Open Access
Add to BibSonomy
Abstract
Current imaging tools are insufficiently sensitive to the early diagnosis of esophageal squamous cell carcinoma (ESCC). The application of polarization-sensitive optical coherence tomography (PS-OCT) to detect tumor-stroma interaction is an interesting issue in cancer diagnosis. In this translational study, we found that en-face PS-OCT effectively characterizes protruding, flat, and depressive type ESCC regardless of animal or human specimens. In addition, the tumor contour and margin could also be drawn and determined on a broad en-face view. The determined tumor margin could be in the proximity of 2 mm to the actual tumor margin, which was proved directly using histology.
© 2022 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
Esophageal cancer is the seventh most common cancer globally, accounting for 572,000 new cases and 509,000 deaths in 2018 [1]. While the prevalence of adenocarcinoma generated from Barrett's epithelium increased in Western countries, esophageal squamous cell carcinomas (ESCC) continue to dominate in the rest of the world. The five-year survival rate of esophageal cancer is poor in advanced diseases, and early diagnosis by endoscopy is the primary key to improving prognosis [2]. With improved endoscopic surface imaging tools, like narrow-band imaging (NBI) and magnification imaging, the diagnostic accuracy of early-stage esophageal cancer has improved compared to traditional white light imaging (WLI). However, the diagnostic accuracy of these endoscopic surface imaging tools is subjective and depends on the doctor’s experience. In addition, the accuracy for differentiating early-stage cancer from inflammation, detecting flat-type tumors, and accurately delineating tumor margins is limited [3].
Optical coherence tomography (OCT) is a back-scattered imaging technology similar to ultrasonography; however, it uses light instead of sound to achieve a higher resolution in the order of magnitude. It allows for obtaining label-free, radiation-free, real-time, qualitative, and quantitative cross-sectional tomographic structural images and three-dimensional reconstructive images with high resolution (10–20 µm) and adequate penetration depth (1.5–2 mm) in human tissue. This technique has been studied to diagnose early-stage cancer, such as oral cancer, esophageal cancer, breast cancer, and bladder cancer [4]. In the esophagus, two commercially available intensity-based OCT systems have shown promising results in the neoplastic detection of Barret’s epithelium [5]. However, the application of these systems in ESCC, the primary form of esophageal cancer worldwide, is rare because of a lack of diagnostic clues in early-stage ESCC, like irregular goblet glands in Barrett’s dysplasia. OCT angiography (OCTA) is an extended mathematical method to produce a mucosal microvascular image, as seen in magnifying endoscopy. This technique has been studied in diagnosing oral and esophageal cancer in animal and human models [6–8]. However, the sensitivity of motion, the requirement for high resolution, and the long acquisition time limit this technique in the extensive field of cancer screening.
Tumor-stromal interaction is a new issue under investigation. Extracellular matrix remodeling, such as loss of normal collagen orientation and the development of desmoplasia, has been observed in squamous cell carcinoma (SCC) [9]. The evolution of polarization-sensitive OCT (PS-OCT) using polarization interferometry with broadband light sources to obtain collagen fiber-enhanced images at the micrometer scale gives a chance to study cancer-associated collagen behavior in-vivo. The study of PS-OCT in delineating tumor tissue is a relatively unexplored but potentially important field based on the local destruction of collagen. Several papers have demonstrated birefringence reduction in tumor tissue in an animal model and human bladder, brain, and skin [10–12]. However, there was still a lack of study to characterize the tumor contour by PS-OCT on a broad en-face view. Our previous animal study introduced the en-face projection of birefringence changes to characterize the oral SCC. The results showed that en-face projection of birefringence changes by slope-fitting method had high sensitivity (100%) and specificity (95%) in the delineation of small (<5 mm) early-stage oral SCC in vivo [13]. However, producing an enface map by a slope-fitting method is time-consuming, highly interfered with by the selected depth, and unsuitable for clinical use. Therefore, in this study, we propose a new standard deviation (SD) en-face projection method and extend the study's aim to characterize the ESCC because of the same carcinogens and field cancerization with oral SCC [14]. The study was initiated from an animal model and was translated into human cancerous specimens. The reality of tumor contour on the en-face projection map was also validated directly by histology.
2. Materials and methods
2.1 Setup and imaging process of the swept-source, polarization-sensitive optical coherence tomography imaging system
The detailed platform setup of the Swept-Source, Polarization-Sensitive Optical Coherence Tomography system was described in our previous paper [13]. The light source was a swept-source laser (AXSUN Technologies Inc., Billerica, MA, USA) with a central wavelength of 1310 nm, spectral bandwidth of 100 nm, and a sweeping speed of 100 kHz. The SS PS-OCT system was constructed with all single mole fibers, and the incident power focused on a scanning spot was ∼7 mW. Through our previous proposed calibration scheme [15], bulk quarter-wave plates can be replaced by fiber optics polarization controllers to generate a circularly polarized light in the sample arm and balance the power of the reflected reference light. This design ensures that the two-dimensional structural and retardation images can be calculated using the same formula and with an accuracy comparable to traditional bulk optics-type, Jones matrix-based CS PS-OCT [16]. The maximum image acquisition rate was 100 frames per second (1000 × 1024 pixels in the x–z plane). Using a mirror with an attenuator as a sample, we measured the axial resolution as ∼15 µm. A 1951 USAF resolution test target (Edmund Optics, Barrington, New Jersey, USA) was used as the standard resolution test sample to measure the lateral resolution, confirming that the lateral resolution of the OCT setup was approximately 7 µm.
