Open Access
Add to BibSonomy
Abstract
Clear cell renal cell carcinoma (ccRCC) is a common histopathological subtype of renal cancer and is notorious for its poor prognosis. Its accurate diagnosis by histopathology, which relies on manual microscopic inspection of stained slides, is challenging. Here, we present a correlative approach to utilize stained images and refractive index (RI) tomography and demonstrate quantitative assessments of the structural heterogeneities of ccRCC slides obtained from human patients. Machine-learning-assisted segmentation of nuclei and cytoplasm enabled the quantification at the subcellular level. Compared to benign regions, malignant regions exhibited a considerable increase in structural heterogeneities. The results demonstrate that RI tomography provides quantitative information in synergy with stained images on the structural heterogeneities in ccRCC.
© 2023 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
Clear cell renal cell carcinoma (ccRCC), a common type of kidney cancer, is associated with poor prognosis and patient outcomes compared to other types of RCC [1]. Designing optimal treatments for patients often requires an accurate diagnosis of ccRCC [2]. Approaches to diagnosing ccRCC based on microscopic histological images have been widely investigated [3,4]. Conventionally, bright-field (BF) microscopy has been used to explore cell morphologies across ccRCC slides using staining reagents such as hematoxylin and eosin (H&E). Although H&E staining directly visualizes the morphologies of nuclei and cytoplasm under BF microscopy, the method relies on qualitative evaluations and inherently suffers from intra-observer and inter-observer variations [5]. A quantitative approach is essential for circumventing these variations.
Quantitative phase imaging (QPI) techniques are label-free quantitative imaging methods that have shown potential for quantifying histopathological features of tissue slides [6–8]. With the advantage of providing optical phase delay of a specimen without using exogenous labeling, QPI techniques have been utilized to investigate diverse tissue slides. Due to the distinctive features of cancers in microscopic images, numerous cancers including prostate [9,10], colon [11–16], and breast cancers [17–20] are widely studied using QPI techniques. In addition to cancers, other notorious diseases such as Alzheimer’s disease [21], Crohn’s disease [22], and pelvic organ prolapse [23] had been investigated using the QPI techniques, providing unprecedented quantitative analyses. Recent technical advances enabled on-chip microscopes which broadened the utility of the QPI techniques in histopathology [24,25]. Furthermore, three-dimensional (3D) QPI techniques have enabled the quantification of 3D (RI) distributions of histopathological specimens with improved spatial resolution and optical sectioning capabilities [26]. Among the 3D QPI techniques, intensity-based 3D QPI techniques have recently emerged as an efficient tool for imaging biological systems without using interferometry [27–30].
In this article, we present quantitative assessments of structural heterogeneities in ccRCC using RI variations across the nuclei and cytoplasm of tissue sections synergistically with H&E stained images. Namely, the structural heterogeneities of subcellular components were obtained from the RI distribution, following a segmentation based on the stained images. To be specific, H&E-stained slides from patients with ccRCC were analyzed using a 3D QPI technique equipped with a color BF microscope. Both the H&E-stained images and 3D RI distribution were measured on the same slide. Subsequently, the areas of the nucleus and cytoplasm were automatically segmented from the measured H&E-stained image using a machine-learning approach. The corresponding areas were then quantitatively analyzed using the measured RI distributions.
We quantified structural heterogeneities, defined as the local standard deviation (SD) of RI values, across segmented nuclei and cytoplasm. Furthermore, we investigated the heterogeneities of intra-nuclear structures using the mean RI, coefficient of variation (CV) of RI, and high-RI proportion within individual nuclei. Compared with benign regions, malignant regions showed a 12.11% increase in the local RI SD and a 35.27% increase in the high-RI proportion in the individual nuclei. We demonstrated that the present synergistic microscopic approach enabled the quantitative analysis of clinical tissue slides from ccRCC patients, and we envision that the present approach can be further extended for quantitative clinical analyses of various types of cancer, computer-aided diagnosis, and providing in-depth subcellular quantitative information for personalized medicine.
