Imaging of human parafoveal area with large field of view in adaptive optics line scanning ophthalmoscope

1. Introduction

The foveal region is responsible for providing the highest resolution vision and contains the highest density of photoreceptors and the finest capillaries in the retina. The photoreceptors are the origin of vision sense, and the surrounding parafoveal capillary network maintains the health and function of the photoreceptors. Lesions can occur in the foveal region at the early stages of retinal or systemic diseases, which makes the imaging of the foveal region critical for studying and diagnosing various retinal diseases and disorders.^1,2 Various tools have been used to study the foveal region in vivo, including high contrast entoptic view, fluorescein angiography (FA), and optical coherence tomography angiography (OCTA),^3,4,5,6 but the resolution was limited by the ocular aberrations.^7,8,9 Combined with adaptive optics, the adaptive optics scanning laser ophthalmoscope (AOSLO) system allows observation of structural and functional details of the parafoveal area at the cellular or subcellular level.^10,11 This advanced imaging system holds great potential in studying the development of retinal diseases and assessing the effectiveness of treatments.^12,13,14

However, due to limitations of the two-dimensional scanning mechanism in the AOSLO system, the fastest scanning mirror worked at a frequency of 16KHz, which resulted in a dilemma between the field of view (FOV) and the imaging speed.¹⁵ Researchers have developed one-dimensional line scanning methods to address this challenge and strike a balance between imaging speed and FOV. Mujat et al.¹⁶ presented a compact adaptive optics line scanning ophthalmoscope (AOLSO) with 5.5^∘ FOV and a 15Hz frame rate. Zhang et al.^17,18 described a high-speed AOLSO with a frame rate of 100–800Hz, and a FOV in the range of 0.3–1.2^∘ to noninvasively characterize erythrocyte movement in human retinal capillaries. Therefore, the AOLSO allowed for high resolution of the retina imaging with a larger FOV or higher speed, but still cannot simultaneously achieve a large FOV and high speed imaging.

To image the retinal parafoveal area with a large FOV and high speed, at the same time, we build an adaptive optics line scanning ophthalmoscope. The nonplanar optical path is optimized to compensate for the astigmatism introduced by the deflection angle between the mirrors, achieving diffraction limitation across the entire FOV of 5^∘. Two cylinder lenses are used for sufficient digitalizing of the imaging light to achieve a high imaging speed. With adaptive optics correction, the AOLSO system allows for cellular resolution retinal imaging and analysis of photoreceptor density. The 5^∘ FOV of AOLSO enables the capture of the parafoveal area in one frame. Additionally, the perfusion map of parafoveal capillaries is generated for further analysis. The AOLSO system not only provides cellular resolution of retinal images, but also has the capability to capture the whole parafoveal region in a single frame with an FOV of 5^∘ and a speed of 60Hz, which offers great potential for noninvasive studying the parafoveal area.

2. Materials and Methods

2.1. Light path setup

The schematic of the AOLSO system is shown in Fig. 1, two laser sources of different wavelengths are introduced into the system for illumination and aberration detection, respectively. Specifically, after the fiber output of the 795nm superluminescent diode (SLD, HP-316, Superlum), the laser beam is collimated by RC1 (RC12APC-P01, Thorlabs), then converted to a line beam by CL1 (ACY254-100-B, Thorlabs) and coupled into the main optical path by BS1 (BSS11, Thorlabs). A 735 nm laser source (SM-735-30, CNI Laser) is collimated by RC2 (RC12APC-P01, Thorlabs) and then coupled into the main optical path by a dichroic mirror DiM (T750lpxr Dichromatic Mirror, CHROMA). The main light path of the system consists of three groups of afocal telescopes made up of spherical mirrors, which alternately form the first pupil position deformable mirror DM (DM97-08, ALPAO), the second pupil position galvanometer mirror (S-9210-X, Sunny Technology) and the pupil position of the human eye.

