Scattering properties and femtosecond laser ablation thresholds of human and canine vocal folds at 776-nm wavelength

Abstract. Ultrafast laser ablation may provide a treatment for vocal fold (VF) scarring. Optical properties of VFs must be known prior to clinical implementation to select appropriate laser surgery conditions. We present scattering lengths of epithelium ℓs,ep, superficial lamina propria ℓs,SLP, and ablation thresholds Fth of human and canine VF tissues. Our experimental approach involves an image-guided, laser-ablation-based method that allows for simultaneous determination of ℓs and Fth in these multilayered tissues. Studying eight canine samples, we found ℓs,ep=75.3±5.7  μm, ℓs,SLP=26.1±1.2  μm, Fth,ep=1.58±0.06  J/cm2, and Fth,SLP=1.55±0.17  J/cm2. Studying five human samples, we found ℓs,ep=42.8±3.3  μm and Fth,ep=1.66±0.10  J/cm2. We studied the effects of cumulative pulse overlap on ablation threshold and found no significant variations beyond 12 overlapping pulses. Interestingly, our studies about the effect of sample storage on the scattering properties of porcine VF show a 60% increase in ℓs,ep for fresh porcine VF when compared to the same sample stored in isotonic solution. These results provide guidelines for clinical implementation by enabling selection of optimal laser surgery parameters for subsurface ablation of VF tissues.


Introduction
Vocal fold (VF) scarring is a primary cause of voice disorders, which affect an estimated 2 to 6 million people in the U.S. alone. [1][2][3] Scar tissue forms as an inflammatory response to external trauma, hindering phonation by reducing viscoelasticity of the vibratory layers of the VF. 4,5 The lamina propria (LP), a subepithelial tissue layer consisting primarily of collagen, elastin, and reticulin fibers, is largely responsible for the vibratory response associated with normal phonation and thus is highly sensitive to scar formation. Current research focuses on restoration therapies for the superficial lamina propria (SLP) including development of various injectable therapeutic biomaterials that could promote a more favorable reparative response. [6][7][8][9][10] These therapies would benefit from the creation of subepithelial voids within the scarred tissue region of interest (ROI), allowing for a "pocket" to localize injected biomaterials for an extended period. Preclinical studies, demonstrating that injection into stiff scar tissue results in poor biomaterial localization and treatment repeatability, highlight the potential therapeutic benefit of subepithelial void formation. [11][12][13] One possible solution utilizes ultrafast laser ablation to create subepithelial voids within the SLP. 14,15 The nonlinear dependence of multiphoton absorption on intensity confines energy deposition to femtoliter sized focal volumes if appropriately high numerical aperture (NA) focusing optics are used. In addition, creation of deep voids becomes feasible when using wavelengths within the optical window of tissues (λ ≈ 700 − 1400 nm), allowing for bulk material dissection without disturbing superficial tissue layers. Prior work by our group successfully demonstrated creation of voids ∼100 μm beneath the tissue surface, well within the SLP of porcine VF with an epithelium thickness of 50 to 80 μm. 14,16 Optical scattering properties of VF tissue must be known before implementing subsurface void formation for biomaterial injection in human patients. The literature only presents a small number of studies discussing the optical properties of human VFs. Among these studies, Mahlstedt et al. 17 utilized a double integrating sphere with Monte Carlo simulations to measure scattering and absorption coefficients over the 400-to 2200-nm wavelength range. Their results indicate a scattering length l s ¼ 83 μm at λ ¼ 700 nm for bulk human VF. Since they homogenized their samples in a precooled mortar prior to measurements, they did not have the ability to differentiate between optical properties of the epithelium and SLP. Additionally, homogenization is known to decrease optical scattering, thus reducing similarity to native tissue. 18 Prior work by our group and others has shown variations in optical properties for different tissue layers, 16,[19][20][21] and a priori knowledge of the optical properties of individual tissue layers is required to implement subsurface void creation for the proposed scarred VF treatment method.
In this paper, we present scattering properties for different tissue layers of freshly excised human and canine VFs and study how tissue storage conditions in isotonic solution can vary these properties. We also present ablation thresholds to create voids below the tissue surface and study how the number of overlapping laser pulses can affect these ablation thresholds. To overcome limitations associated with current tissue scattering measurement methods and enable assessment of multilayered tissues, we use a method based on ultrafast laser ablation. This image-guided method, which we proposed in Ref. 16 and expanded in Ref. 19, allows for simultaneous determination of scattering lengths and ablation thresholds of individual layers in thick, multilayer tissues without requiring any sample processing. Results, obtained from multiple human VF samples as well as canine, which is the animal model of choice for preclinical testing of laryngeal treatments, provide a guideline for designing the necessary laser surgery conditions for future clinical treatment.

