SCANNING VOL. 9999, 1–10 (2016) © Wiley Periodicals, Inc.

Quantitative Differentiation of Normal and Scarred Tissues Using Second-Harmonic Generation Microscopy

1 2,3 4 4 2 MURAT YILDIRIM, KYLE P. QUINN, JAMES B. KOBLER, STEVEN M. ZEITELS, IRENE GEORGAKOUDI, AND 1,5 ADELA BEN-YAKAR 1Department of Mechanical Engineering, The University of Texas at Austin, Austin, Texas 2Department of Biomedical Engineering, Tufts University, Medford, Massachusetts 3Department of Biomedical Engineering, University of Arkansas, Fayetteville, Arkansas 4Department of Surgery, Harvard Medical School, Center for Laryngeal Surgery and Voice Rehabilitation, Massachusetts General Hospital, Boston, Massachusetts 5Department of Biomedical Engineering, The University of Texas at Austin, Austin, Texas

Summary: The aim of this study was to differentiate alignment within the scar. Depth-resolved average voxel normal and scarred hamster cheek pouch samples by density measurements indicated scarred tissues con- applying a quantitative image analysis technique for tained greater (p < 0.005) fiber density (0.72 0.09) determining collagen fiber direction and density in compared to controls (0.18 0.03). In the present study, second-harmonic generation microscopy images. This image analysis of both fiber alignment and density from paper presents a collagen tissue analysis of scarred cheek depth-resolved second-harmonic generation images in pouches of four adult male Golden Syrian hamsters as an epi-detection mode enabled the quantification of the animal model for vocal fold scarring. One cheek pouch increased collagen fiber deposition and alignment was scarred using an electrocautery unit and the other typically observed in fibrosis. The epi-detection geome- cheek was used as a control for each hamster. A home- try is the only viable method for in vivo imaging as well built upright microscope and a compact ultrafast fiber as imaging thick turbid tissues. These quantitative laser were used to acquire depth resolved epi-collected endpoints, clearly differentiating between control and second-harmonic generation images of collagen fibers. scarred hamster cheek pouches, provide an objective To quantify the average fiber direction and fiber density means to characterize the extent of vocal fold scarring in in each image, we applied two-dimensional Fourier vivo in preclinical and clinical research. In particular, this analysis and intensity thresholding at five different non-invasive method offers advantages for monitoring locations for each control and scarred tissue sample, scar treatments in live animals and following the effects respectively. The resultant depth-resolved average fiber of scarring-related treatments such as application of direction variance for scarred hamster cheek pouches steroids or drugs targeting pathways involved in fibrosis. (0.61 0.03) was significantly lower (p < 0.05) than SCANNING 9999:1–10, 2016. © 2016 Wiley Period- control tissue (0.73 0.04), indicating increased fiber icals, Inc.

Key words: hamster cheek pouch, vocal fold scarring, Contract grant sponsor: National Science Foundation; Contract grant collagen fiber alignment, collagen fiber density, second- number: (IDR: CBET-1014953 and Career Award: CBET-0846868); harmonic generation microscopy, scar remodeling, Contract grant sponsor: Cancer Prevention Research Institute of Texas (CPRIT); Contract grant number: RP130412; Contract grant sponsor: image analysis, Fourier transform, ultrafast fiber lasers National Institutes of Health; Contract grant number: K99EB017723. Conflicts of interest: There is no conflict of interest between the Introduction authors. This work was performed at the Mechanical Engineering Department of The University of Texas at Austin. Vocal fold scarring is one of the predominant Address for reprints: Adela Ben-Yakar, Department of Mechanical causes of voice disorders, affecting an estimated 2–6 Engineering, The University of Texas at Austin, 204 East Dean Keeton St., million people in the United State alone (Ramig and ETC Building, Austin, TX 78712. E-mail: [email protected] Verdolini, ’98; Roy et al., 2004, 2005; Best and Received 15 November 2015; Accepted with revision 3 March 2016 Fakhry, 2011; Cohen et al., 2012; Bhattacharyya, 2014). Vocal fold scarring arises as a wound healing DOI: 10.1002/sca.21316 Published online XX Month Year in Wiley Online Library response to injury or inflammation and results in (wileyonlinelibrary.com). collagenous scar tissue which impairs vibration 2 SCANNING VOL. 9999, 9999 (2016) (Rousseau et al., 2004; Tateya et al., 2005, 2006; such as application of steroids or drugs targeting Hirano et al., 2009) by reducing the viscoelasticity of pathways involved in fibrosis. the vocal fold. During the wound healing response, Multi-photon nonlinear imaging microscopies such scar tissue can replace the superficial lamina propria as two-photon autofluorescence microscopy (TPAF) (SLP) along with deeper parts of the lamina propria and second-harmonic generation (SHG) microscopy (LP). The scar tissue predominantly consists of can perform noninvasive and 3D deep tissue imaging collagen and fibronectin, both of which increase the with subcellular resolution using tightly focused ultra- stiffness of the mucosa and can lead to severe short pulses (Zipfel et al., 2003a,b). Second-harmonic impairment in voice production or dysphonia (Hirano, generation microscopy is a coherent two-photon process 2005). Unfortunately, current methods for treating necessitating intense ultrashort laser pulses passing vocal fold scarring are inconsistent and frequently through a highly polarizable