1. Introduction

Laser therapy is used in numerous medical and surgical applications through precise ablation, disruption or thermal treatment of exposed tissues. Most treatments require consistent and controlled dosimetry at precise locations. Therapeutic techniques can benefit from use of relatively long duration laser pulses which are limited by the thermal relaxation time of the treated tissue in order to provide localized treatment while avoiding the damage caused by extremely high peak energies. Typically, there are high levels of inhomogeneity in the optical absorption and scattering properties both within and across tissues in the same individual, as well as between different individuals. Thus, for effective locally confined treatments, a sensitive and accurate means to assess the fluence required to deliver a specified dose to the tissue is needed.

In ocular procedures for glaucoma, such as in selective laser trabeculoplasty, optoacoustic approaches are promising avenues for detecting onset of thermo-mechanically induced gas microbubbles produced from pressure waves because of the thermoelastic expansion of the absorbing medium [1]. In retinal diseases, selective retinal therapy (SRT) is an attractive approach because employment of microsecond pulses allows destruction of the RPE with minimal damage to the photoreceptors and choroid [2]. In all cases, a dosimetric feedback mechanism is highly desirable during treatments of SRT and subthreshold micropulse diode laser therapy, where there is no visual response indicator for the ophthalmologist. Adjustment of laser power and temporal duration has shown potential in the development of titration methods as a dosimetry measure for subvisible tissue effects but such dosimetry can be challenging to perform reliably [3,4]. Although studies have demonstrated better outcomes from these new, effective treatments compared to the more destructive photocoagulation treatments routinely used [5], this barrier of obtaining accurate and personalized dosimetry feedback has precluded their implementation in common surgical practice.

1.1. Photoacoustic sensing for laser treatment dosimetry

Photoacoustic (PA) sensing has been applied in an imaging context in many tissues [6]. As a response feedback mechanism, PA has the advantage that it is directly dependent on the absorbed energy dosage for short laser pulses. Given an incident optical fluence, F₀, the fluence F(z), at a depth of z within an absorbing, low scattering, medium is determined by the Beer-Lambert relation, $F(z )/{F_0} = {e^{ - {\mathrm{\mu }_a}z}}$, and depends on µ_a, the absorption coefficient [7–10]. In the photo thermo-mechanical treatment regime, thermal confinement occurs when the laser energy is absorbed at a rate that exceeds the heat diffusion rate in tissue. In the irradiated volume, heat accumulates because it cannot escape through heat conduction. Thermal confinement happens when the laser pulse duration (pulse width) is much shorter than the time of dissipation of the absorbed energy by thermal conduction (tissue thermal relaxation time) [11]. Under thermal confinement conditions, the temperature rise, ΔT, is dictated by the absorbed optical energy divided by the volumetric heat capacity, C_v, and the density, ρ, of the material,

Microsecond pulses used in SRT as well as other applications [4,17–19], produce pressure waves in the ultrasonic MHz frequency range that can be measured with an acoustic transducer. Laser-induced cavitation formation and collapse produces the desired localized photomechanical damage in SRT and generates measurable pressure waves [20]. This makes photoacoustic monitoring attractive for dosimetry. Consequently, significant research investigations have explored PA measurements for treatment dosimetry [5,20–22] and for detection of cavitation events [8,21,22]. However, the microsecond pulse durations greatly exceed the stress confinement time constant, and thus are not optimal for photoacoustic signal generation. Therefore, sensitivity remains an important problem to address for effective dosimetric applications.

Several factors affect the PA response. For example, for a constant pulse energy the pressure generation efficiency across the acoustic spectrum is higher for shorter pulses (1 ns) than for longer pulses (100 ns) [7]. However, the gain for these shorter pulses is diminished in thicker specimens (i.e. depth >3 cm) since they generate higher frequencies which are more strongly attenuated in aqueous media and tissue. Photoacoustic response also varies with the properties and geometry of the absorbing region, including the absorber cross section, concentration/aggregation of absorbers, the size of the laser spot, the absorption and physical properties of the material, and the speed of sound within them [23]. The detector and electronic amplification circuitry are also factors in the captured frequency response. As a result, the final measured PA signal is a convolution of laser parameters, sample properties and geometry, surrounding tissue acoustic transmission properties, instrument detector and amplification responses, as portrayed in Fig. 1. The PA response also includes acoustic and electronic noise which mixes with the photoacoustic signal, ultimately limiting the sensitivity of the system.

