Glaucoma causes irreversible damage to the retinal ganglion cells and its axons [1
]. Axons of the ganglion cells aggregate in arcuate bundles at the inferior and superior poles of the optic nerve head. These axons disappear during glaucoma progression, causing a characteristic retinal nerve fiber layer thinning at the retinal poles, which can be diagnosed and monitored with retinal imaging systems such as scanning laser ophthalmoscopy (SLO), scanning laser polarimetry (SLP) and optical coherence tomography (OCT).
OCT is an established non-contact method for retinal imaging at unprecedented axial resolution and sensitivity [2
]. Earlier time-domain OCT modalities, however, did not have sufficient accuracy to detect the subtle thinning of the retinal nerve fiber layer that occurs in early-stage glaucoma [4
]. With the introduction of spectral-domain OCT [5
], retinal tomograms could be acquired two-orders of magnitude faster without loss in sensitivity, enabling video-rate OCT retinal imaging [6
]. Intensity-based imaging with SD-OCT has proven highly successful, yet its effectiveness for detecting early-stage glaucoma is challenged by the subtle thickness changes and compositional changes of the retinal nerve fiber layer that accompany the disease. The latter is evident even in healthy eyes in which the glial content of nerve fiber bundles is known to vary significantly (18% to 42% of the cross sectional area in primates) [7
]. Such variation cannot be distinguished with intensity-based SD-OCT.
Polarization-sensitive OCT (PS-OCT) enables simultaneous intensity and depth-resolved measurements of the polarization state in turbid media [8
]. Microtubules in the retinal nerve fiber layer (RNFL) – unlike glial cells– are intrinsically birefringent, and cause a phase retardation in reflection [13
]. PS-OCT measures the double pass phase retardation and nerve fiber thickness simultaneously. It can therefore be used to determine the double pass phase retardation per unit depth (DPPR/UD), which is proportional to the birefringence [14
]. Subsequent to the development of SD-OCT, polarization-sensitive SD-OCT systems were developed for retinal imaging. Studies with these systems on more than 10 healthy subjects in three different labs demonstrated that the RNFL birefringence varies as a function of location around the optic nerve head, indicating a spatial variation in microtubule density [14
]. Lowest values of 0.1°/µm were found temporal and nasal to the optic nerve head, while higher values were found inferior and superior, around 0.35°/µm. Similar results were obtained in a measurement that combined results from OCT and scanning laser polarimetry [21
]. Preliminary measurements on three glaucoma patients indicated that the disease causes a reduction in retinal nerve fiber layer tissue birefringence around the optic nerve head [19
]. PS-OCT has also been used to determine the degree of polarization uniformity (DOPU) of scattering tissue such as the retinal pigment epithelium (RPE) [23
]. A change in the DOPU can be attributed to a change of the fast axis orientation over a short distance, which is most likely caused by scattering. A change in phase retardation over a small distance also causes a change in DOPU, but these changes are relatively small due to the low birefringence of retinal tissue.
PS-OCT is not without limitations. Quantitative RNFL birefringence measurements with PS-OCT have been restricted to the area surrounding the optic nerve head, where the RNFL is thick (approximately 50 µm to 250 µm) and provides a large DPPR signal (up to ~70°) for birefringence measurements. Regions of thin RNFL (<50 µm), such as the macula lutea, have significantly less birefringent tissue and require more sensitive detection. In part because of this reason, birefringence measurements of these thinner regions have yet to be reported with PS-OCT. Another limitation of PS-OCT is its relatively coarse spatial resolution. This limitation stems from two factors. First is the requirement of a small pupil (< 2 mm) so as to optimally balance blur caused by diffraction and ocular aberrations, i.e., maximize image quality. Second, the birefringence is small, and the measurement of small changes in the polarization state requires averaging of Stokes vectors to reduce the influence of speckle noise. The combined effect is a relatively coarse spatial resolution that prevents PS-OCT from probing highly localized changes of birefringence that might exist on a microscopic level in the retina, for example variations in birefringence of adjacent retinal nerve fiber bundles.
The integration of AO into PS-OCT represents a potential solution to these problems. AO permits access to the full retinal reflection that exits a large pupil of the eye (>6 mm). This translates into higher lateral resolution at the plane of focus (~3 µm), a smaller lateral speckle size throughout the entire volume image (~3 µm), and higher collection efficiency for light backscattered from the retina (~6 dB).
The purpose of this paper is to demonstrate the benefit of AO for PS-OCT measurements. To this end, we integrated polarization-sensitive imaging into an existing AO-OCT system. Two different beam sizes were used for imaging the same patch of retina. For AO-PS-OCT measurements a 6.0 mm pupil diameter was used and the beam diameter was decreased to 1.2 mm to mimic a standard PS-OCT system without AO. Differences in signal-to-noise ratio (SNR) were quantified and Stokes vectors were analyzed to determine the benefit of AO for PS-OCT. The speckle size in the human retina was compared for these two systems, using an autocorrelation method. AO-PS-OCT measurements were then performed on the retinal nerve fiber in the macula lutea to quantify the birefringence of thin retinal nerve fiber bundles that cannot be measured with standard PS-OCT. Furthermore, the fast axis orientation and standard deviation of the fast axis orientation were determined and compared with the degree of polarization uniformity (DOPU) in the RPE [27