material with a non- ineffective, and there is currently no accepted centrosymmetric molecular organization, such as treatment for restoring phonation to scarred vocal collagen fibers (Campagnola and Loew, 2003). folds (Woo et al., ’94; Benninger et al., ’96; Kriesel Second-harmonic generation microscopy has shown to et al., 2002; Thibeault et al., 2002; Zeitels and Healy, be a highly functional and noninvasive tool to obtain 3-D 2003; Hirano, 2005; Hansen and Thibeault, 2006; resolved collagen distribution and density in fibrotic Zeitels et al., 2007). tissues (Strupler et al., 2007), human skin (Cicchi et al., In scarred vocal folds, collagen fibers are found to be 2008), human cadaver vocal folds (Miri et al., 2012), more organized and densely packed compared to normal porcine vocal folds (Yildirim et al., 2013), and rat vocal vocal folds (Rousseau et al., 2003, 2004; Yildirim et al., folds (Heris et al., 2015). 2013; Heris et al., 2015). The collagen fiber orientation Different computational methods have been sug- and density are important in determining biomechanical gested to quantify the average fiber direction and fiber properties of biological tissues (Provenzano et al., 2006; density in an image obtained by different imaging Thomopoulos et al., 2006; Hadian et al., 2007; Bayan modalities. The Fourier transform is most commonly et al., 2009; Robinson and Tranquillo, 2009). The gold applied to compute the two-dimensional (2D) power standard method to determine collagen fiber density and spectral density (PSD) of an image (Levitt et al., 2007; direction is to stain tissue sections with picrosirius red or Ayres et al., 2008; Sander et al., 2009; Cicchi et al., Masson’s trichrome. This analysis requires complex and 2010; Sivaguru et al., 2010). Sampling the average destructive sample preparation to stain collagen fibers, power of the PSD at pixels corresponding to different and analysis is typically based on subjective scoring angles can yield an accurate measure of the total fiber systems, which limits comparisons among different orientation distribution in the image (Bayan et al., studies. Thus, there is a need to develop a nondestructive 2009). More complex fiber orientation measurements technique to resolve collagen fibers in three-dimen- also have been demonstrated, such as defining and sional (3D) tissues and automatically quantify fiber tracking fiber objects through energy minimization or density and directionality to provide objective endpoints line propagation algorithms (Mori and Van Zijl, 2002; for evaluating the treatment of vocal fold scarring. Rodriguez et al., 2009; Bas and Erdogmus, 2010). Also, The hamster cheek pouch model has been widely a set of zero-crossing maps was generated to form a used as a model for mucosal disease processes such as stability map from which significant linear patterns carcinogenesis (White et al., ’81; Kingsbury et al., ’97; (fiber directions) were detected (Liu, ’91; Heris et al., Adams et al., 2000; Driver et al., 2005; Meier et al., 2015). In terms of the fiber density, the common method 2007; Vairaktaris et al., 2008). Like the vocal fold is to set a threshold intensity manually for forming a mucosa, there is a squamous epithelium, an underlying binary image and to count the number of pixels whose lamina propria (which is thinner than specialized vocal intensities are equal to 1 (Strupler et al., 2007; fold lamina propria) and then a thin layer of striated Medyukhina et al., 2011; Heris et al., 2015). In addition muscle. The layered structure is analogous to the vocal to determine collagen fiber direction and density, fold although the thickness of the constituent layers is geometrical organization (Medyukhina et al., 2011), thinner in the cheek pouch. The cheek pouch is much and geometrical properties (Cicchi et al., 2010) of the more accessible and has a larger area than the vocal fold. collagen fibers were determined to differentiate normal While the surface area of one vocal fold is a few mm2 in and scarred tissue samples. laboratory rodents, the hamster cheek pouch area that In a parallel effort, we have developed an algorithm can be everted through the mouth is a fold of mucosa that to rapidly and accurately detect fiber orientation (Quinn is about 100 mm2 on each side. Thus, the size of the cheek and Georgakoudi, 2013) and applied it to quantify fiber pouch, the fact that it can be everted for treatment or organization in cutaneous scar tissue (Quinn et al., imaging, and the analogous anatomical organization of 2015). For the fiber density calculations, we select the tissue layers are the main advantages of this model. appropriate intensity thresholds determined by the Using this model, one could track the development of scar Otsu’s method, which calculates the optimal threshold in vivo or follow the effects of scarring-related treatments intensity by dividing the signal and background values M. Yildirim et al.: Quantitative Differentiation of Normal and Scarred Tissues 3 so that their combined variance is minimal (Otsu, ’79; 3 W average power at 1,552 nm (1.5 mJ pulse energy) Provenzano et al., 2006; D’Amore et al., 2010). Thus, and 1 W when frequency doubled to 776 nm (0.5 mJ our method is less susceptible to uncertainties in pulse energy) at 2 MHz with a 600 fs pulse width. Since collagen fiber density measurement due to manually the number of photons generated in two-photon based determined intensity threshold values, which have been microscopies is inversely proportional with duty cycle commonly used, in previous studies (Strupler et al., (multiplication of repetition rate and pulse width), 