Polarization-sensitive optical coherence tomography for renal tumor detection in ex vivo human kidneys

Abstract

Kidney cancer is a kind of high mortality cancer because of the difficulty in early diagnosis and the high metastatic dissemination in treatments. The surgical resection of tumors is the most effective treatment for renal cancer patients. However, precise assessment of tumor margins is a challenge during surgical resection. The objective of this study is to demonstrate an optical imaging tool in precisely distinguishing kidney tumor borders and identifying tumor zones from normal tissues to assist surgeons in accurately resecting tumors from kidneys during the surgery. 30 samples from six human kidneys were imaged using polarization-sensitive optical coherence tomography (PS-OCT). Cross-sectional, enface, and spatial information of kidney samples were obtained for microenvironment reconstruction. Polarization parameters (phase retardation, optic axis direction, and degree of polarization uniformity (DOPU) and Stokes parameters (Q, U, and V) were utilized for multi-parameter analysis. To verify the detection accuracy of PS-OCT, H&E histology staining and dice-coefficient was utilized to quantify the performance of PS-OCT in identifying tumor borders and regions. In this study, tumor borders were clearly identified by PS-OCT imaging, which outperformed the conventional intensity-based OCT. With H&E histological staining as golden standard, PS-OCT precisely identified the tumor regions and tissue distributions at different locations and different depths based on polarization and Stokes parameters. Compared to the traditional attenuation coefficient quantification method, PS-OCT demonstrated enhanced contrast of tissue characteristics between normal and cancerous tissues due to the birefringence effects. Our results demonstrated that PS-OCT was promising to provide imaging guidance for the surgical resection of kidney tumors and had the potential to be used for other human kidney surgeries in clinics such as renal biopsy.

1. Introduction

The kidney cancer that arises in the renal parenchyma is mainly adenocarcinoma, also known as renal cell carcinomas (RCCs) [1]. RCCs comprise a heterogeneous group of cancers with different genetic and molecular alterations [2]. Almost a third of all patients with kidney tumors have metastatic dissemination at clinical diagnosis and nearly half of all patients die from the tumor [3]. There are about 79,000 cases of kidney cancer will be diagnosed and 13,920 cases will die in 2022 according to the most recent estimation from the American Cancer Society. Only 10% of kidney cancer patients present with the classic triad of symptoms: hematuria, flank pain, and palpable masses. Most cancer cases can be diagnosed or incidentally found on magnetic resonance imaging (MRI), computed tomography (CT) scan, or ultrasound [4, 5]. The systemic therapy plan for the kidney tumor such as surgically resection and ablation, percutaneously biopsied and immunohistochemically (IHC), depends on patient characteristics and extent of the cancer [4].

Surgical resection is a main treatment that is currently used for small renal masses with benign or malignant tumors in most patients [6, 7]. With the early detection and pathologic histology of the kidney tumor, the resection of localized tumorous masses is the most effective treatment for patients. To accurately resect the localized kidney tumor at an early stage and maximally decline the risk of distant metastases, the precise detection of tumorous regions and margins plays an important role in the surgery. Standard imaging tests including MRI, CT, and ultrasound are main screening modalities currently utilized for characterizing the mass size, possible abdominal metastases, tumor extension, and venous involvement for staging [2]. Although macroscopic imaging methods such as MRI and CT allow for the evaluation of kidney tumor extension at advanced stage, localizing small kidney tumors at early stage is still difficult because of relatively lower spatial resolutions [8-11]. MRI and CT have no advantage for precise surgical guidance to resect small kidney tumors under retaining patient’s normal kidney tissues. Although ultrasonography was able to provide real-time observation of ablated tissues [12], the main aim of ultrasound imaging was to focus on the diagnosis and treatment monitoring of big kidney tumors at advanced stage [13, 14]. The accurate differentiation and localization of small kidney tumors by ultrasonography were limited due to the low resolution (150 μm in high-resolution ultrasound imaging system) [15]. In summary, the primary challenges that surgeons are facing in renal tumor resection surgeries are: 1) finding an imaging tool for real-time guiding the resection of tumor tissues during the surgery; and 2) obtaining high-resolution images from the real-time imaging to segment tumor margins and profiles. Therefore, there is a critical need to develop a high-resolution imaging modality to provide real-time tumor region and margin distinguishments for renal tumor resection surgery.

Optical coherence tomography (OCT) uses a low-coherence interferometer to produce noninvasive two-dimensional and three-dimensional images with high resolution (~10 μm) for tissue microstructures in vivo and ex vivo [16]. Current OCT techniques have been widely reported in cancer/tumor diagnoses and anti-cancer drug screenings [17-21]. However, one of the limitations of conventional intensity-based OCT system is the lack of the tissue-specific contrast thus it is still often difficult to directly differentiate different tissues [22]. Boer and Nelson el al utilized a polarizer between the low-coherence laser and beam splitter in spectral domain OCT system to build a polarization sensitive OCT system (PS-OCT) that can achieve the differentiation of different internal tissue microstructures based on the optical phase delay and polarization state [23]. Compared to traditional intensity-based OCT modes for structural imaging, PS-OCT provides additional tissue specific contrasts to avoid the ambiguity with image interpretation within internal structures and achieves the quantitative information of different tissues [22]. The biologic tissues with different birefringence property cause unequal propagation speed of differently polarized light. Therefore, PS-OCT is applied to characterize samples by analyzing those changing light polarizations. Because traditional imaging guidance techniques cannot distinguish regions of tumor from admixed contaminant fibrotic stroma, PS-OCT is used to measure the correlation of collagen content in matched histological staining for lung cancer [24]. The study demonstrated that PS-OCT enabled accurate fibrosis detection and distinguished tumor regions with low fibrosis in human lung carcinoma ex vivo. Moreover, Strasswimmer and Duan et al indicated that PS-OCT was a potentially useful tool to examine human and mice skin cancer (basal cell carcinoma) based on the dermal birefringence between normal skin structure and cancerous tissues [25-27]. To improve the accuracy of brain tumor resection, PS-OCT has been demonstrated to precisely delineate the boundary between brain tumor and normal brain tissues due to the capability of PS-OCT for differentiating glioma from white matter [28]. Furthermore, South et al reported that PS-OCT was able to provide enhanced contrast between healthy and cancerous breast tissues indicating PS-OCT as a potential tool for intraoperative tumor margin evaluation [29]. These studies demonstrate that PS-OCT has significant potential for clinical impact in tumor recognition. Because of the ability to distinguish tumors by detecting the birefringence from fibrosis or collagen, PS-OCT can be applied to guide intraprocedural tissue sampling in vivo or achieve rapid biopsy adequacy assessment. Therefore, PS-OCT can serve as a potentially effective imaging tool to differentiate tumor margins and profile tumor regions via fibrosis detection to provide clinical surgical guidance for renal cancer resection.