The light reflected from the retina returns through the main optical path and is separated by the DiM. The imaging light passes through the DiM and BS1, and is then focused by CL2 (ACY254-250-B, Thorlabs) and CL3 (ACY254-75-B, Thorlabs) into the linear CCD detector (spl2048-140k, Basler) to generate images. The optical axis of two cylindrical lenses is arranged orthogonally to optimize the imaging process, this arrangement elongates the point spread function (PSF) of the imaging light along the sensor direction, ensuring sufficient digitization of the optical resolution in the imaging direction, the power of the imaging light is utilized as much as possible.¹⁹

The aberrations light passes through the BS2 and is detected with a custom-made Shack–Hartmann Wavefront Sensor (SHWS, 7.2mm pupil size, $18\times 18$ sub-apertures, $150\times 150$$\mu$m pinch and 5mm focal length lens-let array). The ocular aberrations are Zernike coefficients up to 35th order, which is compensated by the deformable mirror in a closed-loop control. The 80$\mu$m stroke range of the DM makes it possible to reduce the majority of the ocular aberrations. The corrected wavefront and residual aberrations are also displayed to determine whether or not the system has reached the diffraction limit.

2.2. Optimization of optical system

Reflective telescopes in a planar configuration inevitably introduce astigmatism into the optical path. While deformable mirrors can partially compensate for these aberrations, residual astigmatism makes it challenging to achieve diffraction-limited performance across a larger FOV. An effective solution to reduce astigmatism without adding extra elements is to position the mirror in a nonplanar manner.²⁰ As shown in Fig. 1, the last pair of mirrors M5 and M6 is designed to be arranged in a nonplanar configuration, and the distance and angle between all optical elements are optimized to achieve the diffraction limitation.

Spot diagrams on the retinal plane in the illumination path are computed to illustrate the change of focal spot without and with nonplanar design. Figure 2(a) displays the spot diagram of the wavefront sensing light at the retinal plane using the planar optical set, the astigmatic aberration is obvious, and the spot size is much larger than the Airy disc (2.39$\mu$m). Raising the last spherical mirror (M6) vertically introduces vertical astigmatism, which significantly reduces astigmatism in the AOLSO system. As shown in Fig. 2(b), the spots on the retinal plane fall within the Airy disk, which means that the system is diffraction-limited across the full FOV of 5^∘. Through the optimization of the nonplanar arrangement of spherical mirror M6 in the optical path, we have minimized astigmatism and achieved diffraction-limited imaging performance across the entire FOV of the AOLSO system.

In the imaging path, two cylinder lenses are placed before the linear CCD sensor with orthogonal optical axes. This arrangement serves to stretch the PSF and maximize the utilization of the imaging light’s power. Figure 3(a) illustrates the spot diagram at the imaging plane when two cylinder lenses are set in the optical path, the light spot falls within the black circle of the Airy disk. Furthermore, Fig. 3(b) demonstrates the encircled energy around the focal spot with a single achromatic lens compared to the encircled energy with the two orthogonal cylinder lenses, the normalized encircled energy within a 14$\mu$m radius (corresponding to the sensor pixel size) with two cylinder lenses is 0.798, nearly two times of that with a single achromatic lens. This enhancement contributes to superior image quality and enables an imaging speed of 60fps in the proposed AOLSO system.

2.3. Preparations for retinal imaging

A near-infrared wavelength of 795nm is used as illumination for comfort. The power of the light used for retinal imaging is 0.9mW, and the light of 735nm for wavefront detection is 0.03mW. Both power levels are below the safety threshold established by the American National Standards Institute (ANSI).²¹

To evaluate the performance of the AOLSO system, we conduct tests using a model eye. After that, four male participants between 26 years and 31 years of age with no known retinal or systemic conditions, clear ocular media, and a refractive error of less than 4 diopters spherical and 2 diopters astigmatism are imaged with the AOLSO system. The participants’ pupil is dilated with 1.0% tropicamide and 2.5% phenylephrine hydrochloride. An LED dot array presents a blue light point serving as a fixation target to direct the subject’s view direction. During imaging, the dot array is controlled manually for capturing different retinal regions of interest. The captured images are carefully stored for subsequent processing and analysis. This allows us to further evaluate the performance of the AOLSO system and extract valuable insights from the acquired data.

