2.1.

Scanner Design

A 2-D acousto-optic deflector (AOD) was used to generate the scan pattern in both the in vitro and in vivo setups. In an AOD, maximum intensity of diffracted light in the first-order beam occurs when the Bragg condition is satisfied. An acoustic wave propagating through a crystal creates a sinusoidal change of its optical index. The incident collimated laser beam then “sees” an optical diffraction grating, where the acoustic wavelength determines the grating width. Sweeping the acoustic frequency by an electrical signal changes the grating width, yielding a sweeping scan angle and thus, a moving spot. Mounting two AOD crystals perpendicular to each other into one package then creates a 2-D scan pattern.

To control our 2-D AOD (2DS-100-35-.532, Brimrose, Baltimore, Maryland), four electric signals are needed (Fig. 1 ): one frequency modulation (FM) signal per crystal or scan axis, respectively, an amplitude modulation (AM) signal to enable the first-order beam, and a trigger pulse to synchronize the control unit.

Fig. 1

Electrical command signals employed to create one scan pattern with separated lines with the 2-DAOD in our application. ${\mathrm{FM}}_1$ deflects the light fast but with a constant speed (horizontal line scan). A second, slower ${\mathrm{FM}}_2$ steps the spot into the position where the next line is to be produced (vertical frame scan). To switch the first-order beam on and off to sharply define the edges of the scan lines, the AM signal is applied in parallel to both crystals. A trigger pulse synchronizes the control unit and defines repetition rate and number of patterns to be created in one cycle.

A two-channel, arbitrary function generator (Tektronix, AFG320) creates the two custom-programmed FM signals. Utilizing the four fixed memories for arbitrary waveforms of the generator, the user is able to switch back and forth between different characteristics of the scan pattern. The AM signal is generated by a window comparator. Our home-built window comparator switches AM ON, when the ${\mathrm{FM}}_1$ voltage crosses the lower voltage set point that determines the left edge of the scan field. Likewise, it switches AM OFF, when ${\mathrm{FM}}_1$ crosses the higher ${\mathrm{FM}}_1$ reference voltage that corresponds to the right edge of the scan field (Fig. 2 ). This enables the user to switch between different characteristics of the scan pattern without changing its size. For instance, it is sufficient to change the slew rate of the ${\mathrm{FM}}_1$ ramp signal to produce a different scan speed. To define the repetition frequency (frame rate) of the scan pattern and the number of repetitions applied to each irradiation site, an additional digital pulse generator (Tektronix, PFG5105) is used to trigger the entire control unit. With this control design, we were able to produce two basic shapes of a scan pattern: one that is composed of six separated lines (SEP) and a second consisting of 21 interlaced lines (INT) without spacing between lines (Fig. 2).

Fig. 2

Two patterns of the in vivo scanning device. On the left is the separated lines scan pattern (SEP) with cartoons of its accompanying control signals (Fig. 1). The dashed vertical lines mark the low and high reference voltages of ${\mathrm{FM}}_1$ that define the left and right edges of the pattern; they are utilized to switch the AM signal on and off. The pattern on the right is an interlacing scan pattern (INT). It produces, similar to the SEP pattern, three patterns of seven separated lines that are spatially displaced by one line.

Both scanners are aligned in such a way that the central acoustic frequency of the full bandwidth of the 2-D AOD defines the optical axis of each device. The scan angles for the line and the frame scans are produced as half angles around this optical axis to minimize coma, which may occur due to any scanning process, as the beam by default is steered off the optical axis. To focus the first-order beam, a single achromatic lens was placed one focal distance from the midpoint between the two crystals of the 2-D AOD, forming a nearly telecentric optic, to minimize curvature of the scanners’ focal plane.

The spot diameters and beam profiles of the setups were determined on a stationary spot centered in the scan pattern using the knife-edge method along $\left(x\right)$ , at $45\mathrm{deg}$ and perpendicular $\left(y\right)$ to the direction of motion of the line scan. A $200\text{-}\mu \mathrm{m}$ -wide slit was mounted a known distance from the axle on a rotating wheel. While spinning the wheel at a carefully monitored frequency, the rise and fall time of the signal behind the two edges of the slit were measured to determine the spot size. Three measurement runs per orientation were conducted to compute mean diameter and error. For profiling, the average rise time signals were differentiated. The scan speed was measured by scanning the laser spot along a microscope scale $(100\mathrm{div}∕1\mathrm{mm})$ , detecting the transmitted light with a photodiode and measuring the time between two minima produced when the spot runs over a cross-hair of the scale.

