Abstract
Purpose :
To improve the repeatability and reproducibility measurements of vessel density and avascular area in optical coherence tomography (OCT) angiography
Methods :
3x3 mm macular scans of healthy eyes were taken using a spectral OCT system (RTVue-XR, Optovue, Inc.). The Split-Spectrum Amplitude Decorrelation Angiography (SSADA) algorithm produced an en face inner retinal angiogram by segmenting anterior to the outer boundary of inner plexiform layer. An algorithm using the fuzzy C-means method detected vasculature with high decorrelation values. The level-set method refined the separation between vessels and noise artifacts. Then, the extracted vessel binary maps were processed to measure: 1. Vessel area density (VAD): percentage area occupied by vasculature in the selected area, 2. Vessel length density (VLD): length of skeletonized vasculature divided by the area and 3. Avascular area: area of decorrelation values detected below the threshold derived by a function of level set evolution. The intra-visit repeatability and inter-visit reproducibility were calculated in terms of coefficient of variation (CV). The reliability was compared to prior work that used a fixed cutoff to extract vessels and the commercial software (AngioAnalytics) available on RTVue-XR.
Results :
The scans from 32 healthy adults with a signal strength index (SSI) >60 were included. The mean VAD varied in different algorithms. The level-set method measured VAD and FAZ with greater repeatability and reproducibility compared to the other methods. The new parameter VLD was similarly reliable. The FAZ was automatically detected and quantified in subjects with a wide range of SSI, indicating less dependency on signal strength (Fig 1).
Conclusions :
Level-set algorithm for quantification of vessel density and avascular areas in OCT angiography is more reliable compared to fixed cutoffs or currently available commercial software. Further testing in diseased eyes can validate its clinical utility.
This is an abstract that was submitted for the 2016 ARVO Annual Meeting, held in Seattle, Wash., May 1-5, 2016.