Diffuse reflectance spectroscopy for optical characterizations of orthotopic head and neck cancer models in vivo

We demonstrated an easy-to-build, portable diffuse reflectance spectroscopy device along with a Monte Carlo inverse model to quantify tissue absorption and scattering-based parameters of orthotopic head and neck cancer models in vivo. Both tissue-mimicking phantom studies and animal studies were conducted to verify the optical spectroscopy system and Monte Carlo inverse model for the accurate extraction of tissue optical properties. For the first time, we reported the tissue absorption and scattering coefficients of mouse normal tongue tissues and tongue tumor tissues. Our in vivo animal studies showed reduced total hemoglobin concentration, lower tissue vascular oxygen saturation, and increased tissue scattering in the orthotopic tongue tumors compared to the normal tongue tissues. Our data also showed that mice tongue tumors with different sizes may have significantly different tissue absorption and scattering-based parameters. Small tongue tumors (volume was ∼60 mm3) had increased absorption coefficients, decreased reduced-scattering coefficients, and increased total hemoglobin concentrations compared to tiny tongue tumors (volume was ∼18 mm3). These results demonstrated the potential of diffuse reflectance spectroscopy to noninvasively evaluate tumor biology using orthotopic tongue cancer models for future head and neck cancer research.


Introduction
Head and neck cancer (HNC) represents the seventh most common malignancy worldwide and it is an aggressive life-threatening disease associated with high mortality rates [1].The incidence of HNC continues to rise and is anticipated to increase by 30% (∼1.1 million new cases annually) by 2030 [2].Radiotherapy (RT) alone or combined with chemotherapy is used as a primary treatment modality for over 75% of locally advanced HNC patients [3].Despite that many efforts have been dedicated to developing new radiation techniques [4], over 50% of RT-treated HNC patients are still prone to develop radio-resistance (RR) and recurrence [3].It is well known that RT efficacy relies on both the dosimetry and tumor oxygenation levels and would be significantly reduced by tumor hypoxia [5].Though the exact correlation between tumor oxygenation and RT efficacy remains unclear and unestablished, recent studies showed that monitoring tumor oxygenation kinetics may aid in the prediction of local tumor recurrence after RT [6].Moreover, a recent study showed that both tumor tissue optical scattering and oxygenation changed significantly in response to radiation treatment [7].
Given the clinical importance of tumor hypoxia in RT and the high heterogeneity of tumor oxygenation during the RT [6,8], continuous and routine monitoring of the tumor vascular microenvironment parameters is critical for improved understanding of the tumor radiation biology, which may potentially help optimize the RT to achieve the max efficacy.Although various methods have been explored to quantify tumor vascular parameters, their limitations hinder them from being routinely used in clinics [9].For example, Positron Emission Tomography (PET) with hypoxia probes has been considered as a promising clinical imaging tool for tumor hypoxia detection, while PET involves the use of a radioactive tracer and requires long time imaging and data processing with high cost [10].Oxygen-enhanced MRI has been explored to report tumor hypoxia [11], while the MRI technique has relatively low sensitivity and the cost is high.Fluorescence imaging techniques have also been explored to report tumor hypoxia [12], while special hypoxia-sensitive probes are required to quantify the oxygenation.
Alternatively, the diffuse reflectance spectroscopy technique offers a cost-effective and labelingfree approach for tumor vascular parameters measurement [13][14][15][16], and it has been explored extensively to study tumor RT responses [6][7][8].For example, Hu et al. reported the use of diffuse reflectance spectroscopy to monitor oxygenation and perfusion kinetics in response to fractionated RT in HNC xenografts [6].Diaz et al. reported a diffuse reflectance spectroscopy to study changes in vascular oxygenation and total hemoglobin concentration in a matched model of radiation resistance for lung tumors treated with a clinically relevant dose [8].Later, they used the same platform to study the oxygenation and tissue scattering changes of breast tumors treated with a clinically relevant dose [7].The same group also explored the use of their diffuse reflectance spectroscopy to study the radiation-induced reoxygenation in radiation-resistant head and neck tumors [17].In all the above diffuse reflectance spectroscopy studies for HNC research, the subcutaneous tumor model was used to form solid tumors for HNC RT response investigations [6,17].
Orthotopic tumor models, implanting tumors in a region where a particular cancer normally arises, have been developed to provide a more realistic model for cancer inquiries [18,19].In this study, we established an orthotopic HNC tumor model [20] for tongue cancer research.To facilitate the optical measurements on orthotopic tongue tumors, we reported an easy-to-build, low-cost, portable optical spectroscopy platform for vascular parameters measurements on mouse tongue tissues.To quantify tissue absorption and scattering-based parameters from tongue tumors in vivo, we adapted our formerly reported Monte Carlo (MC) inverse model [15] with our new optical spectroscopy platform for tissue optical properties extraction.To verify the new optical spectroscopy device and MC inverse model for accurate estimation of tissue absorption and scattering parameters, tissue-mimicking phantoms with known optical properties were used.Our tissue-mimicking phantom studies demonstrated that the reported optical spectroscopy could quantify the medium absorption and scattering with high accuracy, i.e. less than 6% error.For the first time, we reported tissue absorption and scattering coefficients of mouse normal tongue and tongue tumor tissues, which may serve as references for future studies with a focus on tongue cancer.Our in vivo animal studies showed that mice tongue tumors have significantly different optical parameters compared to normal tongue tissues.Specifically, the tongue tumor tissues have reduced total hemoglobin concentration ([THb]), lower tissue vascular oxygen saturation (StO 2 ), lower oxygenated hemoglobin concentration ([HbO 2 ]), and increased tissue scattering coefficients compared to the normal tongue tissues.Our results also showed that mice tongue tumors with different sizes may have different tissue absorption and scattering-based parameters.Small tongue tumors (volume was ∼60 mm 3 ) had increased absorption coefficients, decreased reduced scattering coefficients, increased [THb], and increased de-oxygenated hemoglobin [Hb] compared to tiny tongue tumors (volume was ∼18 mm 3 ).These studies demonstrated the potential of the easy-to-build, portable diffuse reflectance spectroscopy to noninvasively evaluate tumor biology using orthotopic tongue cancer models for future HNC research.

