236
K. K. Vupparaboina et al.
9.3.2.3 Choroid Localization and Binarized Choroid
At this point, localizing the choroid is left, so that only the choroid layer could be
binarized in exclusion of other layers such a retina and sclera. To this end, choroid
inner boundary (CIB), and choroid outer boundary (COB) is obtained by employing
the SSIM-based methodology described in Sect. 9.2 on raw OCT scans (Fig. 9.15h)
[26]. Subsequently, thus obtained boundaries are used to obtain desired choroid
region from the previous binarized estimate (Fig. 9.15i).
9.3.3 Stromal-Lumial Analysis: Experimental Results
In this section, performance of Vupparaboina et al.’s methodology is presented visà-vis earlier ImageJ-based protocol for different subjects is presented. Subsequently,
Vupparaboina et al.’s algorithm is applied for stromal-luminal analysis of OCT volume data.
9.3.3.1 B-Scans
As alluded earlier, the absence of reference (ground truth) procedure does not allow
quantitative comparison. So the comparison was made visually, and based on expert
opinion. To this end, SD-OCT B-scans acquired from multiple subjects are considered. Those images were analyzed using ImageJ-based protocol reported by Sonoda
et al. [16], as well as the Vupparaboina et al.’s method. While Sonoda et al. manually
extracted the choroid layer, to ensure fair comparison, the same choroid localization employed by Vupparaboina et al.’s method was also used in conjunction with
the ImageJ method. Results for six B-scans from six different subjects (A–F) is
depicted in Fig. 9.16. Results obtained by Vupparaboina et al.’s method appear to
closely resemble intuitive estimation of vasculature. Indeed, subjective interpretation
by multiple clinicians found results of Vupparaboina et al.’s method superior.
9.3.3.2 Volume Analysis
Next volumetric stromal-luminal analysis is performed on each of the eyes of two
healthy subjects (subjects A and B). In particular, 97 OCT B-scans with uniform
separation of 30 µm were acquired from each eye of the subject and evaluated
using the proposed algorithm. Further, bright-dark ratios (ratio between stromal and
luminal regions) were estimated for all the B-scans. Figure 9.17 depicts variation in
bright-dark ratios with scan index for each eye under consideration. Further, overall
bright-dark ratio of each eye is quantified, by first estimating volumes of luminal and
stromal regions. Such volume estimates were in turn obtained using cubic interpolation. Table 9.2 presents volume bright-dark ratios for each eye of the two subjects.
K. K. Vupparaboina et al.
9.3.2.3 Choroid Localization and Binarized Choroid
At this point, localizing the choroid is left, so that only the choroid layer could be
binarized in exclusion of other layers such a retina and sclera. To this end, choroid
inner boundary (CIB), and choroid outer boundary (COB) is obtained by employing
the SSIM-based methodology described in Sect. 9.2 on raw OCT scans (Fig. 9.15h)
[26]. Subsequently, thus obtained boundaries are used to obtain desired choroid
region from the previous binarized estimate (Fig. 9.15i).
9.3.3 Stromal-Lumial Analysis: Experimental Results
In this section, performance of Vupparaboina et al.’s methodology is presented visà-vis earlier ImageJ-based protocol for different subjects is presented. Subsequently,
Vupparaboina et al.’s algorithm is applied for stromal-luminal analysis of OCT volume data.
9.3.3.1 B-Scans
As alluded earlier, the absence of reference (ground truth) procedure does not allow
quantitative comparison. So the comparison was made visually, and based on expert
opinion. To this end, SD-OCT B-scans acquired from multiple subjects are considered. Those images were analyzed using ImageJ-based protocol reported by Sonoda
et al. [16], as well as the Vupparaboina et al.’s method. While Sonoda et al. manually
extracted the choroid layer, to ensure fair comparison, the same choroid localization employed by Vupparaboina et al.’s method was also used in conjunction with
the ImageJ method. Results for six B-scans from six different subjects (A–F) is
depicted in Fig. 9.16. Results obtained by Vupparaboina et al.’s method appear to
closely resemble intuitive estimation of vasculature. Indeed, subjective interpretation
by multiple clinicians found results of Vupparaboina et al.’s method superior.
9.3.3.2 Volume Analysis
Next volumetric stromal-luminal analysis is performed on each of the eyes of two
healthy subjects (subjects A and B). In particular, 97 OCT B-scans with uniform
separation of 30 µm were acquired from each eye of the subject and evaluated
using the proposed algorithm. Further, bright-dark ratios (ratio between stromal and
luminal regions) were estimated for all the B-scans. Figure 9.17 depicts variation in
bright-dark ratios with scan index for each eye under consideration. Further, overall
bright-dark ratio of each eye is quantified, by first estimating volumes of luminal and
stromal regions. Such volume estimates were in turn obtained using cubic interpolation. Table 9.2 presents volume bright-dark ratios for each eye of the two subjects.