The purpose of this study was to use PS-OCT to accurately differentiate the margin of kidney tumors and position multi-dimensional tumorous profiles and regions in renal surgical resection. Herein, we applied a label-free PS-OCT system to qualitatively and quantitatively distinguish and position tumor margins and profiles using several polarization parameters including polarization retardation, optic axis direction, and degree of uniformity to guide surgical resection of tumorous tissues in fresh ex-vivo human kidneys. In particular, we characterized tumor margins of cross-sectional and enface directions by utilizing 2D XZ (front to back vision)/YZ (left to right vision) and XY (top to bottom vision) structural parameter images. We showed that PS-OCT provided accurate and robust differentiation between normal and cancerous kidney tissues. Moreover, we demonstrated the ability of 3D PS-OCT to provide the spatial information of tumor structure within human kidneys based on multi-parameter structural images in different depths and surfaces, offering the availability to position and track the spatial structure of tumor during the kidney resection surgery. Our results demonstrated that PS-OCT is a promising tool to accurately detect and differentiate tumorous tissues from normal tissues within human kidneys. Overall, PS-OCT possesses wide prospects for clinical applications in kidney cancer resection surgeries.

2. Methods and Materials

2.1. Sample preparation and Ethics Committee Approval

This study was approved by the University of Oklahoma and the University of Oklahoma Health Sciences Center Institutional Review Board (IRB) (Study number: IRB #12462 and IRB #14794). All the experiment-used human kidney samples were preserved by hypothermic machine perfusion (HMP) for keeping the kidney sample's physiological status and imaged within 2 days after removing from the donors and patients. Six human kidneys with renal tumors and without other kidney diseases were used in the study. Five locations in each kidney subject were selected for PS-OCT imaging .

2.2. Histology staining

To verify the accessibility and accuracy of PS-OCT imaging for kidney tumor, the regions of human kidney tissues were excised and processed for histological staining after PS-OCT imaging to compare with corresponding PS-OCT results. The resected kidney tissues were fixed with 10% formalin, embedded in paraffin, then sectioned (4 μm thick) and stained with hematoxylin and eosin (H&E) for histological analysis. Sectioning and H&E staining was manipulated and finished by the Tissue Pathology Shared Resource, Stephenson Cancer Center (SCC), University of Oklahoma Health Sciences Center. Histological images were taken by Keyence Microscope BZ-X800 (BZ-X series, Itasca, IL, USA). The H&E staining dye (Hematoxylin cat#3801571 and Eosin cat# 3801616) was purchased from Leica biosystems (Deer Park, IL, USA) and the histology staining was performed using Leica ST5020 Automated Multistainer (Deer Park, IL, USA) following the H&E staining protocol at the SCC Tissue Pathology core.

2.3. PS-OCT schematic

Figure 1 shows a schematic of the integrated PS-OCT system used in imaging the ex vivo human kidneys. The broadband light with a center wavelength of 1300 nm and a spectral bandwidth of 100 nm generates a vertical linearly polarized light. The linearly polarized light is then coupled into the polarization maintaining fibers and is further split into reference and sample arms by a beam splitter (BMS). The linearly polarized light in the reference arm passes through a zero-order quarter-wave plate (QWB-1) with 22.5° orientation and exits with a 45° linear polarization after passing through QWB-1 twice. In the sample arm, the polarized light passes through the zero-order quarter-wave plate (QWB-2) oriented at 45° and is converted into circularly polarized light. The polarized light reflected and scattered by the sample in the sample arm becomes an elliptical polarization state after passing through QWB-2. The recombination of the polarized light in both arms of the system is split into the vertical linearly polarized signal and horizontal linearly polarized signal by two polarization-sensitive beam splitters (PBS), which are detected and processed by two polarization-sensitive channel sensors (CH-1, CH-2) [23, 30, 31]. The sensitivity of the system at 48 kHz A-scan rate was 105 dB, and the axial and lateral resolutions were 5.5 μm and 13 μm in air, respectively.

Figure 1.

System schematic of polarization-sensitive optical coherence tomography (PS-OCT) for kidney tumor imaging. Broadband Light Source, 1300 nm center wavelength linear-polarized light. CRL, circulator. CLM, fiber-to-free-space collimator. BMS, beam splitter. Iris, adjustable iris. QWP-1, quarter-wave plate (22.5° orientation). QWP-2, quarter-wave plate (45° orientation). PBS, polarization-sensitive beam splitter. CH-1, channel-1 sensor. CH-2, channel-2 sensor. Sample Arm of Interferometer, incident circular light – equal light amplitude in both orthogonal polarizations, backscattered and reflected elliptical light – encoded polarization and intensity information.

