Methodologies for Analyzing Retinal Nerve Fiber Layer Thickness/Area Using Spectral Domain Optical Coherence Tomography.

Purpose: Glaucoma is a group of optic neuropathies that can lead to irreversible vision loss, especially when not treated. Having an unknown cause, the diagnosis or progression is dependent on accurate clinical measures of structural and functional signs of the disease. The most common quantifiable retinal structure is the peripapillary retinal nerve fiber layer (RNFL), a surrogate for the total ganglion cell content within the eye. With technologies such as spectral domain optical coherence tomography (SD OCT), the RNFL thickness can be quantified with resolutions approaching 5 µm. Although these measures have been shown to be reliable, the accuracy of thickness measures is related to ocular biometry and the retinal scan path, which have not been investigated systematically. The purpose of this dissertation was to develop methodologies that will improve the precision and accuracy of SD OCT measures of retinal structures. These methods were then applied to investigate morphological changes in the non-human primate ocular hypertensive glaucoma model. Methods: 1) For the first experiment, 16 infant rhesus monkeys were scanned with an SD OCT system, at regular intervals, starting at around 36 ± 6 days of age until 628 ± 27 days. Changes in ocular magnification were investigated using both image registration and a three surface schematic eye. The total retinal thickness, and pit morphology at each scan session were quantified using a semi-automated segmentation algorithm. Individual layers of the retina were manually identified with the aid of reflectivity profiles through the region of interest. 2) The second experiment investigated the influence of anterior segment power change, induced by soft contact lenses, on ocular magnification and RNFL scan path. 15 healthy human subjects were scanned with soft contact lenses ranging from -12 to +8 D in 2 D steps. Changes in ocular magnification were investigated using both image registration and the schematic eye. RNFL thickness was quantified using a custom semi-automated segmentation algorithm that also compensated for major retinal vasculature. RNFL area measures were calculated by multiplying thickness measures from each A-scan by its calculated width. 3) The influence of ocular biometry, and segmentation algorithms on RNFL measures were further investigated in 45 human eyes scanned with two clinical SD OCT instruments. 4) The scaling and RNFL segmentation algorithms developed were applied to investigations of scan path distance from the optic nerve head (ONH) rim margin and RNFL thickness measures. For this experiment, RNFL thickness and area, with and without blood vessel subtraction were measured for elliptical scan paths 300-600 µm from the rim margin in 40 normal rhesus monkeys. 5) For the fifth experiment, RNFL thickness measures from an elliptical scan path 550 µm from the rim margin was investigated along with ONH morphology and total macula volume in 6 ocular hypertensive experimental glaucoma animals. Results: 1) Transverse scaling based on schematic eye calculations correlated with axial length related changes in retinal image size for the 16 infant animals followed longitudinally (R2 = 0.88, slope = 0.98, p < 0.01). Most changes in total retinal thickness was attributable to the outer retinal layers, and were complete around 120 days of age. The Nyquist limit calculated from outer nuclear layer thickness was in agreement with previously published histological data. 2) In human subjects, the region of the retina scanned increased with increase in contact lens power (R2 = 0.94, p < 0.01), and was accurately modeled with the schematic eye (R2 = 0.97, p < 0.01). The relationship between contact lens power and RNFL thickness was attributed to differences in the length of the scan path. When the scan circumference was incorporated, RNFL area was not related to contact lens power (R2 = 0.003, p = 0.47). 3) For the 45 eyes included, there were significant differences in instrument derived RNFL thickness between the two clinical instruments used (mean difference = 6.7 ± 4.8 µm, ICC = 0.62). However, when the same scan path and segmentation algorithm were used, these differences were reduced (mean difference = 0.1 ± 3.1 µm, ICC = 0.92). Global RNFL thickness from both the instrument derived and the custom segmentation were linearly related to axial length (slope = -3 µm/mm, R2 = 0.24, p < 0.01). However, when transverse scaling was incorporated, RNFL area was not related to axial length (R2 = 0.004, p = 0.69). 4) For the 40 eyes of rhesus monkeys scanned, RNFL thickness decreased with increase in scan distance from the ONH rim when custom elliptical scans were used (R2 = 0.61, p < 0.01). In contrast, global RNFL area did not change over this same region (R2 = 0.002, p =0.49). The major retinal vascular contribution accounted for 9.3% of global measures, and was similar for all scan locations. 5) For ocular hypertensive animals, the RNFL area did not change until after the ONH neural rim volume decreased by about 60%. While the percent vascular contribution to the RNFL increased (R2 = 0.64, p < 0.64), the overall vessel area decreased (R2 = 0.33, p < 0.01) with disease progression. Conclusion: Measurements of RNFL thickness by SD OCT are dependent on the optics of the eye, including both anterior segment power and axial length. The results suggest that the RNFL cross sectional area after compensation for major retinal vasculature is an accurate surrogate for retinal ganglion cell axonal content within the eye. As morphological changes in the ONH precede RNFL thinning, it is important to assess both in determining glaucomatous disease and disease stage.Optometry, College o