1. Introduction

The non-invasive detection of autofluorescent signals of the eye is already an important diagnostic tool in ophthalmology [1]. Major source of fundus autofluorescence is the lipofuscin content of the retinal pigment epithelium and the changes of its autofluorescence is a useful indicator in several ophthalmic diseases. One of the major functions of the retinal pigment epithelium is the cyclic phagocytosis and digestion of shed photoreceptor outer segments whose incomplete degradation leads to increased lipofuscin accumulation with aging [2]. Its autofluorescence can be easily detected and it is characterized by excitation wavelengths of 300-600 nm and emission wavelengths of 480-800 nm [3].

Investigations showed that for monitoring disease progression e.g. in retinitis pigmentosa (RP) NIR (Near-Infrared) excited (~800 nm) autofluorescence may have clinical advantages over direct short-wavelength (~450 nm) autofluorescence excitation, mainly because of favorable absorption and scattering properties of tissue at NIR wavelength [4].

Many types of retinal cells in animals were investigated by fluorescence imaging using retrograde tracers, both in vivo and in vitro [5,6]. In one typical experiment retinal ganglion cells of primate eyes labeled by retrograde tracers introduced artificially were visualized in vivo by properly adjusted fluorescent imaging [5].

Examination of the cornea and retina and practical realizations of confocal microscopes for both cornea and retina in examination instruments are available [7]. In humans confocal technology is used to measure autofluorescence involving exciting lasers of visible wavelengths, not harmful to eye.

Two-photon microscopy (TPM) is a novel imaging method for assessing functional and structural information on the subcellular level. One big advantage of two-photon microscopy is its possibility to collect fluorescence information from deep tissue layers without the need for a de-scanned aperture, as in confocal microscopy, thus enabling detection of scattered fluorescence photons. This feature increases signal intensity and image contrast. Furthermore, two-photon microscopes based on acousto-optic scanners allow rapid and pure electronic 3D scanning, down to 600 µm depth in living tissue [8].

Two-photon ophthalmoscopy is already used to investigate structural and functional properties of the animal retina. AOSLO (adaptive optics scanning laser ophthalmoscope) system was used in vivo and in vitro to create two-photon images of the retina with subcellular resolution collecting autofluorescence signals from both RPE cells and neurons [5]. Such systems involve adaptive optics in closed loop to correct for individual eye aberrations and motion to obtain nearly diffraction limited excitation focal spots [5,9].

Our aim was to create the theoretical basis of a novel method for in vivo imaging of the human retina with an acousto-optic two-photon microscope. Here the objective is replaced by the eye, therefore we would call it acousto-optic two-photon ophthalmoscopic instrument. Our approach intends to correct for aberrations and astigmatism of the eye using the combination of an aspheric lens and the electronic astigmatic focusing abilities of the acousto-optic scanner. Our modeling shows that this combination allows for fluorescence imaging with subcellular resolution in a wide range of individual eyes of different mechanical and optical parameters (refractive power and astigmatism). The main advantage of the acousto-optic two-photon instrument is that it is capable for real time 3D scanning, allowing high temporal resolution measurement of neural activity along the axis of the photoreceptors up to the ganglion cells. Moreover, highly precise and fast retinal thickness measurements and morphological investigations are allowed within relative large areas (400 µm diameter) in a single frame. In raster scan the ~30 µs pixel dwell time the full frame rate is 1.4 frames per second, but other type of scans (linear or curved focal spot drift) may increase the speed at full resolution.

In this paper we present modeling results that sustain the feasibility and envisioned parameters of the acousto-optic two-photon ophthalmoscope. This method provides diffraction limited lateral and depth resolution comparable to existing approaches as predicted clearly in our modeling. It also promises optical sectioning with resolution comparable to optical coherence tomography (OCT). The expected frame rate is bigger than that obtainable with adaptive optical systems involving mirror scanners, due to unprecedented scanning speed of the acousto-optic deflectors, when properly driven [8]. Random trajectory scanning allows measuring neural activity (spikes, EPSP’s) along targeted individual neural segments with 30-100 µs temporal resolution. These features together provide unprecedented investigation possibilities of neuron activity and morphology in the retina.

