Computational fluid dynamics (CFD) estimation of thrombus formation in diabetic retinal microaneurysms (MAs)

Abstract

Purpose : Retinopathy severity and vision loss in diabetic patients may be influenced by MAs that resolve, clot, bleed or leak fluid. However, we lack methods to predict the course of an individual MA. CFD modeling of platelet (PLT) aggregation has been shown to reliably predict thrombus in aortic aneurysms. This preliminary study investigated the ability of CFD modeling to estimate the presence of thrombus in diabetic retinal MAs.

Methods : MA structural information was obtained on adaptive optics scanning laser ophthalmoscopy (AOSLO) video imaging. MA boundaries, thrombus and perfusion percentage of the MA body were manually determined and 3D models were constructed. Particle-continuum coupled CFD models simulated PLT transport and aggregation in each MA over 10 cardiac cycles. To estimate for effect of varied hemodynamics on thrombus development, we varied mean blood velocity from U₀(reference)=1.6mm/s to 0.5U₀=0.8mm/s and 2U₀=3.2mm/s.

Results : CFD models of 4 diabetic retinal MAs (Table) simulated changes in thrombus burden as PLTs closer to the MA wall became adhesive and developed into clot that propagated towards the centerline. After 8 cardiac cycles, PLT aggregates generally became stable. Estimated thrombus area ranged from 25.0 to 11.3%. CFD estimated thrombus area at the reference velocity U₀ decreased by 2.7-10.3% in faster flow conditions (2U₀) and increased by 4.0-19.0% under slower flow (0.5U₀). The 3 MAs with highest estimated thrombus area on CFD had visible clot presence and also the lowest perfusion percentages on AOSLO standard deviation maps. The MA with the lowest estimated thrombus area had no visible clot and the highest perfusion percentage on AOSLO. CFD estimated thrombus area versus clinically apparent area of lower perfusion on AOSLO was most similar to that obtained under 0.5U₀ conditions for 3 MAs and most similar to the 2U₀ condition for 1 MA.

Conclusions : These exploratory findings suggest that CFD AOSLO modeling may provide reasonable approximations of clot presence and perfusion percentage within diabetic retinal MAs. However, individual MA variations in blood velocity should be explored further in order to enable more precise estimation of thrombus area. Future studies may also elucidate the role of other parameters such as MA morphology and red and white blood cell contributions to blood flow physiology in the prediction of thrombus progression over time.

This is a 2020 ARVO Annual Meeting abstract.