2.2.

Determination of Threshold Exposure

Both in vitro and in vivo devices were used to determine the effective dose 50% $\left({\mathrm{ED}}_{50}\right)$ for RPE cell damage. ${\mathrm{ED}}_{50}$ describes a probability of 50% that an exposed cell will be killed. ${\mathrm{ED}}_{50}$ was determined by Probit analysis for individual samples in vitro and individual eyes in vivo.

In Probit analysis, the cumulative density distribution of the percentage of targets that do respond to a given dose is plotted versus the dose that was applied. The program (Probit, Version 2.1.1) produces a lognormal fit through the quantal response data, which are represented by 1 for success and 0 for no success. As a result, a probability function versus the dose is being calculated and the ${\mathrm{ED}}_{50}$ determined.¹⁶

The endpoint for experiments on RPE sheets in vitro was the appearance of lines of dead cells according to a viability assay that is described in Sec. 2.3.2. If lines in a lesion were partially damaged, i.e., if it was questionable whether to assign a 0 or a 1 to that lesion, the number of dead cells in a scan line was counted. Since about 20 RPE cells fit into the length of one scan line, lesions with fewer than 10 cells per line constitute a 0 for Probit analysis, and lesions with more than 10 cells per line constitute a 1.

The endpoint for in vivo experiments in rabbit eyes was the appearance of hyperfluorescent lesions in fluorescence angiography, as described in Sec. 2.4.4.

Results are presented for each set of parameters as the average ${\mathrm{ED}}_{50}$ and the absolute deviation of ${\mathrm{ED}}_{50}$ from individual samples or eyes. The slope of the probability distribution, defined as the ratio of ${\mathrm{ED}}_{85}$ over ${\mathrm{ED}}_{50}$ , is also presented in Table 1 for in vitro experiments on sheets of RPE and in Table 2 for in vivo experiments in rabbit eyes.

Table 1

Slopes of the probability distribution of in vitro experiments on sheets of RPE. Here, the slope is defined as the ratio of ED85∕ED50 . All measurements were performed in duplicates and the data listed are the average values.

Table 2

Slopes of the probability distribution of in vivo experiments. Here, the slope is defined as the ratio of ED85∕ED50 . All measurements were performed in duplicates and the data listed are the average values.

2.3.

In Vitro Experiments

2.3.1.

Setup

The in vitro scanner was assembled using the design described earlier. It produced a scan pattern of $300\times 300\mu \mathrm{m}$ with spacing of $60\mu \mathrm{m}$ between line centers. A $5\text{-}\mathrm{W}$ cw laser (VERDI, Coherent, Saint Clara, California) served as laser source for the irradiation. An argon ion laser (INNOVA 90, Coherent) at $488\mathrm{nm}$ , delivered through a multimode fiber, was used for excitation of the fluorescent cell viability probe, as described later. A custom-built compound fluorescence microscope was set up to capture fluorescence images before and after irradiation (Fig. 3 ).

Fig. 3

Schematic of the bench-top in vitro setup. The output beam of the diode-pumped solid-state Verdi was enlarged to the size of the full active aperture of the 2-DAOD $\left(5\mathrm{mm}\right)$ with a Keplerian beam expander. After passing the focusing achromat $(\mathrm{f.l}=150\mathrm{mm})$ at the output side of the AOD, the first-order beam was directed vertically by a dichroic filter onto the sample, which was placed horizontally on a 3-D translation stage. A compound fluorescence microscope employing an infinity-corrected $20\times$ objective with a large working distance was erected on top of the dichroic filter and the sample holder to visualize the sample through a CCD camera. A fiber-coupled white light source illuminated the sample (not shown). Radiation from an argon ion laser $\left(488\mathrm{nm}\right)$ was coupled into a multimode fiber. The fiber output was introduced into the beam path by a flipping mirror for fluorescence excitation of the viability assay Calcein.

2.4.

In Vivo Experiments

2.4.1.