A schematic of the imaging process is shown in Fig. 1. Data were post-processed to get intensity and retardation images using MATLAB (MathWorks, Natick, Massachusetts, USA). Images were presented as two-dimensional (2D) cross-sectional (x–z) planes and an en-face (x–y) plane. After acquiring the interference signal, the shape of the spectrum was modified with dispersion compensation to achieve better resolution [17]. The A-scan signal obtained by inverse Fourier transform of the acquired fringe data can be expressed as
Fig. 1. Scheme of the image processing routine. The cross-sectional imaging is obtained from interference signal after dispersion compensation and fast Fourier transform process. The en-face imaging is obtained by projecting the standard deviation (SD) value calculated from all A-lines within the setting depth (from the surface to 30 pixels depth, about 176 µm).
A step of en-face projection was as follows. First, a gray-level threshold algorithm binarized the 2D intensity structural image and automatically determined the surface's upper border. After determining the surface, the fixed depth (30 pixels, about 176 µm) was set to include epithelial and lamina propria layers. These positions were recorded to correspond to the 2D retardation image. Finally, a standard deviation (SD) of intensity and phase retardation within the setting depth was calculated from all A-lines and was projected to produce en-face intensity and retardation map.
2.2 Mouse model
K14-EGFP-miR-211-GFP transgenic mice, generated by Chang et al., were used in this study. The advantage of this model is the higher induction rate of oral and esophageal SCC. The diminutive dysplastic or cancerous lesions could also be identified easily using fluorescence imaging because of the increased expression of K14-EGFP-miR-211-GFP in the tumor area [18]. The induction protocol included adding 100 µg/mL of the water-soluble 4-Nitroquinoline-1-oxide (4NQO) (Sigma-Aldrich, St. Louis, MO, United States), a carcinogen that produces an effect similar to that of tobacco [19], into the drinking water of six-week-old, K14-EGFP-miR-211-GFP transgenic mice for 12 weeks. The mice were sacrificed 24 weeks after the study started to obtain esophageal specimens. The experimental steps were as follows: The esophagus was cut carefully along the long axis, and the mucosal surface was opened on a corkboard. The suspicious tumor area was found and labeled with ink on bilateral sides with the aid of fluorescence imaging. The PS-OCT scanning (4 × 4 mm in XY plane) was performed between two ink labeling steps. After scanning, the specimens were fixed in 10% neutral buffered formalin and sent to the histological department. The histological sections were done along the two sides labeled with ink. Using this method, we can compare PS-OCT imaging and histology results precisely at the same location. All experiments were approved by the Institution for Animal Care and Use Committee at National Yang-Ming University (ethical code: #1070216).
2.3 Human specimens
The human specimens were collected from the patients who underwent endoscopic submucosal dissection (ESD) due to the diagnosis of high-grade esophageal dysplasia or ESCC without submucosal invasion. The informed consent was obtained before ESD, and all procedures were performed by an endoscopist, doctor Chen PH. After the ESD procedure, the esophageal specimen was flattened and fixed on the corkboard, and PS-OCT scanning was performed on the whole specimen immediately. We collected PS-OCT images in the first three cases and compared them to those obtained in WLI and NBI using commercial endoscopes (GIF-H290, Olympus). In the following two cases, we performed the image processing immediately after scanning and used an en-face retardation map to delineate the tumor margin. Two biopsies (Radial JawTM 4 forceps, pediatric, Boston Scientific) 2 mm outside of the delineated margin and two biopsies 2 mm inside of the delineated margin were performed in each specimen. The accuracy of determining tumor margin by en-face retardation map was validated using histology findings as a reference. The study was approved by the Institutional Review Board of Taipei Veterans General Hospital. (2019-08-005BC)
3. Results
3.1 Mouse model: en-face Intensity and retardation maps of the normal esophagus and esophageal papillary carcinoma in situ
Figure 2 A-D show an example of OCT images of the normal esophagus in a control mouse. There was no tumor on WLI (Fig. 2(B)) and fluorescence image (Fig. 2(A)). The en-face intensity map (Fig. 2(C)) from the square of Fig. 2(B) showed multiple bright linear lines due to the hyper-scattering effect of the longitudinal esophageal folds. In contrast, the en-face retardation map (Fig. 2(D)) was less influenced by these uneven folds. The map showed homogeneous background in red color, which meant a high SD value of phase retardation from epithelium to subepithelial stroma in normal mucosa.