2. Methods

2.1. Advanced eye model

We used an optical model of the eye and ophthalmoscope to assess the possible performance of the novel acousto-optic method in retinal investigation. The questions to be answered:

The basis of our modeling was an optical model of the human eye constructed in Zemax optical design software (Zemax, LLC, Redmond, WA), based on the widely used Liou-Brennan eye model [10,11]. The model was first set up with average standard parameters, as listed in Table 1 [12]. We aligned the optical axis of the incoming light with the visual axis of the eye by properly rotating the eye by 5° (Fig. 1.).

Table 1. Properties of ocular components in the human eye [13]

View Table | View all tables in this article

 

Fig. 1 Zemax optical model shows the eye and the aspheric interface lens. Main parts of the eye are cornea, aqueous, crystalline lens and vitreous. The crystalline lens is composed of two parts; the pupil is attached to the front lens surface.

Download Full Size | PDF

We used two types of illuminations during modeling. The bare eyes were tested with collimated polychromatic light, built up of three main wavelengths resembling natural white light. The two-photon instrument combined with the eye was modeled with near infrared radiation of 800 nm central wavelength. We calculated with a 10 nm optical bandwidth of Gaussian distribution characteristic to an excitation pulse of 100-110 fs length, which is consistent with the transform limited laser pulses of recent two-photon microscopes [5,8]. The aperture of the collimated beam was always matched to the eye pupil size that was used accordingly as aperture stop. By combining this eye model with the model of the instrument we can answer questions 1 and 2, as a fundamental proof of principle of the new method.

In order to test the flexibility of our device – to answer question 3 – we created an eye model set with the use of clinically measured parameters on different eyes. We adopted our modeling for clinically assessed eye parameters from 25 patients (ages 24-52), possessing emmetropic eyes – with mean axial length of 24 mm (range 22.3-24.9 mm). The investigation was performed according to the Code of Ethics of the World Medical Association (Declaration of Helsinki). The parameter ranges measured with LENSTAR LS900 (Haag-Streit Inc., Switzerland) device on these patients are listed in Table 1. However, some parameters necessary for the model have not been measured by the instrument: lens surface curvatures, refractive index values, thickness values for vitreous and lens.

To construct the model for each particular eye, we had to set the parameters that could not be measured. Since the measurements were performed on patients with emmetropic eyes (with spherical refractive errors in the range of −2.00 to 2.00 D), we assumed that their visual acuity and modulation transfer function (MTF) is approaching the average – e.g. of the standard model eye. To create the individual models we introduced all measured parameters into the standard model replacing the original values: corneal radius, anterior chamber depth, axial length and astigmatism. The measured axial length value was used to set the sum of vitreous and lens thicknesses: their sum was determined by subtracting the measured anterior chamber depth plus the corneal and retinal thicknesses from the measured axial length (total eye length). The refractive index of the cornea was left at 1.3375, as assumed also by the measuring device. The refractive index of the lens was set to a “gradient3” surface (Zemax terminology).

The unmeasured standard parameters had to be changed in order to obtain eye models resembling nearly the functionality of the measured eye. We introduced them as variables into an optimization process, as described as follows. The target of the optimization was the spot size on the retina generated from the multi-wavelength white light beam, with pupil size of 2.5 mm. We assumed that the uncorrected best visual acuity of the analyzed patients imposes MTF with high contrast, corresponding to the minimum spot size obtainable with the combination of the measured and optimized components.

During the optimization we used the standard parameters (used e.g. in the eye model of Atchison [12]) as starting values for the algorithm. The optimized parameter values were constrained to the theoretical ranges given in Table 1, taken from reference [13]. Front and back radii, thickness and refractive index of the crystalline lens as well as posterior radius of the cornea were subject of optimization. We fixed the absolute thickness sum of the lens, vitreous and retina, since this sum could be determined from subtraction of the measured anterior chamber depth from the measured axial length.

Optimization of these parameters was continued on each particular eye model until the Strehl ratio on the PSF (point spread function) of the spot on the retina has fallen in the range 0.3-0.9. The obtained ratios were usually between 0.3 and 0.6 (the standard model produced Strehl ratio of 0.61 under the same conditions: 2.5 mm pupil diameter, collimated white light illumination). In these cases the image of the structured view at 0.1 degrees opening angle (letter F – optotype) created by the bare eye models in the visible wavelength range was of acceptable quality, with distinguishable details.