2007; Medyukhina et al., 2011; Heris et al., 2015). moderate repetition rates in the range of 1–10 MHz are The aim of the present work is to compare control and optimal for SHG imaging with reasonable imaging scarred hamster cheek pouch samples by quantifying speeds (Wang et al., 2014). Thus, having 2 MHz collagen direction and density with an automated SHG repetition rate enabled us to image using approximately image analysis technique to see if this method can an order of magnitude less average power compared to differentiate between these two tissue states. This those studies using standard high repetition rate Ti: technique is based on taking the Fourier transform sapphire lasers (Heris et al., 2015). (FT) and calculating the 2D power spectral density of We focused 776 nm excitation wavelength with a depth-resolved SHG images to determine collagen fiber 0.75-NA, 20 air objective (Nikon Plan Apo) to direction. Additionally, our technique utilizes Otsu perform SHG imaging in epi-detection mode. The method for automatic intensity thresholding of the SHG epi-detection geometry is the only viable method for in signal to calculate any differences in collagen fiber vivo imaging as well as imaging thick turbid tissues. density. Overall, this work illustrates the feasibility of Thus, the same objective was used for excitation and utilizing SHG imaging and automated image analysis collection. By collecting in the epi-direction, from a techniques as pragmatic approaches easily applicable in thick turbid tissue, our signal was likely a combination vivo to non-invasively quantify collagen organization in of backward SHG and backscattered SHG, and thus scarred vocal folds and guide surgical interventions. lacked any sensitivity to the directionality of the SHG signal. We compensated the laser power attenuation due to absorption and scattering coefficients of the samples, Materials and Methods as we imaged deeper in the tissue. The laser power was adjusted at each depth according to the optical properties Experimental Setup of specific tissue type and its SHG signal histogram in a way that only 0.1% of overall pixels were saturated in We used our home-built, upright laser-scanning each image. microscope to perform second-harmonic generation Emitted light is collected by a cold mirror (CM1–HT- (SHG) imaging (Fig. 1) using an ultrafast Er-doped fiber 1.00, CVI Laser, Carlsbad, CA) and collection optics laser (Discovery, Raydiance Inc.). This laser provides (CO) placed right behind the objective. A second cold

Fig. 1. Schematic of the upright microscope system for nonlinear imaging. (a) Ultrafast laser pulses at 776 nm from a compact fiber laser system are attenuated using a combination of half-wave plate (l/2) and polarizing cube beam splitter (PCBS). Another half-wave plate is used to adjust the polarization state of the laser. Laser pulses are scanned by a pair of galvanometric scanning mirrors (SM), which is imaged on the back aperture of the 0.75 NA, 20 objective by a pair of scan lens (SL) and tube lens (TL). The samples are placed on a three-axis motorized stage (XYZ) for nonlinear imaging. Emitted light (either TPAF or SHG signal) is collected by a cold mirror (CM1) and collection optics (CO). A second cold mirror (CM2) separates SHG and TPAF signals into different collection paths. 4 SCANNING VOL. 9999, 9999 (2016) mirror (CM2–Di01- R405, Semrock, Lake Forest, IL) computed from real and imaginary parts of the 2D separates the SHG and two-photon autofluorescence discrete Fourier transform and provided the relative (TPAF) signals into different collection paths. We magnitudes of the underlying frequency components of collect the TPAF signal through collection optics A the image. Rapid changes in intensity across an image at (COA) and a laser-blocking filter (FA) into the PMT A a given orientation are reflected by greater power (H10770PA-40, Hamamatsu, Japan) and the SHG signal density values in the orthogonal direction in Fourier through collection optics B (COB) and a laser-blocking space. Because the PSD map is symmetrical about the filter B (FB) into the PMT B (R3896, Hamamatsu, origin located in the center of the map, we assigned each Japan). In our experiments, we adjusted linear polariza- PSD pixel location with polar coordinates relative to the tion angle of the incoming laser beam with using another origin. Average PSD values were computed in discrete half wave plate (HWP) so that we obtained maximum increments of 1˚ within spatial frequencies ranging from SHG signal for each sample to maximize our sensitivity 0.05 to 0.5 pixels1 to obtain orientation distribution to collagen fiber density. Further details of the histograms for each image. The average collagen fiber experimental setup can be found in our previous study direction and directional variance were calculated from (Yildirim et al., 2013). the orientation distribution of each image through vector To measure the resolution of our system, we addition (Sander and Barocas, 2009). Directional suspended 100 nm fluorescent beads (Invitrogen, variance values range from 0 (perfectly aligned fibers) F8803) in agar gel which replicated the expected to 1 (isotropic/random organization). Thus, our method tissue scattering lengths (30–35 mm), and measured the could estimate the level of isotropy of collagen fibers point spread function (PSF) at imaging depths ranging independent of their absolute direction which may be from 50 to 500 mm. The average lateral and axial affected by the laser polarization. full width half maximum (FWHM) of the two-photon Collagen fiber density in the SHG images was PSF were 0.58 0.07 mm