Proper Registration of Macular and Peripapillary OCT Scans Provides Reliable Correlation between Macular and PP Retinal Layer Thickness along Corresponding Nerve Fiber Bundle Directions

Abstract

 

Purpose:

 

The purpose of this study was to determine whether the correspondence between RNFL and GCL retinal layer thicknesses and their macular and peripapillary (PP) locations along nerve fiber bundles can be improved by proper registration of macular and PP OCT images. Two techniques were studied - a conventional approach, in which PP and macular regions are associated to each other without their anatomic registration, and a novel registration approach in which directions of fiber bundles are specifically considered.

 

Methods:

 

20 macular and 20 PP OCT scans (Fig. 1(a), 200 × 200 × 1024 voxels, 6 × 6 × 2 mm³) were obtained from both eyes of 10 glaucoma patients using a Cirrus^TM HD-OCT machine (Carl Zeiss Meditec, Inc., Dublin, CA). In the conventional approach, macular and PP layer thickness correlations were derived from the mean thicknesses of RNFL and GCL at regions 1 - 6 and at regions α and β (1 vs. α, 2 vs. α , 3 vs. α , 4 vs. β, 5 vs. β and 6 vs. β) as depicted in Fig. 1(c). In the registration approach, macular and PP OCT images were semi-automatically registered to form a single 3D dataset as shown in Fig. 1(b). Macular and PP layer thickness correlations were determined from the mean thicknesses of corresponding RNFL and GCL at regions depicted in Fig. 1(d).

 

Results:

 

When employing the conventional approach, the RNFL and GCL layer thicknesses in the macular and PP locations have different correlation coefficients (R) depending on the scanning position. Fig. 2 shows the variation of R values by 0°, 5°, 10° rotations of PP scans. The R values obtained by the registration-based approach are 0.68 ± 0.11 for the RNFL layer thickness and 0.80 ± 0.09 for the GCL layer thickness.

 

Conclusions:

 

The registration-based approach provides higher RNFL and GCL retinal layer thickness correlations along the nerve fiber bundles and seems therefore a preferred approach when determining associations between macula-based and ONH-based retinal layer properties.  

 

Keywords: imaging/image analysis: clinical • image processing • retina