2. Materials and methods
2.1 Sample preparation
Two ccRCC slides were prepared according to standard protocols (Fig. 1(a)). A total of 26 RCC case slides, newly diagnosed at Gangnam Severance Hospital in January 2022, were reviewed by a pathologist (SJS). Among the 26 cases, two small ccRCC cases with low World Health Organization/International Society of Urological Pathology grade (grade 2) were randomly selected. Formalin-fixed paraffin-embedded tissue blocks were obtained from the selected cases. Tissue blocks were sliced to a 4 μm-thickness using a microtome. The slices were carefully placed on 1 mm-thick glass slides and de-paraffinized using xylene. The slides were then stained with H&E, mounted with a mounting medium, which had an RI value of 1.495 measured by a refractometer (R-5000, Atago), and finally covered with coverslips. An experienced pathologist (SJS) whose specialism is uropathology annotated malignant tumors, normal renal parenchyma, fibrous, and blood clot regions after H&E staining. This study was approved by the institutional review board (IRB) of Gangnam Severance Hospital (IRB No. 3-2022-0083) and conducted in accordance with the Declaration of Helsinki.
Fig. 1. Structural heterogeneity quantification of human ccRCC slides. (a) The workflow for quantifying structural heterogeneities of tissue slides. Both BF and RI images were obtained as 30 × 30 tiled images. Nuclei and cytoplasm were segmented based on BF images. The segmentation masks were applied to the RI image to extract the RI distribution of the subcellular components. (b) The setup of PEPSI-ODT. (c) Reconstruction process. From the four intensity images obtained with optimized illumination patterns, a 3D RI tomogram is reconstructed. A BF image is obtained by merging the three images with different LEDs. (d) Machine learning approaches were used to segment nuclei and cytoplasm based on the BF images. (e) RI images of ccRCC slide obtained at various axial positions.
2.2 3D RI tomography
To measure the RI distributions of human ccRCC slides, we employed partial coherence RI tomography using plural efficient patterns for self-interfering optical diffraction tomography (PEPSI-ODT) (Fig. 1(b)) [30]. In PEPSI-ODT, transmitted intensity images of a sample were measured using four optimized Köhler illumination patterns, from which the 3D RI distribution was reconstructed by deconvolution of the intensity images with theoretically estimated point spread functions (Fig. 1(c)). In the experimental setup, a light-emitting diode (LED) with a peak wavelength of 455 nm within a digital micromirror device (DMD) module (E4500MKII, EKB Technologies) was used as the illumination source. The blue wavelength was chosen to avoid overlap with the peak absorption spectrum of H&E staining [31]. The illumination intensity was controlled using a DMD located in the Fourier plane, which was then projected onto the sample using a customized condenser lens. The diffracted light from the specimen was collected using an objective lens (×20, numerical aperture (NA) = 0.70, UCPLFLN20X, Olympus) and measured using a CMOS camera (BFS-U3-28S5M-C, FLIR) located in the image plane. To obtain conventional BF microscopic images, two additional LEDs with peak wavelengths of 520 and 624 nm were also installed. The BF images were obtained by merging the three raw images illuminated by each LED.
The RI images of the ccRCC slides were reconstructed by deconvolution of the raw intensity images obtained using the four optimized illumination patterns with a theoretically estimated point spread function. The RI and BF images were repeatedly reconstructed for every 240 × 240 μm field of view. The theoretically calculated spatial resolutions of the system were 162 nm and 1312 nm for the lateral and axial resolutions, respectively [32].
2.3 3D tiled imaging
For 3D tiling, we programmed the motorized XY stages to move the sample by 200 μm per tile, and a motorized Z stage (ZA05A-W101, Kohzu Precision) to move the objective lens by 547 nm per slice. A total of 30 × 30 tiled images were obtained per slide with an overlap of 40 μm for both BF and RI. The tiled images were stitched using the stitching plugin tool in ImageJ [33].
2.4 Machine learning-based segmentation of subcellular organelles
Machine learning-based segmentation of the cytoplasm, background, and nuclei was performed on the measured BF images to guide quantitative analysis within the RI tomogram. The BF images were merged from the three raw images obtained under the illumination of red, green, and blue LEDs with peak wavelengths mentioned in the methods. To segment the cytoplasm and background, we used Ilastik, an open-source software that provides an interactive tool for image segmentation [34]. For the segmentation of nuclei, we used HoVer-Net, a pre-trained deep learning network [35]. Segmentation masks of the nuclei, cytoplasm, and background were directly applied to the registered RI images for quantitative analysis (Fig. 1(d)). Note that we used the RI image of the focal plane (z = 0) to apply the segmentation masks and to obtain the following results (Fig. 1(e)).