and 2.51 0.35 mm, respec- obtained by computing the number of pixels with tively. We scanned the laser beam in the x- and y-axes intensities exceeding an optimum threshold intensity using a pair of galvanometric mirrors (Cambridge relative to the total number of pixels. To automate the Technologies, Inc.) and swept the focal spot over a determination of this optimum threshold intensity, we 150 150 mm2 field of view (FOV) for 3.3 s to take 10 used Otsu’s method (Otsu, ’79). Otsu’s method is a images of the same FOV. The axial displacement nonparametric and unsupervised method for automatic between consecutive SHG images was 2 mm, and threshold selection for 2-D images using the zeroth- and the imaging depth was increased until the signal to the first-order cumulative moments of the gray-level background ratio (SBR) became 1. The maximum histograms. SHG imaging depth was approximately 120 mmfor To perform a reliable image analysis in terms of the all control and scarred tissue samples, reaching the selections of the region of interest (ROI) without any theoretical maximum imaging depths as will be bias, we chose five different locations in a 1.7 mm2 discussed below. The computer software (MPScan) region for all control and scarred samples. Each ROI was used to reconstruct these signals into 512 consisted of 150 150 120 mm3 tissue block. To 512 pixel images at 3.05 frame per second (fps) with reduce bias in the selection of the ROI, the center of pixel dwell time of 1.25 ms. Since the laser repetition the first tissue block location was randomly selected rate was 2 MHz, each pixel received at least two laser and the center of other four ROIs were selected 1 mm pulses. We therefore averaged 10 images for each apart from the center of the first tissue block in four plane to smoothen pixel-to-pixel fluctuations. directions (east, west, north, and south) (Yildirim et al., 2013). For each tissue block, average fiber direction and density results were calculated by Automated SHG Image Analysis Method averaging corresponding 2-D results over the whole tissue block between depths where collagen fibers first Collagen fiber direction in the SHG images was appeared and signal to background ratio became one. determined in a similar manner to previous Fourier- Then, we averaged all five tissue block results to based analysis approaches (Bayan et al., 2009; Sander calculate overall fiber direction and density for each and Barocas, 2009; Quinn and Georgakoudi, 2013). control and scarred sample. Each 8-bits, gray scale image is composed of pixels that vary spatially in intensity. Rapid changes in intensity are indicative of object edges and thus fiber orientation can Animal Model be quantitatively analyzed through a 2D Fourier transformation of the image into frequency space. Prior We used four adult male Golden Syrian hamsters to transformation, SHG images were apodized using a (Charles River Labs, Wilmington, MA) of 100 to 120 g Hann window to eliminate discontinuities at the edges of body weight, in which scars could be created within an the image. A power spectral density (PSD) was easily accessible mucosal surface. To scar the cheek M. Yildirim et al.: Quantitative Differentiation of Normal and Scarred Tissues 5 pouches, we first anesthetized the animals by injecting Massachusetts General Hospital and The University of a mixture of ketamine (200 mg/kg) and xylazine Texas at Austin. All surgical procedures and housing of (8 mg/kg) intraperitoneally. We then cauterized 5– animals took place at the Massachusetts General 10 mm diameter circular areas of one cheek pouch Hospital. with an electrocautery unit (Conmed Saber 2400, Utica, NY). The hamsters were euthanized after a survival period of 1 month using 0.5 mL intraperitoneal euthasol. Statistical Analysis The scarred and contralateral normal cheek pouches were removed and mounted on rubber test-tube stoppers Paired Student’s t-tests were used to compare the with fine needles to hold the cheek pouch mucosa flat. collagen fiber directional variance and density between After rinsing with saline, the tissue was frozen in scarred and controlled cheek pouches with a ¼ 0.05. isopentane and cooled in liquid nitrogen. The cheek pouch tissue was prepared at Massachusetts General Hospital Voice Laboratory (MGH) in Boston and Results shipped on dry ice to the Ben-Yakar Laboratory at the University of Texas at Austin for bench-top testing. Figure 2 represents SHG nonlinear images used to After delivery, the cheek pouches were stored at 80˚C. characterize the direction and density of collagen fibers For each experiment, we thawed the cheek pouches in in normal and scarred hamster cheek pouches. Qualita- saline solution and covered them with a glass cover slip tive interpretation of the SHG images of normal (Fig. 2a) to flatten their epithelial surface to compensate aberra- and scarred (Fig. 2d) tissue samples revealed that tions with the objective. To ensure the proper collagen fibers were denser and more aligned in scarred identification of the tissue surface with TPAF, 5 mLof samples than control samples. We performed a a solution of 100 nm fluorescent beads (F-8823, quantitative analysis for collagen fiber direction and Invitrogen) in saline was deposited onto the tissue density using the methods described above. In a surface prior to placement of the cover slip. All representative example, the FT results showed that experimentation followed an animal use protocol that scarred tissue sample (Fig. 2e) had a directional variance was approved by the Committee on Animal Care and the of 0.61 and the amplitude of the dominant direction was Institutional Animal Care and Use Committee of the at least fourfold greater than other directional angles. On

