View source for Imaging Strategies for Enhancing Blood Visibility in Oral Tissue

== Introduction ==
Early detection is critical for improving outcomes in oral cancer, motivating a range of noninvasive optical approaches that aim to flag suspicious tissue changes at the point of care. Among these, fluorescence-based imaging is attractive because collagen can emit in the green band when excited with blue/violet light. However, interpreting oral fluorescence images remains challenging because a decrease in measured fluorescence is not specific: reduced signal may arise from malignant change, but it can also reflect benign physiological variation—most notably blood content. In the oral cavity, this ambiguity is especially pronounced because hemoglobin strongly absorbs around 415 nm, so vascularized regions can appear “dark” even in otherwise healthy tissue. 

Alongside this physiological confound are the optics of the measurement itself. Oral tissue is humid and highly reflective, producing strong specular highlights and additional scattering that reduce contrast and can mask the very absorption differences needed to localize blood. Spectrally, the visible range is particularly problematic because hemoglobin absorption overlaps both excitation/reflectance pathways and tissue autofluorescence, whereas the near-infrared (∼700-900 nm) reduces hemoglobin absorption and autofluorescence background while improving penetration. These challenges motivate imaging strategies that suppress surface reflections and better isolate signals attributable to hemoglobin versus underlying tissue structure. 

In this study, we investigate practical imaging strategies for enhancing blood visibility in oral tissue by comparing: 
* cross-polarized reflectance imaging to suppress specularity and glare 
* fluorescence imaging with spectral filtering to isolate emission and increase vascular contrast

By evaluating these approaches in a controlled acquisition setup and analyzing the channels most informative for reflected versus emitted light, we aim to clarify the tradeoffs of each method and inform future systems that more robustly separate blood-driven absorption from tissue changes relevant to diagnosis.

== Background ==
Optical imaging techniques have been increasingly explored for noninvasive evaluation of oral tissue properties and early detection of pathological changes. One fundamental challenge in optical imaging of the oral cavity is the strong absorption of visible light by hemoglobin (Hb), which significantly affects both reflectance and fluorescence signals captured from tissue (Sweeny et al., 2011). Hemoglobin absorbs light strongly in the visible spectrum, particularly near 415 nm, and this property underlies much of the contrast observed in reflectance imaging modalities used to detect vascular features (Sweeny et al., 2011).

Reflectance imaging has been investigated clinically as an adjunct to conventional examination for screening oral lesions. Reflectance spectra of oral mucosa reveal absorption features attributable to oxygenated hemoglobin that correlate with increased microvasculature in neoplastic tissue (Tiwari et al., 2019; Sweeny et al., 2011). Optical devices using specific illumination bands, such as green‑amber light (~540–575 nm), aim to enhance visualization of underlying vasculature by exploiting these absorption characteristics (Tiwari et al., 2019).

Endogenous tissue fluorescence provides an additional source of contrast in optical imaging. Fluorescence arises when endogenous fluorophores absorb excitation light and re‑emit it at longer wavelengths due to the Stokes shift, enabling emission signals to be separated from reflected light (Wu et al., 2018). In oral tissue, structural proteins such as collagen contribute significantly to autofluorescence, and changes in these signals have been associated with pathological transformation (Pavlova et al., 2008; Tiwari et al., 2019). Studies show that dysplastic and malignant tissues can exhibit decreased fluorescence relative to normal tissue, partly due to alterations in fluorophore composition and increased absorption by blood and other chromophores (Pavlova et al., 2008; Tiwari et al., 2019).

Polarization imaging offers a method to reduce specular reflection from tissue surfaces by filtering light based on its polarization state. While polarization techniques have been used in various biomedical applications to probe tissue structure and reduce surface reflectance (Tuchin, 2016), their application in oral tissue imaging remains relatively exploratory. Polarization‑based imaging can improve contrast by suppressing surface reflections that otherwise mask subsurface optical properties, although most existing oral applications focus on polarization‑sensitive optical coherence tomography rather than wide‑field polarization reflectance imaging (Yoon et al., 2015).

Handheld and wide‑field imaging systems that combine reflectance and fluorescence have been evaluated clinically, with mixed results regarding sensitivity and specificity for oral cancer screening (Sweeny et al., 2011; Tiwari et al., 2019). These findings support continued investigation into systematic imaging strategies that isolate absorption and emission effects while minimizing confounding surface reflection, particularly under controlled illumination conditions.