2.4. Measurement

Tumorous tissue and normal tissue from the kidneys were sectioned to be exposed for PS-OCT imaging. The sectioned tissues (cancerous and normal) were kept in the perfusate for keeping tissue physiological functions during the imaging. 2D PS-OCT scanning with a length of 9 mm was used to obtain cross-sectional structure images at 12 different angles (red line arrows in normal tissue (Figure 2A and 2C) and cancerous tissue (Figure 2B and 2D), 0 – 180° with a 15-degree step of clockwise) for labeling tumor margin. 3D scanning with a field of view (FOV) of 9 × 9 mm² (red frames in Figure 2A and 2B) was utilized to obtain the volumetric structure and enface profile of the samples. 1800 × 1800 × 1024 pixels were set on the length, width, and depth of the 3D PS-OCT imaging, as shown in Figure 2E-G. The sampling resolutions were 5 × 5 × 2.5 μm³ in XZ (cross-sectional), YZ (cross-sectional), and XY (enface) directions of a 9 × 9 × 2.6 mm³ volumetric rendering data.

Figure 2.

Scanning and measurement strategy of PS-OCT in 2D and 3D modes. A and C, normal kidney tissue. B and D, cancerous renal tissue. E, 2D YZ cross-sectional slice mode in 3D PS-OCT imaging. F, 2D XZ cross-sectional slice mode in 3D PS-OCT imaging. G, XY enface slice mode in 3D PS-OCT imaging.

2.5. Data and image processing

In this study, the multi-dimensional structural reconstruction of OCT intensity and PS-OCT information was completed by ImageJ Fiji (ImageJ 1.53q, Bethesda, MD, USA) and Amira (Amira 2021.1, Thermo Fisher Scientific, Agawam, MD, USA). The data statistics and analyses were performed using GraphPad (Prism 9.3.1, GraphPad, San Diego, CA, USA) and Python (Python 3.10.1, Python Software Foundation, Fredericksburg, VA) coding. To characterize and quantify the polarization information induced by light-tissue interactions, phase retardation, optic axis orientation, and degree of polarization uniformity (DOPU) based on Stokes vectors and Mueller matrices formalism [32] were used to distinguish tumor regions. Phase retardation and optic axis orientation were determined via the phase difference and the direction of the eigenvalue and eigenvector of the Jones matrix [22], respectively. The phase retardation was aliased into a 0 to π/2 rad range and the optic axis was ranged from −π/2 to π/2 in the angular direction. DOPU was a quantitative measurement of polarization properties of tissues via Stokes vector elements to indirectly quantify the polarization of light. The DOPU value equaled to 1 meaning the single speckle light was fully polarized whereas 0 meaning none polarized. Three Stokes parameters (Q, U, V describing the proportion of polarization state of light corresponding to linear polarization, circularly polarization, and elliptical polarization lights, respectively) were also used to provide the margin detection of tumors. To compare the detection effectiveness between conventional OCT intensity attenuation contrast and PS-OCT, an attenuation coefficient algorithm based on the Beer-Lambert law [33] was applied on OCT intensity images to detect tumor regions. The Dice’s coefficient was used to describe the degree of agreement between the histology and PS-OCT data. A score of 1 presented a complete agreement and a score of 0 presented no agreement. The paired student t-test was performed in the quantitative statistics. A P-value of < 0.05 was employed to present the statistical significance between the comparison.

3. Results

3.1. Cross-sectional PS-OCT structure and tumor margin detection in different angles

To identify PS-OCT being able to detect the border of tumors within human kidneys, PS-OCT was used to obtain cross-sectional images with different angles (0° - 180° with 15° intervals) to distinguish normal and cancerous kidney tissues via polarization and Stokes parameters. Here, we showed representative cross-sectional PS-OCT images of normal and cancerous kidney tissues on the scanning angle of 0°, as shown in Figure 3. In Figure 3A-3G, we observed that normal kidney tissues were highly uniform in OCT intensity and polarization images. In contrast, tumor tissues were clustered and heterogeneous in intensity and polarization structures, as shown in Figure 3H-3N. At the border of tumor tissue, polarization images showed obvious structure changes between normal and cancerous tissues (yellow and white window frames in Figure 3H-3N), providing the accurate detection of tumor margins. We further zoomed in on the structure of tumor borders marked via yellow and white frames in Figure 3H-3N. Although the OCT intensity image was able to approximately position the tumor region (Figure 3H), the border of tumor tissues was still unclear (Figure 3H₁ and 3H₂). Compared to the OCT intensity image, polarization images (retardation, optic axis, Stokes Q, and Stokes U) provided accurate detection of tumor borders with better contrast (yellow and white arrows in Figure 3I₁, 3J₁, 3L₁, 3M₁ and Figure 3I₂, 3J₂, 3L₂, 3M₂).

Figure 3.

Representative cross-sectional PS-OCT images of normal and cancerous kidney tissues in intensity and polarization modes. A-G, OCT intensity and polarization images of normal kidney tissues. H-N, OCT intensity and polarization images of cancerous kidney tissues. H₁-N₁, zoomed in sub-images of the left margin of tumor tissues in OCT intensity and polarization modes. H₂-N₂, zoomed in sub-images of the right margin of tumor tissues in OCT intensity and polarization modes. In retardation parameter images, 0.0 represented no phase retardation of light in tissues and 1.6 represented a π/2 phase retardation of light in tissues. Value 1.6 and −1.6 in optic axis images separately represented the optic axis of backward scattered signal from tissues locate on π/2 and −π/2 directions, value 0 represented the optic axis is on the center of the coordinate (0°). In DOPU parameter images, 1 indicated the single speckle light in tissues was fully polarized and 0 indicated complete depolarization. In different Stokes parameters, Q: Proportion of horizontally polarized light (Q=−1) or vertically polarized light (Q=+1); U: Proportion of light linearly polarized at 45° (U=−1) or linearly polarized at 135° (U=+1); V: Proportion of left-circularly polarized light (V=−1) or right-circularly polarized light (V=+1).