Fig. 2. Representative SHG images, collagen fiber direction histograms, and collagen fiber density plots for control (a–c) and scarred (d– f) hamster cheek pouch samples. The scale bar represents 50 mm. 6 SCANNING VOL. 9999, 9999 (2016)

Fig. 3. Representative results for the calculations of depth-resolved collagen fiber directions for control (a) and scarred (b) hamster cheek pouch samples. the other hand, control tissue sample (Fig. 2b) had a significant fiber alignment between 40 and 100 mm higher variance of 0.70 and more importantly the imaging depth. The average fiber direction was 130 5˚ amplitude of the dominant direction was comparable to which remained approximately constant in the range of most of the other directional angles. After determining 40–100 mm imaging depth. As expected collagen fibers fiber direction in the SHG images, we performed did not appear in the epithelial layer of the tissue automated intensity thresholding to obtain the collagen between 0 and 40 mm depths. Beyond 120 mm, the SHG fiber density. A representative example of this analysis imaging quality degraded substantially as the SBR showed that the fiber density in the scarred tissue sample approached to 1. (Fig. 2f) was fourfold higher than the fiber density in the To evaluate the utility of these quantitative metrics control tissue sample (Fig. 2c). and understand inter- and intra-sample variations in To reveal the distribution of collagen fiber direction fiber density and direction, we performed depth- as a function of imaging depth, we report a depth- resolved fiber direction and fiber density analysis at resolved plot of collagen fiber orientation for the normal five different locations with four different normal and and scarred tissue samples discussed above in Figure 3. scarred tissue samples as shown in Figure 4. The Here, the corresponding average power is also included collagen fiber directional variance (p < 0.005 for as a color bar on the right hand side of the figure. The Samples 1, 3, and 4 and p < 0.05 for Sample 2) was normal tissue sample (Fig. 3a) had no dominant fiber lower in scarred tissues than control ones, indicating all direction for the entire 120 mm imaging depth. On the scarred tissue samples had more aligned collagen fibers other hand, the scarred tissue sample (Fig. 3b) showed than the controls (Fig. 4a). In addition, the fiber densities

Fig. 4. Collagen fiber directional variance and fiber density for four different samples as averaged at five different locations per sample and for 30–35 axial planes for each location. (a) In all samples, directional variance was lower in scarred samples than control samples in a statistically significant manner. The resultant average fiber direction variance was 0.73 0.04 for control and 0.61 0.03 for scarred hamster cheek pouch tissues. (b) Collagen fiber density was three to fivefolds higher in scarred samples than in control samples. The resultant average fiber density was 0.18 0.03 for control and 0.72 0.09 for scarred hamster cheek pouch tissues. The error bars in the plots represent the mean standard deviation for each sample as averaged over 150 FOVs. (p < 0.005, p < 0.05). M. Yildirim et al.: Quantitative Differentiation of Normal and Scarred Tissues 7 in scarred tissues were three to fivefolds higher to their fibril organization. Furthermore, we adjusted the (p < 0.005) than control tissues for all four samples laser excitation power to maximize the SHG signal with (Fig. 4b). Overall, both fiber direction and fiber density saturating 0.1 % of the pixels and to compensate for analysis revealed a statistically significant difference loses due the optical properties of samples at each depth. between control and scarred tissue samples. This approach consequently reduced the effect of intensity variations due to scattering and absorption in our calculations. While forward to backward ratio and Discussion polarization-sensitive SHG methods can provide addi- tional context regarding the molecular organization The development and evaluation of effective treat- within fibers, these are however time consuming ments necessitate understanding and objectively char- measurements and difficult to perform in vivo and 3D acterizing structural changes during vocal fold scarring. thick tissues. Most frequently, they are performed on Density and organization of the extracellular matrix thin sections or ex vivo samples. With epi-detected SHG (ECM) components determine the biomechanical imaging method, we offer a more pragmatic approach function of the lamina propria (Billiar and Sacks, that is easily applicable in vivo. 2000; Sander et al., 2009). Several previous studies have Since routine histology is a destructive method addressed the orientation and density of collagen fibers requiring complex processing steps, SHG microscopy is in the lamina propria in relation to vocal fold scarring. In a valuable alternative to non-invasively determine some of these studies, collagen density has been collagen fiber direction and density for assessing the determined by performing immunohistology and non- progression of vocal fold scarring. The analysis of SHG linear imaging. These studies have shown that collagen microscopy has the ability to provide results that are became thicker, more organized, and denser (two to consistent with the histological evaluation of tissues. In fivefolds) in mature scarred vocal folds (Rousseau et al., our previous work, for example, we found that the 2003, 2004; Yildirim et al., 2013). Early