Investigation of the Clinical Potential of Polarization-Sensitive Optical Coherence Tomography in a Laryngeal Tumor Model

Abstract

BACKGROUND:

The vocal cord tissue consists of three anatomical layers from the surface to deep inside: the epithelium that contains almost no collagen, the lamina propria that is composed of abundant collagen, and the vocalis muscle layer. It is clinically important to visualize the tissue microstructure using a non-invasive method, especially in the case of vocal cord nodules or cancer, since histological changes in each layer of the vocal cord cause changes in the voice. Polarization-sensitive optical coherence tomography (PS-OCT) enables phase retardation measurement to evaluate birefringence of tissue with varied organization of collagen fibers in different tissue layers. Therefore, PS-OCT can visualize structural changes between normal and abnormal vocal cord tissue.

METHOD:

A rabbit laryngeal tumor model with different stages of tumor progression was investigated ex-vivo by PS-OCT. A phase retardation slope-based analysis, which quantifies the birefringence in different layers, was conducted to distinguish the epithelium, lamina propria, and muscle layers.

RESULTS:

The PS-OCT images showed a gradual decrease in birefringence from normal tissue to advanced tumor tissue. The quantitative analysis provided a more detailed comparison among different stages of the rabbit laryngeal tumor model, which was validated by the corresponding histological findings.

CONCLUSION:

Differences in tissue birefringence was evaluated by PS-OCT phase retardation measurement. It is also possible to indirectly infer the dysplastic changes based on the mucosal and submucosal alterations.

Introduction

The larynx, commonly called the voice box, is involved in breathing, phonation, and protecting the airway against food aspiration. The vocal cord of the larynx is composed of five histo-anatomical layers, that is, the mucosa (epithelium with basement membrane), three layers of lamina propria (superficial, intermediate, and deep layer), and the vocalis muscle [1, 2].

Most laryngeal cancers originate in the glottis, which is a thin opening between the vocal cords, while squamous cell carcinomas involve the vocal mucosa. Since the phonatory mechanism requires the vibration of vocal mucosa, voice alterations are noted with the smallest deterioration in the mucosal structure.

Early detection of laryngeal cancer is of utmost importance. Imaging modalities such as computed tomography, ultrasound, and magnetic resonance imaging are favorable for evaluating the tumor size, extent, and nodal metastasis. However, their limited resolution makes it difficult to differentiate the laryngeal cancer tissue from the normal tissue. Optical coherence tomography (OCT) can provide pathological and structural information at the histologic level with its high resolution and sensitivity, thus, demonstrating great potential in cancer detection [3]. OCT has been applied to compare normal and laryngeal cancer tissues [4], where OCT revealed the loss of basement membrane integrity in patients with laryngeal cancer. Additionally, conventional OCT has been utilized to improve the precision of surgical interventions [5]. However, these studies utilized the optical reflectivity (intensity) information, which does not directly represent the collagen organization in the larynx tissues, thus, restricting the application of conventional intensity OCT in the early detection of laryngeal cancer.

Polarization-sensitive optical coherence tomography (PS-OCT) [6] is a functional extension of OCT with the capability of measuring the polarization properties of biological tissues, such as tissue birefringence [7]. Tissue birefringence occurs in fibrous structures such as collagen. When the collagen fibers are highly organized along a wide area of the tissue, the phase retardation generated between the fast and slow axis of the collagen fibers can be measured by PS-OCT. The differences in tissue birefringence in normal and abnormal components of tissues, such as skin and cartilage, can be distinguished by PS-OCT [8, 9]. The collagen fibers in healthy connective tissues are normally well organized, which provides relatively high birefringence. However, the tissue birefringence decreases when the collagen organization is destroyed in conditions, such as burns, physical damage, or tumor (Fig. 1). By using the PS-OCT based-phase retardation images, Fleischhauer et al. [10] observed lower birefringence in tumor tissue than in the normal tissue of laryngeal cancer. Differentiating the tissue layers by birefringence has been reported in human retina [11] and swine articular cartilage [12], using PS-OCT phase retardation measurement. There are several other studies on the laryngeal tissue or its tumor tissue imaging with PS-OCT [13, 14]. Nevertheless, there are no reports on the method to differentiate the tissue layers or quantify the difference between normal and tumor tissues based on the layered structural features.

Fig. 1.