 

Fig. 1. Generation of photoacoustic frequency response.

Download Full Size | PDF

Efforts to optimize photoacoustic generation based on modifying temporal laser shapes have not been extensively investigated. Sheinfeld et al. [24] reported on a pulse formatting method to optimize photoacoustic measurements and enhance the signal from select components in a mixture of PA generating sources. It was demonstrated that by modifying the laser pulse temporal profile to match the time-reversed impulse acoustic response of the PA system, the laser-induced photoacoustic SNR can be improved by approximately 1.5-fold versus square or Gaussian pulses. Furthermore, they demonstrated that if the impulse response is tailored to one PA-generating component of the sample, this signal can be amplified preferentially compared to other emitters present. Gao et al. [25] examined the effects of modulated laser pulses on the PA signal and found optimal modulation frequencies for biological tissues, providing ∼2-fold improvement in response over the range of modulation frequencies tested. Because the optimal frequency was dependent on the type of tissue, they suggested that photoacoustic resonance spectroscopy could be used for identification of tissues, or tissue properties.

The objective of the present work is to investigate the application of amplitude modulation of microsecond pulse profiles on retinal explants to demonstrate a proof of concept of an approach to optimize the PA feedback SNR in a context where longer laser pulses are employed. This presents an important first step relevant to future high SNR SRT dosimetry in vivo, and other similar applications.

2. Materials and methods

2.1. Phantom preparation and characterization

Recent studies report the suitable acoustic and optical properties of polyvinyl chloride plastisol (PVC-P) for the fabrication of PA phantoms [26–30]. Its properties (e.g. the speed of sound, heat capacity, and coefficient of thermal expansion) mimic those of biological tissues [27,30]. Phantoms were therefore prepared from a PVC-P formulation (701-89 Vynaflex plastisol compound, Gripworks). The selected black pigmented PVC-P mixture was chosen for its elevated coefficient of absorption, µ_a, mimicking the highly absorbent pigmented retinal epithelium layer (RPE). It is also efficiently generates a strong repeatable acoustic response.

The plastisol solution was degassed under vacuum for 3 h, while a metallic mold was pre-heated to 155°C. The degassed solution was poured to form a 6 mm thick layer in the heated mold and cured at 155°C for 5 minutes, and then allowed to cool before demolding. Phantoms were cut into 2 cm x 2 cm sections.

The speed of sound in the PVC-P phantom was determined by subjecting the bottom surface of the phantom to 1 ns laser pulses and measuring the time required for the midpoint between compression and rarefaction acoustic waves to reach the hydrophone, and dividing this result by the thickness of the phantom. The heat capacity of the phantom was determined in a differential scanning calorimeter (DSC6200, Perkin Elmer), using a 10 mg sapphire disk as a heat capacity reference and based on the three-run method previously described [30]. Phantom spatial uniformity was evaluated by measuring the laser-induced PA response in constant conditions (laser spot size and fluence) across the phantom surface.

2.2. Retinal explants preparation

Rabbit heads from young pigmented brown rabbits were obtained from a local slaughterhouse. The rabbits were not genetically chosen. Retinal explants were extracted from fresh rabbit eyes, in accordance with the Association for Research in Vision and Ophthalmology (ARVO) Animal Statement and our local institution’s guidelines. Following enucleation, eyes were cut around the iris region. The anterior portion of the eye, the vitreous humor and the retinal layer were removed, and the remaining posterior globe was sectioned to facilitate planar horizontal mounting. The collected explants were kept for a maximum of 2 h at room temperature in phosphate buffer saline for experimentation.

