retrospective observational study evaluating automated screening system on anonymized fundus images. 7-field ETDRS stereoscopic fundus images of 755 eyes from DRCRnet were graded on the ETDRS scale. EyeArt was evaluated on 2 sets of 10,000 Monte-Carlo experiments, each with 10,000 eyes, with first set using 3-retinal monoscopic fields (macula centered, optic nerve centered, and temporal) and the second set using 1-retinal monoscopic field (macula centered). Both sets simulated a screening population: 68% no DR cases, 10% mild NPDR cases, 14% moderate NPDR cases, 4% severe NPDR cases, and 4% PDR cases.
EyeArt results are summarized in table below.
Automated screening using EyeArt is safe and effective as demonstrated by high sensitivity/specificity when analyzing simulated screening populations with monoscopic 3-fields and 1- field per eye (typical in screening programs) against the gold standard grading of 7-field stereoscopic ETDRS fundus images. Sensitivity for treatable DED is high though EyeArt detects CSME only based on surrogate markers (standard practice in DR screening programs that use non-stereoscopic images).
Automated screening with EyeArt achieves high sensitivity and specificity on image sets with 3-fields, 1-field per eye against the gold standard 7-field ETDRS grading.
- Kaushal Solanki, Malavika Bhaskaranand, Chaithanya Ramachandra, and Sandeep Bhat. “Clinical Validation Study of an Automated DR Screening System against 7-Field ETDRS Stereoscopic Reference Standard.” In AAO 2016. USA, 2016.
- Kaushal Solanki, Malavika Bhaskaranand, Chaithanya Ramachandra, and Sandeep Bhat. “Comprehensive Clinical Validation Study of a Fully-Automated Diabetic Retinopathy Screening System Using Color Fundus Images against 7-Field ETDRS Stereoscopic Reference Standard.” In 16th EURETINA Congress. Copenhagen Denmark, 2016.
- Download study PDF