Experimental Approach
To measure the optical properties of excised human and canine VF samples, we used a laser ablation-based method that we recently developed. 19 Using this method, we can simultaneously determine both scattering lengths and ablation thresholds in multiple layers of a heterogenous tissue. The method relies on finding input energies required to initiate ablation at multiple depths in each tissue layer. Assuming only ballistic photons contribute to nonlinear ablation, the threshold fluence at the focal plane of a single-layered tissue can be described by Beer's law: E Q -T A R G E T ; t e m p : i n t r a l i n k -; e 0 0 1 ; 6 3 ; 5 0 9 F th ¼ where E th;surf is the input threshold energy at the tissue surface, w 0 is the 1∕e 2 beam radius, z ab is the ablation depth in tissue, and l s is a direct measure (i.e., not reduced) of the scattering length. If E th;surf is found at two depths, one can then determine the scattering length according to One can determine threshold fluence F th by inserting l s into Eq. (1). For multilayered tissues, Eq. (2) can be used to solve for l s if each tissue layer is considered as a distinct, homogeneous medium. Threshold fluences in deep tissue layers can be found by extending Eq. (1) to consider the properties of multiple superficial layers, in our case epithelium and SLP: where z ep and l s;ep are the thickness and scattering length of the epithelium, and l s;SLP is the SLP scattering length. For homogenous media, damage is expected to occur at a sharp threshold, as is the case for most metals and glass. 22,23 However, this assumption does not hold in tissue, as heterogeneity in tissue scatterer distribution may lead to a broad range of energies over which ablation occurs. To address the effects of tissue heterogeneity on threshold determination, E th;surf is found by recording percent ablation in a given field of view (FOV) for a range of pulse energies near threshold. The pulse energy required for 50% FOV ablation is found by fitting an error function to the collected data: where ΔE is the difference in pulse energies required to reach 25% and 75% FOV ablation. If E th;surf is found at multiple depths in each tissue layer, a nonlinear least squares fit of the natural log of E th;surf versus ablation depth allows for determination of l s , defined as the inverse slope [according to Eq. (2)] and F th [according to Eq. (3)].
The experimental setup includes a home-made, upright twophoton microscope that employs an Er-doped fiber laser (1.0 ps pulse width, frequency doubled to λ ¼ 776 nm, Discovery, Raydiance Inc.) for both imaging and ablation. We use a repetition rate (RR) of 303 kHz for imaging and ablation during experiments. A pair of galvanometric mirrors raster scans the incoming laser beam, which is expanded onto the back aperture of a coverslip-corrected air objective (20×, NA 0.75, Plan Apo, Nikon) and focused on the tissue. Two-photon autofluorescence is epi-collected, filtered (Schott BG39), and detected by a photomultiplier tube (PMT). The galvanometric mirrors, PMT, and stages are controlled by MPScan software suite. 24 To determine the beam waist at the focal plane inside scattering media, we imaged 100-nm fluorescent beads (FluoSpheres™, 2% solids, Invitrogen™), suspended in 2% low-melt-point agar gel. We added 0.95-μm polystyrene beads (Bangs Lab) to the suspension at a concentration of 1.85 × 10 10 beads∕mL to provide an estimated l s ≈ 46 μm, representing the expected scattering length of tissue. Bead solutions were added to agar, vortexed, set within a coverslip mold with spacers of known thickness, and allowed to solidify. Analysis of two-photon images at multiple depths determined 1∕e 2 beam waste radius w 0 ¼ 0.75 AE 0.01 μm. This value was consistent over 10 depths in the 0to 96-μm range, indicating minimal change in spot size due to spherical aberrations and scattering.