Portable diffuse reflectance spectroscopy
An easy-to-build, portable diffuse reflectance spectroscopy system was developed using a white LED source (SOLIS-3C, Thorlabs) and a compact spectrometer (FLAME-T-VIS-NIR, Ocean Optics), along with a low-cost fiber probe (BF19Y2HS02, Thorlabs) following the layout in Fig. 1 A. The fiber probe has 10 illumination fibers and 9 collection fibers evenly and adjacently bundled in the common end (Fig. 1 A bottom right).The diameter of each fiber is 200 µm, therefore the source-detector separations of the probe range from 0.2 mm to 1 mm.Based on former reports [21], the sampling depth of visible diffuse reflectance spectroscopy in a layered skin tumor model was around half of the source-detector separation.Therefore, the estimated optical sensing depth of the fiber probe used in our device can reach up to 0.5 mm when using visible light [22].The fiber probe common end has a diameter of ∼2.5 mm, which makes the fiber probe suitable for optical measurements on mouse tongue as illustrated in Fig. 1(B).To keep the consistency of the contact between the fiber probe and tissue, the fiber probe was mounted to a probe holder (Fig. 1 B, bottom) and then the probe holder was mounted to a 3-axis optical stage (Thorlabs) for precision control of the probe placement on tissue surface.The key optical components used to build the diffuse reflectance spectroscopy device are summarized in Table 1.The total weight of the system is less than 2 lbs which makes the system highly portable.The system can be packaged into a suitcase as shown in Fig. 1(B) for future point-of-care measurements.The diffuse reflectance spectroscopy will be easy to build using the design shown in Fig. 1(A) and all the key components can be easily purchased from Thorlabs.

