Abstract
Purpose :
OCTA does not successfully image all MAs, yet it is unclear what determines MA OCTA visibility. Our goal was to identify diabetic MA wall characteristics on adaptive optics scanning laser ophthalmoscopy (AOSLO) as well as blood flow parameters estimated by computational fluid dynamics (CFD) modeling that are correlated with OCTA MA visibility.
Methods :
Retinal MAs in diabetic eyes were imaged by AOSLO and OCTA (Optovue, software ver 2017.1.0.149) at the same visit. AOSLO images were graded for wall hyperreflectivity (WH), wall deformability (WD), cross-sectional area, clot presence, perfusion percentage and morphology. On OCTA, MAs were assessed as fully visible (FullVis), partially visible (PartVis) or not visible (NoVis). Multimodal images for each MA were registered. CFD modeling estimated shear rate drop (SRD) and wall shear stress drop (WSSD) within a subset of MAs. Statistical analyses adjusted for correlations between follow-up images of individual MAs.
Results :
A total of 162 MAs (45 eyes) were imaged (259 sessions, 97 MAs with >1 follow-up). On OCTA, MAs were 58.7% (N=152) FullVis , 21.2% (55) PartVis and 20.1% (52) NoVis. FullVis MAs had greater perfusion percentage (FullVis mean±SD: 71.4±29.9%, PartVis: 52.8±32.3% NoVis: 41.5±36.4%, p<0.0001) and were also more likely to exhibit WD (WD in 29.8% FullVis vs 13.2% in PartVis or NoVis MAs, p=0.006). No relationship was identified between OCTA visibility and MA WH, area or clot presence. Within MAs that were fusiform or saccular (N=212), FullVis MAs were less likely to be saccular (FullVis: 76.0%, PartVis or NoVis: 90.1%, p=0.01). Of 16 MAs that underwent CFD modeling, FullVis MAs had lower SRD and WSSD than PartVis or NoVis MAs (SRD: 54.0±57.2 vs 59.0±68.6, p<0.0001; WSSD: 19.5±15.5 vs 28.6±24.5, p<0.0001).
Conclusions :
In this study, over 40% of MAs present on AOSLO were not fully visible in superficial or deep capillary plexus OCTA images. OCTA visibility appeared to be more dependent on MA perfusion than size. Thus, OCTA evaluation of MA may particularly underestimate MAs with limited perfusion. Computational modeling may help identify MAs with slower blood flow that may be less visible on OCTA. These data support future investigations to determine the implications of MA OCTA visibility for predicting local neural retinal pathology and visual dysfunction in the diabetic eye.
This is an abstract that was submitted for the 2018 ARVO Annual Meeting, held in Honolulu, Hawaii, April 29 - May 3, 2018.