Purpose:
Spectral domain optical coherence tomography (SD OCT) can provide highly reliable estimates of the shape of the optic nerve head (ONH), which are useful for diagnosing and managing glaucoma and other optic neuropathies. Stereo color photographs are a lower cost, widely used imaging modality that has lacked objective assessment. We validated our automatic algorithm for computing depth maps from stereo color images of the optic disc by comparing them to SD OCT images obtained on the same day.
Methods:
30 pairs of fixed-base stereo color photographs and SD OCT images of the ONH were obtained from both eyes of patients with glaucoma. Dense correspondences between stereo images were found automatically using a scale space based approach in the presence of spatially-varying reflectance, limited illumination, noise, and low contrast or density of the features, which are commonly observed in stereo optic disc photographs. Horizontal disparities between these correspondences formed the disparity map reflecting detailed structures of the ONH, such as cup size, cup depth, neuroretinal rim slope and curvature of the retinal nerve fiber layer (RNFL). For performance evaluation purposes, surfaces of the retinal layer were detected in the raw OCT volume using 3D segmentation. Depth information was recorded as intensities and registered with the reference stereo images to provide the ground truth. The method was validated quantitatively by comparing the same shape obtained from SD OCT images.
Results:
The root mean squared (RMS) difference between normalized structures was 0.1592 (95% CI 0.1264 - 0.1920).
Conclusions:
Automatically computed depth maps from stereo color images of the optic disc faithfully reproduce the shape of the ONH as confirmed by comparing them to SD OCT images. Stereo color imaging of the ONH combined with automated depth mapping is a low cost alternative for SD OCT that has potential for more cost-effective diagnosis and management of glaucoma and other optic neuropathies.
Keywords: image processing • imaging/image analysis: non-clinical • imaging methods (CT, FA, ICG, MRI, OCT, RTA, SLO, ultrasound)