We tested eight canine and five human VF samples. Canine VFs were harvested immediately following euthanization for unrelated research projects. Human VF samples were collected under an Institutional Review Board (UT Southwestern)approved tissue collection protocol. The tissue was excised from VFs unaffected by tumor in total laryngectomy specimens immediately after the larynx was removed in surgery. For both canine and human samples, we excised the specimen for testing as a rectangular piece of VF mucosa from the mid-VF by carefully dissecting the mucosa off the underlying muscle. The mucosal specimens were placed in phosphate-buffered saline (PBS) on ice and overnight shipped to UT Austin. The samples were kept at 4°C until testing, which typically took place within 24 to 48 h of tissue harvest. For measurements, we set the VFs in a Petri dish with 170-μm glass coverslip and submerged them in PBS to prevent desiccation. Fluorescent beads deposited onto the tissue surface aided in detection of the epithelium. Prior to ablation experiments, we determined the epithelial thickness by obtaining z-stacks in three different areas adjacent to the ROI. The RR of the laser was changed from 303 kHz to 2 MHz during z-stack image acquisition to acquire higher quality images before starting the experiments to determine average epithelium thickness with high certainty.
To measure optical properties of VF samples, we ablated a 40 × 40 μm 2 FOV in a single frame by scanning 512 lines at 1.11 frames per second (fps), resulting in N ≈ 301 pulse overlap. We found the number of spots available in our ablation FOV by dividing the FOV size (40 × 40 μm 2 ) by our beam area (πw 2 0 ) and used this with the RR of the laser to calculate N according to the following: E Q -T A R G E T ; t e m p : i n t r a l i n k -; e 0 0 5 ; 3 2 6 ; 1 0 2 Journal of Biomedical Optics 085005-2 August 2019 • Vol. 24 (8) We acquired images over a slightly larger FOV of 80 × 80 μm 2 at 1.11 fps to cover the entire pre-and postablated areas to determine the size of the ablated region. The pulse energy required for 50% ablation E th;surf was found at eight depths per sample: four depths in the epithelium and four depths in the SLP. Ablation curves were generated at each depth by titrating over a range of pulse energies and recording the percent of FOV ablated at each energy. We acquired images as the average of 10 images and performed smoothing using a Gaussian filter (window size 10 × 10) prior to binarization. We utilized Otsu's method to threshold before and after images and applied the same threshold to each image pair to calculate ablation percentage as the ratio of pixels below threshold after ablation to pixels below threshold before ablation within FOV.

Human and Canine Vocal Fold Scattering Length and Ablation Threshold Study
Example two-photon autofluorescence images, taken before and immediately following ablation in canine VF #1, are shown in Fig. 1(a) for a range of pulse energies at a depth of 36 μm. Only one depth in a single specimen is displayed for brevity. We could observe in all two-photon images a smooth transition from cellrich epithelium to LP with randomly organized collagen fibers, which are seen in normal SLP. 16 Data shown in Fig. 1(c) represents pulse energies and corresponding FOV ablation percentage at z ep ¼ 36 μm in canine VF #1. Collected data were fit according to Eq. (4), and the dotted red line indicates the point of 50% ablation where we defined E th;surf ¼ 43.9 AE 1.5 nJ with ΔE ¼ 5.5 AE 1.6 nJ. Although curves similar to Fig. 1(c) were generated for every depth in each tissue sample, a single dataset is displayed for brevity.
We repeated this procedure to determine E th;surf at the seven remaining depths on the VF of canine #1. Figure 1(d) shows the semilogarithmic plot of E th;surf versus ablation depth from which l s was deduced according to Eq. (2) and F th;ep and F th;SLP were then found using Eqs. (1) and (3), respectively. The process was repeated for the seven remaining canine tissues and five human specimens. Results are tallied in Table 1.
Uncertainty in individual l s;ep and l s;SLP measurements represents the standard error in fitting parameters for a weighted nonlinear least squares estimation of parameters. Weights were taken as the inverse of the variance for each approximated E th;surf value. Uncertainty in E th;surf and beam waist (w 0 ) measurements were propagated forward to determine uncertainties presented for F th;ep , and uncertainty in F th;SLP was estimated using uncertainty propagation to account for uncertainty in z ep , l s;ep , l s;SLP , and E th;surf . Uncertainty of averaged values represents the root-mean-square error of contributing data points.