Setup

For the in vivo experiments, a slit-lamp adapted laser scanner was developed. The emission $\left(532\mathrm{nm}\right)$ of a cw laser (VERDI V-10, Coherent) was coupled into a polarization maintaining single-mode fiber (PM fiber) (HB450, Fibercore Limited, United Kingdom). The output power of the Verdi laser was attenuated by combination of a $\lambda ∕2$ -plate and a polarizing beamsplitter cube and coupled into the fiber. An additional $\lambda ∕2$ -plate matched the polarization state with the fast axis of the PM fiber. Using an attenuator allows the laser to run at constant power at all times and minimizes pointing instabilities. PM fiber was necessary to preserve diffraction-limited beam characteristics and because linear polarization is required for optimal operation of the AOD. Its use limited the power that could be delivered to the setup to about $300\mathrm{mW}$ . With a typical diffraction efficiency of two AODs $(\approx 64\%)$ , the highest possible power entering the eye was $185\mathrm{mW}$ .

The scanner design described in Sec. 2(A) was mounted with its mechanical components, collimating and focusing optics vertically on top of the slit lamp (SL130, Zeiss, Oberkochen, Germany). The PM fiber output and the collimator were mounted on a six-axis stage ( $x$ , $y$ , $z$ , tip, tilt, and rotate) to match the Bragg condition and the preferred polarization direction of the 2-D AOD. A mirror, placed between the two lenses of the stereo objective of the slit lamp, directed the first-order beam horizontally away from the objective into the eye, running along the optical axis of the slit lamp (Fig. 4 ).

Fig. 4

Schematic of the slit-lamp adapted in vivo setup. The output power of the Verdi laser was controlled by an attenuator. An additional $\lambda ∕2$ -plate rotated the polarization state of the light into the fast axis of the PM fiber to maintain the highest possible stability of the linear polarization orientation at the output side of the fiber. A single achromatic lens collimated the beam, leaving the PM fiber to the size of the active aperture of the AOD. The focusing lens was placed after the AOD in a nearly telecentric way. The mirror directing the irradiation light away from the slit lamp was placed between the two lenses of the objective of the slit lamp and the two mirrors of the double-slit illumination.

To achieve the same size of the scan pattern on the retina as used in the in vitro experiments, the scan field in air was enlarged to a width and height of $450\mu \mathrm{m}$ , with $90\text{-}\mu \mathrm{m}$ spacing between line centers for the SEP pattern. The energy to the rabbit eye was applied through a Goldman contact lens, which together with the optics of the rabbit eye, magnifies all distances on the retina by a factor of 0.66,¹⁸ making the scan area $300\times 300\mu \mathrm{m}$ , with $60\text{-}\mu \mathrm{m}$ spacing, on the retina.

3.1.

In Vitro Experiments

With the knife-edge method, the spot diameter on the sample was measured to be $20\mu \mathrm{m}$ . The scan field was nearly square with a length of the scan lines and a height of the pattern of $300\mu \mathrm{m}$ .

Selective damage of RPE cells is feasible using a laser scanning device that employs a Gaussian laser beam. Calcein viability assay performed after irradiation showed separate lines of dark cells in the fluorescence images, resembling the applied scan pattern. Close to ${\mathrm{ED}}_{50}$ , the widths of these lines were one or two RPE cells thick. Cells in between the scan lines remained viable, suggesting selectivity within the RPE monolayer (Fig. 6 ).

Fig. 6

Calcein fluorescence images after irradiation with $3\mu \mathrm{s}$ (all 10 repetitions with $100\mathrm{Hz}$ ). Power increases from (a) to (c). Below ${\mathrm{ED}}_{50}$ , (a) few exposed cells are damaged. At about ${\mathrm{ED}}_{50}$ , (b) lines of dead cells are one or two cells wide; not all cells are damaged. Above ${\mathrm{ED}}_{50}$ , (c) all exposed cells were damaged and widening of the lines was observed. Clearly separated lines of dead cells, resembling the applied scan pattern, suggest selectivity of the method within the RPE monolayer.

The ${\mathrm{ED}}_{50}$ power for cell damage decreased with the increase of the number of scans from 10 to 100 repetitive exposures [Fig. 7(a) ]. Increasing the dwell time from $3\text{to}10\mu \mathrm{s}$ (i.e., slowing down the velocity of the flying spot) also decreased the necessary power to damage the cells [Fig. 7(b)]. The slopes of the dose response of individual samples were all smaller than 1.1 (Table 1).