2.3. Experimental setup and data analysis

The experimental setup configuration is shown in Fig. 2(a). A pulse programmable fiber laser (MOPAW, 532 nm) [31] was employed using custom-programmable temporal pulse shapes from 3 ns to 2 µs in duration. Amplitude modulated pulses were created as a single laser pulse by programming the temporal amplitude profile of our fiber laser. Our approach is to apply high frequency amplitude modulation to these microsecond pulses as our modulation scheme to create various modulated pulses, in some cases resembling pulse trains as shown in Fig. 2(a) inset. Thus, high frequency content was introduced into microsecond pulses of the type employed in SRT [5]. The laser beam was focused onto the sample (PVC-P phantom or retinal explants shown in Fig. 2(b) and 2(c), respectively) by a long working distance objective lens. Beam displacement was achieved using a pair of motorized galvanometer mounted mirrors (Thorlabs, GVSM002). An acousto-optic modulator (AOM) (Intra action, AFM-1102A1) was used to control the laser treatment pulse energy. Pulse energy was determined with a joulemeter tap (Coherent, J-10Si-Le) and the pulse shape was monitored using a fast photodetector (Electro-Optics Technology Inc., GaAs PIN detector ET-4000). Laser spot size was measured to be 28 µm as imaged by a CCD camera using 1/e² intensity threshold for a non-saturated image of the laser beam transverse profile. Similar to previous studies, acoustic waves were captured with hydrophone [32] model HNC-1000 (Onda) coupled to an AH1100 20dB amplifier (Onda) with a 20 MHz bandwidth. In some experiments, a HeNe laser at 633 nm (Melles Griot, 5 mW) was used as a reflection probe to confirm the presence of cavitation events [32] (data not shown). The signal was detected by an avalanche photodetector (ThorLabs, APD 120A2). A LabVIEW program was designed to scan samples with the treating beams, control the treating laser pulse energy and acquire the PA signals for each treatment zone. Two electronic cards (NI PCIe-6323 and GaGe CSE1222-4G) were used to control the galvanometric mirrors, the acousto-optic modulator, and acquire data from the hydrophone, the power meter, and the avalanche photodiode.

 

Fig. 2. (a) Schematic representation of the experimental setup. (c-b) Photographs of the hydrophone positioned under the microscope objective in the proximity of a (b) phantom or (c) a retinal explant.

Download Full Size | PDF

Photodetectors and amplifier data were stored on disk. Photoacoustic signal amplitude, PA power spectral density, and acoustic wave spectrogram calculation as well as image reconstruction were obtained with custom MATLAB scripts.

3. Results

3.1 Phantom properties

The fabricated phantom, shown in Fig. 3(a), was used as a simple, homogeneous, stable substrate to test different laser pulse shapes. Its properties were measured to validate its similarity to reported biological tissue and designed to match tissue acoustic properties. The speed of sound in the phantom was found to be approximately 1520 m/s. The measured specific heat capacity was 1.1 J/gK at 35°C which is similar to the previously reported value of 0.85 for PVC without plasticizer [33]. Comparing to tissues, this is significantly lower than most soft tissues with high water content (3-4 J/gK), though similar to cortical bone [34] sufficient for validation purposes. The PA response uniformity over the phantom surface was measured to be 10% relative standard deviation (Fig. 3(b)).

 

Fig. 3. (a) Photoacoustic phantom (a) and averaged PA response (b) for a fixed energy of 0.8 mJ. The measured standard deviation shows photoacoustic signal uniformity (n = 25 and 24 PA signals, respectively).