3.2. Enface PS-OCT structure and tumor profile detection in different depths

Since cross-sectional PS-OCT images were able to distinguish tumor borders and locate tumor regions, we employed 3D PS-OCT modes to obtain the volumetric structure of kidney tumors and sectioned enface images from different depths to show tumor profiles within human kidneys. In Figure 4, we showed enface images of PS-OCT in intensity and polarization states at 245^th, 305^th , 405^th , and 485^th slices for tumor tissues and 305^th slice for normal tissues. Similar to cross-sectional images, PS-OCT enface images showed that normal kidney tissues maintained highly uniform structures, as shown in Figure 4A-4G. PS-OCT enface images of kidney tumor tissues (Figure 4I-4N) in different depths indicated that tumor tissues were heterogeneous. The border and region of kidney tumors could be clearly differentiated based on the polarization changes. Particularly, PS-OCT images showed the specific distribution and profile of kidney tumor tissues. We marked the specific tumor tissue that showed significantly different polarization differences within kidney tumors (white arrows in retardation, blue arrows in optic axis, yellow arrows in DOPU, pink arrows in Stokes-Q, green arrows in Stokes-U, and red arrows in Stokes-V). These tumor tissues with different polarization states prominently distinguish tumor borders and locate tumor regions. Although kidney tumor regions were approximately located on intensity images (Figure 4H), the borders between normal and tumor tissues were still challenging to identify compared to that in PS-OCT images. The precise border detection from PS-OCT was used to improve the accuracy of kidney tumor resection and minimize the unnecessary removal of normal kidney tissues.

Figure 4.

Representative enface projection PS-OCT images of kidney normal and tumor tissues in different depths. A-G, enface PS-OCT images of kidney normal tissues at 305^th slice. H-N, enface PS-OCT images of kidney tumor tissues at 245^th, 305^th, 405^th, and 485^th slices.

3.3. PS-OCT structure and histology verification for tumor tissue distinguishment

PS-OCT provided the polarization information based on the birefringence of fibrosis tissues or collagens within human kidney tumors. To mark the tumor region out from the kidney tissue, we used the H&E histology as a gold standard to evaluate the accuracy of polarization states for extracting the polarization information of tumor tissues. Figure 5A showed the H&E histology of the tumor tissue, which clearly showed the tumor region and border from normal kidney tissues. The tumor profile offered by the H&E histology was used to verify the results obtained from PS-OCT data. We calculated the probability distribution of polarization and Stokes parameter values from normal and cancerous tissues in Figure 5H. The average between two adjacent peaks of probability distribution was set the threshold for distinguishing tumor areas from normal tissues (Figure 5H). The polarization thresholds were 0.60, −0.55 & 0.69, 0.38, −0.35 & 0.45, −0.35 & 0.45, and −0.35 & 0.45 in retardation, optic axis, DOPU, Stokes-Q, Stokes-U, and Stokes-V parameters, respectively, as shown in Figure 5B-G. These polarization thresholds effectively separated tumor regions from normal kidney tissues and achieved a good match with the tumor region provided by the histology. Moreover, we used the intrinsic optical attenuation coefficient to detect the kidney tumor region for comparing the accuracy of tumor detection by PS-OCT. As shown in Figure S1, the intensity profiles (Figure S1-C) along the axial scan lines in normal and tumorous OCT images (Figure S1-A and S1-B) showed significantly attenuated trends. The probability distribution of the attenuation coefficient (Figure S1-D) indicated that tumor tissues had two peaks and one peak overlaid with the peak of normal tissues. The mean value of the attenuation coefficient (2.04 mm⁻¹) between the peak probability in tumor (1.39 mm⁻¹) and normal (2.69 mm⁻¹) tissues was utilized to separate tumor tissues. Figure S1-E and S1-F showed the attenuation coefficient extracted images corresponding to Figure S1-A and S1-B. In Figure S1-H, we showed the enface structure of the tumor based on the attenuation coefficient. Compared to PS-OCT images, the attenuation coefficient could not match well with the histology (Figure S1-G).

Figure 5.

Enface histology and PS-OCT images of tumor region extraction in human kidney tumor samples. A, H&E histology of the kidney tumor sample. B-G, extracted enface polarization images of retardation, optic axis, DOPU, Stokes-Q, Stokes-U, and Stokes-V parameters based on the thresholds. H, Statistic histogram of the probability distribution of polarization and Stokes parameter values for normal and cancer tissues. Color-bars and values corresponded to the polarization state range of the corresponding polarization and Stokes parameter values. Dark frames on color-bars represented the threshold of the corresponding polarization parameter for distinguishing tumor and normal tissues. The corresponding polarization state ranges separated by the threshold were labeled the corresponding normal and tumor tissues. Sample size, N=6.

Furthermore, we applied the Dice’s coefficient to describe the degree of agreement between histology and PS-OCT data. Figure 6 showed the process of calculating the degree of agreement between histology and PS-OCT data and the comparison between PS-OCT and the traditional attenuation contrast. We extracted tumor profiles (Figure 6J-O) from the enface polarization images of retardation, optic axis, DOPU, Stokes-Q, Stokes-U, and Stokes-V parameters (Figure 5B-G and Figure 6B-G). Meanwhile, the tumor profile from the attenuation contrast image was also extracted (Figure 6H and 6P), and the tumor profile from the H&E staining was labelled by the pathologist (Figure 6A and 6I). Next, we overlaid the extracted H&E staining tumor profile and PS-OCT tumor profiles and the attenuation contrast tumor profile (Figure 6Q-W). The Dice coefficient was calculated among each of the overlaid images. As shown in Table 1, we found that PS-OCT data presented good agreements with the histology (Dice’s coefficient > 0.86), while the attenuation coefficient method showed a smaller Dice score, which confirmed that PS-OCT offered better detection for kidney tumors. With polarization images, we observed that tumor borders were particularly highlighted in the imaging of optic axis, Stokes-U, and Stokes-V parameters (white arrows), indicating that kidney tumor borders primarily induced more changes in optic axis , 45°/135° linearly polarized light, and the circularly polarized states. Moreover, tumor regions were mainly highlighted via retardation, Stokes-U, and Stokes-V, which presented that tumor tissues induced the primary change of phase delays, 45°/135° linearly polarized light, and the circularly polarized states. DOPU primarily quantified the tumor region but lacked the quantification of tumor borders. The followed separation and differentiation of tumor borders and regions were based on the same polarization and Stokes thresholds.

