SD-OCT image contrast and background color settings influence identification of retinal structures

Abstract

 

Purpose

 

To evaluate the impact of contrast and color setting for the assessment of retinal structures/morphological alterations with SD-OCT.

 

Methods

 

248 single Heidelberg SD-OCT-B-scans of 62 patients with various macular diseases were analyzed by 4 equally experienced retina specialists. Two readers were used to interpret white on black images in daily practice and 2 were used to black on white images. One scan of each patient was extracted in 4 different contrast/color settings: white+normal(w+n)=black on white image with predefined contrast of 9 the from the Spectralis software, white+high(w+h)=black on white image with highest possible contrast of 16, black+normal(b+n)= white on black image with predefined contrast of 12 the from the Heidelberg software and black+high(b+h)=white on black image with highest possible contrast of 16(Figure 1). The images were randomly mixed and each of the 4 readers analyzed images in the same order for 34 parameters e.g. presence of geographic atrophy(GA), subretinal fluid(SRF), CME, reticular/cuticular drusen, epiretinal membrane(ERM), pigment epithelium detachment(PED) as well as the integrity of ELM, ellipsoid band and RPE. Inter-reader correlation (Fleiss kappa) in respect to the different contrast/color settings for each variable was performed to investigate whether one setting is superior. It was further evaluated whether interpreting either white on black or black on white images in daily practice has an impact on the inter-reader accordance in respect to the contrast/color setting.

 

Results

 

The presence/absence of ERM and PED was better evaluated in the black color setting compared to the white and better in high contrast compared to normal contrast. Both showed the highest inter-reader accordance in b+h setting(κ=0.6+κ =0.62) and the lowest in w+n setting(κ=0.39+κ=0.44). The presence/absence of SRF was by far best evaluable in the b+h contrast setting(κ=0.78). However, the presence of GA was best evaluable in the w+n setting(κ=0.62 vs. b+h:κ=0.48). The examiners used to interpreting white on black images had the highest accordance when looking at this color setting, whereas the readers used to black on white SD-OCT did better on high contrast images irrespective of color.

 

Conclusions

 

Contrast as well as color background seem to have an impact when evaluating retinal structures/morphological alterations on SD-OCT.

 

 

 

Keywords: 688 retina • 550 imaging/image analysis: clinical