3. Results

3.1. AO performance

A model eye is first imaged with the AOLSO system, which consists of a 17.6mm focal length achromatic lens and a reflective diffusing target representing the retina. Figure 4(a) displays the results of wavefront sensing on the model eye. With AO corrections, most of the aberrations are effectively compensated. Notably, the low-order aberrations, such as defocus and astigmatism (3rd to 5th terms), are major contributors to image blurring, accounting for approximately 85% of the root mean square (RMS) error. These low-order aberrations are significantly reduced after AO correction. The largest high-order aberration in the system, spherical aberration (9th term), also experiences a significant decrease. Figure 4(b) illustrates the improvement in the system’s performance, the RMS value of the system decreases from approximately 0.308$\mu$m to around 0.052$\mu$m in 0.3s. The residual RMS aberration is lower than $\lambda$/14, which indicates that the system meets the diffraction limit based on the Maréchal criterion.

The PSF of the AOLSO system is also obtained with this model eye, which allows us to assess the performance of AO correction in another way. Figures 5(a) and 5(b) show the changes in PSFs without (AO off) and with (AO on) AO correction. When the AO correction is not applied, the light spot appears irregular in shape and the light intensity is not well focused. The gray value distribution across the spot indicates that the full width at half maximum (FWHM) of the light spot is approximately 29.3$\mu$m. With the AO correction, significant improvements are observed. The shape of the light spot becomes regular, and the light energy is concentrated, the majority of the light focuses within one pixel (5.86$\mu$m). This indicates a drastic improvement in the focusing of the light spot, resulting in an ideally Gaussian-distributed light spot energy. The experimental results clearly demonstrate that the AOLSO system operates at the diffraction limit.

3.2. Retina imaging

The AOLSO system is then used for retinal imaging. During the imaging process, the subject is instructed to stare at a fixed target, allowing the system to accurately locate the areas of interest. Figure 6 presents a series of images to demonstrate the impact of AO corrections. Figure 6(a) shows a single retinal image with an FOV of 5^∘ without AO correction, while Figs. 6(b)–6(d) represent slices from Fig. 6(a) at different eccentricities. In these images, nothing can be distinguished except only a few vessel shadows, the overall brightness is poor, and the retinal features appear distorted due to ocular aberrations. Figure 6(e) shows the same imaging area as Fig. 6(a) after AO correction, Figs. 6(f)–6(h) display the same sections from Fig. 6(e) at the same location with Figs. 6(b)–6(d). The significant improvement in image quality could be distinguished from the images directly. The average gray value of the image increases from 32.3 to 80.1, indicating a substantial enhancement in brightness. Furthermore, fine retinal structures, such as capillaries and photoreceptors are clearly visible after AO correction, and the signal-noise ratio of the images is improved from 6.40 to 8.95. In addition to the visual assessment of retinal images, a quantitative measurement called the contrast value is introduced to objectively evaluate the quality of the images,²² which is defined as

where M and N are the pixel number of the image, equal to 2048 in the AOLSO system. $I(i,j)$ represents the gray value at each pixel, and $Ī$ represents the mean gray value of the whole image. The calculated contrast values in Figs. 6(a) and 6(e) are 13.25 and 43.28, respectively. The improved images enable a more accurate analysis and evaluation of retinal structures, contributing to a deeper understanding of the human eye’s microvascular and cellular characteristics.

To further quantify the effects of AO correction, Fig. 6(i) presents the Zernike coefficients of ocular aberrations from the 3rd to 35th order with and without AO correction. Without AO correction, all orders of aberration are large, especially the lower-order aberrations, and the coma aberration emerges as the largest high-order aberration in the human eye. With AO correction, most of the aberrations are effectively corrected. The RMS value of the aberrations improves from approximately 0.247$\mu$m to around 0.058$\mu$m. The residual RMS aberration is lower than $\lambda$/14, which indicates that the system meets the diffraction limit for human retina imaging.

Indeed, the FOV limitation in the traditional AOSLO system necessitates the capture of dozens or even hundreds of images to study the fine structure of the fovea and parafoveal area, which spans approximately 8.5^∘.^23,24,25 In contrast, the AOLSO system offers a more efficient alternative. Only several images are required for parafoveal area analysis with its 5^∘ FOV and captured easily with an imaging speed of 60fps.

