Abstract
Purpose: :
To evaluate diabetic macular edema (DME) levels from green channel (GrCh) fundus images (simulating red free) as compared to full color (Red/Green/Blue, RGB) images.
Methods: :
From the Diabetes Control and Complications Trial/Epidemiology of Diabetes Interventions and Complications Study (DCCT/EDIC) 100 digital image sets stratified by DME presence and severity were selected from previously graded color images. Original images were tonally optimized to extract the GrCh for monochrome display. Reading center graders independently evaluated the GrCh images according to modified ETDRS procedures for presence, location and area of DME. Assessment included duplicate grading (with adjudication when required) to replicate the procedures applied to the color image evaluation. These two independent gradings allowed evaluation of grading reproducibility.
Results: :
Both GrCh graders agreed within 0.5 disc areas (DA) on total DME area (within the ETDRS grid) in 79 out of 95 eyes, within 1.0 DA in 89 out of 95 eyes and within 2.0 DA in 94 out of 95 eyes. The two gradings yielded an intra-class correlation of 0.80 with a mean difference between gradings of 0.13, (SD = 0.52), p=0.01. After post hoc comparison of discrepant RGB and GrCh sets, another grader judged that subtle areas of edema, which comprised the majority of eyes with DME in the sample, were better seen in GrCh than in full RGB.
Conclusions: :
Advantages of red free images over color images for assessing diabetic retinopathy (DR) severity have been reported previously. This was attributed to greater contrast of DR lesions against the retinal pigment epithelial (RPE) background in GrCh compared to RGB. The high reproducibility of grading DME from GrCh in our current study is attributed to detection of altered retinal transparency from DME and heightened perception of stereo effect due to better visibility of small intra-retinal cues in the otherwise transparent retina. Currently, most clinical trials collect color images, fluorescein angiograms and OCTs in order to evaluate various aspects of diabetic macular edema. Red free images may allow improved evaluation of DR and DME with less patient discomfort due to the lower flash and may be a practical alternative to RGB in DR clinical trials or tele-ophthalmology DR screening.
Clinical Trial: :
http://www.clinicaltrials.gov NCT00360893
Keywords: clinical research methodology • diabetes • edema