3. Results and discussion
3.1 RI distribution of ccRCC and its histological features
To visualize the clinical features of ccRCC in both RI and BF images, we imaged approximately 5 × 5 mm regions of each slide obtained from two human patients, which in total include more than 80,000 nuclei, by stitching 30 × 30 tiled images (Fig. 2(a–c)). The RI distribution of ccRCC slides ranged from 1.46 to 1.52, which is reasonable according to the previous results [26]. Within this range, the RI images visualized the anatomical structures and tissue compositions, which corresponded well with the BF images. In the normal renal parenchymal region (i), the nuclei stained blue in the BF image are also clarified in the corresponding RI image. In the malignant tumor region (ii), we identified the well-known cytoplasmic void structures found in ccRCC cells in both the BF and RI images [36]. Additionally, we identified a blood clot region composed of red blood cells (iii), and a non-tumor fibrous region with circular vessels (iv) using both imaging modalities. The cytoplasmic void structures and the circular vessels are indicated with yellow arrows in (ii), and (iv) respectively. The magnified images of the structures are presented in (v), and (vi) respectively.
Fig. 2. Wide-field images of RI and BF. 30 × 30 tiled images were stitched to obtain the wide-field images. (a) A total number of patients, regions of interest, and the number of contained nuclei. The stitched BF (b) and RI (c) images. Detailed images of the stitched BF image (a) (i–vi) and RI image (b) (i–vi). Representative regions are magnified to visualize normal renal parenchymal (i), malignant tumor (ii), a blood clot (iii), and non-tumor fibrous (iv) regions. A yellow arrow in (ii) indicates the void structures of ccRCC cells and is magnified in (v). A yellow arrow in (iv) indicates the circular vessels and is magnified in (vi).
3.2 Quantification of structural heterogeneity
Inspired by the structural distinction between the tumor and normal regions in the RI images, we quantitatively compared the RI distribution between the malignant and normal regions of the ccRCC slides. Specifically, we quantified the local RI SD in each region of two slides obtained from two different patients (Fig. 3). We mapped the distribution of the local RI SD within a sliding window of 16 × 16 μm. The window moved with a step of 8 μm. To obtain the local RI SD at diverse cellular levels, we used three RI images: 1) nuclei, 2) cytoplasm and 3) both segmented images. The numbers of segmented nuclei in patient #1 were 17,484 and 14,090 for the malignant and normal domains, respectively. In patient #2, 33,143 malignant nuclei and 18,879 normal nuclei were segmented.
Fig. 3. Quantification of the structural heterogeneities and statistical analyses. (a, k) Wide-field BF images of both patients. (b, l) Annotation masks delineated by the pathologist based on the BF images. (d, n) Representative images of the malignant (i) and normal (ii) regions. Quantified local RI SD (e–g) and the corresponding statistical comparison between malignant and normal domains (h–j) of patient #1. Quantified local RI SD (o–q) and the corresponding statistical comparison between the domains (r–t) of patient #2. In each box plot, colored lines in the middle of the box indicate the median, the ends of the box represent the first and third quartiles, and the whiskers span a 1.5 interquartile range from the ends. Unpaired two-sided Student’s t-test was used to calculate the P values. P < 0.001 was considered statistically significant. ***: P < 0.001.
We compared the local RI SD between the malignant and normal regions. The annotation masks were delineated by an experienced pathologist based on the H&E stained images (Fig. 3(a), 3(b), 3(k), and 3 l). The representative images of the malignant and normal regions for each patient are presented (Fig. 3(d), and (3n)). After mapping the local RI SD (Fig. 3(e–g), and (3o–q)), the histograms and boxplots were obtained from the marked domains for statistical analysis (Fig. 3(h–j), and (3r–t)). The malignant domains displayed increased RI SD in both patients. Although this tendency was consistent in both the nuclei and cytoplasm, the extent of the increase varied.
For patient #1, the local RI SD in the malignant domain increased by 4.76% and 16.7% in the nucleus and cytoplasm, respectively. The nuclei showed the smallest increase in RI SD, with average values of 0.0022 ± 0.0007 and 0.0021 ± 0.0007 in the malignant and normal domains, respectively. The RI SDs of the cytoplasm were 0.0028 ± 0.0006 and 0.0024 ± 0.0005 for malignant and normal domains, respectively. When the nuclei and cytoplasm were considered together, the computed RI SD was 0.0029 ± 0.0006 for the malignant domain, which indicated a 16% increase compared to the normal domain (0.0025 ± 0.0005).