The LENSTAR device provided astigmatism values of the cornea together with the orientation of the astigmatism axes. We inserted the astigmatism into the eye model by setting the anterior surface of the cornea to irregular type, and introducing the astigmatism values into the proper place in the parameter list. The angle of the astigmatic axis was introduced by corresponding rotation of the anterior corneal surface relative to the rest of the eye.

The eye models constructed in this way were tested one by one in ZEMAX program for visual performance using PSF, MTF and image analysis. Only those models were further analysed that provided the above defined PSF’s. With this method 19 eyes were transferred to the complete optical model of the instrument.

2.2. Interface lens design

Parametric optimization was used to develop a lens with aspheric surfaces that works as an optical interface between the instrument and the eye (Fig. 1.). It mainly corrects for the chromatic aberration occurring at the near infrared wavelengths used in the two-photon instrument.

We have used the standard eye parameter set [12] for calculation of the aspheric matching lens. Collimated near infrared beam was used for illumination, the eye pupil diameter was set close to maximum (6 mm). The optimization target was the minimum spot size and the maximum peak intensity, which are the most important parameters for retina investigation.

The optimized lens was then tested with the eye models constructed with the parameters obtained from measurements, as explained in the previous section. The performance of the lens with the different eyes was compared upon the ratio of the observed peak intensity at the detection plane and the theoretical maximum peak intensity of a perfect imaging system working at the diffraction limit – Strehl ratio.

The interface lens should be placed close to the cornea surface, the distance between lens and eye varying individually. This distance is optimized for each individual eye model, to obtain best focal spot at the excitation wavelength; its average value is around 3 mm. The lens functions as the last element of the instrument’s optical chain – the aperture matching telescope creates the image of the scanner on the back aperture of the lens.

The lens thickness along the axis is 2mm, its material is SFL57 glass (Schott). It’s aperture is set to 10 mm. Detailed parameters are in Table 2.

Table 2. Geometrical parameters of the aspheric interface lens (even asphere)

View Table | View all tables in this article

2.3. Optical modeling of the setup

During this analysis we inserted the models of the 19 eyes constructed partially from clinically measured parameters unchanged into the target position of the instrument model, behind the aspheric interface lens (Fig. 2.). We optimized the following control values to obtain the optimum spot size and field of view with each eye parameter set:

 

Fig. 2 Schematic setup of our system. This is a schematic representation of an existing operational 3D acousto-optic TPM – figure shown in [30] - in which we intend to replace the sample with the human eye replacing the objective with the aspheric interface lens and the eye. In the setup’s optical model the human eye is represented by the Liou-Brennan eye model.

Download Full Size | PDF

We assumed to apply linear acoustic chirp in each deflector that can shift the focus of the full optical system [8,14,15]. The acoustic chirp parameters were used mainly to position the focus on the retina and to compensate for the eye’s astigmatism.

Detailed description of an existing system is given in Ref [8]. The optical model of this system was used during the analysis by replacing the objective with the eye models. The light source is a femtosecond laser with pulse bandwidth of 10 nm (around 110 fs at 800 nm central wavelength). A Faraday isolator is used to avoid coherent back reflection into the laser. Two AO deflectors control the z focusing of the beam in two perpendicular (x-z and y-z) planes. (AO lenses). A 2D AO scanner unit (2D-AO scanner and drift compensator) performs x-y scanning and drift compensation during z scanning. Two optical telescopes, each formed from two lenses, are used to imagine the AO lenses to the AO scanner and the AO scanner to the interface lens and eye, respectively. The fluorescence emitted by the sample is detected by photomultiplier tubes (PMT).

2.4. Astigmatism compensation

In human eyes regular astigmatism is a normal physiologic feature, and essentially means that the incoming light is focused to different depths in different meridional cross sectional planes along the optical axis and there are two perpendicular planes in which the focal length takes minimum and maximum value, respectively. These planes are oriented with respect to the horizontal and vertical planes in the eye at an angle that varies individually. The focusing power of the acousto-optic arrangement can be set separately for two perpendicular planes aligned parallel to the local x and y axes of the scanner (z being the longitudinal axis). Hence, when the planes of astigmatism with minimum and maximum power are aligned with the x-z and y-z planes of the scanner, the acousto-optic focusing power can be set to compensate the difference in focal lengths.