Download Full Size | PDF

3.2. Modulated microsecond pulses and photoacoustic SNR improvement in phantom

The photoacoustic SNR ratio was then evaluated for 0.6 µs pulses with various superimposed amplitude modulations applied to the pulse. Square wave amplitude modulation with 50% duty cycle were applied to maintain both the peak power (2X the unmodulated pulse) and overall energy equal for each of the tested modulated pulse shapes. Laser pulse amplitude modulation frequencies were chosen at relatively regular intervals to test different regions of the hydrophone response, ranging from 2.5 − 28 MHz. In addition to modulated 0.6 µs pulses, non-modulated 3 ns pulses were also assessed in the same experiment to provide a source of high frequency content that could provide insight as to the frequency response of the system. Photoacoustic measurements were also taken with the laser blocked, to allow estimation of the frequency content of instrumental noise sources independent from the generated PA signal. The average frequency responses of the 3 ns pulses and the noise data collected are plotted in Fig. 4(a). It is observed that higher SNR is present in the 3 − 7 MHz range as well as the 27 − 29 MHz range, and low SNR is seen below 2 MHz, between 10 − 27 MHz, and above 30 MHz.

 

Fig. 4. (a) Averaged frequency response of 3 ns duration pulses (black) and the noise data (grey). (b) Signal to noise ratio obtained with different modulated pulse shapes. Representative pulse shapes are illustrated below the x-axis, where all overall pulse envelopes are 600 ns in length, with the exception of the 3 ns pulse. Pulse energies and modulation frequency are listed in the x-axis labels. Bandpass frequency filtered data are defined in the figure legend. Each data point is the average of 8 PA measurements.

Download Full Size | PDF

Collected from the oscilloscope, the non-filtered PA amplitude signal demonstrated very little improvement in the SNR for any of the modulated pulses, with a maximal SNR improvement over a non-modulated 600 ns square pulse being 1.7-fold for the 2.5 MHz pulse (Fig. 4(b)). However, by applying bandpass filters to the data centered on the frequency of the pulse modulation, much larger gains could be observed in the regions expected to have higher SNR based on the 3 ns pulse results. Maximal SNR was obtained for the 4.2 MHz modulated pulses, when the 3.2 − 5.2 MHz bandpass filter was applied, resulting in a 7-fold increase in SNR as compared to the non-modulated 0.6 µs square pulse of same energy with the same filter (or a 14-fold increase over the SNR of the raw signal). While the modulated pulses are expected to have a 2-fold increase in peak excitation power over the non-modulated square pulse due to the duty cycle (50% vs. 100%), this is not sufficient to explain the increase in SNR. In fact, a 3 ns pulse was found to have only a 2.2-fold increase in SNR compared to the 4.2 MHz modulated 600 ns pulse (both signals have optimal SNR using 4.2 MHz bandpass filter) despite a difference of approximately 100-fold in peak power. For modulated pulses at frequencies that were expected to have low SNR (14 MHz), the effect of the pulse modulation was minimal, with the PA response for both filtered and non-filtered results being comparable to the 0.6 µs square pulse.

3.3 Modulated microsecond pulses and photoacoustic SNR improvement in phantom: 2D mapping

From knowledge that modulation frequencies ranging from 3 − 7 MHz gave the best SNR results in our system, we imaged the same section of a PA phantom with non-modulated and 5 MHz-modulated microsecond pulses of the same energy, 0.4 µJ/ pulse (see Fig. 5).

 

Fig. 5. PVC-P phantom image as opbtained with non-modulated (a) and modulated (b) laser pulse obtained with a 0.1 − 100 MHz (left) and a 4.5-5.5 MHz (right) filter window. Filtered (dark blue) form as well as the envelope (black) of the acoustic wave (light blue) for a given image coordinate are represented for each images.

