Fig 1.
A violin plot of RNFL thickness distribution of Gnorm—normal eyes, GOHT—OHT eyes, Gpre—pre-perimetric glaucoma, Gper—perimetric glaucoma.
The white dot represents the median value of the RNFL thickness in each group, the thick gray bar in the center represents the interquartile range, while the thin gray line shows the rest of the distribution, except outliers. On each side of the gray line is a kernel density estimation to show the histogram shape of the RNFL data. Wider sections of the violin plot represent a higher histogram probability value and the thinner parts represent a lower probability.
Fig 2.
a) an original mean image calculated from a video sequence after frame-to-frame registration, b) binary mask of segmented blood vessels, c) binary mask of ONH.
Fig 3.
a) PAA(x,y) spatial distribution, b) PAA(x,y) masked by the blood vessels (black areas). The colorbar shows PAA value. The green cross depicts the center of ONH and the red ring the approximation of the ONH border.
Fig 4.
An example of PAA spatial map with masked blood vessels and depiction of the analyzed region in a) 360° circular area (PAA360), b) temporal semi-circular area (PAAtemp), and c) nasal semi-circular area (PAAnasal). Blood vessels are excluded from calculation of the PAA parameter.
Fig 5.
Image showing the annulus pattern in the PAA map and the meaning of the relative parameters wr (relative width) and dr (relative distance).
The blue color depicts the parameters. The red arrow indicates the ONH radius for a particular eye, normalized to 1 (red circle). Different situations can occur when searching for the optimal values of the proposed parameters. When dr<r, the annulus is inside the ONH, while when dr>r, the annulus is placed outside the ONH region. Parameter wr controls the thickness of the annulus, which influences how much of PAA surface will be considered for calculation.
Fig 6.
Values of correlation coefficient as a result of full-search optimization for optimal parameter settings.
Left: Images showing the result of full-search computation maximizing the average value of R; the lightest areas correspond to the highest values of R. Right: The values of R drowned in isolines as a contour plot. The most yellow color depicts the region with approximately 95% of maximum correlation. Results are shown for a) the 360° area, b) the temporal area, and c) the nasal area.
Table 1.
Comparison of estimated optimal parameter values and correlation coefficients.
Fig 7.
Statistical relation between RNFL thickness and PAA values for each eye in the dataset.
Values of PAA were obtained with the optimal parameters setting (Table 1) maximizing the correlation coefficient. The graphs show particular relations between RNFL thickness and a) PAA360 (R = 0.530), b) PAAtemp (R = 0.557), c) PAAnasal (R = 0.332), d) PAAONH (R = 0.459).
Fig 8.
Boxplots showing distribution of the values of mean RNFL thickness (red color) and PAA (blue color) in individual glaucoma groups (Gnorm-Gper) for all calculated areas of interest.
a) 360° circular area, b) temporal semi-circular area, c) nasal semi-circular area, and d) entire ONH area. The values were calculated with the optimal parameter setting as shown in Table 1.