Fig. 7

Summary of in vitro experiments. (a) ${\mathrm{ED}}_{50}$ power reduced with the number of repetitions applied to each treatment site as shown for $3\text{-}\mu \mathrm{s}$ exposures at 100 and $500\mathrm{Hz}$ . (b) ${\mathrm{ED}}_{50}$ reduced with exposure time as shown for 10 repetitive exposures. All measurements were performed in duplicates. Therefore, the data points are the average ${\mathrm{ED}}_{50}$ , and error bars represent the absolute deviation.

Morphological changes were not observed under white light illumination except at power levels significantly above threshold; they were more likely to be seen when scans were performed with $500\text{-}\mathrm{Hz}$ repetition frequency. A dependence of the ${\mathrm{ED}}_{50}$ on the repetition frequency was not observed. However, reversible shrinking of the tissue occurred during irradiation with $500\text{-}\mathrm{Hz}$ repetition frequency, when the sample was exposed to power levels three times above the damage threshold $\left({\mathrm{ED}}_{50}\right)$ while scanning 100 repetitive $10\text{-}\mu \mathrm{s}$ exposures.

3.2.

In Vivo Experiments

3.2.1.

Scanner characterization

The $1∕{\mathrm{e}}^2$ spot diameter was measured, by means of detecting the rise and fall time of the rotating knife edge, to be $27.5\mu \mathrm{m}$ in air (Fig. 8 ). The spot size on the rabbit retina was calculated to be $18\mu \mathrm{m}$ after taking into account the Goldman contact lens and the optics of the rabbit eye.¹⁸ Differentiation of the rise and fall time signals from the knife-edge measurement suggested that the beam was nearly Gaussian (Fig. 9 ). Discrepancy between the calculated diffraction-limited spot diameter of $19.2\mu \mathrm{m}$ and measured real diameter of $27.5\mu \mathrm{m}$ indicates a times-diffraction-limit factor ${\mathrm{M}}^2$ of about 1.4.

Fig. 8

Beam profile in air along $\left(\mathrm{a}\right)=x$ and perpendicular $\left(\mathrm{b}\right)=y$ to the direction of motion of the spot in the focal plane. The beam is $27.5\mu \mathrm{m}$ in diameter with a nearly Gaussian irradiance profile. Ratio of the measured spot diameter $(1∕{\mathrm{e}}^2)$ of $27.5\mu \mathrm{m}$ over a calculated diffraction-limited diameter of $19.2\mu \mathrm{m}$ indicates ${\mathrm{M}}^2$ of 1.4.

Fig. 9

Beam diameter around the focus of the system derived from the rotating knife-edge measurements. Data were fit to diffraction-limited beam propagation considering an ${\mathrm{M}}^2$ of 1.4. With this beam propagation, irradiance varies by only 10% over a distance of $+∕-200\mu \mathrm{m}$ around the focus.

Translation of the rotating knife edge in small increments through the beam waist along the optical axis led also to determination of the beam propagation around the focus. In our device, the irradiance applied to the target site varied by only 10% over a range of $\pm 200\mu \mathrm{m}$ on either side of the focus (Fig. 8). The long depth of focus is the result of system design that uses low numerical aperture optics.

3.2.2.

${\mathrm{ED}}_{50}$ Threshold determination

RPE cells can be damaged selectively in vivo as indicated by leakage of the fluorescein dye in the form of the scan pattern with concurrent lack of visible coagulation of the neurosensory retina. Figure 10(a) is a fundus photograph of a rabbit minutes after irradiation. Two columns of marker lesions that served as orientation points are visible. Selective lesions were placed between columns of markers and were ophthalmoscopically not visible (i.e., no retinal whitening) [Fig. 10(a)]. Ophthalmoscopically invisible lesions can be visualized by fundus angiography, where fluorescein can pool into the eye in those areas, where the blood ocular barrier has been compromised by laser mediated RPE cell damage. Thus, Fig. 10(b) shows the fluorescence image of the same site of Fig. 10(a) one hour post irradiation. Here, the marker lesions appear dark in the center with a bright rim, and previously nonvisible selective lesions show up as hyperfluorescent. Thus, angiographic ${\mathrm{ED}}_{50}$ was determined by the angiographic visibility of lesions.