'''Oral Tissue Reflectance Challenges:'''
Imaging oral tissue presents specific optical challenges due to the combined effects of blood absorption and surface properties. Illumination with 415 nm light is strongly absorbed by hemoglobin, which can reveal vascular structures, but the naturally humid and reflective surface of the lips produces intense specular reflections. In addition, scattering within the superficial tissue layers diffuses the light, reducing image contrast and masking the true absorption patterns of blood. As a result, specular highlights can dominate the captured signal, making it difficult to distinguish variations caused by hemoglobin from those caused by surface reflections or scattering. To obtain accurate measurements of blood absorption, imaging conditions must minimize or remove these confounding effects, isolating the reflectance arising directly from the tissue and allowing reliable interpretation of hemoglobin-related contrast.
[[File:Specularity.png|thumb|325 px|centre|Figure 1: Illustration how light interacts with a particle]]

== Methods ==
We compared two imaging configurations designed to increase the visibility of blood-driven absorption in oral tissue: (1) cross-polarized reflectance imaging to suppress surface glare, and (2) fluorescence imaging with spectral filtering to isolate emitted light. Both setups used a DSLR camera (Nikon D500) and a fixed geometry in which the camera and illumination source were placed at approximately 45° relative to the target (lower lip) or a Macbeth Color Checker (MCC) used for reference and setup validation.
[[File:labsetup.png|thumb|325px|centre|Figure 2: Dr. Joyce Farrell as the test subject]]

=== Cross polarization methodology ===
Cross-polarization was implemented with two polarizers: one mounted on the illumination source and one mounted on the camera lens. Images were acquired in parallel orientation, which allows specular reflections to reach the sensor, and crossed orientation, which suppresses specular reflection to better reveal subsurface reflectance and absorption patterns. Linear polarization was treated as preferable for this application because it provides sharper differentiation between blocked surface glare and transmitted subsurface scatter than circular polarization.

The camera (Nikon D500) and light source were positioned at ~45° relative to the lips or MCC, with the subject centered along the illumination and imaging axes; this geometry intentionally permits specular reflections to occur so that the effect of cross-polarization on suppressing glare can be observed. One polarizer on the light source was kept fixed, while the polarizer on the camera lens was rotated to capture two images (parallel vs. crossed).
[[File:CP-Image-Nikon.png|thumb|325px|centre|Figure 3: A photo of a Macbeth Color Checker that has been cross-polarized and processed by the Nikon D500 software]]

'''Is cross-polarization using circular polarizers less effective than cross-polarization using linear polarizers?''' 
With circular polarizers, the polarization change happens more gradually through tissue; less of the returned signal actually switches from one handedness to the other; and more good signal is rejected along with the bad signal.

'''What if you only use a polarizing filter on the camera?'''
If unpolarized light (regular light) is shined at the tissue and only put a polarizing filter on the camera, a one-sided measurement is being done. The camera only sees light with one specific polarization orientation. This means most of the signal is lost because a polarizer blocks roughly 50% of unpolarized light; surface reflections and scattered light still interfere with deeper tissue imaging; and the light is not being filtered and rejected in a controlled way. 

The '''Dichromatic Reflectance Model (DRM)''' provides a conceptual framework for understanding the light scattering seen inside the tissue, shown in Figure 3. According to the DRM, the reflected light from tissue is the sum of a specular surface reflection, which preserves polarization, and a diffuse subsurface component, resulting from scattering and absorption within the tissue. Crossed polarizers effectively remove the surface reflection, but internally scattered light is depolarized and contributes to image blur. This explains why cross polarization enhances subsurface contrast while not fully eliminating scattering effects inside the tissue (Wandell, 2025).
[[File:CP-Diagram.jpg|thumb|400px|centre|Figure 4: Diagram of cross-polarization methodology. The blue arrows inside the tissue represent scattering of light within the tissue itself.]]

=== Fluorescence imaging methodology ===
Fluorescence imaging used ~415 nm excitation to elicit green-band emission (approximately 500–600 nm) from oral tissue (e.g., collagen fibers). Blood appears dark because hemoglobin absorbs both the excitation and portions of the emitted light (including around 540 and 580 nm), increasing vessel contrast against brighter fluorescing tissue.