Download Full Size | PDF

For each image pixel, the acoustic wave was digitized. Then a 3 × 3 pixel moving average was applied. Acoustic signals were digitally filtered and the signal envelope was evaluated as the absolute value of its Hilbert transform. Images were constructed using the peak-to-peak amplitude of the Hilbert transform. Examples of the filtered form as well as the envelope of the acoustic wave for a given coordinate are represented for each image in Fig. 5 to show the signal processing leading to the image construction. Phantom images obtained with the non-modulated microsecond pulse (Fig. 5(a)) suffer from poor SNR (for both 0.1 − 100 MHz and 4.5 − 5 MHz filter window). However, phantom images obtained with 5 MHz-modulated pulses (Fig. 5(b), left) show a much better SNR ratio. This is mainly due to the 50% duty cycle affecting the peak power as discussed above. When applying a filter window centered on the modulation frequency (filtering all the noise outside of the laser modulation), the SNR can be further increased. In fact, the SNR obtained with a modulated-µs pulse is similar to that obtained with a single 10 ns pulse, a pulse duration commonly used for PA imaging (see Table 1).

Table 1. SNR for different pulse formats.

View Table

3.4. Photoacoustic frequency-domain cavitation detection in retinal tissue

In addition to SNR improvement, we examined pulse modulation for detection of cavitation or other non-linear events. When using a pulse modulation with a narrow spectral band (i.e. sinusoidal modulation), an acoustic wave is observed in a similar spectral band as long as the pulse energy remains in the linear thermoelastic regime. When a non-linear event occurs, such as a cavitation, ultrasound frequencies outside of the modulation band are generated. Thus, using spectral analysis, one can extract cavitation occurrence. Figure 6 shows the spectrograms of the PA signal obtained from a retinal explant with different laser pulse energies.

 

Fig. 6. (a) Spectrograms of the PA signals obtained from a leporine retinal explant with different laser pulse energies (0.09 − 0.59 µJ). As laser pulse energy increases and cavitation threshold is reached (around 0.33 µJ), frequencies outside of the modulation bands appear. Normalized laser pulse format is represented in blue (lower section). Spectral densities shown integrated over 4.5-5.5 MHz (b) and 0.1-3.5 MHz (c) spectral windows for all the laser pulse energies used in (a).

Download Full Size | PDF

When this energy is sufficient to generate a measurable acoustic wave, the acoustic signal shows a clear band at the modulation frequency (5 MHz) and very little signal outside of this band (see spectrograms for 0.18 µJ/pulse and 0.25 µJ/pulse in Fig. 6(a)). For energies beyond cavitation threshold (confirmed via cavitation detection using He-Ne probe, data not shown), energy is transferred by the cavitation event to other frequencies, reducing the spectral content at the laser modulation frequency. Note that cavitation microbubbles also shield the laser energy absorption in the focal area as expected and previously observed [32], and the thermoelastic signal at 5 MHz is also reduced (see spectrograms for 0.33 µJ, 0.42 µJ and 0.59 µJ in Fig. 6(a)). Figure 6(b) and 6(c) shows the spectral density integrated over different spectral windows as a function of time. The spectral density within the 0.1 − 3.5 MHz band clearly shows the distinction between cases under and above cavitation threshold (Fig. 6(c)).

3.5. Mapping treatment outcome in retinal tissue

We then applied this improved SNR and non-linear event detection method, enabled by pulse modulation, to probe laser treatment in a rabbit retinal explant model. We treated the retinal explant with 2 µs laser pulses modulated at 5 MHz (the laser pulse temporal profile was the same as in Fig. 6(a)). The treatment beam was scanned on a retinal sample containing two regions with distinct absorption characteristics due to the presence or absence of the RPE cellular layer (Fig. 7(a)), and the generated PA signal was monitored. Note that for a given scan, a constant treating spot size was used and scanned over the area (sample from 7a) to build, point by point, a PA response map.

 

Fig. 7. Photoacoustic laser treatment mapping in retinal explants. (a) Image of the scanned retinal explant. (b) Map of the photoacoustic signals recorded as the treating laser pulse was scanned over the area shown in (a). Three scans were performed in the same area with increasing treating pulse energy (right to left). Spatial scale bar in (a-b) is 50 µm. (c) Spectrograms of the PA response induced by a laser pulse (0.67 µJ, 5 MHz, 2 µs pulse duration) before, during and after treatment pulses for 3 sample coordinates represented in (b). Temporal scale bar in (c) is 1 µs.