This is performed by chirping the acoustic waves differently in the deflectors performing scanning in the perpendicular x-z and y-z scanning, respectively. If the chirp parameters of the four deflectors can be expressed with the values ax1, ax2, ay1, ay2 respectively, (x1 stands for the first deflector deflecting in the x-z plane, y1 for the first deflector deflecting in the y-z plane, etc., ax1, ay1 etc. being the derivatives of the applied acoustic frequencies with time e.g. ax1 = dfx1/dt) the differences ax1-ax2 and ay1-ay2 determine directly the focal line’s z position in the x-z and y-z planes separately [8]. Hence by selecting the chirp parameters properly, focal plane shift (setting the focal plane exactly to the retina) and astigmatism compensation can be achieved simultaneously. The aperture of around 15 mm imposes 400 kHz chirp (difference of the minimum and maximum frequency within the aperture) for 10 µm focal length difference. Compared to the 50-60 MHz total bandwidth of the deflectors the focal length difference can be compensated in a very large amount (1000 µm or more).

During adaptation the axes of the eye astigmatism, introduced into the eye model as described above, should be aligned with the x and y axes of the scanner (deflection direction of the consecutive deflectors). This is performed by rotating the complete eye model about the scanner’s optical axis with the corresponding amount. The acousto-optics can cancel astigmatism if the astigmatic axes are parallel to the deflection axes. In practice this operation can be performed by mounting the scanner and the last optical elements of the instrument (tube lens and interface aspheric lens) on a rotational mount and rotating about the z axis during adjustment until optimal resolution is reached.

2.5. Ex vivo human retina preparation and imaging

All experiments were approved by the local Scientific and Research Ethics Committee (#58-1/2006). The eye of a 41 years old male donor with no known eye diseases was enucleated within 2 hours after death under aseptic conditions. The eyeball was rinsed in Betadine (Egis Pharmaceuticals PLC, Hungary) and the cornea was removed immediately thereafter. The retina together with the adjacent pigment epithelium and choroid was dissected free from the remaining eyecup and immersion fixed in 4% paraformaldehyde in 0.1 M phosphate buffer (PB) for 30 minutes at room temperature. Following fixation, the retina was washed three times in 0.1 M PB. Approximately 5 mm x 15 mm small retinal pieces containing both the pigment epithelium and the choroid were cut out from the fixed specimen and analyzed by two-photon microscopy.

The specimens were measured with FEMTO2D (Femtonics Ltd., Hungary) two-photon microscope. The different pictures were selected from z-scans across the samples performing more than 200 µm focal plane shift.

3. Results

The baseline model was used to optimize the parameters of the interface lens between the instrument and the eye. This model gives realistic resolution in the visible wavelength range, particularly at 475-650 nm. Figure 3(a) shows spot distribution on axis generated by the model at these wavelengths with a collimated entrance beam aperture of 2.5 mm that fills the pupil completely. Figure 3(b) shows the spot distribution at 800 nm at 6 mm pupil opening, obtained with the same collimated beam and otherwise the same eye parameters; note that the spot is hardly recognizable. The interface lens is needed to obtain diffraction limited spot size at 800 nm with largest pupil diameter. This lens corrects for the aberration of the rays entering at the sides of the pupil relatively far from its axis. The lens also corrects the chromatic aberration introduced at 800 nm, far from the ideally adapted 500-600 nm wavelength range. We designed a lens with one aspheric and one spherical surface made of N-SF57 glass in order to compensate for these aberrations. The spot distribution on axis provided by this lens is shown in Fig. 3(c). The spot is fully corrected and approaches the diffraction limit.

 

Fig. 3 a.) Polychromatic spot distribution on axis generated by the model at three standard visible wavelengths with an entrance beam aperture that fills the 2.5 mm diameter pupil completely. b.) Spot distribution in the same eye model at 800 nm and 6 mm pupil opening, obtained with the same collimated beam. c.) Spot distribution with the same eye parameters at 800 nm and 6 mm pupil opening corrected with the aspheric interface lens.