A 415 nm short-pass filter was placed on the light source to ensure excitation wavelengths below 415 nm reached the tissue. A 475 nm long-pass filter was placed on the camera lens to block reflected excitation light while passing emitted fluorescence in the 500–600 nm range.

As in cross-polarization, the camera and blue LED were positioned at ~45° relative to the lips or MCC with the subject centered on the illumination/imaging axes. Images were then acquired for analysis focused on the fluorescence band.

'''The effect of fluorescence imaging''': 
* '''Enhanced blood vessel contrast:''' Blood absorbs both the 415 nm excitation and the emitted green fluorescence, causing blood vessels to appear darker against bright fluorescent tissue.
* '''Improved subsurface visibility:''' Fluorescence originates from collagen below the surface, reducing the influence of surface reflection compared to reflectance imaging.
* '''Selective signal isolation:''' The 475 nm long pass filter blocks reflected excitation light, allowing cleaner measurement of fluorescence emission.

[[File:FluorescenceDiagram.png|thumb|400px|centre|Figure 5: Diagram showing fluorescence imaging methodology]]

=== Image formats and preprocessing ===
To examine the impact of camera processing, both “unprocessed” (raw-like, maximum post-processing flexibility) and “processed” (camera-applied settings such as white balance/sharpening and compression) imagery were compared.
[[File:Unprocessed.png|thumb|325 px|centre|Figure 6: Examples of unprocessed images]]
[[File:Processed.png|thumb|325 px|centre|Figure 7: Examples of processed images]]

=== Use of MatLab ===
All quantitative viewing/inspection was performed on the unprocessed raw sensor images in MATLAB using a custom script, s_crossPolarizationAndFluorescence.m, which reads Bayer-patterned sensor data, parses image metadata including the CFA pattern, and outputs either an RGB rendering or a single color channel for analysis.
[[File:Raw matlab.png|thumb|325 px|centre|Figure 8: A raw image rendered in MatLab]]

== Results ==
===Blue vs. green channels===
In our analysis, the green channel primarily captures emitted fluorescence in the 500–600 nm range, largely from collagen fiber fluorescence. Blood vessels tend to appear darker in this channel because hemoglobin absorbs the 415 nm blue excitation light, which reduces the amount of fluorescence that can be generated and detected. In contrast, the blue channel is dominated by reflected/scattered excitation light (roughly 400–500 nm) rather than fluorescence, so it often has higher background noise from surface reflections and typically shows weaker blood-related contrast compared to the green channel because hemoglobin absorption is more pronounced at green wavelengths than at blue in this context.
=== Images from the Lower Lip Experiments ===
<gallery widths=300 heights=200>
File:CP-BlueChannel.png|thumb|325 px|centre|Figure 9: Cross-polarization reducing specularity and glare
File:F-GreenChannel.png|thumb|325 px|centre|Figure 10: Fluorescence showing better mapping and localization of hemoglobin
File:Reflected-light.png|thumb|325 px|centre|Figure 11: For reflected light, specify the blue channel
File:Emitted-light.png|thumb|325 px|centre|Figure 12: For emitted light, specify the green channel
</gallery>

== Conclusions ==

The results of this project demonstrate that cross polarization and fluorescence imaging each provide complementary insights into oral tissue, contributing to the overall goal of visualizing blood-related structures while mitigating confounding factors.

Cross-polarization imaging proved more valuable for structural assessment than direct hemoglobin detection. By suppressing surface specular reflection, this method enhances visualization of tissue architecture and collagen organization. Such structural features can serve as diagnostic indicators independent of blood absorption, providing a clearer view of tissue morphology. However, internal scattering within tissue limits the ability of polarization alone to resolve deeper structures, as cross polarization primarily suppresses surface-reflected light while light scattered internally remains depolarized and reduces image sharpness.

Fluorescence imaging effectively highlights both metabolic and structural abnormalities and is sensitive to the presence of hemoglobin. By exploiting collagen autofluorescence and the absorption of excitation and emission light by blood, fluorescence provides strong contrast for vascular structures. This makes it particularly useful for mapping blood distribution and identifying regions where hemoglobin attenuates the emission signal.