Download Full Size | PDF

The obtained PA feedback maps are shown in Fig. 7(b). Although the energy was kept constant for a scan, the PA response for each area/pixel in the map varied due to varying adsorption properties. Figure 7(c) represents the spectral signature of the acoustic wave for 3 different coordinates (pulse energy = 0.67 µJ, pulse energy density = 109 mJ/cm²) from which we can extract the treatment outcome. Such spectral analysis clearly shows the generation of ultrasound outside of the modulation band (characteristic to cavitation occurrence) visible in coordinates where absorption was high.

In general, PA feedback images provide a clear distinction of the two regions above cavitation threshold. However, below cavitation threshold, the PA signal is hard to distinguish from noise (see Fig. 7(b)). Subthreshold (thermoelastic) feedback can be improved using spectral analysis. Filtering the signal around 5 MHz allows extracting the thermoelastic portion of the signal and enables recovery of anatomical information (or the level of absorbed laser energy, as seen in Fig. 8(a)). Additionally, mapping the signal around the cavitation-induced frequency band (Fig. 8(b)) is interpreted as the cavitation-related non-linear PA response and allows mapping of the desired photomechanical treatment effect. Both thermoelastic and photomechanical damage maps could be used as sensitive dosimetry feedback in laser treatment.

 

Fig. 8. Photoacoustic mapping with spectral analysis. (a-b) Photoacoustic images for different treatment laser pulse energies with spectral filtering window of 4.5-5.5 MHz (a) and 2-3.5 MHz (b). (c) Binary map showing the presence (white pixels) or absence (black pixels) of signal at cavitation-induced frequency. We used a SNR of 3 (or 4.77 dB) as a threshold for cavitation signal detection. Scale bar is 50 µm in (a − c).

Download Full Size | PDF

For visualization purposes, we also mapped cavitation occurrence in a binary map where white pixels represent cavitation locations (thus photomechanical damage), and where black pixels represent location where no cavitation occurred (Fig. 8(c)). We assumed that we were in the presence of a cavitation when the signal around the cavitation-induced ultrasonic frequency band SNR reached a reasonable value of 3 (or 4.77dB). For each pixel the SNR was defined using the averaged energy E_S and E_N (energy of signal and energy of noise, respectively) of the monitored signal, s(t), and noise, n(t) (ultrasound recorded before the laser pulse occurrence) as per the following relations [35],

In Eqs. (4) and (5), T represents the analysis window duration, which is 1.5 times the laser pulse duration in the present case. This assumption was observed to optimize the amount of false negative and false positive results when compared to visual cavitation assessment.

4. Discussion

4.1 SNR improved photoacoustic feedback

The present work demonstrates that, in a method similar to frequency-domain PA spectroscopy, pulse modulation can be used to optimize the laser-induced acoustic feedback SNR. We have demonstrated with a homogeneous tissue PA phantom that our approach can yield a greater than 7-fold improvement in SNR over non-modulated square pulses of the same duration and pulse energy. Also, while microsecond square laser pulses of 0.4 µJ yielded poor phantom images, modulated pulses with the same duration and energy provided clear phantom images (Fig. 5). The spectral response of the hydrophone appears to be the key determinant for the SNR of the PA spectrum. However, this is not the only factor. Tissue attenuation rates are often more complex than distilled water. Multiple absorbing geometries may have more complicated effects on the emitted PA spectrum and the use of flatter response detectors and amplifiers can lead to cases where the sample dictates the optimum frequency for detection. In fact, it has been demonstrated that certain geometries, sizes and concentrations of absorbing elements within samples can produce different frequency responses compared to others [23,25] allowing some level of selectivity when targeted with these frequencies. In these cases, it is useful to provide optimal pulse shapes to produce the highest possible SNR with the lowest laser energy. This can be achieved by properly taking advantage of the differential impulse response of the targets of interest. Moreover, in future implementations, the use of modulated pulses could be paired with highly sensitive resonant acoustic transducers to help photoacoustic mapping of selective retinal therapy outcomes in real time [36].