Download Full Size | PDF

We examined the spot distribution variation with off axis angles during scanning. Unfortunately, this system introduces residual angular dispersion proportional to the scanning angle, which causes elongation of the focal spot at increasing off axis coordinates [8]. With the optimized interface lens we could maintain the lateral resolution within 3 µm range in a circular pattern of 200 µm radius on the retina, centered at the fovea. The simulated spot distribution and size are changing with the intercept position on the retina according to Fig. 4.

 

Fig. 4 a.) Dependence of the exciting spot diameter (primary vertical axis) and of the Strehl ratio (secondary vertical axis) on the distance between the focal point and the optical axis, measured along the x axis (x coordinate in µm). Dx is the spot size parallel to the x axis, Dy is the size measured perpendicular to the x axis. b.) Spot distribution at different distances from the center of the scanned area on the retina.

Download Full Size | PDF

We present the simulated imaging properties of our eye models constructed from the measured eye parameters on two examples with significantly different measurement results in Fig. 5. The simulation results of the reference eye model constructed with standard literature parameters is also shown for comparison [12].

 

Fig. 5 Modeling results comparing eye performances of the standard model and different patients. Top row: PSF at centre of the fovea with the bare eye, visible polychromatic light; Second row: MTF + image of a 0.1 mm structure (letter F) under cicumstances of the top row; Third row: PSF at 800 nm; (10 nm bandwidth) in the eyes placed in the instrument, where the acousto-optics and the interface lens compensate for the defects. a.) Results with the artificial model - axial length 23.93 mm. b.) Results with a measured patient’s eye of 22.6 mm axial length c.) Results with an astigmatic patient’s eye of 24.9 mm axial length.

Download Full Size | PDF

The upper row contains point spread functions (PSF) generated with visible white light (wavelength components 486 nm, 587 nm and 656 nm) without any intermediate element from a light beam of circular aperture. The imaging ability of the eye models was tested with an optotype of 0.1 mm height (letter F). The images of the optotypes generated in white light with the sample eye models are shown as inlets in the second row. The imaging is also characterized by the attached MTF curves that show that the spatial resolution of these eyes is around average human eye resolution and somewhat better than that of the standard model eye. The bottom row contains the PSF-s generated by the eyes from a NIR laser beam of 10 nm bandwidth (785 nm – 805 nm) of circular aperture, delivered by the acousto-optic instrument to the eye. The PSF spots in the upper and bottom rows correspond to focal spots produced by the eyes in the center of the fovea. Notice the asymmetric shape of the PSF-s generated by the bare eye models constructed based on measured data. The reason of that is the astigmatism being present in most human eyes. The compensation ability of the acousto-optic scanner removes efficiently the effect of astigmatism as shown in the PSF-s in the bottom row. We tested the theoretical range of astigmatism compensation and found that it is well beyond the physiological range (up to 10 Diopters).

The compensated PSF-s are nearly diffraction limited in all eye models derived from measured parameters, at least in the central region of the fovea.

Acousto-optic focusing enables spot size conservation over more than 500 µm of in vivo tissue depth [8], therefore we expect effective measurements in most in vivo samples in a volume of about 400µm lateral size (diameter) and 500 µm depth. One possible and straightforward application of our analysis method is to measure the local cell morphology (shape and density of different cell types) of the retina (Fig. 6.). This experimental figure presents possibilities to clearly distinguish different cell types –photoreceptors, ganglion cells, pigment epithelial cells – based on laser induced two-photon fluorescence.

 

Fig. 6 Visualization of an example two-photon measurement performed on human retina. a.) Photoreceptor inner and outer segments b.) Retinal pigment epithelium c.) Ganglion cell layer d.) Inner nuclear layer. The images were collected ex vitro according to paragraph 2.5.

Download Full Size | PDF

4. Discussion

4.1 Imaging technique

Our first task was to achieve a single coupling lens between the instrument and the eye for majority of eyes to obtain practically useful two-photon resolution. This is necessary to develop an instrument that we can use practically with any eye without substantial modifications. Focusing differences between the eyes with different refractive errors (e.g. myopia, hyperopia) can be almost completely compensated by the focusing capability of the acousto-optic scanner. Additional degrees of freedom accessible during in situ adjustment are axial distances between the coupling lens and instrument, and between the lens and the eye, as well as the relative positions (decenter, angle) of the instrument and eye axes. These parameters can be changed during a regular examination situation. On modeling level the mechanism of this adjustment during the examination is equivalent to an optimization process in ZEMAX, with the adjustable parameters set as variables and the optimization merit function set to minimize the focused spot size.