Monte Carlo inverse model for diffuse reflectance spectral data processing
Our formerly reported scalable MC inverse model [14] was adapted with our new diffuse reflectance spectroscopy system for data processing as decried previously [23].In general, the MC model assumes oxygenated hemoglobin and deoxygenated hemoglobin as absorbers, and cellular components as scatterers.The tissue absorption coefficient was calculated by utilizing the widely used extinction coefficients reported previously [24], and the tissue scattering was calculated using Mie theory for spherical particles [25].The MC inverse model adaptively fits the measured reflectance spectrum to the MC simulated spectrum until the sum of squares error between the two is minimized, then the absorption and scattering over the interested wavelength bands can be extracted.Specifically, the extracted tissue absorption spectrum will be the sum of absorption spectra of oxygenated hemoglobin and de-oxygenated hemoglobin with different concentrations.The extracted tissue scattering spectrum will be proportional to the scattering spectrum of a reference solution with pure spherical particles (07310-15, Polysciences).The MC-extracted absorption spectra can be further processed to extract tissue vascular parameters using a fitting procedure with a linear combination of the extinction spectra of oxy-hemoglobin and deoxy-hemoglobin [26].Since the MC model works on an absolute scale while the tissue measurements are relative to a reflectance standard, a reference phantom with known optical properties was required to scale the tissue optical properties accurately [16].

Tissue-mimicking phantom studies
Tissue-mimicking phantoms were created to verify the diffuse reflectance system and MC inverse model for accurate measurement of absorption and scattering properties [16].Dehydrated human hemoglobin (H0267, Sigma-Aldrich) was used as an absorber, while Polystyrene spheres (07310-15, Polysciences) were used as a scatterer.Phosphate Buffered Saline (PBS) 1x (Fisher Scientific) was used to suspend the spheres and hemoglobin for the liquid phantoms.Two sets of tissue-mimicking phantoms with different initial scattering levels were prepared.The sphere stock was diluted with PBS to generate two phantoms with an initial average reduced scattering of 5 cm −1 and 10 cm −1 (between 450-650 nm), respectively.Within each phantom, 6 increasing concentrations of hemoglobin were added to generate average absorption coefficients of 1.05-7.52cm −1 (between 450 −650 nm).Hemoglobin concentration was increased by adding aliquots of the stock hemoglobin solution with a known absorption coefficient spectrum that was determined by a spectrophotometer.After each addition of hemoglobin stock solution, diffuse reflectance spectra were measured from the phantom.Diffuse reflectance from each phantom was processed using the MC inverse model to extract the absorption and scattering values of the other 11 phantoms.The phantom that generated the lowest error was selected as the reference phantom for the in vivo spectral data processing.

Animal experiments
To demonstrate the feasibility of the diffuse reflectance spectroscopy system for optical measurements on orthotopic tongue tumors in vivo, a pilot animal study was conducted according to a protocol approved by the University of Kentucky Institutional Animal Care and Use Committee (IACUC).Male or female athymic nude mice (nu/nu, Jackson Laboratory) with an age of 8 to 10 weeks were used for these studies.A total of 20 animals were assigned to (1) the control group (N = 9); and (2) the tumor group (N = 11).The animals assigned to the tumor group received an injection of SCC-61 cells (0.05 mL of cell solution with a concentration of 2-4 × 10 7 cells/mL) in the tongue under isoflurane anesthesia.The SCC-61 cell line was derived from an HNC tumor located at the base of the human tongue [27], which serves as a proper candidate for forming orthotopic tongue tumors in small animals.The mice were returned to the cage and monitored continuously for two weeks after the tumor cell injection.On day 10 after the tumor injection (or tumor diameter is ∼3 mm), mice were anesthetized with isoflurane for the diffuse reflectance spectroscopy study.Specifically, all optical measurements were performed on animals under anesthesia using mixture of isoflurane and room air (1.5%, v/v), and a heating pad was used to help mice maintain body temperature.Depending on the tumor size, three to five measurements were conducted on multiple locations of the tongue tissue region to get an average spectrum for each animal.