===Limitations and Future Directions===
Several aspects of this study highlight opportunities for improvement and future investigation. Fluorescence imaging was not combined with cross polarization, which would have further suppressed surface reflection and improved isolation of subsurface signals. Integrating these techniques in future work could maximize vessel contrast and provide clearer separation of structural and hemoglobin-dependent features.

The Dichromatic Reflectance Model (DRM), found in Chapter 5, section 7 of the Foundations of Image Systems Engineering book by Dr. Brian A. Wandell, provides a useful framework for understanding the limitations of cross polarization in this context. According to the DRM, reflected light consists of two components: a specular surface reflection and a diffuse subsurface component caused by scattering and absorption within the tissue. Cross polarization effectively suppresses the surface term because it preserves polarization, but internally scattered light becomes depolarized and contributes to image blur (Wandell, 2025). Consequently, polarization alone cannot fully remove the effects of subsurface scattering, and complementary imaging strategies or computational corrections are required to better isolate structural and absorption features.

Additionally, sample diversity was limited, with only one subject used for fluorescence imaging and three subjects for cross polarization. This restricts the generalizability of the results and prevents robust statistical conclusions. Expanding the study to include more subjects with varied tissue characteristics would strengthen future analyses.

Overall, while this study demonstrates the individual benefits of cross polarization and fluorescence imaging, the DRM highlights the inherent limits of polarization for internal scattering. Future studies integrating these modalities and expanding the sample set will provide a more comprehensive understanding of optical imaging strategies for oral tissue.

== References ==
Pavlova, I., Williams, M., El-Naggar, A., Richards-Kortum, R., & Gillenwater, A. (2008). Understanding the Biological Basis of Autofluorescence Imaging for Oral Cancer Detection: High-Resolution Fluorescence Microscopy in Viable Tissue. Clinical Cancer Research : An Official Journal of the American Association for Cancer Research, 14(8), 2396–2404. https://doi.org/10.1158/1078-0432.CCR-07-1609

Sweeny, L., Dean, N. R., Magnuson, J. S., Carroll, W. R., Clemons, L., & Rosenthal, E. L. (2011). Assessment of Tissue Autofluorescence and Reflectance for Oral Cavity Cancer Screening. Otolaryngology, 145(6), 956–960. https://doi.org/10.1177/0194599811416773

Tiwari, L., Kujan, O., & Farah, C. S. (2020). Optical fluorescence imaging in oral cancer and potentially malignant disorders: A systematic review. Oral Diseases, 26(3), 491–510. https://doi.org/10.1111/odi.13071

Tuchin, V. V. (2016). Polarized light interaction with tissues. Journal of Biomedical Optics, 21(7), 071114. https://doi.org/10.1117/1.jbo.21.7.071114 

Wandell, B. A. (2025, September 25). Foundations of Image Systems Engineering. Github.io. https://wandell.github.io/FISE-git/#image-systems-engineering-at-stanford

Wu, C., Gleysteen, J., Teraphongphom, N. T., Li, Y., & Rosenthal, E. (2018). In-vivo optical imaging in head and neck oncology: basic principles, clinical applications and future directions. International Journal of Oral Science, 10(2), 10. https://doi.org/10.1038/s41368-018-0011-4

Yoon, Y., Won Hyuk Jang, Xiao, P., Kim, B., Wang, T., Li, Q., Ji Youl Lee, Chung, E., & Ki Hean Kim. (2015). In vivo wide-field reflectance/fluorescence imaging and polarization-sensitive optical coherence tomography of human oral cavity with a forward-viewing probe. Biomedical Optics Express, 6(2), 524–524. https://doi.org/10.1364/boe.6.000524

Figure 1: https://www.microtrac.com/es/base-de-datos-de-conocimientos/static-light-scattering/

== Appendix I ==
'''Google Drive Links'''
*Images: https://drive.google.com/drive/folders/1Ag5cgZCwH47aWsyPj9bw0VDWQZA30WPj?usp=sharing
*MatLab: https://drive.google.com/drive/folders/1ClfkDamncIVvAc4tG6Txndqy_Ym_-eTO?usp=sharing

== Appendix II ==
René took the majority of the photos with Joyce in the lab. Rubina also took some photos and processed the images with the MatLab script. Both worked equally on the presentation slides and the write-up.