Retinal adaptive optics imaging with a pyramid wavefront sensor

Abstract

The pyramid wavefront sensor (P-WFS) has replaced the Shack-Hartmann (SH-) WFS as the sensor of choice for high-performance adaptive optics (AO) systems in astronomy. Many advantages of the P-WFS, such as its adjustable pupil sampling and superior sensitivity, are potentially of great benefit for AO-supported imaging in ophthalmology as well. However, so far no high quality ophthalmic AO imaging was achieved using this novel sensor. Usually, a P-WFS requires modulation and high precision optics that lead to high complexity and costs of the sensor. These factors limit the competitiveness of the P-WFS with respect to other WFS devices for AO correction in visual science. Here, we present a cost-effective realization of AO correction with a non-modulated P-WFS based on standard components and apply this technique to human retinal in vivo imaging using optical coherence tomography (OCT). P-WFS based high quality AO imaging was successfully performed in 5 healthy subjects and smallest retinal cells such as central foveal cone photoreceptors are visualized. The robustness and versatility of the sensor is demonstrated in the model eye under various conditions and in vivo by high-resolution imaging of other structures in the retina using standard and extended fields of view. As a quality benchmark, the performance of conventional SH-WFS based AO was used and successfully met. This work may trigger a paradigm shift with respect to the wavefront sensor of choice for AO in ophthalmic imaging.

Fig. 1.

Scheme of the AO-OCT system with a non-modulated P-WFS and a SH-WFS. For simplification only one galvanometer scanner and only 2 of the 4 beams created by the P-WFS are drawn. PC: polarization controller, Pol.: Polarizer, (Pol.) BS: (polarizing) beam splitter. The dashed blue and yellow lines mark pupil and focal planes of the system, respectively

Fig. 2.

Cut outs of the four pupil plane images produced by the P-WFS without scanning in a) and with scanning in b). The images were obtained in a model eye under the presence of system aberrations and a slight defocus (overall wavefront error RMS 270 nm, measured with the SH-WFS).

Fig. 3.

Data pipeline implemented for the non-modulated P-WFS depicted in a): Cut outs of the four pupil plane images obtained from the P-WFS after background subtraction and intensity thresholding are shown in b), followed by the output of the digital binning routine (set amount: 10 × 10 pixels) in c). Only the illuminated pixels within the studied pupil in d) are used according to the pupil image numbering for the computation of the two data maps in e) via the standard approach of slope-like P-WFS data definition. The data was obtained in vivo.

Fig. 4.

P-WFS calibration data for computation of the sensor response to 20% stroke at actuator Nr. 25 (left box) and target computation for closed-loop focus shifting with the P-WFS (right box). The reference P-WFS data maps are shown in a) next to the absolute P-WFS response maps in b) and the relative P-WFS response maps in c). The P-WFS data maps measured for a defocus wavefront of 0.4 µm root mean square (RMS) amplitude are shown in d), followed by the target data maps computed through a linear polynomial fit of the measured data in e) and the polynomial fitting error in f). The defocus wavefront was measured with the SH-WFS, where g) shows the wavefront estimate and h) the corresponding Zernike decomposition (including the first 15 modes according to the Noll definition [42], except the piston mode).

Fig. 5.

In vivo AO-OCT images of cone photoreceptors obtained with AO correction based on the P-WFS. The representative images were retrieved from a single data volume recorded at the fovea of a healthy volunteer (28 years, female, left eye). The field of view is approximately 0.94° x 0.99°. The en-face projection in c) was created by depth integration over the cone photoreceptor layers and is accompanied by the 2D Fourier transform (FFT) which shows a Yellott’s ring. The radius of the Yellott’s ring indicates the spatial frequency of the cone mosaic that corresponds to the row to row spacing of the cones in the imaged areas. The white arrow indicates the approximate location of the fovea centralis (estimated by the highest density of cones). The blue dashed line highlights the location of the single B-scan shown in b), where the yellow box highlights a single cone photoreceptor and the red arrows mark the limits of the en-face integration range. A sketch of the outer retinal bands as known from histology is provided in c).

Fig. 6.

Different cell types in the outer retinal layer recorded at an eccentricity of 14° temporal / 6° superior in a healthy volunteer (29 years, male, right eye) with a FoV of 0.95° x 1° using AO-OCT with the P-WFS. The en-face images, a) to e), were obtained from a single data volume by integration over different depth ranges in the outer retinal band and have the respective 2D FFTs as an inset. The averaged B-scan in g) shows a projection of the entire volume along the slow scanning direction with the arrows indicating the integration ranges for a) to e). For the following en-face projections, specific retinal cell types can be identified as main contributor in terms of reflected signal: a) junction between inner and outer segments of cone photoreceptors (IS/OS), b) cone outer segment tips (COST), c) rod outer segment tips (ROST), d) retinal pigment epithelium cells (RPE). The structure in e) presumptively corresponds to the distal part of the RPE or the Bruch’s membrane (dRPE / BM). The composite in f) is a false color image of COST (red) and ROST (green). In the single B-scan in h), only the outer retinal bands are shown and a selection of the hyper-reflective spots, which form the cell mosaics in a)-d) and the structure visible in e), is marked. The location from which h) was extracted is highlighted in the en-face images with a blue dashed line

Fig. 7.

Different structures of the inner the retina visualized for a healthy volunteer (29 years, male, right eye) at 14° temporal / 6° superior with AO-OCT using the P-WFS. The focus was set to the nerve fiber layer and the FoV is 0.82° x 0.78°. The representative images are extracted from a volume obtained by averaging 12 registered volumes that were recorded within ∼40 sec. The en-face images a)-h) were obtained by depth integration over parts of or in vicinity of the following retinal layers: a) inner limiting membrane (ILM), b) nerve fiber layer (NFL), c) ganglion cell layer (GCL), d)-g) inner plexiform layer (IPL), h) outer plexiform layer (OPL). The integration ranges are indicated by color-coded arrows in the averaged B-scan in i) which was obtained by projection of the full data volume in the slow imaging direction. The dashed lines in the en-face images mark the location of the 5 adjacent B-scans averaged in j) in the data volume. k) is a sketch of the retinal layers and cells as known from histology

Fig. 8.

Quantitative comparison of the P-WFS and the SH-WFS by retrieving AO imaging performance metrics from en-face AO-OCT images of the cone photoreceptor mosaic recorded in the central fovea of healthy volunteers: a) shows the common regions of interest (RoIs) defined as close as possible to the fovea centralis. For the RoIs, histograms in b) and power spectra (2D FFT) in c) are obtained. The power spectra are radially averaged over the angles marked in c) and fitted with a linear polynomial in d). The final performance metrics are the standard deviation of the RoIs provided by the histograms and the values obtained by integrating the difference between the radial power spectrum averages and the respective fits over the spatial frequency ranges Range 1 and Range 2.

Fig. 9.

Comparison of the P-WFS and SH-WFS sensitivity. The ratio between the residual wavefront RMS error after AO loop convergence (measured with the SH-WFS at 40 ms exposure time) and the benchmark RMS value is plotted for different exposure times. Several intensity thresholds are considered for the center of gravity computation of the SH-WFS focal spots and the P-WFS pupil image masking. The benchmark RMS value corresponds to the residual wavefront RMS value after AO loop convergence obtained for the respective sensor setting at 40 ms exposure time. The dashed horizontal line marks the performance required for diffraction limited imaging according to the Maréchal criterion.