Purchase this article with an account.
M. Prakash, J. K. Sun, J. D. Cavallerano, T. J. Murtha, N. N. Vaidya, P. A. Silva, S.-E. Bursell, L. P. Aiello; A Comparison of the Automated Retinal Imaging System (ARIS) and ETDRS Protocol Color Stereoscopic Retinal Photography in Lesion-Level Assessment of Diabetic Retinopathy. Invest. Ophthalmol. Vis. Sci. 2008;49(13):4229.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
ETDRS 7-standard field 35 mm color stereoscopic retinal photography (StdPh) is the "gold-standard" for determining diabetic retinopathy (DR) severity. ARIS (model 110, Prescott, AZ) acquires 7 sequential color stereoscopic digital images approximating the ETDRS photo fields by an automated technician-run process. We performed lesion level correlation between these two imaging modalities in patients with diabetes mellitus (DM).
Clinical exam by a retina specialist was followed by a one-time acquisition of ARIS and ETDRS protocol retinal images, subsequently graded by independent masked readers. Kappa values are weighted.
Type I DM was present in 56% of the 105 patients (209 eyes) evaluated. Mean age was 50 years (range 19-78), DM duration 23 years (1-66), HbA1c 7.7% (5.1-12.0) and 52% were male. By StdPh there was no DR in 11%, 5% questionable DR, 43% mild NPDR, 10% moderate-severe NPDR, and 31% PDR. Three (1.5%) StdPh and 59 (29%) ARIS image sets were partially ungradable; however, DR severity could not be determined in only two (1.0%) StdPh sets and six (3.0%) ARIS image sets. Exact agreement between StdPh and ARIS occurred in 70% (k=0.83) and agreement within 1 level (1L) was 91%. Exact agreement for macular edema was 82% (k=0.43). Analysis of NVD revealed exact and 1L agreement in 96% (k=0.70) and 100%, respectively. Two "questionable" NVDs noted on StdPh were not seen on ARIS. ARIS found more severe NVD in 4 eyes. Lesion comparisons between ARIS and StdPh are shown. Low lesion prevalence may have reduced sensitivity and PPV.
ARIS imaging compared favorably with StdPh for determining overall clinical DR level. High sensitivity for H/MA suggests ARIS may be useful for assigning retinopathy levels with respect to this lesion. In general, ARIS individual lesion grading had high specificity and NPV, indicating lesions graded as "not present" by ARIS were unlikely to be missed. Thus, ARIS imaging may be particularly useful for establishing absence of diabetic pathology.
This PDF is available to Subscribers Only