Abstract
Purpose :
Compare the performance of three novel algorithms that aim to quantify the focal capillary dropout in patients with DR with the available method of vessel density (VD).
Methods :
Metrics associated with focal capillary dropout (FCD) may provide major clinical value by indicating the eyes at increased risk for progression to more severe stages of retinopathy.Three new methods were developed in the scope of this study that aim to detect regions associated with FCD in the superficial retinal plexus. The first method (perfusion method) aims to detect holes in the OCTA perfusion slab using morphological filters.The second, texture method, aims to detect the changes in the texture of the OCTA slab associated with regions with FCD.The regions detected by the third method, fusion method, correspond to the regions simultaneous detected by the other two FCD methods.The associated FCD metric for each method correspond to the total of pixels of the detected areas.The results obtained with data, acquired using the CIRRUS™ HD-OCT 5000 with AngioPlex® OCT Angiography (ZEISS, Dublin, CA) (angiography 3x3), from 142 eyes from subjects with DR (68±7 years) categorized by ETDRS severity groups were compared with the results obtained with the data from 45 healthy eyes.A ROC curve was used to compare the performance of the three methods with the VD reported by the Density Exerciser software (version 10.0.0.12787).
Results :
From the analysis of the ROC curve (Figure 1) obtained when the task was to distinguish the DR eyes from the normal eyes the best FCD method (the perfusion method) got an AUC equal to 0.81 and for the VD the AUC was equal to 0.75. The perfusion and the fusion method show similar behavior, especially in the region with high sensitivity. When the task was to distinguish the ETDRS 35-47 group from healthy plus the ETDRS 10-20 group the best FCD method (the fusion method) got an AUC equal to 0.76 and for the VD the AUC was equal to 0.74 (Figure 2).
Conclusions :
The results obtained using the methods to identify zones of FCD do not appear to significantly improve the discrimination of DR eyes comparing with generalized VD suggesting that in the initial stages of diabetic retinal disease the capillary dropout is generalized.
This abstract was presented at the 2019 ARVO Imaging in the Eye Conference, held in Vancouver, Canada, April 26-27, 2019.