Optical measurement and data analysis
All tissue-mimicking phantoms and animals were measured using the optical spectroscopy device introduced in Fig. 1(A).Diffuse reflectance spectra were acquired from 450 nm to 650 nm with an integration time of 70 ms.All diffuse reflectance spectra were calibrated using a 20% reflectance standard (Spectralon, Labsphere).The MC model was used for spectral data processing to estimate the absorption and scattering parameters of tissue-mimicking phantoms and animal tissues.The extracted optical parameters from phantom studies were compared with their corresponding true values.The processed diffuse reflectance spectra from mice were used to estimate absorption coefficients and scattering coefficients spectra.The absorption spectra of the animal tissues were further fitted with a linear combination of the extinction spectra of oxy-hemoglobin and deoxy-hemoglobin to quantify tissue StO 2 and hemoglobin contents [26].
Student's t-test was used to generate p-values.MATLAB (Mathworks, USA) was used to perform all data processing and statistical analysis.

Monte Carlo model for accurate quantification of medium absorption and scattering
Figure 2(A)-(B) showed the comparison of the diffuse reflectance spectra measured on the two sets of tissue-mimicking phantoms and the corresponding MC simulated diffuse reflectance spectra.The high agreement between the MC simulated spectra and the measured spectra enables the accurate estimate of the absorption and scattering coefficients of the tissue-mimicking phantoms.Figure 2(C)-(D) showed the comparison between the MC-extracted absorption and scattering coefficients and the expected values for the phantoms.The average percent error was 6.0% and 5.7%, respectively, for the absorption and scattering coefficients estimation using the MC model when the diffuse reflectance spectra between 480 nm to 650 nm were used for the inversion.Figure 2(E) showed the reference phantom that generated the smallest error and its comparison with the MC fitted spectrum, while Fig. 2(F) showed the corresponding expected absorption and scattering coefficients of the reference phantom.

Diffuse reflectance spectra of normal tongue tissues and tongue tumors
Figure 3 provided a summary of the diffuse reflectance spectra measured on normal mice tongue tissues and mice tongue tumor tissues.Figure 3(A) showed representative photos of a mouse normal tongue and a mouse tongue tumor, that were measured using diffuse reflectance spectroscopy.Figure 3(B) showed all calibrated diffuse reflectance spectra measured on normal tongue tissues, while Fig. 3(C) showed all calibrated diffuse reflectance spectra measured on tongue tumor tissues.Comparing Fig. 3(B) and (C), we noticed higher variations of the spectra intensities in the tumor group compared to the normal tissue group.We also observed spectra shape differences around the 550 nm band between the normal tissue group and tumor tissue group, which is likely caused by the different oxygenation levels in the two types of tissues.Figure 3(D) showed the comparison of the averaged diffuse reflectance spectra between the normal tongue tissue group and the tumor tissue group.As shown in Fig. 3(D), the average diffuse reflectance spectra intensity from the tumor group is significantly higher compared to that from the normal tissue group between 450-600 nm band, while their average intensity is comparable between 620-700 nm band.

Absorption and scattering properties of normal tongue tissue and tongue tumors
Figure 4 showed the MC estimated tissue optical properties of both mice normal tongue tissues and tongue tumor tissues.Figure 4(A) showed the absorption coefficients for all normal mice tongue tissues, while Fig. 4(B) showed the absorption coefficients for all mice tongue tumor

Vascular parameters of normal tongue tissue and tongue tumors
Figure 5 showed the optically measured vascular parameters for both normal tongue tissues and tongue tumor tissues.These tissue vascular parameters were estimated from the MC-extracted absorption spectra as shown in Fig. 4(A)-(B).Figure 5(A) showed that tongue tumors had significantly lower StO 2 compared to normal tongue tissue, which was expected as HNC tumors were typically hypoxic.Figure 5(B) showed that tongue tumors had significantly lower [THb] compared to normal tongue tissues.Figure 5(C) showed that tongue tumors had significantly lower [HbO 2 ] compared to normal tongue tissue, while Fig. 5(D) showed that tongue tumors had comparable [Hb] compared to normal tongue tissue.Figure 5(E) showed representative diffuse reflectance spectra measured on normal tongue and tongue tumor tissues along with the MC-fitted spectra that were used for the vascular parameter estimations.The high agreement between the MC fitted spectra and the measured diffuse reflectance spectra enables accurate estimations of the tissue vascular parameters shown in Fig. 5.