Study design. A Normal laryngeal wall with normal pattern of birefringence and slope of phase retardation; B fibrosis or degeneration by tumor, which has the abnormal pattern in slope of phase retardation according to extent of muscular injury

In this study, a rabbit laryngeal tumor model, with different stages of tumor progression, was investigated by PS-OCT, ex-vivo. The PS-OCT images showed a gradual decrease in birefringence from the normal tissue to advanced stage tumor tissue, which was histologically supported. A phase retardation slope-based analysis [12] was further conducted on the PS-OCT data to distinguish the epithelial layer from the lamina propria. The slope analysis quantifies the birefringence in different layers and provides a detailed comparison among the different stages of laryngeal tumor.

Materials and methods

Sample preparation

Animal experiments were performed in virtue of the Guide for the Care and Use of Laboratory Animals [15]. The Kosin University College of Medicine had the approval for the animal study from the Animal Care and Use Committee. PS-OCT imaging was performed on three laryngeal tissue specimens: a normal tissue, an early stage tumor tissue, and a late stage tumor tissue. After tumor inoculation as per the study groups, the rabbits were euthanized with intravenous pentobarbital, followed by collection of laryngeal tissues. Laryngeal tissue specimens were cut into approximately 6 × 6 cm squares, fixed in 10% neutral buffered formalin, and embedded in paraffin. PS-OCT imaging was performed on the bulk tissue. For histology, hematoxylin and eosin (H&E) stained serial sections (4 μm thick) were examined by microscopy [16].

Imaging system

The laryngeal tissue specimens were imaged by a PS-OCT system (system details shown in Ref [6]). It contained a swept-source (Axsun Technology Inc., Billerica, MA, USA), with center wavelength of 1.06 μm, full width at half maximum of 110 nm, and a scanning rate of 100 kHz. With a passive polarization delay unit, two orthogonal polarization states were obtained to illuminate the sample. The polarization sensitive features (e.g., birefringence property) were measured by a Jones-matrix-based method [6]. The average power on the sample was around 2 mW. The depth resolution of the system was ~ 8.1 μm, and the lateral resolution was ~ 19.2 μm. The PS-OCT images were measured with the illumination light almost perpendicular to the tissue surface. The imaging volume size was around 4.5 × 4.5 mm in lateral domain and 1.5 mm in depth.

Quantification method

Based on the slope of the PS-OCT phase retardation, a quantification analysis was conducted to investigate the invasion progression of the laryngeal tumor model. A preliminary PS-OCT phase retardation slope-based segmentation analysis was conducted to distinguish the three laryngeal tissue layers–the epithelium, lamina propria, and vocalis muscle. The thickness of the epithelium of the three tissues was compared.

Results

The PS-OCT imaging results of the rabbit larynx tissues are shown in Fig. 2. The images in the three columns, from left to right, correspond to the representative results from the normal tissue, the early stage tumor tissue, and the late-stage tumor tissue, respectively. The en-face intensity OCT images of the three specimens are shown in Fig. 2A–C. Figure 2A shows the normal larynx surface. Figure 2C shows a circular shadowed region on the right side, probably related to the late stage tumor. In Fig. 2B, the early stage tumor invasion is not confirmed. Figure 2D–F show the representative B-scan intensity OCT images of the normal tissue, the early stage tumor tissue, and the late stage tumor tissue, respectively. Figure 2E shows that the tumor placed in the Reinkes’ space is pushing the mucosa upward as it grows and the lower level on the left side shows that the layer is missing due to the tumor. There is a raised area in the right part of the intensity OCT B-scan in Fig. 2F, which matches with the circular shadow in Fig. 2C. Apart from the shadow and the raised area, there is no clear evidence of tumor progression. Figure 2F indicates considerable thinning of the mucosa caused by the enlarged tumor. Meanwhile, the corresponding phase-retardation B-scan images in Fig. 2G–I show detailed features related to the tumor progression. From the rainbow color-map setting, the phase retardation level of the three groups can be identified; the tighter band pattern indicates higher tissue birefringence, or larger phase retardation slope. Thus, the rainbow-like band pattern in the phase retardation image in Fig. 2G (entire image), and in Fig. 2H (right side) indicates relatively strong tissue birefringence. Since faster color change represents higher birefringence, we can observe that the image measured from the early cancer stage renders less birefringence than the normal tissue. The image measured from the late stage cancer case shows minimal birefringence.

Fig. 2.