Fig. 10

View through the slit lamp (a) minutes after irradiation and corresponding fluorescence angiographic image (b) taken $1\mathrm{h}$ after irradiation of the same site with (c) additional corresponding sketch of the site. Selective lesions were placed between marker lesions (M), in which photoreceptors have been purposely coagulated by applying a very slow scan speed. Selective $7.5\text{-}\mu \mathrm{s}$ exposures using separate lines were performed at (1) $1\times {\mathrm{ED}}_{50}$ and (2) $2\times {\mathrm{ED}}_{50}$ , and with interlaced lines also at (3) 1 ${\mathrm{ED}}_{50}$ and (4) 2 times ${\mathrm{ED}}_{50}$ . (a) All selective lesions were ophthalmoscopically invisible minutes post irradiation—the view in the lower left side of the image is partially obstructed by reflections (r) of the double-slit illumination from the contact lens. (b) Selective lesions become visible in fluorescence angiography as separated hyperfluorescent lines, where irradiation had been performed with the SEP scan pattern. Lesions close to the right column of markers were produced using the INT scan pattern; consequently, no separate fluorescence is visible.

For the separated lines scan pattern, the cell damage ${\mathrm{ED}}_{50}$ decreased with exposure time from $93\mathrm{mW}$ with $7.5\text{-}\mu \mathrm{s}$ to $66\mathrm{mW}$ with $15\text{-}\mu \mathrm{s}$ exposure. This corresponds to an increase of the ${\mathrm{ED}}_{50}$ radiant exposure from $340\mathrm{mJ}∕{\mathrm{cm}}^2\text{to}482\mathrm{mJ}∕{\mathrm{cm}}^2$ (Fig. 11 ). ${\mathrm{ED}}_{50}$ (both in terms of power and radiant exposure) also decreased with increasing number of exposures. With the available maximum power of the system, we were not able to reach the ophthalmoscopic damage threshold, as indicated by visible retinal coagulation, for all irradiation parameters except for 100 repetitions of $15\text{-}\mu \mathrm{s}$ exposures with an interlaced scan pattern (Fig. 11 and Table 3 ).

Fig. 11

Influence of the exposure time and the number of exposures of the SEP pattern on (a) the angiographic ${\mathrm{ED}}_{50}$ power and (b) ${\mathrm{ED}}_{50}$ radiant exposure. Angiographic ${\mathrm{ED}}_{50}$ radiant exposures were calculated using Eq. 4. All measurements were performed in duplicates. Therefore, the data points are the average ${\mathrm{ED}}_{50}$ and error bars represent the absolute deviation. Ophthalmoscopically visible lesions could not be produced for these parameters with laser power of up to $185\mathrm{mW}$ .

Table 3

Dependence of angiographic ED50 and therapeutic window (TW) on the scan pattern applied [interlaced lines (INT) versus separated lines (SEP)]. For 7.5-μs exposures, no visible coagulation of retinal tissue was achieved with the maximally available power of 185mW . For 15μs , using the interlacing scan pattern, retinal whitening occurred at a power of a factor 1.7 higher than the angiographic ED50 . For the same exposure time applied through the SEP scan pattern, we were not able to coagulate photoreceptors; our maximal available power of 185mW is a factor of 3.2 higher than the angiographic ED50 of that set of parameters. Thus, the therapeutic window for this case would be larger than 3.2.

Parameter	Angiographic ED50 (mW)	Ophth. ED50 (mW)	TW
$7.5\mu \mathrm{s}$ , $N=10$ , SEP	92	$(>180)$	—
$7.5\mu \mathrm{s}$ , $N=10$ , INT	69	$(>18)$	—
$15\mu \mathrm{s}$ , $N=100$ , SEP	57	$(>18)$	$(>3.2)$
$15\mu \mathrm{s}$ , $N=100$ , INT	45	$=75$	1.7

Irradiating with the INT scan pattern, which uses interlacing lines without spacing between lines, reduced the threshold power required for cell damage compared to separated lines (Table 3). Ophthalmoscopic threshold was reached for the most “invasive” parameter used—100 repetitions with $15\text{-}\mu \mathrm{s}$ exposure time. For this case, the ophthalmoscopic ${\mathrm{ED}}_{50}$ was 1.7 times the angiographic ${\mathrm{ED}}_{50}$ . The ratio of ophthalmoscopic ${\mathrm{ED}}_{50}$ over angiographic ${\mathrm{ED}}_{50}$ is commonly referred to as the therapeutic window (TW).

