6. Applications of Adaptive Optics Retinal Imaging
Adaptive optics retinal imaging has demonstrated the capability to image the living human retina at microscopic resolution. The improvement in retinal image contrast and resolution allows the direct observation of retinal microstructures giving the researcher the opportunity to analyze their integrity and/or pathological abnormalities. An AO ophthalmoscope provides en face images of the retinal layers showing photoreceptors, retinal vessels and nerve fiber bundles ( Figure 3). Advances in image processing methods are required in order to maximize the value of the retinal data acquired and to make the AO retinal imaging technology accessible to the clinical ophthalmic community. Accurate automated routines are indeed mandatory when large quantities of data need to be analyzed.
Figure 3
AO retinal images with adaptive compensation of the photoreceptor layer (A), vasculature (B) and retinal nerve fiber layer (C). Scale bars represent 50 μm. The blood vessels form a three-dimensional network across the inner retinal layers. All the images shown in this review have been acquired using a flood-illumination AO retinal camera (rtx1, Imagine Eyes, Orsay, France).
Currently, the majority of studies have focused on the generation and analysis of AO images of photoreceptor cells, including cones and rods [67, 72–75]. Over the last few years, efforts have been made to develop reliable methods to measure the cone density as a function of retinal eccentricity in populations of healthy adults [76–81]. In general, data on populations of healthy eyes are fundamental for characterizing the density and the spacing distribution as well as the brightness of healthy photoreceptor cells in vivo ( Figure 4). This will allow measurement of the normal ranges which can be compared to pathological photoreceptors, even in the early stages of retinal diseases. An increasing number of studies [67, 82–90] are showing the distribution of cone photoreceptor density and spacing in adults. The in vivo measurements of cone density [67, 82–85] have shown good agreement with histologic data from cadaver eyes [86–90]. Using an AO scanning laser ophthalmoscope (AOSLO), researchers [82–85] found the cone density to drop on average from 120,000 cones/mm2 at 0.1 mm to 20,000 cones/mm2 at 1.0 mm from the foveal center. In general, the cone density values at the same eccentricities of the nasal and temporal retina within 2 mm eccentricity from the fovea were found to be within 10% of each other both in ex vivo and in vivo studies [84–87, 91]; a 10% higher density of cones along the horizontal than the vertical meridian was, in general, found. A synopsis of the average cone density from previous AO studies in populations of healthy subjects is shown in Table 1. The parafoveal cone density showed a moderate to high inter-individual variation with coefficient of variation (defined as the ratio of the standard deviation to the mean) values ranging between 12% and 20% [82–86, 91]. Discrepancies between studies could be due to different factors, such as the inclusion of subjects with different ages or eyes with different axial lengths and refractive corrections, the instrument used for biometry, the different model eye used to estimate the retinal image size, the use of foveal center or foveal fixation as reference point to define retinal eccentricities and the sampling window area used to count cones.
Figure 4
AO montage of the photoreceptor mosaic in a 46 year old subject. The asterisk shows the foveal center. The black and grey boxes enclose two high-magnification images of the photoreceptor layer from 240–1,440 μm eccentricities from the foveal center. The center-to-center distance between cones has been shown to increase with greater eccentricities from the fovea; the cone density, accordingly, declines with increasing eccentricity from the fovea. Scale bars represent 50 μm.
Table 1
Cone density estimates in histology and AO retinal imaging studies taken at increasing eccentricities from the foveal center.
