Abstract
Purpose: :
To explore integration of ocular coherence tomography (OCT) images into grading of diabetic macular edema (DME) from stereo color fundus photos (SFP), in order to combine benefits of both.Background: Retinal thickening (RT) evaluation provides important morphological parameters for DME research. Historical studies graded DME using SFP. Many contemporary DME studies use OCT data as the primary variable, because it objectively measures RT. However, faulty scan centration, boundary detection errors, and sketchy sampling of peripheral macula (radial scan pattern) undercut some measurements. Grading SFP still provides specific DME information (foveal proximity and area of RT) unavailable from Stratus OCT. The ETDRS combined SFP data into an overall severity scale highly correlated with visual acuity. DRCRnet (MD Davis, ARVO 2005) found that the ETDRS scale and OCT central thickness both correlated with vision (Spearman correlations of 0.43 and 0.45), but were not identical to each other (0.63 correlation). To address drawbacks of SFP grading (suboptimal stereo effect or other quality problem; or confounding from epiretinal membrane), we supplemented ETDRS DME grading (T Gardiner et al., Curr Eye Res 2006) with concurrent examination of OCT scans.
Methods: :
Stereo color fundus images already evaluated in two DME studies were stratified by ETDRS scale (RE Gangnon, ARVO 200), from group 1 with no DME to group 5 with severe central involvement. We randomly selected 100 eyes to obtain 10 eyes in group 1A-1B, 10 in 1C, 20 in group 2, 10 in 3A ,10 in 3B, 20 in group 4, 10 in 5A, and 10 in 5B - enriching 3B to 5B with eyes having subretinal fluid and/or vitreoretinal abnormalities by OCT. Two experienced graders evaluated DME from SFP, first independently of OCT (sOCT) and then using OCT as an adjunct (cOCT). SFP grading results were compared between the two approaches. Data from 56 eyes (favoring central DME) are complete, and data on another 44 eyes (with milder DME) will be presented also.
Results: :
In 56 eyes graded sOCT, DME was central in 49, with mean area of 4.59 disc areas (SD 4.15, min-max of 0-14.20 DA). Graded cOCT, DME central involvement and area had similar distributions, but individual eyes often shifted. Central RT grade changed 1+ step in 31 eyes, and DME area estimate changed in 51 eyes by an average of 1.86 disc areas.
Conclusions: :
Adding information from OCT to SFP grading may maximize information from the two modes, improving data accuracy by reducing errors due to image quality and confounding lesions in either. Combined SFP/OCT assessment may help develop more a more nuanced description of DME with data from HD OCT.
Keywords: diabetic retinopathy • macula/fovea • imaging methods (CT, FA, ICG, MRI, OCT, RTA, SLO, ultrasound)