Fig. 2. Mouse model: OCT images in the normal esophagus (A-D) and esophageal papillary carcinoma in situ (E-K). (A) Fluorescence image of the normal mucosa. No green spot is found. (B) The white light image (WLI) of normal mucosa. Two random areas are labeled with ink. (C) En-face intensity map from the square in (B): multiple hyper-scattering lines are seen due to longitudinal folds on the mucosa. (D) Corresponding en-face retardation SD map shows a high SD value of retardation (red color) on normal mucosa with less hyper-scattering interference. (E) Fluorescence image shows a green spot indicating the tumor. (F) WLI shows a protruding tumor (black arrow). (G) En-face intensity map from the square in (F) shows heterogeneous scattering on the tumor. (H) Corresponding en-face retardation SD map shows a low SD value of retardation (blue color) on the tumor. (I) Histology from the dotted line of (H) shows dysplasia involving the full thickness of the squamous epithelium with some invaginated lamina propria into epithelium (white arrowheads), which is compatible with carcinoma in situ (CIS). (J) Cross-sectional image of retardation from the dotted line of (H). Thick epithelium with low retardation in the tumor area. Some high retardant projection in the epithelium (red circle) is compatible with collagen-abundant invaginated lamina propria in histology. Other high retardation signals below the red circles are from the cork sieve for fixed sample use. (K) Cross-sectional image of intensity OCT from the dotted line of (H). See Visualization 1 and Visualization 2 for 3D data of Fig. (K) and (J).
Figures 2 E-K show a case of esophageal papillary carcinoma in situ (CIS). A protruding tumor (black arrow in Fig. 2(F)) with a bright, green spot on the fluorescence image (Fig. 2(E)) was labeled with ink. The en-face intensity map (Fig. 2(G), from the square of Fig. 2(F)) showed a tumor with heterogeneous scattering. On the corresponding en-face retardation map (Fig. 2(H)), a tumor showed low SD of retardation (blue color), which was compatible with studies of birefringence reduction in cancer mass [10–12]. However, some area with high SD of retardation in the tumor (red spots) was also found, which presented the heterogenicity of a tumor. From the histology (Fig. 2(I), corresponding to the dotted line in Fig. 2(H)), an increased thickness of the dysplastic epithelium with some invaginated lamina propria in the epithelial layer (white arrows) was noted. The proliferation of lamina propria with abundant collagen explained the heterogeneous retardant change within a tumor, where high retardant spikes in the epithelium were also seen in the cross-sectional retardation image (red circles in Fig. 2(J)). These results supported that SD en-face projection not only presented the low retardant properties of the tumor but also showed inhomogeneous characteristics in tumor development. The 3D images of phase retardation and intensity can be seen in Visualization 1 and Visualization 2.
3.2 Mouse model: en-face Intensity and retardation map of flat type dysplasia
Figure 3 shows an example of the flat-type dysplastic lesion in a mouse model, the most common type of early-stage ESCC in humans. On the fluorescence image (Fig. 3(A)), a faint green signal is found near the left esophageal edge, but no significant lesion could be seen under WLI (Fig. 3(B)). Heterogeneous scattering was noted on the en-face intensity map (Fig. 3(C), from the square of Fig. 3(B)), but it was challenging to identify the tumor area. On the corresponding en-face retardation map (Fig. 3(D)), scattered distribution of blue, low SD retardant area (white arrows, suspicious tumor area) was found in almost part of the scanning area and was close to the bilateral inks (asterisks in Fig. 3(D)). In histology images (Fig. 3(E), area corresponding to the dotted line in Fig. 3(D)), multiple skip dysplastic epitheliums (white arrows) intervened with normal epithelium distributed from the left to right side near the bilateral inks (asterisks). The distribution of tumor area was more compatible with the en-face retardation map than the en-face intensity map and fluorescence image. Figure 3(F) shows the magnified histological image of the dysplastic epithelium. Cross-sectional images of OCT intensity and retardation corresponding to the histology were shown in Fig. 3(G) and 3(H), respectively.
Fig. 3. Mouse model: Optical coherence tomography images of flat type esophageal dysplastic lesion. (A) Fluorescence image: a faint green signal is seen at the left edge of the esophagus. (B) WLI: no significant tumor can be seen. (C) En-face intensity map from the square in (B): heterogeneous scattering is noted, and it is challenging to identify the tumor. (D) The corresponding en-face retardation SD map presents the scattered distribution of blue, low retardant area (white arrows, indicated tumor area) from left to right edge. The tumor is close to the bilateral inks (asterisks). (E) Histology from the area corresponding to the dotted line in (D): multiple skip dysplastic epithelium (white arrows) intervened with normal epithelium is found from left to right side. The distribution of dysplasia is close to the bilateral inks (asterisks). The histology tumor distribution is compatible with it on the en-face retardation map. (F) Magnified histological image of dysplastic epithelium. (G) Cross-sectional image of intensity OCT corresponding to (F). (H) Cross-sectional image of retardation corresponding to (F).