A grid of $3\times 3$ images is captured within several minutes in our AOLSO system. As depicted in Fig. 7(a), these images are then stitched together to create a single image with a FOV of $9\times 9^{\circ }$. Figures 7(b)–7(e) show magnified images at different eccentricities picked from the stitched image. The photoreceptor cells remain distinguishable within the stitched parafoveal area. The cells have a large density at small eccentricities, as the eccentricity increases, the cell density decreases, and the visibility of the cells diminishes.

3.3. Photoreceptor cell density

The cell density of the photoreceptor is a crucial physiological parameter that is closely linked to retinal pathological changes. The photoceptor cell density of four subjects is calculated with the AOLSO system for quantitative analysis of the parafoveal area. Regions at different eccentricities are selected for calculation, in which there is optimal contrast of the photoreceptors and minimal or no shadowing from the overlying retinal capillaries. The selected region size is $50\times 50$ pixels, which corresponds to an area of $35\times 35$$\mu$m on the retina.

The calculated results are shown in Fig. 8, it should be noted that the unit is set to millimeters to facilitate comparison with the previous results.²⁶ As expected, the highest cell density is observed near the fovea, with a gradual decrease in density as the eccentricity increases, which aligns with previous histological research. The measured results closely match the reported results when the eccentricity is less than 1mm, while the measured results are smaller when the eccentricity exceeds 1mm, and this difference between the two results became more pronounced as the eccentricity increased. The maximum difference was observed at an eccentricity of approximately 1.32mm. The cell density is different individually, and the discrepancy between results remains within reasonable limits.

3.4. Visualization of parafoveal capillaries

The parafoveal area is home to the tiniest capillaries in the retina, which play a crucial role in maintaining the health and function of the retinal tissue for central vision. Therefore, clear visualization and distinction of parafoveal capillaries are valuable and helpful in the early detection of some pathological changes in the retina. As shown in Fig. 9(a), the AOLSO system captures the high-resolution images of parafoveal capillaries without the need for fluorescent injections, and the parafoveal capillary vasculature is generated using perfusion map method,²⁷ as shown in Fig. 9(b), which provides valuable insights into the structure of the parafoveal capillary vasculature with high contrast and resolution.

The size of the foveal avascular zone (FAZ) is calculated from the extracted vasculature of all subjects, which has a mean area of $289882.625\pm 47091.24$$\mu {\mathrm{\text{m}}}^2$. To further analyze the capillary vasculature, the effective diameter is determined as the diameter of a circle with an equal area, calculated using the formula $d_{\mathrm{\text{eff}}}=2\times \sqrt{\mathrm{\text{Area}}∕\pi }$. The calculated mean effective diameter is approximately $0.607\pm 0.245$mm. The calculated size and effective diameter of the FAZ area fit well with results obtained from the AOSLO system, OMAG, and fluorescent angiography.⁶

4. Discussion

We develop an AOLSO system that combines adaptive optics technology and line scanning ophthalmoscope to visualize and analyze the parafoveal photoreceptor and capillaries.

The AOLSO system utilizes reflective optical elements of spherical mirrors in pairs. These mirrors serve two important purposes: first, they maintain the optical components conjugate to the eye pupil; and second, they compensate for astigmatism caused by the deflection angle between the mirrors. The last pair of afocal spherical mirrors is nonplanar arranged, in that way, the astigmatism is compensated and the spot diagram across the full FOV of 5^∘ is optimized to diffraction limited. There are other methods to reduce systemic astigmatism, such as using long focal-length spherical mirrors²⁸ or placing all mirrors at right angles.²⁹ However, these approaches would increase the size of the optical path. The other optimization for the optical path is the use of two cylinder lenses in front of the detector, which serves to stretch the PSF and maximize the utilization of the imaging light’s power, and an imaging speed of 60fps is achieved in this AOLSO system.

