Purchase this article with an account.
Luís Mendes, Inês Marques, Luis De Sisternes, Maria Helena Madeira, Torcato Santos, Warren Lewis, Mary K Durbin, Jose G Cunha-Vaz; Comparison of different methods for the identification of capillary closure in diabetics retinopathy using OCTA. Invest. Ophthalmol. Vis. Sci. 2021;62(8):2475.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Retinal vessel metrics indicating capillary closure (CC) have shown potential clinical value by identifying eyes at different severity levels and at increased risk for disease progression to more severe stages. We compare the performance of 33 metrics computed based on optical coherence tomography angiography (OCTA) for identification of CC in different ETDRS groups.
OCTA data from 84 healthy eyes (70±4.8 years) and 78 eyes of diabetic patients (67±7.5 years), ETDRS 10-20 (24 eyes/patients), 35 (31 eyes/patients) and 43-53 (23 eyes/patients), were processed with using different methods: 1) Abnormal intercapillary spaces (AIS) measured using the method proposed by Mendes et al. (2019); 2) Skeletonized vessel density (SVD) and 3) perfusion density (PD), both computed using the Zeiss Meditec Density Exerciser (version 10.0.12787), 4) Entropy, 5) 10 metrics related with the histogram of the slab, 6) Four metrics based on Frangi filter (FF), 7) Coarseness metric of the Tamura filter (CTF) and, 8) 14 features based on Haralick texture features that include the contrast (CT).The performance of these methods was tested in the superficial capillary plexus (SCP), in the deep capillary plexus (DCP), and in the full retina (FR). Data were acquired using CIRRUS HD-OCT 5000 with AngioPlex® OCTA (ZEISS, Dublin CA) using the acquisition protocol angiography 3x3.
The best separation between the eyes with type 2 diabetics and the control group was obtained in the SCP, with the FR also obtaining a competitive performance. In the SCP the metrics that show better performance were the AIS and the VD (table 1) with a value of AUC equal to 0.89 [CI 95% 0.84-0.94] and 0.85 [CI 95% 0.79-0.91], respectively. The values of these metrics on the ETDRS groups 10-20, 35 and 43-53 show a progressive increase in CC which is correlated with the disease severity (table 2).
On AngioPlex the measurement of CC shows better performance in the SCP using AIS or the SVD. These methods improve the discrimination of eyes with diabetic retinopathy with different degrees of severity (ETDRS groups 10-20, 35 and 43-53).
This is a 2021 ARVO Annual Meeting abstract.
Top 5 AUC values obtained when the task was to perform separation between pathological eyes and control group. The rank of the metric is also presented (#).
Mean and standard deviation of the top 5 CC metrics for the SCP in the control group and pathological eyes by ETDRS grade.
This PDF is available to Subscribers Only