3.3 Human specimen: en-face Intensity and retardation maps of early-stage squamous cell carcinoma
Figure 4 shows a case of esophageal CIS in humans after ESD. The en-face OCT images were compared to current standard endoscopic images of ESCC, the WLI (Fig. 4(A)), and NBI (Fig. 4(B)). In WLI, the specimen looks semi-transparent because this is a very early stage cancer limited in the epithelial layer. The resected tissue only includes thin epithelium, underlying lamina propria, muscularis mucosa, and part of the submucosal layer. The central hyperemic mucosa on WLI and brown mucosa on NBI indicate the cancer area. From the en-face intensity (Fig. 4(C)) and retardation SD map (Fig. 4(D)), the tumor contour was compatible with NBI and even more apparent than the WLI. The low SD retardation change with some heterogeneous high SD retardation areas in human ESCC (white arrowheads in Fig. 4(D)) was compatible with findings in the animal model (Fig. 2). However, there were also some blue color spots at the margin of the specimen (yellow arrows in Fig. 4(D)), which confused the main tumor area. These artificial interferences were due to the burning effect and thermal markers (red arrows in Fig. 4(A)) during the ESD procedure and will not be seen in the in-vivo study of the human esophagus in the future. The histology from the tumor center showed depressed mucosa with CIS (Fig. 4(E)). Cross-sectional images of OCT intensity and retardation from the dotted line in Fig. 4(D) were shown in Fig. 4(F) and 4(G), respectively. The 3D images of intensity and retardation can be seen in Visualization 3 and Visualization 4. The images of another human CIS case are shown in Fig. S1.
Fig. 4. Optical coherence tomography images of esophageal carcinoma in situ in humans. (A) hyperemic mucosa on the center of a specimen on the WLI. (B) A brown tumor (black arrowheads) is seen on a narrow band image (NBI). (C) A well-defined tumor with a low SD intensity value is seen (white arrowheads). (D) On the en-face retardation SD map, a well-defined tumor with a low SD value of retardation is also seen (white arrowheads). (E) The histology from the center of the tumor reveals depressed mucosa with CIS. (F) Cross-sectional image of intensity OCT from the dotted line in (D). (G) Cross-sectional image of retardation from the dotted line in (D). See Visualization 3 and Visualization 4 for 3D data of Fig. (F) and (G).
3.4 Human specimen: comparing tumor contour between en-face retardation map and histological reconstruction
In Fig. 5, we directly compared the tumor contour on the en-face retardation map with histological reconstruction. On NBI (Fig. 5(A)), only a slightly brown color could be seen at the specimen center, and the tumor margin was unclear. On the en-face retardation map (Fig. 5(C)), the blue tumor with a clear margin could be circled in a black line. After OCT scanning, multiple histological sections were performed on a 2-3 mm interval (red dotted lines I-VI in Fig. 5(C)) and were arranged orderly to compare the tumor distribution with the en-face retardation map. The tumor area in histology slides was circled by the pathologist Dr. YC Yeh (Fig. 5(D), yellow squares). Comparing Fig. 5(C) and 5(D), the distribution of the tumor area was matched well. In particular, in dotted red line V, the skip distribution of the tumor could be seen consistently in two images. The magnified histological image of the primary tumor shows CIS (Fig. 5(B)).
Fig. 5. Comparing tumor contour between en-face retardation map and histological reconstruction. (A) NBI of the post-endoscopic submucosal dissection (ESD) esophageal specimen. The tumor contour is unclear. (B) Histology shows CIS. (C) En-face retardation SD map shows a tumor contour (circled by a black line). According to the red dotted line I-VI, multiple histological sections with 2-3 mm intervals are performed. (D) Orderly arranged histological sections. Tumor areas are marked in yellow squares. The distribution of the tumor areas is consistent with the en-face retardation map and histological reconstruction.
3.5 Example of the success of determining tumor margin on en-face retardation map
The final step was to directly prove histological evidence's accuracy in determining tumor margin on the en-face retardation map. As in previous images, the tumor margin was more apparent on the en-face retardation map (circled by the black line in Fig. 6(B)) than on the NBI (Fig. 6(A)). After OCT scanning, we performed image processing immediately and printed the en-face retardation map. First, the printed map was put on the specimen to delineate the tumor margin. Then, we cut the map along the delineated margin and did the multiple biopsies (white arrow on Fig. 6(C)) 2 mm inside and outside the determined margin. Total two biopsies outer the margin (black asterisks in Fig. 6(B)) and two biopsies inner the margin (red asterisks in Fig. 6(B)) were performed on the specimen. Figure 6(D) shows the histological images of these biopsies. At the two sites marked with black asterisks, the histology showed normal epithelium without malignant change, while the histology revealed CIS at the two sites labeled with red asterisks. Figure 6(E) is another example of an en-face retardation map of esophageal CIS. Similarly, Biopsies on two sites located 2 mm outside the determined margin (black asterisks on Fig. 6(E)) and those on two sites located 2 mm inside the determined margin (red asterisks on Fig. 6(E)) were done. The histological results in Fig. 6(F) showed no malignancy on the sites marked with two black asterisks, while CIS was determined on those labeled with two red asterisks.