To evaluate the AOLSO performance, we conduct experiments on both model eyes and human eyes. The AOLSO system is able to complete AO correction in 0.3s, which ensures minimal disruption during imaging procedures, enabling efficient and seamless operation. The effectiveness of our AO correction is evident from the significant decrement in the RMS value of ocular aberrations and the measurement of PSF. Prior to the AO correction, the RMS value was approximately 0.31$\mu$m. With the AO correction, the RMS value is significantly reduced to approximately 0.05$\mu$m. The light spot shape becomes regular, and the light energy is concentrated. These results suggest that the AOLSO system has successfully achieved the diffraction limit, where ocular aberrations are minimized, and image quality is optimized.

In retina imaging of human eyes, with the help of AO correction, the brightness of the images increases from 32 to 80 and the contrast value is improved from 13.25 to 43.28. As a result, retinal photoreceptors and capillaries become clearly visible across the whole frame. Additionally, the AOLSO system captures retinal images with 5^∘ FOV and 60Hz frame rate, the high-resolution images of the parafoveal area could be stitched with a few images, which facilitates the analysis of photoreceptor density and parafoveal capillary vasculature. The trends of photoreceptor density fit well with existing histological research that the cell density decreases gradually with eccentricity. By generating a perfusion map of the parafoveal capillaries, the diameter of the FAZ area and the density of capillaries within the parafoveal region are calculated. The clear visualization of retinal cells and capillaries, accurate photoreceptor counting, and calculation of capillaries density have the potential for advancing retinal imaging and research.

While the AOLSO system effectively corrects ocular aberrations through adaptive optics, providing sufficient resolution to discern fine structures such as photoreceptors and capillaries across the full 5^∘ FOV with an imaging speed of 60fps, there are a couple of areas where improvements could be made. One limitation is that the current implementation of the AOLSO system utilizes a one-dimensional line scanning method, which slightly compromises the contrast and resolution along the line direction. This limitation could be mitigated by incorporating a structural illumination method, which would enhance the contrast and further improve the imaging capabilities of the system. Another consideration is that the extraction of parafoveal vasculature relies on high-quality retinal images. In cases where the images have low contrast, it can be challenging to generate accurate perfusion maps. This limitation underscores the importance of ensuring high image quality and contrast during the imaging process to obtain reliable and informative results.

5. Conclusion

By combining the one-dimensional line scanning method with adaptive optics, we construct an AOLSO system that achieves cellular resolution retinal images with a 5^∘ FOV and an imaging speed of 60fps. Experimental results demonstrate the effectiveness of the AOLSO system in capturing human retinal images. With AO correction, the images exhibit improved contrast and resolution, enabling clear visualization of photoreceptors, capillaries, and other clinically significant retinal targets. Benefiting from the high imaging speed of 60fps, only $3\times 3$ images are captured, it took a few minutes to stitch a montage image with an FOV of $9\times 9^{\circ }$, which reveals the capillary vasculature surrounding the fovea in a single frame. Furthermore, the AOLSO system not only allows for the precise counting of cone cells around the fovea, but it also enables analysis of the FAZ area and calculation of capillary density. This functionality is instrumental in facilitating functional imaging and quantitative analysis of the parafoveal region.

The clear visualization of retinal fine structures with a large FOV and a high imaging speed, such as photoreceptors and capillaries, the accurate photoreceptor counting and calculation of capillary density, will improve the efficiency of diagnosis and increase the feasibility in future clinical applications. Furthermore, the AOLSO system’s potential for functional fundus imaging will open up new possibilities for understanding the dynamic aspects of retinal function and hold promise for future applications in research and clinical settings.

Acknowledgments

This work is supported by the National Natural Science Foundation of China under Grant No. 62075235, National Key R&D Program of China under Grant No. 2021YFF0700700; Gusu Innovation and Entrepreneurship Leading Talents in Suzhou City under Grant No. ZXL2021425; Youth Innovation Promotion Association of the Chinese Academy of Sciences under Grant No. 2019320; Innovation of Scientific Research Strategic Priority Research Program of the Chinese Academy of Sciences under Grant No. XDA15021304.

Conflicts of Interest

The authors declare that there are no conflicts of interest relevant to this paper.

ORCID

Wen Kong  https://orcid.org/0000-0002-1679-7187

Yiwei Chen  https://orcid.org/0000-0002-3638-2309

Guohua Shi  https://orcid.org/0000-0002-6836-1617

Yi He  https://orcid.org/0000-0003-0114-6143