Abstract
Purpose :
Optical coherence tomography angiography (OCTA) has the potential to detect retinal ischemia. To quantify the degree of ischemic pathology most studies have been using built-in tools of the OCTA device. The aim of this study was to develop and compare different device independent analysis approaches for the quantification of ischemia in OCTA.
Methods :
6x6mm en face-OCTA (Zeiss Angioplex, Carl Zeiss AG, Jena, Germany) images of the superficial plexus of 17 healthy eyes and 20 eyes with different ischemic retinal diseases were analysed retrospectively using self-developed matlab scripts. Included were patients with diabetic retinopathy (n=9), central (n=5) and branch retinal vein occlusion (n=4), hypertensive retinopathy (n=1), and occlusive retinal vasculitis in sarcoidosis (n=1). Prior to image analysis the global brightness was normalised across images. Four different analysis approaches were applied to each OCTA image: 1) mean grey value of all pixels, 2) number of pixels above a threshold grey level of 50 (8 bit), 3) total blood vessel length, based on a skeletonise algorithm, 4) a vector approach, where from each non-vessel pixel the length of a vector to the next blood vessel was measured. This approach contained three subgroups: a) the shortest vector to the next blood vessel (SV), b) the mean length of all vectors originating from this pixel (MV), and c) the longest vector to the next blood vessel (LV). For each subgroup the number of pixels above a given vector length was counted: SV>2, MV>6, and LV>14 pixels. The area under the ROC-curve (ROC-AUC) was calculated for each approach.
Results :
Mean (group 1-3) or median (group 4) values were significantly different between healthy and ischemic retinae for all approaches (p<0.001; Wilcoxon-Mann-Whitney Test). Discrimination between healthy and ischemic was best in the shortest-vector-length group (4a) with a ROC-AUC of 0.95. A specificity of 100% was achieved at 80% sensitivity for this test. The ROC-AUC of all other approaches was (in declining order): total blood vessel length (0.947), mean vector length (0.941), longest vector length (0.938), number of pixels above a threshold grey level of 50 (8 bit; 0.918), and mean grey value (0.906).
Conclusions :
All approaches produced significantly different results between healthy and ischemic retinae. The shortest-vector approach showed the best discrimination.
This is an abstract that was submitted for the 2018 ARVO Annual Meeting, held in Honolulu, Hawaii, April 29 - May 3, 2018.