In its present form our model does not include aberration correction with adaptive optics. Modeling results show that in principle the dynamic and astigmatic focusing capabilities of the acousto-optic device can mitigate the need for the adaptive optics, which is a very expensive device. It is important to note that our model does not take into account the motion of the eye and of the pupil, which in practice may indicate the use of adaptive optics indeed [9]. However, this is not a fundamental limitation, since the system can be combined with a reflective adaptive optics device when further modeling or experimental results indicate; only the aperture of this and the acousto-optics should be matched with a telescope.

The baseline eye model that we used is a generally accepted model based on the Liou-Brennan eye model [10,12]. Basic property of this model is that at the central wavelength of the visible spectrum, 588 nm exactly focuses the rays to the fovea, with a 2.5 mm pupil diameter. However the diffraction limit restricts the focused spot size to >3.5 µm at the central wavelength at this pupil diameter, thus our optical resolution would be very poor at this pupil size. In the clinical praxis pupil diameters of 6-8 mm can be routinely obtained, that result in a theoretical minimum Airy disk diameter of 1.13 µm at the wavelength of 588 nm. At 800 nm, however, the Airy disk size is of 1.6 µm, being now the practical limit of the optimum diffraction limited lateral spot size obtained with the two-photon exciting laser beam.

The cells to be resolved are mainly RPE cells of around 10-15 µm diameter in the foveal area (and reaching up to 60 µm size at periphery) [16]. Thus our expected spot with ~1.6 µm lateral and ~4 µm axial sizes will allow us to resolve individual cells. Based on previous measurements diameters of foveal cones vary between 2.5 to 10 µm depending on the location relative to the fovea center [13]. Rod’s diameters vary from 3 µm to 5.5 µm increasing towards the periphery. Hence theoretically our optical resolution makes possible distinguishing individual cells and measuring their activities separately.

We modeled effectively the potential of the electronic compensation offered by the acousto-optic lenses that provided electronically variable focus in perpendicular x-z and y-z planes. We could maintain the lateral resolution within the 1.6 µm range in a circular pattern of 200 µm radius on the retina, centered at the fovea (Fig. 4). Simulation also shows that the quality of the spot does not change considerably during the focal change range of 200 µm necessary to penetrate through the retina performed by the acousto-optics.

Two-photon imaging has the advantage that the NIR excitation radiation of 800-900 nm is very less absorbed than it would be the one photon excitation of 400-510 nm required for fundus autofluorescence excitation. Here absorptive effects of crystalline lens, RPE melanin as well as macular pigments are eliminated that all strongly absorb below 500 nm [17]. Since 90% of the excitation radiation may pass until the surface of the retina according to the predictions, our targeted method provides high efficiency in excitation. The two-photon excitation of fluorescence needs 10⁵-10⁶ bigger intensity than single photon excitation, therefore the use of ultrashort pulsed lasers with 80-100 fs pulse length is necessary. However the average power levels for excitation with a mode-locked laser and a cw laser are comparable (the excitation on a tissue surface needs maximum around 10 times bigger average intensity for two-photon than for single photon) and this will change in the favor of two-photon when penetrating deeper in the tissue. We expect that absorption difference of the eye between near infrared (two-photon) and blue (single photon) spectral ranges will allow better image quality and higher signal to noise ratio with the new ophthalmoscope than with conventional fundus fluorescence imaging.

Non-sequential modeling shows that 10% of the generated fluorescence radiation will exit the eye and 8% will pass the coupling aspheric lens with antireflection coating optimized for the peak fluorescence at 640 nm. The spectral composition of the fluorescence radiation was assumed Gaussian between 480 and 800 nm [17]. In principle all this radiation can be captured with high sensitivity PMTs. Within the minimum spot size obtainable in the model we can keep the power density below 4 mW/cm² if we apply a total power of 2 *10⁻⁶ W. This power is sufficient for efficient excitation of both artificial and autofluorescent markers. This limit allows the operation of the ophthalmoscope as an instrument without producing light hazard.