4.2 Cavitation detection at optimal damage confinement in retinal explants

In laser-based medical procedures, it is desirable to monitor cavitation threshold as an indicator of cellular damage [37]. As such, one can optimize damage confinement in treated tissue and improve therapy outcomes. When modulating the treatment laser pulse, the generated acoustic pressure has a spectral signature around the modulation frequency. Given that the very fine spectral response induced by a modulated treatment laser pulse differs from the cavitation-induced frequency response, spectral analysis can recover information and map the nature of the light-tissue interaction in the treatment site (as seen in the cavitation map of Fig. 8(c)). The energy densities used where significant cavitation was determined by spectral analysis (0.4 µJ; 65 mJ/cm²) is similar to the cavitation thresholds identified in previous studies [38]. Using this strategy, one can extract quantitative data from the PA feedback signal.

4.3 Implications for photoacoustic feedback dosimetry in ophthalmic applications

In applications such as SRT, one can envision that improving the SNR for sub-cavitation threshold PA transients, with modulated laser pulses, provides a flexible method for higher sensitivity laser dosimetry. The photoacoustic signal can potentially provide both pre-treatment optical absorption mapping as well as enable treatment and therapy feedback solutions, such as in SRT. Moreover, this can be achieved sharing the same wavelength distribution (with respect to absorption and scattering properties) and same instrument (optical path/alignment). Increased SNR of PA feedback could be particularly relevant in treatments which require sub-cavitation threshold/thermoelastic regime laser fluences, as these regimes necessitate a high sensitivity indicator for damage. Since these treatments typically use microsecond pulses, there is a definite prospect to improve the PA feedback using the modulation strategies described here [39]. In a broader application context, this approach of high SNR photoacoustic absorption mapping achieved through laser pulse formatting may hold interesting opportunities in dosimetry or treatment feedback evaluations such as through PA examination of fluid movement inside the eye in treatment of glaucoma, as recently described [40].

In this study using retinal explants, we have demonstrated nearly an order of magnitude improvement in PA SNR with our pulse modulation approach. Nevertheless, its application to more complex clinical models, in which both light and ultrasound waves must travel through the whole eye, remains to be investigated and validated with suitable in vivo instrumentation. Nevertheless, its application to more complex clinical models, in which both light and ultrasound waves must travel through the whole eye, remains to be validated. We previously demonstrated in melanosome models that in longer laser pulses (> 100 ns) cavitation threshold radiant exposure and dynamics above the threshold strongly depend on the pulse format [32] whereas we observed that continuous pulses had similar effects on the cellular damage threshold of retinal explants [41]. This is because pulse modulation in the picosecond time scale is faster than the mechanical response of the system. These results suggest that sub-microsecond laser pulse shaping could be exploited to optimize precision in numerous applications of laser-directed microcavitation and control cellular damage thresholds with longer pulses. In concert with our modulated PA dosimetry scheme presented here, multimodal treatment with highly sensitive PA feedback may be feasible. However, more studies must be done with retinal explant and clinically relevant models to demonstrate that modulated laser pulses (with higher peak powers) have similar outcomes in a clinical context.

5. Conclusion

Photoacoustic feedback detection is a promising method for SRT dosimetry [21,22]. However, microsecond laser pulses are not optimal for high amplitude PA wave generation since their pulse durations are above the stress confinement time constant. Here, we demonstrated that a pulse programmable fiber laser, generating pulse shapes at modulation frequencies that match the maximal response of the acoustic system significantly improves the PA signal to noise ratio. This technique allows for a greater ability to extract dosimetric signals out of a noisy background as the acoustic response can be subjected to a tight frequency bandpass filter in post-processing, eliminating the vast majority of system noise. The present work establishes that modulated pulses can lead to higher subthreshold PA feedback and more effective cavitation detection thereby potentially providing a more advanced approach for mapping laser treatment outcomes.