We determined epithelium thickness z ep by measuring the depth of transition from the cell-rich epithelium to the densely packed collagen fibrils of the SLP. We measured z ep in three areas on each sample near ablation data collection ROIs. The presence of two distinct tissue layers is substantiated by our ablation threshold data, where we observed an abrupt increase in the pulse energies required to initiate ablation near z ep , indicating a medium change. Scattering is likely lower in the epithelium due to sparse distribution of cell nuclei and other subcellular scatterers, whereas tightly packed collagen and elastin fibrils distributed throughout the SLP are highly scattering, leading to increased surface pulse energies required to initiate ablation. Further, uncertainty is higher in E th;surf determined for epithelium than it is in E th;surf of the SLP. This can be explained by regularly distributed fibrous protein networks in SLP that help to maintain tissue structure postablation, making it easier to determine percentage of FOV removed and thus E th;surf with higher certainty. Ablation curves follow the fit defined by Eq. (4) very well, and small step increases in pulse energy (typically 3 and 8 nJ increments for epithelium and SLP, respectively) near the point of 50% ablation helped to achieve a well-defined threshold. Although low variance was observed in scattering properties within species, there was a noticeable difference between the scattering lengths of human and canine tissue. We obtained an average canine l s;ep ¼ 75.3 AE 5.7 μm and an average human l s;ep ¼ 42.8 AE 3.3 μm. The shorter scattering length of human VF limited experiments to the epithelial layers, as the pulse energies required to ablate at deeper depths exceeded the available laser power.
In general, the width of the ablation threshold curve ΔE increased with depth, likely caused by the accumulated influence of tissue inhomogeneity at deeper depths. We found the average fluence threshold for human epithelium (F th;ep ¼ 1.66 AE 0.10 J∕cm 2 ) to be slightly larger than those for canine epithelium and SLP (F th;ep ¼ 1.58 AE 0.06 J∕cm 2 and F th;SLP ¼ 1.55 AE 0.17 J∕cm 2 ).

Porcine Pulse Overlap Study
We utilized the same experimental methodology presented above on freshly excised inferior porcine VFs to study how varying the number of overlapping pulses N affects our l s and F th measurements. We tested four different pulse overlap conditions by varying FOV and/or frame rates: N ¼ 301 (FOV ¼ 40 × 40 μm 2 , frame rate ¼ 1.11 fps), N ¼ 75 (FOV ¼ 80 × 80 μm 2 , frame rate ¼ 1.11 fps), N ¼ 12 (FOV ¼ 120 × 120 μm 2 , frame rate ¼ 3.05 fps), and N ¼ 4 (FOV ¼ 200 × 200 μm 2 , frame rate ¼ 3.05 fps). We obtained two porcine VFs from whole porcine trachea purchased from a local slaughterhouse and tested within 6 h of animal sacrifice. Epithelial thickness was measured in multiple locations along the ROI as done in the human/canine VF study. Four depths were examined in the epithelium for the four different pulse overlap conditions. We tried to use identical pulse energies between pulse overlap conditions to allow for direct comparison of each condition for a given depth. Table 2 summarizes epithelial scattering lengths and ablation thresholds for two porcine VF samples measured at four different pulse overlap conditions. We performed analysis of covariance (ANCOVA) using Minitab statistical software to determine if a significant difference existed between regression coefficients (i.e., slope of regression line, l −1 s;ep ) and constants (i.e., y intercept, related to F th;ep ) generated for all four datasets in each sample. The test statistic used for hypothesis testing follows the method outlined by Clogg et al. 25 where p-value ≤ 0.05 represents significance.
We saw no significant difference in l s;ep measurements for all four pulse overlap conditions, and no significant difference in F th;ep for N ¼ 301, 75, and 12. We did see a significant difference in F th;ep when comparing N ¼ 301 (p ¼ 0.05) and N ¼ 75 (p ¼ 0.04) to N ¼ 4 in the first VF sample, and a significant difference in F th;ep when comparing N ¼ 301 (p ¼ 0.01), N ¼ 75 (p ¼ 0.01), and N ¼ 12 (p ¼ 0.04) to N ¼ 4 in the second VF sample. Figure 2(a) displays the data for all four pulse overlap conditions tested on VF sample #2. The y intercept for N ¼ 4 data is higher on average than it is for N ¼ 301, 75, and 4, resulting in a larger F th;ep for N ¼ 4. On average, the percent FOV ablated was less at N ¼ 4 for a given input energy and target depth, confirmed by counting pixels below threshold in the ablated areas using MATLAB. Images shown in Fig. 2(b) illustrate this point; a smaller percentage of the FOV is removed during ablation at N ¼ 4 for identical pulse energies, suggesting that incubation may play a role in lowering the fluence threshold for N > 4. Importantly, the exponential increase in input energy required for ablation at each depth scales at the same rate for all N values tested, demonstrating that l s;ep may be estimated with high certainty, regardless of the FOV or frame rate employed. We determined an average porcine l s;ep ¼ 59.9 AE 4.1 μm, which is within 20% of our previous findings. 