PS-OCT images of the laryngeal tissue with different health stages. A PS-OCT en-face result of the normal (healthy) tissue specimen; B and C PS-OCT en-face result of the tumor tissue specimen, in early and late stage, respectively; D-F intensity OCT images; and G-I PS-OCT images. In B and C, the red circles indicate the tumor area. En-face images matched OCT images with 1:1 ratio. In G-I, the red arrows indicate the positions for measuring the phase retardation A-line shown in Fig. 3

Figure 3 shows the corresponding phase retardation A-lines for the three tissues. The A-lines are averaged over a window of surrounding 5 B-scans and 10 A-lines, followed by an 8-pixel medium filter along the Z-axis (the depth direction from the tissue surface to deep region). The multi-segment linear fitting of A-lines helps to differentiate the layers. The phase retardation A-lines and the corresponding fitted lines shown in Fig. 3 indicate that the healthy specimen shows fast accumulation in phase retardation, the early stage tumor specimen shows a medium slope accumulation, and the late stage tumor specimen shows a relatively slow slope of phase accumulation. The three curves in Fig. 3A, in color blue, green, and cyan, respectively, represent the phase retardation A-line selected from the positions marked by red arrows, from left to right, in Fig. 2G; similarly, the curves in Fig. 3B and C correspond to the arrows in Fig. 2H and I, respectively.

Fig. 3.

Slope analysis results of the PS-OCT phase retardation. A-C indicate the slope analysis results from the three tissues shown in Fig. 2D–F, respectively. In each case, the blue, green, and cyan curve represent the phase retardation A-line from the positions as marked by the red arrows in Fig. 2G–I, from left to right. The red dashed lines are the slope-based analysis results of the corresponding phase retardation A-lines

From the multi-segment linear fitting, three segments can be differentiated, which correspond to the three layers of the tissue. Due to the localization difference, the selected positions of the phase retardation A-lines could affect the results, particularly in the early stage tumor where different grades of invasion can be seen in different areas. This was avoided by averaging the three selected A-lines.

The averaged analysis results, namely the averaged slope and layer thickness of each segment, are summarized in Tables 1 and 2, respectively. For 12 rabbits (4 normal, 4 early-stage, and 4 late-stage), a total of 12 specimens were used to obtain averaged slope rates from Tables 1 and 2. The thickness of the epithelium (Layer-1) increases from the healthy specimen to the early stage and late stage tumor. The normal tissue has relatively high slope value in the lamina propria layer (Layer-2), with the average at $8.56\times {10}^{-3}$ rad/µm. The late stage tumor has a small average slope value of $2.60\times {10}^{-3}$ rad/µm in the lamina propria layer, and the early stage tumor shows a medium average slope value of $4.53\times {10}^{-3}$ rad/µm. In Tables 1 and 2, from the healthy tissue to the late stage tumor, the epithelial thickness was doubled, while the phase retardation slope of the lamina propria layer was found to drop threefold. As the laryngeal tumor model progresses from normal to early, and late stage, the thickness of the epithelium layer monotonically increases, and the tissue birefringence in the lamina propria layer monotonically decreases.

Table 1.

Layer thickness values of the three tissue groups

Layer thickness (μm)	Healthy tissue (μm)	Early stage tumor tissue (μm)	Late stage tumor tissue (μm)
Layer-1	87.31 ± 16.57	136.35 ± 24.86	180.61 ± 9.03
Layer-2	452.11 ± 9.51	331.48 ± 18.30	502.35 ± 18.37
Layer-3	173.43 ± 58.45	255.96 ± 159.36	449.69 ± 127.24

Table 2.

Phase retardation slope values at different layers of the three tissue groups

Phase retardation slopes (${10}^{-3}$ rad/μm)	Healthy tissue (${10}^{-3}$ rad/μm)	Early stage tumor tissue (${10}^{-3}$ rad/μm)	Late stage tumor tissue (${10}^{-3}$ rad/μm)
Layer-1	1.27 ± 0.32	0.47 ± 0.31	0.46 ± 0.15
Layer-2	8.56 ± 0.21	4.53 ± 0.90	2.60 ± 0.44
Layer-3	2.67 ± 1.56	1.07 ± 0.36	0.20 ± 0.10

To validate our results, a histological study was conducted on the same specimens used for PS-OCT measurement. Figure 4 shows histological sections of the three tissues. Figure 4A shows the histology image of the normal tissue. Figure 4B shows several small clusters of tumor cells associated with necrosis as marked by the yellow-circled region. This phenomenon indicates early stage tumor. Figure 4C shows a large tumor area marked by a yellow circle. The diameter of the tumor area is approximately 2 mm, which indicates late stage of the tumor [17].

Fig. 4.

Histology images of the laryngeal tissue with different healthy stages. A Histology image of the healthy specimen; B histology image of the specimen of the early-stage tumor tissue; C histology image from the specimen of the late-stage tumor tissue. In B and C, the yellow circles represent several small clusters of tumor cells associated with necrosis and a large tumor area, respectively. Scale bar: 2 mm

Discussion

Most laryngeal cancers originate in the glottis, while squamous cell carcinomas involve the vocal mucosa. Several common habits including the use of tobacco and heavy chronic consumption of alcohol have been attributed as risk factors.