Fig. 6. Accuracy of en-face retardation map to predict tumor margin. (A) NBI of post-ESD esophageal CIS. The tumor contour is unclear. (B) En-face retardation SD map of the specimen. A black line circles the suspicious tumor margin. Two biopsies are performed outer the margin (black asterisks) and two inner the margin (red asterisks). (C) Biopsy protocol. The en-face retardation map is cut along the delineated tumor margin and put on the specimen. The biopsies (white arrow, Radial JawTM 4 forceps, pediatric, Boston Scientific) are performed 2 mm inside and outside the tumor margin. (D) Histological results of biopsies. Two biopsies from sides labeled with black asterisks show normal epithelium without residual malignancy. Two biopsies from sides labeled with red asterisks reveal CIS. (E) En-face retardation SD map of another specimen. The same, two biopsies 2 mm outside the margin (black asterisks) and two biopsies 2 mm inside the defined margin (red asterisks) are performed. (F) Histological results of biopsies. Two biopsies from black asterisk sites show no residual malignancy. Two biopsies from red asterisks sites reveal CIS.
4. Discussion
The critical result of this study is to find out a new imaging model for characterizing early-stage ESCC. Our previous experiments validate using an enface birefringence map to detect small and early-stage oral malignancy with high accuracy and consistency [13]. However, this slope-fitting method is time-consuming, highly interfered with by the selected depth range, and not suitable for clinical use, especially in human specimens where there is a significant need for efficient evaluation of the tumor margin during surgery. On the contrary, by directly calculating the SD/maximum/sum value through the averaged data of nearby A-lines, projection methods can construct an enface map quickly, saving at least ten times for calculation. Fig. S2 demonstrates the slope-fitting method and several projection methods (SD/maximum/sum projection) from an esophageal CIS in a human who underwent endoscopic submucosal dissection surgery (i.e., the case of Fig. 4). The tumor contour is relatively consistent between the slope-fitting method (Fig. S2A), SD calculation (Fig. S2B) and NBI image (Fig. 4(B)), compared to maximum projection (Fig. S2C) or sum projection (Fig. S2D). The computational time to produce this en-face map (tissue size: 18 × 16mm; Personal computer with macOS Big Sur system, Intel Core i7 Central processing unit (CPU) processor with a base processor frequency of 2.5GHz; Algorithms in MATLAB (The MathWorks Inc.)) is 128 minutes by slope-fitting method and 11 minutes by SD calculation. In addition to the speed, SD calculation results in a tumor contour with less interference from the depth range than slope-fitting. (Fig. S3). Besides, the SD projection method can not only reflect the low retardation accumulation within the tumor area (e.g., a flat A-line and thus low SD value at point i of Fig.S2), but may also show inhomogeneous characteristics in the tumor development (e.g., a low phase accumulation but large local variations lead to a higher SD at point ii compared to point i in Fig. S2). This inhomogeneous phenomenon may be explained by the inhomogeneous proliferation of lamina propria during cancer development, like in Fig. 2(I). However, the real pathological meaning of these inhomogeneous SD changes still needs further investigation.
Although endoscopy is a popular method to screen and diagnose ESCC, endoscopic operative skills and image judgment need a long tanning period to improve diagnostic accuracy and consistency. Therefore, research in developing a diagnostic system with objective diagnostic parameters and artificial intelligence aid is a major issue in the future. OCT is an interesting one because of its characteristics of high resolution and real-time imaging with quantitative properties. However, before the development of quantitative diagnostic parameters, finding out the useful qualitative parameters of OCT in cancer diagnosis was the first step and is still under investigation. Detection of subtle morphological change between cancer and normal tissue by high-resolution intensity OCT image is a commonly used parameter. However, this method is still not accurate enough and is often confused between early-stage cancer and benign hyperplasia and inflammation. Tumor-stroma interaction is a new era of cancer study. Evidence has shown that stromal changes can appear as early as in dysplasia and early-stage cancers. [20–22]. The development of PS-OCT to detect collagen desmoplasia, the major component of the stroma, got initial success in diagnosing breast, bladder, brain, and skin cancers [10–12]. Further studies even found the ability of PS-OCT to delineate cancer components from surrounding benign inflammatory and fibrotic tissue, which is a major limitation of most imaging tools, like endoscopy [23,24]. Although these advances, the application of PS-OCT in digestive tract cancer is still lacking. Thus, we design serial animal and human studies to investigate the feasibility of PS-OCT in digestive SCC, including the oral cavity and esophagus. In the first pilot animal study, we found that en-face PS-OCT can detect small (<5 mm) and early-stage oral SCC in-vivo with high sensitivity of 100% and specificity of 95% [13]. In this esophageal study, we also found the consistent imaging result with the previous study that low retardation change in depth can be detected as early as in dysplasia and early-stage ESCC. Unlike papillary tumors in oral SCC, early-stage ESCC is often flat and depressed, and it is difficult to be detected by current imaging tools. This study found that en-face PS-OCT was as effective as detecting protruding, flat, and depressed type ESCC. The PS-OCT was less affected by hyper-scattering interference from the uneven mucosal fold than the intensity-OCT, as shown in Fig. 2. This advantage was significant in detecting flat-type tumors, like in Fig. 3. This study also successfully translated the animal results into human specimens. In Fig. 4 and supplementary Fig. 1, the contour of early-stage ESCC on the en-face retardation map is consistent with NBI imaging.