Work | Subjects (N.); age (range, years); AxL (range, mm); sampling window area (μm or pixels); model eye | Cone density (average, cones/mm2) as a function of retinal eccentricity (range, μm) along the horizontal meridian | |||
---|---|---|---|---|---|
230–360 μm | 400–540 μm | 720–890 μm | 1,000–1,350 μm | ||
Curcio et al.[86] | 7; 27–44 yrs; AxL not reported; variable sampling windows; anatomical schematic eye | Nasal/Temp: 60,000–55,000 | Nasal/Temp: 40,000 | Nasal/Temp: 26,000 | Nasal/Temp: 20,000 |
Li et al.[83] | 18; 23–43 yrs; 22.9–28.3 mm; adaptive windows (adjusted to contain 150 cones); Gullstrand schematic eye model | All meridians: 60,000–45,000 | Not reported | Not reported | Not reported |
Chui et al.[82] | 11; 21–31 yrs; 22.8–27.5 mm; 150 × 150 pixels window; standard reduced eye model | Not reported | Nasal/Temp: 41,000 | Nasal/Temp: 27,000 | Nasal/Temp: 15,000 |
Chui et al.[84] | 4; 24–54 yrs; AxL not reported; 22 × 22 μm window; model eye not reported | Not reported | Temp: 30,000 | Not reported | Temp: 15,000 |
Song et al. [85] | 10; 22–35 yrs; 22.1–26.1 mm; 50 × 50 μm window; Indiana model eye | Nasal: 59,700–50,000 Temp: 59,200–50,500 | Nasal: 43,700–37,800 Temp: 41,200–37,300 | Nasal: 29,100–24,200 Temp: 28,100–24,100 | Nasal: 19,100–16,800 Temp: 19,900–16,300 |
Lombardo et al. [91] | 12 (24 eyes); 24–38 yrs; 22.6–26.6 mm; 50 × 50 μm window; Gullstrand schematic eye model | Nasal/Temp: 49,400 | Nasal/Temp: 38,500 | Nasal/Temp: 30,000 | Not reported |
AO retinal imaging in healthy subjects also makes it possible to better understand the sampling limit of resolution of the cone mosaic in vivo. Analysis of the spatial distribution of the cone photoreceptors provides new information on the physical aspects of visual sampling of the human eye. In a recent work, our group [91] found that the mean Nyquist limit sampling of resolution of the cone mosaic (Nc) was 33 ± 2 cycles/degree (c/deg) at 260 μm eccentricity, declining to 26 ± 2 c/deg at 600 μm eccentricity in a population of twelve young adults (age range: 24–38 years; 24 eyes; axial length of the eye (AxL) range: 22.61–26.63 mm). Authors previously found comparable results for young adults [83, 84, 92]. The Nc was calculated to be 34 c/deg at 1° eccentricity (approximately 270 μm) by Chui et al.[84]. Coletta and Watson [92] estimated Nc to range between 50 and 42 c/deg at the fovea and between 24 to 22 c/deg at 4° eccentricity (approximately 1.1 mm) in a population of subjects with spherical equivalent error (SEr) ranging between 0 D and −14 D. In a population of 18 healthy subjects (age range: 23–43 years; 18 eyes: AxL 22.86–28.31 mm), Li et al.[83] found that cone density tended to decrease with increasing axial length at eccentricities between 100 and 300 μm from the foveal center, with no statistically significantly differences between emmetropes and myopes at the fovea. In a population of 11 healthy subjects (age range 21–31 years; 11 eyes), cone density in moderate myopes (up to−7.50 D) was found to be significantly lower than in emmetropes within 2.00 mm eccentricity from the fovea [84]. The spatial vision of the cone mosaic reduces with increasing axial length [91]: the lower Nyquist limit sampling of resolution of the cone mosaic in myopes than emmetropes has been postulated to be caused by retinal stretching, occurring at the posterior pole of myopic eyes, due to the eye's increased axial length [84, 91, 92]. A higher difference between myopes and emmetropes has been demonstrated when the acuity limits were expressed in retinal units (c/mm) rather than in angular units (c/deg). The discrepancy between resolution and Nc expressed in retinal acuity was primarily considered a result of psychophysical and neural factors, including the neural sampling rate (i.e., the retinal ganglion cells receptive field density) [93–95].
AO retinal imaging has also been used to study the waveguide and reflectance properties of cone photoreceptors in vivo[96–104]. Variation of brightness between adjacent cones was observed, even when imaging the retina of healthy subjects. Differences in reflectance between adjacent cones were seen both at the boundaries of retinal vessels and in areas devoid of vessels. While intra-retinal scattering was considered as the primary source of the higher reflectance of cones that reside beneath the vessels in comparison with adjacent cones, this phenomenon could not explain the variation in brightness between adjacent cones in areas devoid of vessels. Various hypotheses have been made to find the possible factors that influence the intensity variation of the cone mosaic: (1) the differences in reflectivity could be caused by molecular differences within the cones that are due to phototransduction [96]; (2) the reflectance variation could be based on the cone outer segment length [97–100, 102–104]: fluctuations in reflectivity indeed could be due to changes in the outer segment length, related to disk shedding. (3) The cause of the variation could also be related to the pointing direction of cones and the angle of incidence of the light entering the pupil [105–108]. (4) Finally, technical factors should be considered to contribute to differences in reflectance between adjacent areas of cones, including the light source (i.e., laser or SLED emitter) and/or the wavelength of light used to illuminate the retina, the axial resolution of the imaging system (confocal or flood-illumination) and the imaging process [100, 101, 104]. The relative contribution of the sources of reflection within the cones was shown to be not constant with time [98, 100, 102, 103] and the photopigment density has been found to decrease with increasing eccentricity from the fovea up to 2° eccentricity [109].