Figure 6.

Comparison of the degree of agreement between H&E staining and polarization-extracted PS-OCT images. A, H&E staining histology. B-G, polarization-extracted images of retardation, optic axis, DOPU, Stokes-Q, Stokes-U, and Stokes-V parameters. H, the intrinsic optical attenuation contrast. I, the labelled tumor profile from H&E staining by the pathologist. J-P, extracted tumor profiles from retardation, optic axis, DOPU, Stokes-Q, Stokes-U, Stokes-V, and attenuation contrast parameters. Q-W, the overlaid images between the extracted H&E staining tumor profile and PS-OCT parameters and the attenuation contrast.

Table 1.

Dice’s coefficient of tumor regions at the matching between PS-OCT (retardation, optic axis, DOPU, Stokes-Q, Stokes-U, Stokes-V, and attenuation coefficient) and the histology.

	Retardation	Optic axis	DOPU	Stokes-Q	Stokes-U	Stokes-V	Attenuation
Histology	0.886	0.867	0.889	0.886	0.890	0.880	0.802

We further validated the thresholds by evaluating the difference in various polarizations between normal and tumorous tissues. Five regions were selected from each 2D cross-sectional PS-OCT frame in 30 normal and tumor samples for the data statistics. In Figure 7A-7N, the attenuation coefficient (Figure 7A and 7H) and PS-OCT data showed that normal tissues had a high uniformity, but tumor tissues displayed a high heterogeneity. Figure 7O-7U showed the statistical plots corresponding to the attenuation coefficient, polarization, and Stokes parameters to describe the difference of tissue values in normal and tumorous kidneys. Although the attenuation coefficient between normal and tumor tissues showed a significant difference, the difference in the probability of the tissue changing degree was not as obvious as in PS-OCT data. Within the comparison of polarization parameters, Stokes-U and Stokes-V parameters had a lower probability of the degree of tissue changing in kidney tumors. This result indicated that the efficacy of using PS-OCT data to distinguish tumor tissues was higher than the attenuation coefficient.

Figure 7.

Statistical plots in the probability of the degree of the tissue changing in normal and tumor tissues. A-G, attenuation coefficient, retardation, optic axis, DOPU, Stokes-Q, Stokes-U, and Stokes-V in normal tissues. H-N, attenuation coefficient, retardation, optic axis, DOPU, Stokes-Q, Stokes-U, and Stokes-V in tumor tissues. O-U, corresponding statistical histograms of attenuation coefficient, retardation, optic axis, DOPU, Stokes-Q, Stokes-U, and Stokes-V in normal and tumor tissues. Significant value: ***, p < 0.001. ****, p < 0.0001. N = 6.

Based on the validated thresholds, we further extracted the tumor borders from cross-sectional polarization images, as shown in Figure 8. We zoomed in on the sub-figures cropped from the two sides of tumor tissues in OCT intensity and PS-OCT polarization images. In Figure 8B₁-8G₁ and 8B₂-8G₂, we clearly observed that tumor borders were differentiated by the polarization differences that were marked by different colors in polarization and Stokes parameters, which was unable to be directly differentiated in the OCT intensity images (Figure 8A, 8A₁, and 8A₂). After applying the thresholds, we extracted the tissue structure with different polarization changes at the tumor border, as shown in Figure 8B_1-2-6G_1-2 and 8B_2-2-6G_2-2. Polarization-extracted PS-OCT images could precisely locate tumor regions within the human kidney.

Figure 8.

Representative cross-sectional polarization-extracted PS-OCT images and OCT intensity based on histology verification at tumor borders. A-G, cross-sectional tumor images in OCT intensity, retardation, optic axis, DOPU, Stokes-Q, Stokes-U, and Stokes-V parameters. A₁-G₁, left tumor border images in OCT intensity, retardation, optic axis, DOPU, Stokes-Q, Stokes-U, and Stokes-V parameters. A₂-G₂, right tumor border images in OCT intensity, retardation, optic axis, DOPU, Stokes-Q, Stokes-U, and Stokes-V parameters. B_1-1,2,3-G_1-1,2,3, polarization-extracted left tumor border images in OCT intensity, retardation, optic axis, DOPU, Stokes-Q, Stokes-U, and Stokes-V parameters. B_2-1,2,3-G_2-1,2,3, polarization-extracted right tumor border images in OCT intensity, retardation, optic axis, DOPU, Stokes-Q, Stokes-U, and Stokes-V parameters.

3.4. 3D PS-OCT structure and tumor tissue distinguishment

To show the spatial structure of the kidney tumor, we further provided the 3D PS-OCT intensity and polarization images. Figure 9A showed the 3D intensity structure of the kidney tumor but the tumor border was not clear. The intensity difference on the surface may be caused by the uneven structure of the surface. Based on the polarization thresholds, we separated the 3D tumor region from kidney normal tissues, as shown in Figure 9B-G. We observed that different polarization parameters marked the main tumor regions and borders and highlighted different tissue distribution within the spatial structure of the tumor. The 3D polarization structure of the kidney tumor from Stokes parameters showed the degree of changes in different polarization states (linear, circular, and elliptical). Figure 9B₁-G₁ showed the complete 3D polarization structure of the kidney tumor including kidney tumor and normal tissues. With the 3D tumor structure overlaid images between polarization and intensity images (Figure 9B₂-G₂), tumor borders and regions could be clearly distinguished and positioned from normal kidney tissues.

