Abstract
Purpose: :
Develop an automatic method for adjusting retinal layer thickness maps from SD OCT based on the presence of retinal vessels. Determine how such a correction affects the measured thickness variability of the macular nerve fiber layer (MNFL) and ganglion cell layer (GCL) across normal subjects.
Methods: :
28 macula-centered 3-D OCT volumes (200×200×1024 voxels, 6×6×2 mm3) were obtained from the right eye of 28 normal subjects (mean age of 41.0 ± 8.9 years) using a CirrusTM HD-OCT machine (Carl Zeiss Meditec, Inc., Dublin, CA). Seven surfaces were automatically segmented in 3-D using an extended version of our previously reported algorithm (Haeker et al., IPMI 2007: 207-218). 2-D thickness maps were created for the MNFL and GCL (based on surfaces 1, 2, and 3). The vessels were automatically segmented from each OCT volume on an extracted projection image near the RPE plane (Niemeijer et al., ARVO 2008). Vessel regions were temporarily "removed" from the computed thickness maps. Interpolation of surrounding thickness values into these "removed" vessel regions resulted in a vessel-corrected thickness map. The mean and standard deviation of the thickness values across subjects were statistically compared between the original thickness and vessel-corrected thickness maps.
Results: :
The standard deviation decreased from 8.6 to 7.9 µm (p < 0.001) for the MNFL thickness and from 10.3 to 9.2 µm (p < 0.001) for the GCL thickness. The standard deviation decreased from 14.0 to 12.4 µm (p < 0.001) for the MNFL thickness and from 11.6 to 9.3 µm (p < 0.001) for the GCL thickness in specific locations where vessels were located.
Conclusions: :
Correcting for vessel locations can significantly reduce the measured thickness variability of macular retinal layers in normal subjects. Abnormal thickness of these layers has been shown to correlate with glaucoma and diabetic retinopathy, and thus, any corrections that narrow the normal range will provide more precise normative values for diagnosis and management.
Keywords: image processing • imaging/image analysis: clinical • imaging/image analysis: non-clinical