We observed consistent tendencies for patient #2. The local RI SD increased by 13.0% and 9.68% in the malignant nucleus and cytoplasm, respectively. For nuclei, the malignant domain had a greater local RI SD (0.0026 ± 0.0007) than the normal domain (0.0023 ± 0.0007). In the cytoplasm, the malignant area also exhibited a greater RI SD (0.0034 ± 0.0007) than the normal area (0.0031 ± 0.0008). Considering both subcellular structures, the calculated RI SDs were 0.0036 ± 0.0007 in the malignant domain and 0.0032 ± 0.0008 in the normal domain, indicating a 12.5% increase in the malignant region. Overall, the increase in local RI SD indicated greater structural heterogeneity in the malignant regions of the ccRCC tissues.
3.3 Quantification of intra-nuclear structural heterogeneity
To quantitatively assess the previously reported structural alterations of nuclei caused by cancer [37,38], intra-nuclear structural heterogeneities were further analyzed. To be specific, we obtained the statistical properties including the mean RI, CV of RI, and the high-RI proportion from individual nuclei (Fig. 4(a–d)). We defined a high-RI proportion as the percentage of a nuclear area that exhibited RI above the threshold. The high-RI proportion is related to nucleolar structures according to a previous report [39]. The threshold values for patients #1 and #2 were set to 1.4981 and 1.5002, respectively, by averaging the RI values of eight manually selected nucleoli. In the analysis, nuclei showing 0% of the high-RI proportion were excluded.
Fig. 4. Quantification of the intra-nuclear structural heterogeneities and statistical analyses. (a) BF images of a normal and malignant nucleus. (b) Binary masks for the corresponding nuclei in (a). (c) The binary masks were applied to the RI images to extract the RI distributions of the individual nuclei. (d) A threshold (RI ≥ 1.5002) is applied to the segmented RI distribution in (c). After the threshold application, the number of pixels was counted to calculate the high-RI proportion. (e–j) Statistical comparison between the normal and malignant regions of mean RI, CV, and high-RI proportion. Unpaired two-sided Student’s t-test was used to calculate the P values. P < 0.001 was considered statistically significant. ***: P < 0.001
We statistically compared the three properties of malignant and normal nuclei (Fig. 4(e–j)). These properties showed a consistent increase in the malignant nuclei for both patients. For patient #1, the mean RI of individual nuclei increased slightly by 0.013% in malignant nuclei (1.4976 ± 0.0020) compared with normal nuclei (1.4974 ± 0.0016). The CV of RI increased by 2.370% in malignant nuclei (0.138 ± 0.0388%) compared to that in normal nuclei (0.135 ± 0.0417%). Among the three parameters, the high-RI proportion showed the greatest increase (15.21%) in malignant nuclei (40.78 ± 28.46%) compared with normal nuclei (35.39 ± 24.24%). In total, 13,970 and 17,168 nuclei were analyzed for the normal and malignant domains, respectively.
The identical analyses for patient #2 also displayed increases in all three statistical properties, while the extent of increases was greater. The mean RI of individual nuclei increased slightly by 0.734% in malignant nuclei (1.4996 ± 0.0021 and 1.4985 ± 0.0019 for malignant and normal nuclei, respectively). The CV of RI showed an 8.867% increase (0.1633 ± 0.0458% and 0.1500 ± 0.0435% for malignant and normal nuclei). The high-RI proportion indicated a 55.31% increase (41.19 ± 26.57% and 26.52 ± 23.98% for malignant and normal nuclei, respectively). A total of 17,968 and 32,523 nuclei were analyzed for the normal and malignant domains, respectively. The consistent increase of the high-RI proportion in both patients implies an enlargement of nucleoli, which was observed in previous reports [37,38].
4. Conclusion and discussion
In summary, we performed quantitative analysis using RI tomography and H&E stained images correlatively to assess the structural heterogeneities of ccRCC. To perform high-throughput image analysis, we established a tiled acquisition strategy in PEPSI-ODT and a machine learning-based method for automatically segmenting subcellular compositions in the tissue. The segmented RI information provided consistent evidence that the malignancy of ccRCC correlates with increasing RI heterogeneity at the subcellular scale.