Effect of tumor size on the optical parameters of tongue tumors
We observed high diversity in the optical parameters for tongue tumor tissues, therefore we further analyzed the tumor spectral data based on the tumor size.The tumors were further grouped into a small tumor group (volume was ∼60 mm 3 ) and a tiny tumor group (volume was ∼18 mm 3 ), though the tumors are always larger than the fiber probe common end.Figure 6(A) showed the representative photos of a tiny tongue tumor and a small tongue tumor.Figure 6(B) showed the comparison of the averaged diffuse reflectance spectra between the small tongue tumor group and the tiny tumor group.Interestingly, the averaged diffuse reflectance spectra intensity from the small tumor group is significantly lower compared to that from the tiny tumor group across the  whole wavelength band.Figure 6(C) showed that small tongue tumor tissues had significantly higher absorption coefficients compared to tiny tongue tumor tissues, and Fig. 6(D) showed that small tongue tumor tissues had significantly lower reduced scattering coefficients compared to tiny tongue tumor tissues.Figure 6(E) to (H) showed the tissue vascular parameters for small tongue tumors and tiny tongue tumors.Figure 6(E) showed that small tongue tumors had lower StO 2 compared to tiny tongue tumors, but not statistically significant.Figure 6(F) showed that small tongue tumors had comparable [HbO 2 ] compared to tiny tongue tumors.Figure 6(G)-(H) showed that small tongue tumor tissues had significantly higher [Hb] (p = 0.01) and [THb] (p = 0.03) compared to tiny tongue tumor tissues.