Figure 9.

3D and slice diagrams of OCT intensity and polarization of tumor regions and borders at different depths for human kidney tumors. A, 3D intensity-based OCT image of the kidney tumor. B-G, polarization-extracted 3D structures of the kidney tumor in retardation, optic axis, DOPU, Stokes-Q, Stokes-U, and Stokes-V parameters. B₁-G₁, 3D complete polarization images including the kidney tumor and normal tissues in retardation, optic axis, DOPU, Stokes-Q, Stokes-U, and Stokes-V parameters. B₂-G₂, 3D overlaid images between the polarization and the OCT intensity-based structure. H-M, slice diagrams of 3D polarization images in retardation, optic axis, DOPU, Stokes-Q, Stokes-U, and Stokes-V parameters. H₁-M₁, slice diagrams of 3D overlaid images between the polarization and the intensity images in retardation, optic axis, DOPU, Stokes-Q, Stokes-U, and Stokes-V parameters at different depths. H₂-M₂, 2D enface overlaid images between the polarization and the intensity images in retardation, optic axis, DOPU, Stokes-Q, Stokes-U, and Stokes-V parameters. Depth slices were at 245^th, 305^th, 405^th, and 485^th, respectively.

To further display the tumor region and border in different depths within the human kidney, we showed the overlaid polarization and intensity images of the tumor at different depths. Figure 9H-M showed the slice structure of the tumor at 245^th, 305^th, 405^th, and 485^th depths in 3D PS-OCT polarization images. At different depths, tumor tissues showed different tumorous regions and borders. Figure 9H₁-M₁ showed the overlaid tumor structure with the 3D polarization images and intensity images at the same depths. To clearly show the tumor region and border from overlaid images at different depths, we further displayed the overlaid enface polarization tumor tissue structures with the axial-projection enface OCT intensity structures at the corresponding depths. From these overlaid enface images, we observed that polarization and Stokes parameters from PS-OCT data not only displayed the tumor border and region at different depths but also provided the information on tissue distribution within the tumor via the polarization difference. Particularly, the tumor border (red arrows in retardation, yellow arrows in optic axis, green arrows in Stokes-Q, blue arrows in Stokes-U, and white arrows in Stokes-V, Figure 9H₂, 9I₂, 9K₂-L₂) was highlighted at different depths based on the different sensitive degree to polarization and Stokes parameters, which could be effective to guide doctors to precisely position tumor regions from normal tissues during the surgery.

Figure 10 showed the comparison of the tumor region among polarization states and Stokes parameters at different depths. With these zoomed in structures of tumor tissues, we observed that different polarization and Stokes parameters revealed different tissue structures within the kidney tumor. At the same depth, DOPU (Figure 10C-C₁-C₂, 10I-I₁-I₂, 10O-O₁-O₂, and 10U-U₁-U₂) mainly highlighted tumor tissues but did not show internal tumor borders. Compared to retardation, optic axis, Stokes-Q, Stokes-U, and Stokes-V in Figure 10M₂, 10N₂, 10P₂, 10Q₂, and 10R₂, DOPU (Figure 10O₂) were unable to distinguish the internal tumor margin tissues. Moreover, polarization images (retardation, optic axis, and DOPU) did not show the same tumor profiles from sub-figures at different depths. Meanwhile, Stokes parameters also showed the ununiformed tissue distributions and borders within the tumor. These differences in polarization and parameters indicated that the polarization images could be used to precisely locate and distinguish kidney tumor regions and borders in a complementary manner.

Figure 10.

Comparison of tumor regions among polarization states and Stokes parameters at different depths. A-F, tumor regions in different polarization and Stokes parameters at 245^th slice. G-L, tumor regions in different polarization and Stokes parameters at 305^th slice. M-R, tumor regions in different polarization and Stokes parameters at 405^th slice. S-X, tumor regions in different polarization and Stokes parameters at 485^th slice. A₁-X₁ and A₂-X₂ were zoomed in sub-figures corresponding to the green (A₁-X₁) and red (A₂-X₂) frames in A-X at 245^th, 305^th, 405^th, and 485^th slices.

4. Discussion

The regular treatment for kidney tumors was surgical resection which required completely removing cancerous tissues to avoid recurrence and maximumly keeping normal tissues. Therefore, accurate identification of the tumor region and boundary played an impactful role in the surgical removal of kidney tumors. Although many conventional imaging systems such as CT, MRI, and ultrasonography could provide large FOV for tumor diagnosis, the low spatial resolution prohibited the surgery from precisely resecting tumorous tissues from surrounding normal tissues. Moreover, CT and MRI faced the challenge of providing imaging guidance during the surgery in real-time which was one of the most critical parts of clinical surgical removals. OCT had been demonstrated as a promising tool to distinguish normal and cancerous tissues in brain, kidney, oral, and skin tumors [19, 34-36]. However, the intensity-based OCT structure images lacked tissue-specific contrast, which resulted in limited differentiation of the boundary between normal and cancerous tissues to further map the profile of tumors for complete surgical removal. Particularly, the border between normal and cancerous tissues consists of partial tumorous and normal tissues, as well as transferring and degenerating tissues, which was difficult to induce enough change of structure and tissue formation that could be detected by current imaging modalities.