The present results reflect previously known physiological properties of ccRCC. For instance, the increased high-RI proportion within the cancer nuclei implies enlarged nucleoli. Such alteration has been accounted for by severe mutations in their nuclei [40] and aggressive proliferation via active ribosomal biogenesis [40,41]. Therefore, our quantitative assessment of intra-nuclear heterogeneities is highly relevant to the physiological characteristics of the ccRCC.
The use of 3D QPI in histopathological studies has a technical advantage compared to 2D QPI. Because the 2D QPI measures the optical phase delay without optical sectioning capability, it may induce artifacts originating from uneven surfaces of the samples. Optical sectioning using 3D QPI can alleviate the problems. However, 3D QPI is limited in time consumption, because it requires axial scanning. Recent technical advances enabled rapid RI imaging without mechanical scanning [42].
To broaden the applicability of the present method in preclinical research and diagnostic applications in clinics, follow-up studies for standardization should be considered. Variations originating from the tissue preparation protocols, including fixing, embedding, and staining have to be specified through multi-institution studies. We stress that the optical element of the proposed method is suitable for standardization, as RI measurement does not accompany instrument-specific errors.
The present method can be applied to other histopathological analyses. Immediate follow-up studies should include the classification of RCC subtypes exploiting the proposed method. Further correlative utilization of RI tomography will also facilitate the quantification of various cancer-induced alterations in subcellular scales. For instance, immunohistochemical staining will enable quantitative cancer stage determination through a similar correlative analysis [43]. Moreover, such correlative studies will provide insights for RI-only analysis, promoting the realization of purely label-free histopathology. With these advantages, we believe that the proposed method can broaden its utility across various histopathological studies for automated diagnosis and classification, thereby improving precision medicine.
Funding
National Research Foundation of Korea (2015R1A3A2066550, 2022M3H4A1A02074314); Tomocube Inc.; Ministry of Science and ICT, South Korea (2021-0-00745, COMPA2022-SRETC-S03-3, N11210014, N11220131); the Korea Health Technology R&D Project through the Korea Health Industry Development Institute (KHIDI), funded by the Ministry of Health and Welfare, Korea (HI21C0977).
Acknowledgments
The authors thank Young Seo Kim and Herve Hugonnet for providing aids to prepare samples and imaging acquisition.
Disclosures
The authors declare no conflicts of interest.
Data availability
Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request.
References
1. J. C. Cheville, C. M. Lohse, H. Zincke, A. L. Weaver, and M. L. Blute, “Comparisons of Outcome and Prognostic Features Among Histologic Subtypes of Renal Cell Carcinoma,” Am. J. Surg. Pathol. 27(5), 612–624 (2003). [CrossRef]
2. S. R. Prasad, P. A. Humphrey, J. R. Catena, V. R. Narra, J. R. Srigley, A. D. Cortez, N. C. Dalrymple, and K. N. Chintapalli, “Common and Uncommon Histologic Subtypes of Renal Cell Carcinoma: Imaging Spectrum with Pathologic Correlation,” RadioGraphics 26(6), 1795–1806 (2006). [CrossRef]
3. S. A. Fuhrman, L. C. Lasky, and C. Limas, “Prognostic significance of morphologic parameters in renal cell carcinoma,” Am. J. Surg. Pathol. 6(7), 655–664 (1982). [CrossRef]
4. D. G. Skinner, R. B. Colvin, C. D. Vermillion, R. C. Pfister, and W. F. Leadbetter, “Diagnosis and management of renal cell carcinoma A clinical and pathologic study of 309 cases,” Cancer 28(5), 1165–1177 (1971). [CrossRef]
5. M. Al-Aynati, V. Chen, S. Salama, H. Shuhaibar, D. Treleaven, and L. Vincic, “Interobserver and intraobserver variability using the Fuhrman grading system for renal cell carcinoma,” Arch. Pathol. Lab. Med. 127(5), 593–596 (2003). [CrossRef]
