Fig 1.
OCTA en face images of the superficial and deep retinal layers of one healthy eye.
The image order from left to right is: Original image, binarized images using the manual thresholding and the Huang, Li, Otsu, Moments, Mean, Percentile algorithms.
Table 1.
Mean ± standard deviation of binarization thresholds, vessel densities and skeleton densities obtained using the different algorithms for images of the superficial and deep retinal layers.
Fig 2.
Inter-algorithm repeatability: Relative Vessel Density (a) and Vessel Skeleton Density (b) measurement differences obtained using a manual approach and six automated algorithms to binarize images of the superficial and deep retinal layers. Some outliers are not shown due to the scaling of the diagram. Values over 1.5 interquartile range below the first quartile or above the third quartile were defined as outliers; the horizontal lines indicate the 1st quartile, median and 3rd quartile. VD = Vessel Density, VSD = Vessel Skeleton Density.
Fig 3.
ROC curves (healthy versus macular pathology) of the binary logistic regression formulae based on the respective Vessel Density and Vessel Skeleton Density as well as age obtained by the manual approach and the automated algorithms in the superficial (a, c) and deep retinal layer (b, d).
Table 2.
Additional repeatability parameters.
Table 3.
Results of ROC analysis of binary logistic regression formulae based on the respective Vessel Density and Vessel Skeleton Density as well as age in the superficial and deep retinal layers per algorithm.
Table 4.
Principles of the binarization methods used in this study.