In this study, we demonstrated that PS-OCT was able to accurately detect tumor-associated fibrosis and distinguish tumorous tissues from normal tissues based on tissue birefringence. Compared to intensity-OCT in both cross-sectional and enface images, our results indicated that polarization parameters of retardation, optic axis, and DOPU, as well as Stokes parameters of Q, U, and V, provided a more precise boundary differentiation between tumorous and normal tissues. The existed tubules/lesions within the kidney (Figure 3H and 3H₁, white frame) caused confusion for distinguishing the border in OCT intensity images, while PS-OCT images (Figure 3I-N and 3I₂-N₂) could precisely locate the boundary between cancerous and normal tissues (Figure 3I_ii-N_ii). The enface structure of OCT intensity images was able to map the profile of the tumor region, however, the precise differentiation of the border between tumorous and normal tissues could only be detected by PS-OCT images based on the tissue birefringence, especially PS-OCT images, as shown in Figure 4J, 4M, and 4N (blue, green, and red arrows). The capability of precise differentiation of tumorous tissue from normal adjacent tissues by PS-OCT could be taken advantage of to accurately find the tumor border in the mixed field of non-adequate tumorous, normal, and degenerative tissues in clinical renal cancer removal surgeries. This technique could maximumly retain the normal kidney tissue for essential human physiological function and metabolism and completely resect the tumor tissue. In terms of imaging speed for clinical translation, the acquisition time of each B-scan for single channel is ~6ms for our OCT system with an A-Scan rate of 48 KHz, which is consistent with the imaging speed for guiding mouse brain tumor removal in a previous study [37]. The image acquisition time of each frame in clinical ophthalmology surgeries is ~ 9-15ms [38, 39], thus our image acquisition speed is comparable with those reported in the clinical studies. However, it takes 75.814s to obtain seven 9 × 9 × 2.6 mm³ (1800 × 1800 × 1024 pixels) 3D images (including intensity, retardation, optic axis, DOPU, Stokes-Q, Stokes-U, Stokes-V). With the recently reported ultrafast OCT laser design up to MHz [40], the ultrahigh-speed 3D OCT image acquisition can reach multi-cm³ fields per second in the future study.

In the study of quantitative distinguishing tumors from surrounding normal tissues, the intrinsic optical attenuation coefficient has been widely applied to detect tumor regions [20, 35, 41]. There was a significant difference in the attenuation coefficient between cancerous and normal renal tissues [35, 42, 43], which was employed to distinguish tumor regions from normal kidney tissues. We calculated the attenuation coefficient in tumorous and normal tissues and found that tumor tissues had a smaller attenuation coefficient than that of normal tissues, as shown in Figure S1D. Our results were consistent with that reported in [40] that tumor tissues had a smaller attenuation coefficient, while we also notice there were studies in [36] and [39] indicated that tumor tissues had a larger attenuation coefficient compared to normal renal tissues. Therefore, the application of the attenuation coefficient for differentiating kidney tumor boundary still faced the limitation because of the difference of renal tumors in different patients. In this study, we used polarization parameters (retardation, optic axis, and DOPU) and Stokes parameters (Q, U, and V) to make a comparison with the attenuation coefficient via the Dice coefficient and the histology verification. We found that PS-OCT images provided a higher agreement of the tumor region detection with the histology than the attenuation coefficient method (Table 1). The attenuation coefficient could roughly detect the main tumor region, but the tumor boundary information was missed. In contrast, Figure 6 showed that PS-OCT detections effectively located the border between tumorous and normal tissues which was highly consistent with the histology. Thus, our results confirmed that PS-OCT was able to offer more accurate and consistent detection of kidney tumor boundary compared to the attenuation coefficient.

In addition to distinguishing tumor regions from normal tissues, PS-OCT also provided the 3D information of tumors to create the volumetric structure visualization in real-time for surgical guidance. 3D spatial visualization showed significant advantages in tumor localization compared to standard 2D slice visualization for surgical guidance [44, 45]. The 3D structure of kidney tumors from PS-OCT scanning allowed multi-parameter spatial distinguishment from normal tissues, which was significantly more accurate than that of traditional intensity-based OCT structures. 3D polarization parameter visualizations directly located tumor borders and zones in spatial structures that could clearly guide surgeons to resect tumors or check the effect of surgical resection. Moreover, given the constrained depth of penetration inherent to PSOCT, its application in clinical settings to guide the excision of kidney tumors along the interface between healthy and tumorous tissues necessitates a sequential approach. Given that tumor resection by surgeons entails a multistep process rather than a single-cut procedure, PSOCT can serve as a valuable tool for progressively identifying the resection area. This involves conducting imaging after each successive resection, thereby facilitating a step-by-step approach towards achieving complete tumor removal. PS-OCT was also capable to visualize the distribution of fibrosis and tumor tissues within kidneys, which provided further benefits for guiding tumor samplings in vivo in clinical medical research. The visualization of the internal distribution of tumorous tissues based on PS-OCT could assist surgeons to acquire the most representative tumorous tissues, which was critical to conducting oncology clinical assays and gene sequencing. Molecular and genetic analysis of the tumor tissue in patients was significant for anti-metastasis and anti-recurrence therapies, which was also crucial for post-surgery drug screening and therapy [46-48]. Therefore, our results demonstrated that PS-OCT was the potential to play a key role in the guidance of the tumor sampling and assay of renal cancer. This study also provided the quantitative evaluation of tumorous renal tissue from retardation, optic axis, DOPU, Stokes-Q, Stokes-U, and Stokes-V parameters to demonstrate the availability of PS-OCT in the precise detection of kidney tumor boundary. Although the intrinsic optical attenuation contrast of kidney tumors showed there was a significant difference between tumor and normal tissues (Figure 7O), the difference of tissues from kidney tumors and normal tissues was not as obvious as in PS-OCT parameters (Figure 7P-U). Tubules, arterioles, and fibrosis within kidneys were also able to cause the same attenuation contrast as tumorous tissues [49, 50], Figure S1E showed that there were still low attenuation coefficient tissues within normal kidney samples, which might cause relative larger errors to distinguish tumor borders. In contrast to the attenuation contrast, PS-OCT polarization parameters had more significant differences between tumorous and normal kidney tissues. The classification probability of PS-OCT parameters for normal samples was close to 0 which was significantly different from tumorous samples. This character indicated that PS-OCT has obvious advantages in the kidney tumor border distinguishment compared to the attenuation contrast. However, larger-scale studies need to be performed in the future to further validate the findings of this study prior to clinical surgeries.