The Probit slopes for all individual eyes were smaller than 1.1, indicating a steep dose response of the irradiated cells (Table 2).

4.1.

In Vitro Experiments

In vitro results showed that it is possible to selectively destroy RPE cells by means of a scanned cw laser spot with a Gaussian irradiance profile. Selectivity is demonstrated by alternating lines of dead and surviving cells that resemble the applied scan pattern (Fig. 6). The ${\mathrm{ED}}_{50}$ power decreased with increasing exposure time [Fig. 7(b)]. This is expected, as more energy is deposited in the tissue, the longer the laser radiation is applied. ${\mathrm{ED}}_{50}$ power in addition decreased with the increasing number of repetitive exposures applied to the irradiation site [Fig. 7(a)]. No significant difference in ${\mathrm{ED}}_{50}$ between the tested repetition frequencies of 100 and $500\mathrm{Hz}$ was found. However, morphological changes that are visible even without fluorescence illumination seemed to be more likely to appear with $500\text{-}\mathrm{Hz}$ repetition frequency. In addition, shrinking of the tissue was observed at a factor of about 3 above ${\mathrm{ED}}_{50}$ for $10\text{-}\mu \mathrm{s}$ exposures at a repetition frequency of $500\mathrm{Hz}$ , suggesting that this combination of parameters leads to significant background temperature increase. Consequently, all in vivo experiments were performed at $100\text{-}\mathrm{Hz}$ repetition frequency.

4.2.

In Vivo Experiments

In vivo experiments in rabbits demonstrated that RPE cells can be damaged selectively, as shown by fluorescein leakage that resembles the scan pattern with concurrent absence of ophthalmoscopically visible coagulation. RPE cell damage was achieved with moderate laser power $(<100\mathrm{mW})$ . For the most invasive parameter, the therapeutic window is 1.7. The fact that we were not able to visibly coagulate neurosensory retina with all other parameters suggests that the therapeutic window is larger for those cases.

The angiographic ${\mathrm{ED}}_{50}$ power decreased with increasing dwell time. At the same time, the ${\mathrm{ED}}_{50}$ radiant exposure increased. As the dwell time increasingly exceeds the thermal confinement time of the RPE cells $(\sim 5\mu \mathrm{s})$ , more energy has to be applied to compensate for heat diffusion.

Surprisingly, we were not able to visibly coagulate the neurosensory retina with our maximal available power of $185\mathrm{mW}$ when applying the SEP scan pattern, even for an exposure time of $15\mu \sec$ , which exceeds the thermal relaxation time of the RPE. This can be explained by the small spot of irradiation. When a single RPE cell is exposed, heat diffusion can occur in three dimensions. Even when a line is irradiated, 2-D heat diffusion away from the cylinder is more efficient than when a large area of RPE is irradiated at a time. In the latter case, heat can only diffuse axially, leading to higher temperature at the neurosensory retina.

Heat diffusion also explains why the threshold for the interlacing lines (INT) scan pattern is lower than for the SEP pattern for the same scan speed. In the most extreme case, we were able to visibly coagulate the neural retina by applying 100 repetitions of the INT pattern with $15\text{-}\mu \mathrm{s}$ exposure time at a factor of 1.7 above ${\mathrm{ED}}_{50}$ (Table 3). In the interlaced lines scan pattern, adjacent lines are created with a time delay of $3\mathrm{ms}$ . During this time, cells adjacent to the first line of the pattern will have been heated by heat diffusion from the exposed, neighboring cells, so that the returning spot adds its energy to their already increased temperature. In addition, the duty cycle of the interlace pattern is well above 50%, so that the heating approaches a 2-D sheet rather than separated cylinders. Thus, heat diffusion into the neural retina is more pronounced than it would be with the separated line scan pattern.

4.3.