stages of vocal scarred tissue samples showed regular and dense pattern fold scarring, on the other hand, showed low or high of collagen fibers as compared to control samples when density and disorganized collagen fibers (Thibeault they were stained with Masson–Trichrome (Yildirim et al., 2002; Rousseau et al., 2003). These differences in et al., 2013). This high directional and high density collagen density among previous studies may be collagen fiber trend was consistent in both histology and attributed to the fact that there is a different timeline SHG microscopy results for all four samples. When for remodeling stages of vocal fold scarring where there combined with the presented analysis metrics, SHG is an ongoing process of collagen synthesis. Similarly, imaging could provide a means to non-destructively wound-healing studies suggest that as remodeling monitor vocal fold scarring in research studies and, progresses (20 days to 1 year), collagen fibers become given future development of endoscopic probes, in the thicker and more organized (parallel arrays) with clinic. eventual crosslinking (Ehrlich, 2000). Nonlinear imaging modalities such as TPAF and Unlike collagen fiber density, fiber direction has not SHG are well suited for high-resolution in vivo imaging. been thoroughly evaluated in previous vocal fold Second-harmonic generation microscopy specifically scarring studies. One recent study, however, character- resolves collagen fibers because of their non-centro- ized the microstructure of scarred rat vocal folds using symmetric characteristics. On the other hand, both forward collected SHG microscopy of thin slices and elastin and collagen fibers in the extracellular matrix found the dispersion of the collagen fiber directionality autofluoresce, reducing the specificity of TPAF for to be lower than the one in uninjured control tissues collagen fibers. Since SHG is an inherent nonlinear (Heris et al., 2015). Similarly, in our study, the average imaging modality, it can provide deep tissue imaging of fiber directional variance values were significantly unstained tissues. SHG microscopy provided approxi- lower (p < 0.05) in scarred tissues (0.61 0.03) com- mately 120 mm imaging depth within hamster cheek pared to control ones (0.73 0.04), indicating increased pouches in this study. The maximum imaging depth in fiber alignment in the scar. Lower fiber directional SHG microscopy is highly dependent on the optical variance values in scar tissues are also consistent with properties of tissues. It has been shown that the other measurements in mature scars from a rat model of maximum imaging depth for TPAF can vary between cutaneous burns (Quinn et al., 2015). 3–4 scattering lengths near 800 nm excitation wave- In general, intensity threshold based techniques to lengths for different tissue types (Theer and Denk, 2006; calculate collagen fiber density, depend on the chosen Durr et al., 2011). Since the scattering length of hamster intensity threshold. To minimize the effects of intensity cheek pouches has been found to be 30–35 mmat threshold, our approach counts all the collagen pixels 776 nm wavelength (Yildirim et al., 2013), the maxi- above a certain threshold rather than taking the mean mum imaging depth should be in the 100–150 mm range, intensity of all those pixels. Therefore, it is less corresponding well with our experimental results. Thus, susceptible to intensity variations among fibers related SHG microscopy is a good candidate for clinical 8 SCANNING VOL. 9999, 9999 (2016) evaluation of scarring in the superficial lamina propria collagen fiber density demonstrated, as expected, of the vocal fold. Even though optical coherence that the average fiber densities in scarred tissues tomography (OCT) has been used in the clinical setting, were remarkably higher than that in control tissues. there remains a need for new tools such as SHG The method we presented here will guide development microscopy with higher specificity to collagen, and finer of turn-key ultrafast fiber laser-assisted treatment resolution to discriminate individual fibers within the methods for sub-epithelial image guided surgeries, scar. similar to proposed scarred vocal folds treatments On the other hand, SHG has a fundamental imaging currently under development in our lab (Hoy et al., depth limit due to gradually increasing out-of-focus 2008, 2011, 2012, 2014; Yildirim et al., 2013; (background) signal when increasing imaging depth to Ferhanoglu et al., 2014). compensate for the losses due to scattering. Recently, the growing interest in performing high-resolution, deep tissue imaging has galvanized the use of longer Acknowledgments excitation wavelengths and three-photon based techni- ques to improve maximum imaging depth of SHG. Our The authors would like to thank Raydiance Inc., for group showed that the maximum imaging depth in the use of their Discovery fiber laser, Kaushik porcine vocal folds can be improved by three times by Subramanian and Chris Martin for editing the text, performing third-harmonic generation (THG) micros- and Dr. David Kleinfeld and his research group for the copy at 1552 nm excitation wavelength compared to use of the MPScan. SHG microscopy at 776 nm wavelength (Yildirim et al., 2015). While providing the ability to image at depths beyond 400 mm, THG signal can originate from both References collagen and elastin fibers and thus provides a limited specificity to extract 3-D collagen fiber orientation. Adams J, Heintz P, Gross N, et al. 2000. Acid/pepsin promotion of carcinogenesis in the hamster cheek pouch. Arch Otolaryngol To summarize, we utilized