Discussion
Diffuse reflectance spectroscopy is well suited for tissue vascular characterizations on small animals in vivo [14] by probing a tissue volume [28].Subcutaneous tumor models (or flank tumor models) have been extensively used to study tumor angiogenesis [29], tumor hypoxia [30], and therapeutic responses [19,31].Orthotopic tumor models have also been developed to provide a more realistic model for cancer inquiries [18,19].Orthotopic HNC tumor models [32] have been reported recently for oral cancer studies [20].All these tumor models can be easily integrated with optical spectroscopy techniques for cancer research [14].Tongue cancer represents one of the most common HNC type that develop in mouth, which is the primary interest in our study.To the best of our knowledge, we are not aware of any study that explored the use of optical spectroscopy on orthotopic HNC models.In this study, we established an orthotopic tongue tumor model [20] and demonstrated the use of an optical spectroscopy for functional characterizations of orthotopic tongue tumors.In this exploratory study, we measured two types of tongue tumors (defined as "small" and "tiny" tumors) as one way to identify the suitable tumor sizes that may represent the most meaningful tumor biology for in vivo optical characterizations.For the first time, we reported the tissue absorption and scattering coefficients of mice normal tongue tissues and tongue tumor tissues, which may serve as proper references for future MC studies with a focus on tongue cancer.Our in vivo animal studies showed that mice tongue tumors have significantly different optical parameters compared to normal tongue tissues.We also captured that animal tongue tumors with different sizes had significantly different tissue absorption and scattering-based parameters.It should be noted our optical spectroscopy can be also useful for investigations on other types of human epithelial cancers such as oral cancer, cervical cancer, skin cancer et al.
In typical optical spectroscopy platforms, a fiber probe with single or multiple fibers was commonly used for light delivery and collection [33].Generally, the distance between the illumination and collection fibers determines the optical probing depth in tissue.A former study showed that the sampling depth of visible diffuse reflectance spectroscopy in skin tumor models was around half of the source-detector distance [21].A fiber probe with a typical centersurrounding geometry design will sense tissue primarily from a certain depth as it has a fixed source-detector separation.The use of the commercially available fiber probe (BF19Y2HS02, Thorlabs) in this study has source-detector distances varied from 100 µm to 1 mm (Fig. 1).Consequently, the fiber probe used in our optical spectroscopy device can sense the tissue regions ranging from 50 µm up to 0.5 mm in depth when using visible light.The thickness of tongue epithelium is around 90 µm on average for 2-month old mouse [34].Therefore, the fiber probe implemented in our optical spectroscopy system will provide sufficient sensing depth for optical measurements of tongue tumors formed in the mouse tongue tissues.The design of a fiber probe geometry may affect the measured diffuse reflectance intensities, while the estimated tissue optical properties should be similar as long as the optical fiber probes have same optical sensing depth.As long as an optical spectroscopy system enables measurements with sufficient signal to noise ratio, there is no need to standardize the light power for diffuse reflectance measurements.This is because all diffuse reflectance spectra measured on tissue samples will be calibrated by a spectrum of a standard puck measured at the same time to count the LED throughput change with time, which was negligible in this study.
Figure 5 showed that the tongue tumors had lower absorption coefficients and increased tissue scattering coefficients compared to normal tongue tissues.This trend reported here is different compared to what we observed in flank tumor models where the flank tumors usually had increased absorption but reduced scatterings compared to the normal skin tissues [15].This might be due to the inherent structure and component differences between tongue tissue and skin tissue.We observed that the scattering coefficient of tongue tumors is significantly higher than those from the normal tongue tissues, which is consistent with the findings reported in a former mouse tongue imaging study using a hyperspectral imaging system [35].This scattering properties difference may be explained by the fact that tongue tumor tissues have more collagen and irregular fibers [36] than normal tongue tissues, thus the source of stronger scattering [37].Our optical vascular parameters showed that tongue tumors had significantly lower StO 2 compared to normal tongue tissues as expected given solid HNC are typically hypoxic [10].We further compared the [HbO 2 ] and [Hb] for normal tongue tissues and tongue tumors to reveal the source of tumor hypoxia.Figure 5 data showed that tongue tumors had significantly lower [HbO 2 ] but comparable [Hb] compared to normal tongue tissue, which suggested that the oxygen supply might be not adequate for the tumor region compared to normal tongue tissues.Our study primarily focused on the measurements of vascular StO 2 , [HbO 2 ], and [Hb] for tumor characterizations using simple diffuse reflectance spectroscopy, while it should be noted that blood flow could be also a potential powerful biomarker for tumor therapeutic applications [38] and it can be rapidly measured by diffuse speckle spectroscopy techniques [39].
In flank tumor models, tumors can be classified as a tiny tumor (diameter is ∼ 3 mm), a small tumor (diameter is ∼6 mm), and a middle tumor (diameter is ∼8 mm) to represent the tumors at different development stages [14,40].Following the same guideline, we use the tiny tongue tumor to represent an early-stage tongue tumor, while use the small tongue tumors to represent advanced tongue tumors given the mouse tongue is relatively small.We noted that the tumors developed in the tongue region were not always spherical.Tumor volume can be estimated using the formula (V = (W × W × L)/2) introduced by Faustino-Roch et al. [41].The average tumor volume for tiny tongue tumor was ∼18 mm 3 , and the average tumor volume for small tongue tumor was ∼60 mm 3 .Figure 6 showed that small tongue tumors had significantly different tissue absorption and scattering-based parameters compared to tiny tongue tumors.Small tongue tumors had significantly decreased reduced scattering coefficients and significantly increased absorption coefficients compared to tiny tongue tumors.Interestingly, the reduced scattering