PS-OCT also provided the visualization of internal tissue distributions within kidney tumor including microstructures and tissue categories. Based on the H&E staining, we found that the tumor is circumscribed with prominent fibrous pseudocapule (black arrows in Figure S2A). There are cavity areas (cyan arrows in Figure 11A) and papillary structures (black arrows in Figure 11B) existed in RCCs. Therefore, we could suggest that the tumor type was probably a papillary RCCs because tumor cells form many papillary structures that have a fibrovascular core (representatively indicated by orange arrows in Figure S2B and black circle in Figure S2C) and surrounded by a layer of cuboidal or columnar tumor cells [51, 52]. We could also observe the same cavity areas (cyan arrows) and papillary structures (black arrows) from the corresponding PS-OCT images as shown in Figure 11A₁-₆ and B₁-₆. This indicated that PS-OCT was able to detect kidney tumor microstructures for further distinguishing tumor types. It also matched the different tissue structures that we observed at different depths within the tumor (Figure 10). Moreover, H&E staining allowed us to observe the tumor borders compromised by fibrous tissue, smooth muscle fibers (pink arrows in Figure 11C and D, yellow arrows in Figure S2D, E, and F), and myofibroblasts (green arrows in Figure 11D, Red arrows in Figure S2D and E). We found that the corresponding PS-OCT images (Figure 11C₁-₆ and D₁-₆) showed the same components of smooth muscle fibers (pink arrows) and myofibroblasts (green arrows). By the existence of the papillary edema (Figure S2C and D), foamy macrophages (Figure 11B_i and B_ii) in papillary cores, and a thick fibrous capsule (Figure 11C and D, Figure S2A, D, and F), we suggested that the tumor subtype belonged to Type-1 papillary RCCs [52, 53]. In PS-OCT data, we clearly observed critical structures and components provided by H&E staining which was employed to clarify the type of kidney tumors clinically. This result demonstrated that PS-OCT had the potential to provide a clinical classification of RCCs based on the detection of the internal microstructure and tissue distribution of tumors. Compared to histology, PS-OCT had the advantage of real-time imaging and in vivo scanning to save time for diagnosis and planning treatment. Additionally, we found that PS-OCT polarization and Stokes parameters had different sensitiveness to different microstructures and tissues, which could be used for the characterized analysis of the specific type of RCCs and longitudinal tracking of therapeutic effects such as chemotherapy and radiotherapy.

Figure 11.

Comparison of the recognition of tumor internal microstructures between H&E staining and PS-OCT images. A-D, representative H&E staining of internal tissues and borders of kidney tumors. A_i-D_i and A_ii-D_ii, enlarged microstructures corresponded to representative histology images. A₁-A₆, B₁-B₆, C₁-C₆, and D₁-D₆, parameter images of retardation, optic axis, DOPU, Stokes-Q, Stokes-U, and Stokes-V from PS-OCT corresponding to representative H&E staining. Cyan arrow, cavity, or empty area. Yellow arrow, tumor cells. Black arrow, papillary structures. Pink arrow, smooth muscles. Green arrow, myofibroblasts on the fibrous pseudocapsule. Scale bar is 350 μm.

One limitation to notice is that the PS-OCT system in this study is based on single input polarization state, thus the polarization parameters (phase retardation, optic axis, and DOPU) are all cumulative values [54]. The cumulative phase retardation only indicates the phase retardation between the principal polarization states along the complete optical path through the tissue rather than the phase retardation effect at a single depth location [55, 56]. Therefore, the ‘local’ polarization information was not able to be quantitatively provided. One the other hand, since the local birefringence mode was proportional to the amount of the actual birefringent signal per pixel in the OCT images, the local polarization signal suffers from a lower sensitivity [57, 58]. Compared to the local polarization state, the cumulative polarization mode provided higher sensitivity and required much simpler system, therefore was always preferred for binary decisions [55, 58, 59]. Considering the key to the surgical guidance for renal tumor resection is the differentiation between normal and cancerous tissues to avoid residual tumor tissues, we hereby applied the cumulative polarization parameters to detect tumor margins and regions.

5. Conclusion

The structure and polarization information provided by PS-OCT was shown to offer a significant improvement in contrast between renal cancer and normal kidney tissues. PS-OCT polarization parameters provided depth-resolved, cross-sectional, and spatial structure information to achieve real-time distinguishment of kidney tumor borders and regions. We demonstrated that PS-OCT imaging could achieve more accurate detection and differentiation of tumor margins and zones compared to the intrinsic optical attenuation contrast. The results indicate that PC-OCT was able to visualize the tissue distribution of kidney tumor for assisting tumor sampling and surgical guidance. The result of our study provides important information for translating PS-OCT to in vivo clinically surgical resection guidance in the future.

Highlights.

• The polarization-sensitive optical coherence tomography (PS-OCT) system was firstly applied to provide imaging guidance for the surgical resection of renal tumors on 6 ex-vivo human kidneys.

• The polarization parameters provided significantly higher detection accuracy for tumor margins compared to the conventional optical intensity and attenuation contrast method.

• The tumor type and tissue distribution of kidney cancer could be characterized and analyzed noninvasively by PS-OCT polarization and Stokes parameter images.

Data availability

The data presented in this paper are not publicly available but may be obtained from the authors upon reasonable request.