Fourier-based orientation Head Neck Surg 126:405–409. analysis and intensity thresholding to determine colla- Ayres CE, Jha BS, Meredith H, et al. 2008. Measuring fiber gen fiber direction and density, respectively, in 3D SHG alignment in electrospun scaffolds: a user’s guide to the 2D image stacks of control and scarred hamster cheek fast Fourier transform approach. J Biomater Sci Polym Ed 19:603–621. pouches. Our results are promising in that both fiber Bas E, Erdogmus D. 2010. Piecewise linear cylinder models for direction variance (p < 0.05) and fiber density 3-dimensional axon segmentation in brainbow imagery. I S (p < 0.005) are significantly different between control Biomed Imaging, IEEE: New York. p 1297–1300. and scarred hamster cheek pouches. We believe that this Bayan C, Levitt JM, Miller E, Kaplan D, Georgakoudi I. 2009. Fully automated, quantitative, noninvasive assessment of automatized SHG image analysis method can provide collagen fiber content and organization in thick collagen gels. significant feedback on fiber density and alignment in J Appl Phys 105:102042. normal, scarred, and surgically or medically treated Benninger MS, Alessi D, Archer S, et al. 1996. Vocal fold scarring: current concepts and management. Otolaryngol vocal folds. Also, one could test scar treatments such as Head Neck Surg 115:474–482. biomaterials, drugs or growth factors which will Best SR, Fakhry C. 2011. The prevalence, diagnosis, and eventually be used to measure treatment outcome in management of voice disorders in a National Ambulatory humans. Medical Care Survey (NAMCS) cohort. Laryngoscope 121: 150–157. Bhattacharyya N. 2014. The prevalence of voice problems among adults in the United States. Laryngoscope 124: Conclusion 2359–2362. Billiar KL, Sacks MS. 2000. Biaxial mechanical properties of the native and glutaraldehyde-treated aortic valve cusp: part The aim of this study was to develop an automatized II-a structural constitutive model. J Biomech Eng 122: SHG image analysis method to differentiate normal and 327–335. scarred hamster cheek pouch samples by quantifying Campagnola PJ, Loew LM. 2003. Second-harmonic imaging collagen fiber direction and density. We utilized microscopy for visualizing biomolecular arrays in cells, tissues and organisms. Nat Biotechnol 21:1356–1360. Fourier-based methods and intensity thresholding to Cicchi R, Sestini S, De Giorgi V, et al. 2008. Nonlinear laser determine collagen fiber directional variance and imaging of skin lesions. J Biophotonics 1:62–73. density, respectively. The resultant depth-resolved Cicchi R, Kapsokalyvas D, De Giorgi V, et al. 2010. Scoring average fiber direction variance was 0.73 0.04 for of collagen organization in healthy and diseased human dermis by multiphoton microscopy. J Biophotonics 3: control and 0.61 0.03 for scarred hamster cheek 34–43. pouches. This analysis showed that the fiber direction in Cohen SM, Kim J, Roy N, Asche C, Courey M. 2012. Prevalence scarred samples was less random than in control and causes of dysphonia in a large treatment-seeking population. Laryngoscope 122:343–348. samples. Depth-resolved collagen fiber density was Driver M, Upadhyay UD, Shapshay SM, Wang Z. 2005. Laser- 0.18 0.03 for control and 0.72 0.09 for scarred assisted low-dose retinoic acid in oral cancer chemopreven- tissues, respectively. This fourfold difference in tion. Laryngoscope 115:283–286. M. Yildirim et al.: Quantitative Differentiation of Normal and Scarred Tissues 9

Durr NJ, Weisspfennig CT, Holfeld BA, Ben-Yakar A. 2011. Provenzano PP, Eliceiri KW, Campbell JM, et al. 2006. Collagen Maximum imaging depth of two-photon autofluorescence reorganization at the tumor-stromal interface facilitates local microscopy in epithelial tissues. J Biomed Opt 16:026008– invasion. BMC Med 4:38. 026008. Quinn KP, Georgakoudi I. 2013. Rapid quantification of pixel- D’Amore A, Stelia JA, Wagner WR, Sacks MS. 2010. wise fiber orientation data in micrographs. J Biomed Opt Characterization of the complete fiber network topology 18:046003. of planar fibrous tissues and scaffolds. Biomaterials 31: Quinn KP, Golberg A, Broelsch GF, et al. 2015. An automated 5345–5354. image processing method to quantify collgen fibre Ehrlich HP. 2000. Collagen considerations in scarring and organization within cutaneous scar tissue. Exp Dermatol regenerative repair. Basic Clin Dermatol 19:99–114. 24:78–80. Ferhanoglu O, Yildirim M, Subramanian K, Ben-Yakar A. Ramig LO, Verdolini K. 1998. Treatment efficacy: voice 2014. A 5-mm piezo-scanning fiber device for high speed disorders. J Speech Lang Hear Res 41:S101–S116. ultrafast laser microsurgery. Biomed Opt Express 5: Robinson PS, Tranquillo RT. 2009. Planar Biaxial behaviour of 2023–2036. fibrin-based tissue-engineered heart valve leaflets. Tissue Eng Hadian M, Corcoran BM, Han RI, Grossmann JG, Bradshaw JP. Part A 15:2763–2772. 2007. Collagen organization in canine myxomatous mitral Rodriguez A, Ehlenberger DB, Hof BR, Wearne SL. 2009. Three- valve disease: an X-ray diffraction study. Biophys J 93: dimensional neuron tracing by voxel scooping. J Neurosci 2472–2476. Meth 184:169–175. Hansen JK, Thibeault SL. 2006. Current understanding and Rousseau B, Hirano S, Scheidt TD, et al. 2003. Characterization of review of the literature: vocal fold scarring. J Voice 20: vocal fold scarring in a canine model. Laryngoscope 110–120. 113:620–627. Heris HK, Miri AK, Ghattamaneni NR, et al. 2015. Microstruc- Rousseau B, Hirano S, Chan RW, et al. 2004. Characterization of tural and mechanical characterization of scarred vocal folds. chronic vocal fold scarring in a rabbit model. J Voice J Biomech 48:708–711. 