Cell Damage Mechanism

The mechanism for selective RPE cell damage was originally thought to be thermal damage.⁸ Lin and Kelly were able to show that formation of microbubbles around the melanosomes is the main mode of cell killing for pico- and nanosecond pulses. They found that a single formation of a microbubble might be sufficient for cell killing.^{20, 21} Brinkmann demonstrated that cell damage by $3\text{-}\mu \mathrm{s}$ pulses originates from bubble formation around melanosomes that ruptures the cell structure.²² Roegener, Brinkmann, and Lin verified that RPE cells died when a single bubble was detected within a train of $6\text{-}\mu \mathrm{s}$ -long pulses.²³ Schuele demonstrated, using an opto-acoustic technique, that cells are damaged mainly by cavitation for $5\text{-}\mu \mathrm{s}$ exposures, while for $500\text{-}\mu \mathrm{s}$ pulses, all cells die without formation of intracellular microbubbles. He detected a mixed damage effect for $50\text{-}\mu \mathrm{s}$ exposures, with most cells being killed without microbubble formation.²⁴ In a related project, we measured bubble formation by detection of backscattered irradiation light, and found it to be the predominant cell damage mechanism for pulse durations up to $20\mu \mathrm{s}$ .²⁵

These bubbles are small cavitation events that are confined to the vicinity of the absorbers near the center of the scan line where the temperature is highest.¹⁵ At threshold, the mechanical stress caused by expansion and rapidly ensuing collapse of the microbubbles is capable of rupturing the cell membrane or damaging intracellular organelles, which subsequently leads to cell death. They are, however, not capable of destroying Bruch’s membrane or photoreceptors. In our experiments, this is indicated in vitro by loss of Calcein fluorescence and an unaltered morphological appearance of the cells after exposure. In vivo subtle damage to the RPE layer manifests itself in fluorescein leakage with concurrent absence of visible photoreceptor damage and absence of bleeding from the choriocapillaris into the ocular space. Roider demonstrated by means of microperimetry in a clinical study that no blind spots appeared due to the laser treatment,¹³ which means that photoreceptor function has not been compromised by either bubble formation or heat diffusion.

4.4.

Outlook

Our results suggest that controlled injury of the RPE cells can be achieved using a laser scanning device without thermal coagulation of the retina. A scanning device can be made more compact and less expensive than the existing system that generates multiple pulse trains. The required (cw) laser power is moderate, allowing for future treatment devices to be built with a solid-state laser source mounted directly on a slit lamp. Furthermore, the scanning approach is more versatile compared to any pulsed approach, because a variety of scan patterns and scan parameters can be performed under software control. By adjusting the scanning speed, it is even possible to perform both selective targeting and conventional (nonselective) laser coagulation using the same device, as we have shown by placing ophthalmoscopically visible marker lesions in the rabbit fundus using reduced scanning speeds.

With our current setup, power delivery to the eye was limited to $185\mathrm{mW}$ due to the use of a single-mode fiber. As a consequence, exposure time was limited to relatively long exposures. Future setups will incorporate a compact laser source of about $1\text{-}\mathrm{W}$ output power mounted directly onto the slit lamp. This will allow for higher power to be delivered to the eye and thus enable exploration of shorter exposure times. Additionally, it will alleviate the need for single-mode fiber coupling and make the setup more robust. For patient treatment, an active feedback device that monitors proper function of the scanning process will be required to avoid overexposure of the eye in case of a failure of the scanner. However, we note that for AO scanners, most likely failure modes will result in the absence of the first-order laser beam, and therefore no light will reach the eye.

As mentioned in the Introduction, the actual treatment effect is presumably achieved by the regeneration of RPE cells to form a new, healthy blood retinal barrier. Although lesions created by the SEP scanning pattern can be expected to heal faster because of the presence of viable cells between the scan lines, further studies are needed to determine whether the smaller total number of cells damaged by the SEP pattern is sufficient for treating symptoms such as an existing edema or drusen. In addition, histological examination is needed to verify whether the absence of visible retinal coagulation means that the outer segments of the photoreceptors are indeed intact. These studies are currently underway.

Due to the small diameter of the spot, dosimetry and accurate focusing in a clinical setting are crucial. Our scanning device was designed to create a long depth of focus. Using a NA of 0.02, focusing inaccuracies of 200 and $500\mu \mathrm{m}$ lead to a loss of irradiance of only 10 and 30%, respectively, in our scanner (Fig. 9). In our rabbit experiments, focusing was not a problem as long as irradiation was performed in the central region of the fundus, where selective treatment is most useful. Due to the expected large therapeutic window, treatment in a clinical setting can be performed well above ${\mathrm{ED}}_{50}$ . However, to ensure efficient and safe treatment in a clinical setting, immediate feedback over the treatment outcome is desirable. At present, treatment outcome is assessed with fluorescence angiography one hour post-irradiation. We are currently developing a system that monitors cell death during treatment by detecting bubble formation, alerting the physician when the endpoint is reached.