The second important finding of this study is to validate the ability of en-face retardation SD imaging in determining tumor margin by histological evidence. Although the Lugol or NBI stain is frequently used, there is still little evidence regarding how to accurately determine the tumor margin in early-stage ESCC [25]. The accuracy of these methods is especially doubted in inflammatory mucosa and tumor with lateral submucosal extension [26]. For example, the indefinite tumor margin is observed in Figs. 5–6 even under a new generation of Olympus 290 NBI system. Thus, the resection margin should be enlarged during the ESD procedure to avoid an unfree margin, which may increase the risk of post-ESD esophageal stricture. The same problem is seen in oral SCC. The studies showed that up to 30-80% of surgical margin was inadequate under current imaging approaches [27,28]. In our previous studies, we found that the tumor diameter measured on the en-face birefringence map was closed to that measured on histology [13]. In this study, we further validated this finding in ESCC. In Fig. 5, the skip tumor could be seen consistently between en-face retardation SD map and histological reconstruction. In Fig. 6, the defined tumor margin on en-face retardation map was also identical to histology, which was directly validated from eight biopsy specimens. These findings proved the potential of en-face PS-OCT to determine tumor contour, margin and guide the treatment in the future.
From the serial works of our group, we prove the feasibility of PS-OCT in characterizing and delineating oral and esophageal SCC. The SD projection of retardation provides an en-face image with a fast time and pleasing tumor contour, which is more likely a practical and feasible approach for universal and wide-applicant clinic purposes. The next step is to develop the probe-type scanner for clinical use. Integrating probe-type scanners in the endoscopic system can provide real-time scans of the esophagus like an ultrasound probe with higher structural resolution and polarization information. There are still several limitations in our study. First, this is a descriptive study only. More cases, including normal human specimens, will be needed to quantify the retardation change and determine the diagnostic cut-off value. However, this goal cannot be achieved now because there is still a lack of probe-type PS-OCT system for in-vivo study and the ex-vivo collection of normal human esophageal specimens is not feasible due to ethical issues. Second, the ideal analytic method of PS-OCT images and the pathological meaning of retardation changes need more investigation. Other quantities [29–32] may also be valuable and need to be explored in further studies, such as the optical irregularity index (OII), the coefficient of variation (CV), the phase accumulation change (PAC), and phase homogeneity index (PHI), the short-range slope fitting and energy/entropy, the degree of polarization uniformity (DOPU), the SD of the local deviation of retardation, SD/total phase accumulation, etc. However, for clinical use, especially in human specimens, there is a significant need for efficient evaluation of the tumor margin during surgery. Direct projection methods, such as directly calculating the standard deviation (SD)/maximum/sum value through the averaged data of nearby A-lines, can quickly construct an enface map. This work used the retardation SD projection method in this human study for its consistent contrast performance and efficiency compared to the conventional retardation slope-fitting method. Third, the design of this study is ex-vivo. However, the retardation changes of ESCC in this study are consistent with our previous in-vivo study of oral cancer [13]. Therefore, we think it is reasonable to translate this result into an in-vivo study. Fourth, some artificial blue colors at the margin of the human ESD specimen interfere with the tumor margin. This blue color is caused by the thermal damage of the stroma during the ESD procedure, but this interference will not appear in the in-vivo diagnosis of ESCC.
5. Conclusion
This translational study validates the ability of en-face PS-OCT to characterize early-stage ESCC in mouse models and human specimens. The en-face retardation SD map can image tumors with good contrast and contour regardless of protruding, flat, or depressed type cancer. In addition, the en-face retardation SD map can further determine tumor margin precisely with direct evidence from histology.
Funding
National Health Research Institutes (NHRI-EX111-11018EI).
Acknowledgments
We thank Yi-Fen Chen for inducting the animal models and the team of Ching-Liang Lu of the endoscopic center in Taipei Veterans General Hospital to support the place for human study.
Disclosures
The authors declare no conflicts of interest.
Data availability
The data supporting this study's findings are available from the corresponding author upon reasonable request.
Supplemental document
See Supplement 1 for supporting content.