18:116–124. Hirano S, Minamiguchi S, Yamashita M, et al. 2009. Histologic Roy N, Merrill RM, Thibeault S, et al. 2004. Prevalence of voice characterization of human scarred vocal folds. J Voice 23: disorders in teachers and the general population. J Speech 399–407. Lang Hear Res 47:281–293. Hirano S. 2005. Current treatment of vocal fold scarring. Curr Roy N, Merrill RM, Gray SD, Smith EM. 2005. Voice Opin Otolaryngol Head Neck Surg 13:143–147. disorders in the general population: prevalence, risk Hoy CL, Everett WN, Yildirim M, et al. 2012. Towards factors, and occupational impact. Laryngoscope 115: endoscopic ultrafast laser microsurgery of vocal folds. 1988–1995. J Biomed Opt 17:0380021–0380028. Sander EA, Barocas VH. 2009. Comparison of 2D fiber network Hoy CL, Ferhanoglu O, Yildirim M, et al. 2014. Clinical ultrafast orientation measurement methods. J Biomed Mater Res A laser surgery: recent advances and future directions. IEEE J 88:322–331. Sel Top Quant 20:242–255. Sander EA, Stylianopoulos T, Tranquillo RT, Barocas VH. 2009. Hoy CL, Durr NJ, Chen PY, et al. 2008. Miniaturized probe for Image-based multiscale modeling predicts tissue-level and femtosecond laser microsurgery and two-photon imaging. network-level fiber reorganization in stretched cell- Opt Express 16:9996–10005. compacted collagen gels. Proc Natl Acad Sci USA 106: Hoy CL, Ferhanoglu O, Yildirim M, et al. 2011. Optical design 17675–17680. and imaging performance testing of a 9.6-mm diameter Sivaguru M, Durgam S, Ambekar R, et al. 2010. Quantitative femtosecond laser microsurgery probe. Opt Express 19:10536– analysis of collagen fiber organization in injured tendons 10552. using Fourier transform-second harmonic generation Kingsbury JS, Cecere W, Mang TS, Liebow C. 1997. imaging. Opt Express 18:24983–24993. Photodynamic therapy for premalignant lesions in DMBA- Strupler M, Pena AM, Hernest M, et al. 2007. Second harmonic treated hamsters: a preliminary study. J Oral Maxillofac Surg imaging and scoring of collagen in fibrotic tissues. Opt 55:376–381. Express 15:4054–4065. Kriesel KJ, Thiebault SL, Chan RW, et al. 2002. Treatment of Tateya T, Tateya I, Sohn JH, Bless DM. 2005. Histologic vocal fold scarring: rheological and histological measures of characterization of rat vocal fold scarring. Ann Otol Rhinol homologous collagen matrix. Ann Otol Rhinol Laryngol Laryngol 114:183–191. 111:884–889. Tateya T, Tateya I, Sohn JH, Bless DM. 2006. Histological study Levitt JM, Hunter M, Mujat C, et al. 2007. Diagnostic cellular of acute vocal fold injury in a rat model. Ann Otol Rhinol organization features extracted from autofluorescence im- Laryngol 115:285–292. ages. Opt Lett 33:3305–3307. Theer P, Denk W. 2006. On the fundamental imaging-depth limit Liu ZQ. 1991. Scale space approach to directional analysis of in two-photon microscopy. J Opt Soc Am A Opt Image Sci images. Appl Opt 30:1369–1373. Vis 23:3139–3149. Medyukhina A, Vogler N, Latka I, et al. 2011. Automated Thibeault SL, Gray SD, Bless DM, Chan RW, Ford CN. 2002. classification of healthy and keloidal collagen patterns based Histologic and rheologic characterization of vocal fold on processing SHG images of human skin. J Biophotonics scarring. J Voice 16:96–104. 4:627–636. Thomopoulos S, Marquez JP, Wienberger B, Birman V, Genin Meier JD, Enepekides DJ, Poirier B, et al. 2007. Treatment with 1- GM. 2006. Collagen fiber orientation at the tendon to bone alpha,25-dihdroxyvitamin D3 (vitamin D3) to inhibit carci- insertion and its influence on stress concentrations. J Biomech nogenesis in the hamster buccal pouch model. Arch 39:1842–1851. Otolaryngol Head Neck Surg 133:1149–1152. Vairaktaris E, Spyridonidou S, Papakosta V, et al. 2008. The Miri AK, Tripathy U, Mongeau L, Wiseman PW. 2012. Nonlinear hamster model of sequential oral oncogenesis. Oral Oncol laser scanning microscopy of human vocal folds. Laryngo- 44:315–324. scope 122:356–363. Wang K, Horton NG, Charan K, Xu C. 2014. Advanced fiber Mori S, Van Zijl PCM. 2002. Fiber tracking: principles soliton sources for nonlinear deep tissue imaging in and strategies-a technical review. NMR Biomed 15: 468–480. biophotonics. IEEE J Sel Top Quant 20:50–60. Otsu NA. 1979. A threshold selection method from gray-level White FH, Gohari K, Smith CJ. 1981. Histological and histograms. IEEE Trans Syst Man Cybern 9:63–66. ultrastructural morphology of 7,12 demethylbenz(alpha)- 10 SCANNING VOL. 9999, 9999 (2016)

anthracene carcinogenesis in hamster-cheek pouch epithe- Zeitels SM, Blitzer A, Hillman RE, Anderson RR. 2007. Foresight lium. Diagn Histopathol 4:307–333. in laryngology and laryngeal surgery: a 2020 vision. Ann Otol Woo P, Casper J, Colton R, Brewer D. 1994. Diagnosis and treatment Rhinol Laryngol Suppl 198:2–16. of persistent dysphonia after laryngeal surgery: a retrospective Zeitels SM, Healy GB. 2003. Laryngology and phonosurgery. analysis of 62 patients. Laryngoscope 104:1084–1091. N Engl J Med 349:882–892. Yildirim M, Ferhanoglu O, Kobler J, Zeitels SM, Ben- Zipfel WR, Williams RM, Webb WW. 2003a. Nonlinear magic: Yakar A. 2013. Parameters affecting ultrafast laser multiphoton microscopy in the biosciences. Nat Biotechnol microsurgery of subepithelial voids for scar treatment 21:1368–1376. in vocal folds. J Biomed Opt 18:118001–118001. Zipfel WR, Williams RM, Christie R, et al. 2003b. Live tissue Yildirim M, Durr NJ, Ben-Yakar A. 2015. Tripling the maximum intrinsic emission microscopy using multiphoton-excited imaging depth with third-harmonic generation microscopy. native fluorescence and second harmonic generation. Proc J Biomed Opt 20:096013–096013. Natl Acad Sci USA 100:7075–7080.