In this study, tissue birefringence in the different layers of vocal cord and different stages of tumor progression was studied using PS-OCT. The vocal cord tissue contains three layers: the epithelial layer, the lamina propria, and the vocalis muscle [10]. Due to lack of collagen in the epithelial layer, the phase retardation slope is relatively small. In the lamina propria or muscle layer, there are abundant collagen or muscle cells, which results in relatively high phase retardation slope. Tumor progression can significantly alter the collagen organization depending on different stages. In the early stage, the tumor is localized in the epithelium, which may become thickened due to cancer cell proliferation, while the birefringence in the deeper layers remains unaffected. A thickening of epithelium is observed on the right side of the phase retardation image in Fig. 2E. An increase of the thickness of epithelium layer is also observed, which requires further validation. In the middle stage, the tumor invades the lamina propria, rendering a reduced phase retardation slope, which is consistent with this study. In the late stage, tumor invades and alters the collagen organization over a varied depth range, thus reducing phase the retardation in both, the lamina propria and the muscle layers, as shown in Fig. 3C and in Table 2.

The trend noted in the three tissues of our model is as shown in Fig. 5. Figure 5C shows collapsed tissue structure, unlike in Fig. 5A, which is normal. The normal tissue in Fig. 5D and the late stage in Fig. 5F have a homogeneous slope, distinctly different from each other. However, when the tumor and normal tissues coexist, it is important to distinguish the clinical boundaries between the two. In Fig. 5E that shows an early stage with unaffected mucous membrane, the PS-OCT confirms the transitional tissue state as the slope of retardation changes. PS-OCT can accurately show the degree of transitions from normal to early stage and early to late stage. Additionally, PS-OCT can show changes in the small field of view (in mm), thus potentially beneficial in determining the tumor margins [18]. The previous PS-OCT studies were confined to surface changes; while in this study, we observed the possibility to note changes within the deeper tissues.

Fig. 5.

A-C PS-OCT images, D-F PS-OCT phase retardation analysis of normal, early, and late stage, respectively. A 1–3: normal; B 1: lesion, 2: transition, and 3: normal; C 1–3: lesion; D-F red arrows indicate the layer thickness

This study has its limitation. Most laryngeal cancers originate from the vocal mucosa, which penetrates the basement membrane [19]; however, when a cancer model is developed, the tumor cannot be implanted on the surface. We produced a cancer model by injecting the tumor into the submucosa, just below the mucosa. As the stage advances, the tumor enlarges and invades the mucosa, which is the reverse of the actual clinical process. In our tumor model, the changes of basement membrane and epithelial layer were observed during the tumor progression. Although it does not resemble the true tumor formation, we can indirectly infer the dysplastic changes based on the mucosal and submucosal alterations. The surface changes are observed in Fig. 2C, changes in the basement membrane being visible in the submucosa. We could observe changes in the tissues deeper than the Reinkes’ space and compare the tumor infiltrating patterns with the normal tissue. When normal and tumor transition tissues are simultaneously noted, as in Fig. 4B, it is important to distinguish the boundaries; the significance of PS-OCT for the same has been verified.

Although early detection of the laryngeal cancer is crucial, its detection is possible when mucosal changes occur, whether in early or advanced stage. The biopsy specimen shows mucosal changes, however, does not reveal the boundary status. OCT can show the varied submucosal invasion, mucosal thinning secondary to a tumor, and tissue destruction due to an enlarged tumor.

In conclusion, a rabbit laryngeal tumor model, with different stages of tumor progression, was investigated ex-vivo by PS-OCT. Different collagen organizations in the different tissue layers of vocal cord causes varied tissue birefringence, which can be evaluated by the PS-OCT phase retardation measurement. Thus, a phase retardation slope-based analysis can be conducted to distinguish these layers. The thickness variation of the different segments may reflect the tumor progression, while the slopes from the different segments represent the birefringence in different layers, which helps in a more detailed comparison among the different stages of the rabbit laryngeal tumor model. This analysis demonstrates that during the progression from healthy to late stage tumor, the epithelium doubles in thickness and the slope of phase retardation of the lamina propria drops threefold, which cannot be observed by intensity OCT images. The slope analysis of the three-layer laryngeal model could render more details related to the tissue progression, which may offer a clinical guidance in distinguishing different grades of laryngeal cancer.

Compliance with ethical standards

Conflict of interest

The authors declare no conflict of interest.

Ethical statement

The animal studies were performed after receiving approval of the Institutional Animal Care and Use Committee (IACUS) in Kosin University College of Medicine (IACUC approval No. KMAP 15–07).