3D-printed phantoms for measuring lateral resolution and contrast performance of ophthalmic adaptive optics imaging systems

Abstract

Adaptive optics (AO) imaging of the human retina is an emerging clinical technique that confers the highest possible spatial resolution of this tissue in vivo. To support consistent imaging performance across time and devices, we have designed and fabricated a robust and user-friendly phantom that can determine lateral resolution and contrast with 3D-printed microstructures approximating cone photoreceptors. We have carefully characterized multiple copies of this phantom via microscopic imaging and metrology, and we also demonstrated the phantom's utility with two different AO imaging modalities. Our results indicate that this phantom, with its associated ground truth data, can be a very effective qualitative and quantitative evaluation tool for ophthalmic AO devices.

Fig. 1.

Phantom design. (a) 3D rendering of phantom arrays and bars, with insets showing detailed architecture within three individual arrays. (b) 2D map of phantom structures with exact dimensions. The numbers next to each ‘A’ indicate the design value of c-c spacing of that array. The three arrays detailed in (a) are identified by the corresponding-colored arrows.

Fig. 2.

3D printing with DLW-2PP. (a) Cartoon of DLW-2PP process. (b) Graphical simulation and (c) camera snapshots through the objective lens of steps during the printing of one phantom array. Scale bars are 20 μm.

Fig. 3.

Model eye. (a) Front side, showing lens mounted in the center. (b) Back side, showing back of glass diffuser mounted in the x-y translator.

Fig. 4.

FFT analysis of simulated phantom array images with 5 µm c-c spacing. Intensity images (a,c) and normalized 2D-FFT amplitude (b,d) without blurring (a,b) and with Gaussian blurring (c,d). Line profiles through the intensity images at the dashed lines in (a,c) are shown in (e), and line profiles through the positive horizontal spatial frequencies of the FFT amplitude images at the dashed lines in (b,d) are shown in (f). 10% Michelson contrast is calculated for the line profile through the blurred image, which corresponds to a normalized FFT amplitude of ∼0.06 at the 0.2 µm^-1 spatial frequency peak.

Fig. 5.

Scanning electron microscopy of phantoms. (a) P2. (b) P1-A9. (c) P2-A4. (d) P3-A2.75. Colored squares are to help indicate location in the overall phantom layout, but (b) and (d) show images from the other two phantoms. Scale bars are 100 µm in (a) and 10 µm in (b)-(d). Yellow arrows point to fabrication flaws: deformation in (b), gap in (d).

Fig. 6.

Reflected light microscopy of phantom arrays. (a) P1-A9 imaged with the focus at the micromirror plane. (b) P1-A9 imaged with the focus moved down to the glass surface, showing strong reflections of the glass through the array because of incomplete contact. (c,d) Images of P1-A3.5 focused at the (c) micromirror plane and (d) glass surface show no glass reflections within the array, indicating complete contact with the glass. (e) P2-A4. (f) P3-A2.75. (a), (e), and (f) correspond to (b)-(d) in Fig. 5. The same fabrication flaws are pointed out with yellow arrows. All scale bars are 10 µm. (g) Micromirror brightness coefficient of variation (CV) for all arrays from all phantoms.

Fig. 7.

Center-to-center spacing within all arrays of all three phantoms. (a) Mean measured spacing vs. design spacing for each array. Error bars represent one standard deviation of the measurements. (b) Horizontal and vertical spacing errors vs. design spacing.

Fig. 8.

Adaptive optics imaging of phantom with 1° field of view. (a,c,e) AO-SLO. (b,d,f) AO-OCT. (a,b) Innermost 9 arrays and 5 bars of P3. (c,d) P1-A5 (lower left) and -A9 (upper right). (e,f) P2-A4.5 (lower right) and -A8 (upper left). Colored squares indicate corresponding arrays across the images. Scale bars are 20 µm.

Fig. 9.

Contrast vs. c-c spacing plots with images from points of discrepancy. (a) AO-SLO and (b) AO-OCT contrast measured from all arrays in all three phantoms. (c) AO-SLO and (d) AO-OCT contrast separated by horizontal and vertical directions for the smallest nine arrays, averaged over the three phantoms. (e) AO-SLO contrast averaged over the three phantoms at two time points. The insets show images of P3-A6 at the two time points. Error bars in (c-e) represent one standard deviation. (f,g) SEM and AO-OCT images of P3-A9 and P2-A9, corresponding to the arrowed points in (b). (h,i,j) AO-SLO images of P3-3.5, P2-4, and P2-4.5, corresponding to arrows in (c).