19 The negligible variation in F th;ep for N ≥ 12 is in agreement with Rosenfeld et al., 26 which states that F th reaches a steady state value for N > 20 in dielectric materials such as glass where ablation initiates as optical breakdown. In tissue ablation, there is an additional regime of ablation where low-density plasma can lead to ROS-initiated photodamage. 27 Our irradiance thresholds (1.8 × 10 12 W∕cm 2 ) were close to this regime where we would also expect a direct dependence of threshold on the number of overlapping pulses as observed in our earlier findings for laser surgery in severing axons. 28 Our results, however, do not show as large of dependence on N like Bourgeois et al., which may be due to the less deterministic nature of the highly heterogenous VF tissue architecture. Our results indicate that initiation of ablation at N ≥ 12 requires 10% to 15% less energy than a few pulses, whereas a further increase of N does not offer any significant decrease in F th;ep . We, therefore, propose N ¼ 10 to 20 to be an optimal number in clinical practice to achieve the desired fast (10 to 20 min per VF) surgical speeds. 29

Storage Condition Study
To study the effects of tissue storage conditions on scattering length and ablation threshold determination, we measured l s;ep and F th in fresh porcine VF samples and compared these values to l s;ep and F th measured in the same samples after storage in PBS. As our human and canine samples were shipped in PBS to prevent dehydration, we deemed it necessary to understand how the measured properties relate to those of native tissue. We extracted VF from whole trachea of freshly sacrificed porcine immediately prior to testing to limit dehydration, then repeated measurements on VFs after 2 h of storage in PBS at room temperature. Table 3 summarizes the epithelial scattering lengths and ablation thresholds for two porcine VF samples measured before and after storage in PBS. The ANCOVA test described above showed there was a significant difference between scattering length measurements before and after storage (p ¼ 0.01, p ¼ 0.05 for samples 1 and 2, respectively) and a significant difference between the ablation thresholds before and after storage only in sample 2 (p ¼ 0.02). On average, l s;ep is 60% longer in fresh samples when compared to PBS-stored samples.
Increased scattering in PBS-stored samples may be caused by the larger difference between refractive index of PBS (n ¼ 1.333) and refractive index of the epithelium (n ¼ 1.43 for human skin at λ ¼ 800 nm) when compared to the refractive index of interstitial fluid (n ¼ 1.365). 30,31 Prior work by Genina et al. 32 agrees well with our results; they showed scattering coefficients (l −1 s ) of rat skin increased by ∼50% when stored in an isotonic 0.9% solution of NaCl. Another confounding factor could be dehydration of tissue samples, which is known to produce an optical clearing effect by reducing the refractive index mismatch between constituent tissue scatterers. 33 To avoid this effect, we tested samples immediately after excision to mitigate potential dehydration.
These results suggest that scattering length measurements made on PBS-stored human and canine VFs may underestimate the actual scattering lengths of native tissue, as PBS storage is necessary for sample shipment. Hence, we conclude that reduced scattering in intact VF epithelium may lead to lower pulse energies required to ablate tissue within the SLP. This has important implications in surgery, as the lower peak powers required to penetrate epithelium may increase the maximum ablation depth by reducing the probability of unwanted nonlinear effects such as self-focusing, which has been shown to shift the ablation plane toward the surface during deep tissue ablation. 34

Conclusions
We determined scattering properties and ablation thresholds of excised human and canine VFs using a nonlinear image-guided, ablation-based method. We found input pulse energies E th;surf required to ablate 50% of the targeted FOV at increasingly deeper depths, fit these data to an exponential curve predicted by Beer's law, and extracted l s and F th . We observed low-interspecies variance in scattering properties for both tissue types, yet a substantial difference in l s;ep of human and canine VFs. We studied the cumulative pulse overlap effects on two porcine VF samples and observed no statistically significant change in F th;ep for N > 12, and no effect on l s;ep for any pulse overlap tested, providing evidence that this methodology can be applied at a range of spatial scales and scanning rates. Further, we believe surgically relevant speeds are obtainable for N ¼ 10 to 20, providing lowered ablation thresholds and rapid subsurface void creation. We also studied effects of PBS storage on measurements and determined scattering lengths are 60% longer in fresh VFs when compared to PBS-stored tissue, suggesting that ablation and subsequent void formation in the SLP may be possible at lower pulse energies in a clinical setting. These results can be used to define an optimal operating regime and predict ablation performance for precise microphonosurgical treatment.

Disclosures
Authors declare no conflicts of interest.