4.2 Autofluorescence in the eye

Photoreceptor dysfunction and degeneration accompanied by atrophy of pigment epithelium are marked by excessive accumulation of lipofuscin [18]. In case of retinitis pigmentosa, the disease might be revealed by ring or arc shaped perifoveal zones of high autofluorescence [19]. The structural and functional significance of the hyperautofluorescent ring as well as its prognostic and monitoring value have been already established [20]. Robson found high correlation between the radius of the high density ring of increased autofluorescence with the amplitude of the pattern electro-retinogram P50 component in patients with both normal and decreased visual acuity [21].

The RPE is damaged in certain disorders, such as age-related macular degeneration (AMD) and retinal dystrophies and its altered function can be detected as either abnormally high or abnormally low autofluorescence [22]. Autofluorescence is also helpful in distinguishing benign nevi from malignant ocular melanomas, as melanomas usually have elevated levels of lipofuscin [23]. The amount of the lipofuscin can be used as a marker of the disease’s progression and useful in the evaluation and management of AMD, retinal dystrophies, as well as tumors in the back of the eye [24,25].

Although endogenous fluorescence provides morphological, spectral and lifetime contrast that can indicate disease states in tissues, the currently used diagnostic methods allow a highly limited detection quality only [26]. The model presented here allows an alternative method for the investigation of autofluorescence at the cellular level which may have major diagnostic and even therapeutic benefits in the near future.

4.3 Comparison between TPM and SLO/OCT techniques

Compared to scanning laser ophthalmoscopy (SLO) two-photon microscopy in its present form is rather a fundamental research tool. TPM provides primarily lower lateral optical resolution than SLO, mainly because of the longer optical wavelength, but its axial resolution is only limited by the diffraction (4 µm) in contrast to SLO (~100 µm), hence the axial resolution of TPM is more than ten times bigger than that of SLO [27]. SLO can be combined with OCT that provides better axial resolution (<2 µm) than TPM, but this combination cannot be operated with sufficient temporal resolution to catch neural excitation and propagation events in the retina to perform functional imaging [28,29].

However, since acousto-optic two-photon microscopy provides real time 3D imaging, this is a huge advantage in getting structural and functional information from the different layers of the target. It is intended primarily to directly monitor the correlation between the visual stimulation and the activity of hundreds of cells (photoreceptors, bipolar and ganglion cells) recorded nearly simultaneously in vivo, exhausting different scanning modes with fast sampling. It is also meant to directly follow the path of the neural signals generated in the receptors, making use of the high scanning speed that provides sufficient temporal resolution for this task. These latter features are not reachable with different SLO techniques, since their temporal resolution do not allow efficient sampling of neural signals and potential changes, whereas represent affordable potential of the two-photon method.

In the most general imaging operation the exciting laser spot is moved through the selected volume and the detected fluorescence signal is attributed to the targeted coordinate for each scanned spot. Maximum one million different spots can be addressed with the present acousto-optic scanner, the acquisition time for such an image is around 30 s (point switching + sampling time = 30 µs/pixel).

As a clinical diagnostic tool, the main future advantages of the TPM reside also in its capability to provide functional imaging, but these features cannot be fully evaluated yet. Some particular parameters, like less absorption of the excitation light (light gain at least on the inward path compared to SLO) may provide images with better contrast, whereas particular information is gained from the fluorescence nature of the signal, being useful in the follow-up and diagnosis of a number of diseases, as depicted above.

7. Conclusion

The presented modeling results sustain the feasibility and envisioned parameters of the acousto-optic 2-photon instrument intended for functional retinal investigations. Main advantage of this instrument compared to existing approaches is the diffraction limited lateral and depth resolution predicted by the modeling results. This is achieved by the unique capability of the acousto-optic scanner to correct focusing errors of the human eye. Our predicted optical resolution makes possible distinguishing individual cells and measuring their activities separately. Due to the unprecedented scanning speed of the acousto-optic deflectors the expected frame rate is at least ten times higher than that obtainable with systems involving mirror-scanners. The technology promises real time measuring of neural activity in individual neurons, neural segments and cell assemblies with 30-100 µs temporal and subcellular spatial resolution.