and absorption levels of small tongue tumors are almost comparable to those in normal tongue tissues if comparing the data shown in Fig. 4 and Fig. 6.This difference in scattering levels between small and tiny tongue tumors may be explained by the collagen fiber arrangement in tumors.A recent study [36] reported that well-differentiated oral squamous cell carcinoma (WDSCC) and normal mucosa had denser packing, while the moderate-differentiated oral squamous cell carcinoma (MDSCC) had loosing packing.The study also showed that WDSCC and normal mucosa had comparable collagen fiber thickness, which is significantly larger than that in MDSCC.Taken together, these findings may support our observations that small tongue tumors had similar scattering compared to normal tongue tissues, but significantly lower scattering compared to tiny tongue tumors.Different from flank tumor models, small tongue tumors might be already at an advanced stage in the tongue tissue region given the mouse tongue is relatively small.The small tongue tumors may have started to necrotize which leads to both functional and morphological changes in the tissue.Small tongue tumors had even lower StO 2 compared to tiny tongue tumors, though it is not statistically significant, which suggested heavier hypoxia in small tongue tumors as expected.Figure 6 showed that small tongue tumors had significantly higher [Hb] but comparable [HbO 2 ] compared to tiny tumor tissues, which is different from that discussed above between the normal tongue tissue and tongue tumor tissues.These data suggested that small tongue tumors and tiny tumors may have similar inadequate oxygen supply but different oxygen consumption.Small tongue tumors might have higher oxygen consumption compared to tiny tongue tumors.As discussed above, tongue tumors with different sizes investigated in this study may have significantly different functions and morphology.Both functional and morphological changes will change the absorption and scattering of tongue tissues, thereby affecting the measured diffuse reflectance spectra.From cancer biology inquiry perspective, it will be challenging to standardize the optical measurements or normalize the data based on the tumor sizes.Instead, we would recommend researchers carefully monitor the tumor size and characterize the tumors at certain sizes for HNC investigations.Nevertheless, these data showed that orthotopic tongue tumor optical absorption and scattering parameters could be significantly different when they were measured at different stages.Considering the small mice tongue size and the optical properties differences between different tissue types, tiny tongue tumors might be more appropriate for optical investigations compared to small tongue tumors in the future if orthotopic tongue tumor models are used for HNC research.
In phantom data processing, the wavelength range of 450 nm-650 nm was used to help us identify the best reference phantom spectrum that can be used for future in vivo data processing.In all in vivo animal data processing, any wavelength range within 450 nm-650 nm can be used for the MC inversion model processing, however the wavelength range of 480 nm-600 nm was selected for the animal data processing to achieve the highest accuracy across the whole wavelength range (480 nm-600 nm) for all twenty animals.For the first time, we reported the tissue absorption coefficients and reduced scattering coefficients for mice tongue tissues and tongue tumor tissues.We observed higher absorption levels but slightly lower scattering levels for normal tongue tissue compared to skin tissues [15].Unfortunately, we could not find any relevant report regarding tongue tissue optical properties for us to directly compare.However, we confirmed the high agreement between the MC simulated spectral data and measured diffuse reflectance data for each measurement on all animals, which ensured the accurate estimation of tongue tissue optical properties.The reference phantom used in the scalable MC inverse model may also slightly affect the absolute tissue optical properties, while the relative differences between different tissue types would be the same regardless of the selection of reference phantoms as tested by us.The tissue optical properties reported here will still serve as good references for future optical spectroscopy investigations with a focus on tongue cancer.
Several imaging techniques including endoscopy, MRI, PET-CT, et al. have been used in clinics for HNC diagnosis, while the current gold standard for HNC diagnosis still largely depends on histopathology via an invasive procedure, tissue biopsy.Our portable diffuse reflectance spectroscopy may offer a cost-effective, labeling-free, easy-to-access approach for tumor function characterizations, which could be useful for HNC diagnosis and therapeutics applications in clinics in the near future.In this study, we demonstrated the feasibility of our portable optical spectroscopy for optical characterizations of orthotopic tongue tumor in vivo.In our future study plan, we will utilize our techniques to study the effect of tumor oxygenation on HNC RT efficacy using orthotopic tongue tumor models.All optical components used in our optical spectroscopy can be purchased from Thorlabs, therefore any researchers can build their own spectroscopy platforms easily for cancer research if interested.MC inverse model was used for accurate estimation of tissue absorption and scattering coefficients in this present study, while we will plan to explore less expertise-dependent data processing techniques so the techniques will be potentially easier to use.Nevertheless, we demonstrated the use of a simple diffuse reflectance spectroscopy platform for optical measurements of orthotopic tongue tumor models.These demonstrations showed that the reported optical platform could be a useful tool for HNC research using orthotopic HNC models.To maximize the ease of and accessibility in characterizing tissue function in vivo, it is highly significant to develop new optical spectroscopy and imaging tools with high-portable [42] and low-cost footprints, allowing one to quantify tissue functional endpoints in vivo with easy access, in the aim of advancing many critical biomedical inquires.In this study, we reported an optical spectroscopy platform that is highly portable and low-cost, thereby potentially providing point-of-care vascular measurements on tumor models with high user access in the future, which will significantly advance cancer research.It should be noted that our system can be further miniaturized by using a more compact LED source (MNWHL4, Thorlabs) or custom fabricated LED source with a lower cost [43] for future translational cancer applications.