References
1. F. Bray, J. Ferlay, I. Soerjomataram, R. L. Siegel, L. A. Torre, and A. Jemal, “Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries,” CA Cancer J. Clin. 68(6), 394–424 (2018). [CrossRef]
2. A. Shin, Y. J. Won, H. K. Jung, H. J. Kong, K. W. Jung, C. M. Oh, S. Choe, and J. Lee, “Trends in incidence and survival of esophageal cancer in Korea: Analysis of the Korea Central Cancer Registry Database,” J. Gastroenterol. Hepatol. 33(12), 1961–1968 (2018). [CrossRef]
3. H. Mashimo, S. R. Gordon, and S. K. Singh, “Advanced endoscopic imaging for detecting and guiding therapy of early neoplasias of the esophagus,” Ann. N. Y. Acad. Sci. 1482(1), 61–76 (2020). [CrossRef]
4. J. Wang, Y. Xu, and S. A. Boppart, “Review of optical coherence tomography in oncology,” J. Biomed. Opt. 22(12), 1–23 (2017). [CrossRef]
5. M. S. Smith, B. Cash, V. Konda, A. J. Trindade, S. Gordon, S. DeMeester, V. Joshi, D. Diehl, E. Ganguly, H. Mashimo, S. Singh, B. Jobe, M. McKinley, M. Wallace, Y. Komatsu, S. Thakkar, F. Schnoll-Sussman, R. Sharaiha, M. Kahaleh, P. Tarnasky, H. Wolfsen, R. Hawes, J. Lipham, H. Khara, D. Pleskow, U. Navaneethan, P. Kedia, M. Hasan, A. Sethi, J. Samarasena, U. D. Siddiqui, F. Gress, R. Rodriguez, C. Lee, T. Gonda, I. Waxman, S. Hyder, J. Poneros, K. Sharzehi, J. A. Di Palma, D. V. Sejpal, D. Oh, J. Hagen, R. Rothstein, M. Sawhney, T. Berzin, Z. Malik, and K. Chang, “Volumetric laser endomicroscopy and its application to Barrett's esophagus: results from a 1,000 patient registry,” Dis. Esophagus 32(3), doz029 (2019). [CrossRef]
6. H. C. Lee, O. O. Ahsen, K. Liang, Z. Wang, M. Figueiredo, M. G. Giacomelli, B. Potsaid, Q. Huang, H. Mashimo, and J. G. Fujimoto, “Endoscopic optical coherence tomography angiography microvascular features associated with dysplasia in Barrett's esophagus (with video),” Gastrointestinal Endoscopy 86(3), 476–484.e3 (2017). [CrossRef]
7. P. H. Chen, C. H. Wu, Y. F. Chen, Y. C. Yeh, B. H. Lin, K. W. Chang, P. Y. Lai, M. C. Hou, C. L. Lu, and W. C. Kuo, “Combination of structural and vascular optical coherence tomography for differentiating oral lesions of mice in different carcinogenesis stages,” Biomed. Opt. Express 9(4), 1461–1476 (2018). [CrossRef]
8. P. H. Chen, Y. J. Chen, Y. F. Chen, Y. C. Yeh, K. W. Chang, M. C. Hou, and W. C. Kuo, “Quantification of structural and microvascular changes for diagnosing early-stage oral cancer,” Biomed. Opt. Express 11(3), 1244–1256 (2020). [CrossRef]
9. Y. Z. Lim and A. P. South, “Tumour-stroma crosstalk in the development of squamous cell carcinoma,” Int. J. Biochem. Cell Biol. 53, 450–458 (2014). [CrossRef]
10. Y. Q. Xiong, J. Tan, Y. M. Liu, Y. Z. Li, F. F. You, M. Y. Zhang, Q. Chen, K. Zou, and X. Sun, “Diagnostic accuracy of optical coherence tomography for bladder cancer: A systematic review and meta-analysis,” Photodiagn. Photodyn. Ther. 27, 298–304 (2019). [CrossRef]
11. J. Wang, Y. Xu, K. J. Mesa, F. A. South, E. J. Chaney, D. R. Spillman Jr., R. Barkalifa, M. Marjanovic, P. S. Carney, A. M. Higham, Z. G. Liu, and S. A. Boppart, “Complementary use of polarization-sensitive and standard OCT metrics for enhanced intraoperative differentiation of breast cancer,” Biomed. Opt. Express 9(12), 6519–6528 (2018). [CrossRef]
12. T. Marvdashti, L. Duan, S. Z. Aasi, J. Y. Tang, and A. K. Ellerbee Bowden, “Classification of basal cell carcinoma in human skin using machine learning and quantitative features captured by polarization sensitive optical coherence tomography,” Biomed. Opt. Express 7(9), 3721–3735 (2016). [CrossRef]
13. P. H. Chen, H. Y. Lee, Y. F. Chen, Y. C. Yeh, K. W. Chang, M. C. Hou, and W. C. Kuo, “Detection of Oral Dysplastic and Early Cancerous Lesions by Polarization-Sensitive Optical Coherence Tomography,” Cancers 12(9), 2376 (2020). [CrossRef]
14. D. P. Slaughter, H. W. Southwick, and W. Smejkal, “Field cancerization in oral stratified squamous epithelium; clinical implications of multicentric origin,” Cancer 6(5), 963–968 (1953). [CrossRef]
15. H. Lin, M. C. Kao, C. M. Lai, J. C. Huang, and W. C. Kuo, “All fiber optics circular-state swept source polarization-sensitive optical coherence tomography,” J. Biomed. Opt 19(2), 021110 (2013). [CrossRef]