Conclusion
We reported an easy-to-build, portable diffuse reflectance spectroscopy device along with a Monte Carlo inverse model to quantify tissue absorption and scattering-based parameters of orthotopic tongue cancer models in vivo.For the first time, we reported the tissue absorption and scattering coefficients of mouse normal tongue tissues and tongue tumors.We also characterized the tissue vascular parameters for both normal tongue tissues and tongue tumor tissues and we showed that tongue tumors had different optical parameters compared to normal tongue tissues.These results demonstrated the potential of diffuse reflectance spectroscopy to noninvasively evaluate tumor biology using orthotopic tongue cancer models for future head and neck cancer research.

Fig. 1 .
Fig. 1. (A) Schematic of the portable optical spectroscopy platform.(B) The actual photo of the portable optical spectroscopy platform and the potential use for diffuse reflectance measurements on a mouse tongue.

Fig. 2 .
Fig. 2. Comparison between the MC fitted spectra and diffuse reflectance spectra measured from (A) relatively low scattering phantoms and (B) relatively high scattering phantoms.(C) Comparison between the MC extracted absorption coefficients and the expected absorption coefficients.(D) Comparison between the MC extracted reduced scattering coefficients and the expected reduced scattering coefficients.(E) The spectrum of the reference phantom that generated the lowest error in extracting scattering and absorption for all phantoms.(F) The corresponding expected absorption coefficients and reduced scattering coefficients for the reference phantom.

Fig. 3 .
Fig. 3. (A) Representative photos of a normal mouse tongue and a mouse tongue tumor.Diffuse reflectance spectra measured on normal mice tongue tissues (B) and mice tongue tumor tissues (C).(D) Comparison of the averaged diffuse reflectance spectra measured on the normal tongue tissues and tongue tumor tissues.

Fig. 4 .
Fig. 4. MC extracted absorption coefficients of normal mice tongue tissues (A) and tongue tumor tissues (B).(C) The comparison of the averaged absorption coefficients spectra for normal tongue tissues and tongue tumor tissues.MC extracted reduced scattering coefficients of normal mice tongue tissues (D) and tongue tumor tissues (E).(F) The comparison of the averaged reduced scattering coefficients spectra for normal tongue tissues and tongue tumor tissues.

Fig. 5 .
Fig. 5. MC extracted tissue vascular parameters including (A) StO 2 , (B) [THb], (C) [HbO 2 ], and (D) [Hb] for both mice normal tongue tissues and tongue tumor tissues.(E) Comparison between the MC fitted spectra and diffuse reflectance spectra measured on a normal mouse tongue tissue and a mouse tongue tumor tissue.

Fig. 6 .
Fig. 6. (A) Representative photos of a tiny tongue tumor and a small tongue tumor formed in mice.(A) Comparison of the averaged diffuse reflectance spectra for tiny tongue tumors and small tongue tumors.(C) The comparison of the averaged absorption coefficients spectra and (D) the averaged reduced scattering coefficients spectra for tiny tongue tumors and small tongue tumors.MC extracted tissue vascular parameters including (E) StO 2, (F) [HbO 2 ], (G) [Hb], and (H) [THb] for both small tongue tumors and tiny tongue tumors.