16. C. Hitzenberger, E. Goetzinger, M. Sticker, M. Pircher, and A. Fercher, “Measurement and imaging of birefringence and optic axis orientation by phase resolved polarization sensitive optical coherence tomography,” Opt. Express 9(13), 780–790 (2001). [CrossRef]
17. M. Wojtkowski, V. J. Srinivasan, T. H. Ko, J. G. Fujimoto, A. Kowalczyk, and J. S. Duker, “Ultrahigh-resolution, high-speed, Fourier domain optical coherence tomography and methods for dispersion compensation,” Opt. Express 12(11), 2404–2422 (2004). [CrossRef]
18. Y. F. Chen, C. C. Yang, S. Y. Kao, C. J. Liu, S. C. Lin, and K. W. Chang, “MicroRNA-211 Enhances the Oncogenicity of Carcinogen-Induced Oral Carcinoma by Repressing TCF12 and Increasing Antioxidant Activity,” Cancer Res. 76(16), 4872–4886 (2016). [CrossRef]
19. X. H. Tang, B. Knudsen, D. Bemis, S. Tickoo, and L. J. Gudas, “Oral cavity and esophageal carcinogenesis modeled in carcinogen-treated mice,” Clin. Cancer Res. 10(1), 301–313 (2004). [CrossRef]
20. C. S. Farah and S. A. Fox, “Dysplastic oral leukoplakia is molecularly distinct from leukoplakia without dysplasia,” J. Oral Dis. 25(7), 1715–1723 (2019). [CrossRef]
21. K. Bhattacharjee, H. C. Girish, S. Murgod, A. M. J. Alshame, K. Shyamala, and V. N. Nayak, “A Comparative Immunohistochemical Study of Presence and Distribution Pattern of Stromal Myofibroblast in Oral Dysplasia and in Different Grades of Oral Squamous Cell Carcinoma,” J. Int. Soc. Prev. Community Dent. 8(5), 451–456 (2018). [CrossRef]
22. R. Sharma, S. Rehani, M. Mehendiratta, P. Kardam, M. Kumra, Y. Mathias, J. Yadav, and K. Sahay, “Architectural Analysis of Picrosirius Red Stained Collagen in Oral Epithelial Dysplasia and Oral Squamous Cell Carcinoma using Polarization Microscopy,” JCDR 9, EC13-16 (2015). [CrossRef]
23. L. P. Hariri, D. C. Adams, M. B. Applegate, A. J. Miller, B. W. Roop, M. Villiger, B. E. Bouma, and M. J. Suter, “Distinguishing Tumor from Associated Fibrosis to Increase Diagnostic Biopsy Yield with Polarization-Sensitive Optical Coherence Tomography,” Clin. Cancer Res. 25(17), 5242–5249 (2019). [CrossRef]
24. R. Patel, A. Khan, R. Quinlan, and A. N. Yaroslavsky, “Polarization-sensitive multimodal imaging for detecting breast cancer,” Cancer Res. 74(17), 4685–4693 (2014). [CrossRef]
25. W. F. Chen, Q. L. Li, L. Q. Yao, M. D. Xu, S. Y. Chen, Y. Q. Zhang, Y. S. Zhong, L. L. Ma, W. Z. Qin, J. W. Hu, M. Y. Cai, and P. H. Zhou, “[Narrow band imaging versus iodine staining for margin determination of early esophageal cancer during endoscopic submucosal dissection],” Zhonghua Wei Chang Wai Ke Za Zhi 16, 1138–1141 (2013).
26. N. Kakushima, H. Ono, M. Tanaka, K. Takizawa, Y. Yamaguchi, and H. Matsubayashi, “Factors related to lateral margin positivity for cancer in gastric specimens of endoscopic submucosal dissection,” J. Dig. Endosc. 23(3), 227–232 (2011). [CrossRef]
27. R. W. Smits, S. Koljenovic, J. A. Hardillo, I. Ten Hove, C. A. Meeuwis, A. Sewnaik, E. A. Dronkers, T. C. Bakker Schut, T. P. Langeveld, J. Molenaar, V. N. Hegt, G. J. Puppels, and R. J. Baatenburg de Jong, “Resection margins in oral cancer surgery: Room for improvement,” Head Neck 38(S1), E2197–E2203 (2016). [CrossRef]
28. M. A. Varvares, S. Poti, B. Kenyon, K. Christopher, and R. J. Walker, “Surgical margins and primary site resection in achieving local control in oral cancer resections,” Laryngoscope 125(10), 2298–2307 (2015). [CrossRef]
29. R. L. Golden, “Osler's legacy: the centennial of The Principles and Practice of Medicine,” Ann. Intern. Med. 116(3), 255–260 (1992). [CrossRef]
30. S. Nebelung, U. Marx, N. Brill, D. Arbab, V. Quack, H. Jahr, M. Tingart, B. Zhou, M. Stoffel, R. Schmitt, and B. Rath, “Morphometric grading of osteoarthritis by optical coherence tomography–an ex vivo study,” J. Orthop. Res. 32(10), 1381–1388 (2014). [CrossRef]
31. X. Zhou, F. Eltit, X. Yang, S. Maloufi, H. Alousaimi, Q. Liu, L. Huang, R. Wang, and S. Tang, “Detecting human articular cartilage degeneration in its early stage with polarization-sensitive optical coherence tomography,” Biomed. Opt. Express 11(5), 2745–2760 (2020). [CrossRef]
32. D. Zhu, J. Wang, M. Marjanovic, E. J. Chaney, K. A. Cradock, A. M. Higham, Z. G. Liu, Z. Gao, and S. A. Boppart, “Differentiation of breast tissue types for surgical margin assessment using machine learning and polarization-sensitive optical coherence tomography,” Biomed. Opt. Express 12(5), 3021–3036 (2021). [CrossRef]