6. G. Popescu, Quantitative phase imaging of cells and tissues (McGraw-Hill Education, 2011).
7. K. Lee, K. Kim, J. Jung, J. Heo, S. Cho, S. Lee, G. Chang, Y. Jo, H. Park, and Y. Park, “Quantitative phase imaging techniques for the study of cell pathophysiology: from principles to applications,” Sensors 13(4), 4170–4191 (2013). [CrossRef]
8. Y. Park, C. Depeursinge, and G. Popescu, “Quantitative phase imaging in biomedicine,” Nat. Photonics 12(10), 578–589 (2018). [CrossRef]
9. S. Sridharan, V. Macias, K. Tangella, A. Kajdacsy-Balla, and G. Popescu, “Prediction of prostate cancer recurrence using quantitative phase imaging,” Sci. Rep. 5(1), 9976 (2015). [CrossRef]
10. H. K. Roy, C. B. Brendler, H. Subramanian, D. Zhang, C. Maneval, J. Chandler, L. Bowen, K. L. Kaul, B. T. Helfand, C.-H. Wang, M. Quinn, J. Petkewicz, M. Paterakos, and V. Backman, “Nanocytological field carcinogenesis detection to mitigate overdiagnosis of prostate cancer: a proof of concept study,” PLoS One 10(2), e0115999 (2015). [CrossRef]
11. J. K. Zhang, Y. R. He, N. Sobh, and G. Popescu, “Label-free colorectal cancer screening using deep learning and spatial light interference microscopy (SLIM),” APL Photonics 5(4), 040805 (2020). [CrossRef]
12. S. Gladstein, D. Damania, L. M. Almassalha, L. T. Smith, V. Gupta, H. Subramanian, D. K. Rex, H. K. Roy, and V. Backman, “Correlating colorectal cancer risk with field carcinogenesis progression using partial wave spectroscopic microscopy,” Cancer Med. 7(5), 2109–2120 (2018). [CrossRef]
13. J. E. Chandler, L. Cherkezyan, H. Subramanian, and V. Backman, “Nanoscale refractive index fluctuations detected via sparse spectral microscopy,” Biomed. Opt. Express 7(3), 883–893 (2016). [CrossRef]
14. H. Subramanian, H. K. Roy, P. Pradhan, M. J. Goldberg, J. Muldoon, R. E. Brand, C. Sturgis, T. Hensing, D. Ray, A. Bogojevic, J. Mohammed, J.-S. Chang, and V. Backman, “Nanoscale cellular changes in field carcinogenesis detected by partial wave spectroscopy,” Cancer Res. 69(13), 5357–5363 (2009). [CrossRef]
15. P. Wang, R. Bista, W. Khalbuss, W. Qiu, S. Uttam, K. Staton, L. Zhang, T. Brentnall, R. Brand, and Y. Liu, “Nanoscale nuclear architecture for cancer diagnosis beyond pathology via spatial-domain low-coherence quantitative phase microscopy,” J. Biomed. Opt. 15(6), 066028 (2010). [CrossRef]
16. S. Uttam, H. V. Pham, J. LaFace, B. Leibowitz, J. Yu, R. E. Brand, D. J. Hartman, and Y. Liu, “Early prediction of cancer progression by depth-resolved nanoscale mapping of nuclear architecture from unstained tissue specimens,” Cancer Res. 75(22), 4718–4727 (2015). [CrossRef]
17. M. Takabayashi, H. Majeed, A. Kajdacsy-Balla, and G. Popescu, “Disorder strength measured by quantitative phase imaging as intrinsic cancer marker in fixed tissue biopsies,” PLoS One 13(3), e0194320 (2018). [CrossRef]
18. Z. Wang, G. Popescu, K. Tangella, and A. Balla, “Tissue refractive index as marker of disease,” J. Biomed. Opt. 16(11), 1 (2011). [CrossRef]
19. H. Majeed, T. H. Nguyen, M. E. Kandel, A. Kajdacsy-Balla, and G. Popescu, “Label-free quantitative evaluation of breast tissue using Spatial Light Interference Microscopy (SLIM),” Sci. Rep. 8(1), 6875 (2018). [CrossRef]
20. P. Wang, R. Bista, R. Bhargava, R. E. Brand, and Y. Liu, “Spatial-domain low-coherence quantitative phase microscopy for cancer diagnosis,” Opt. Lett. 35(17), 2840–2842 (2010). [CrossRef]
21. M. Lee, E. Lee, J. Jung, H. Yu, K. Kim, J. Yoon, S. Lee, Y. Jeong, and Y. Park, “Label-free optical quantification of structural alterations in Alzheimer’s disease,” Sci. Rep. 6(1), 31034 (2016). [CrossRef]
22. A. Bokemeyer, P. R. Tepasse, L. Quill, P. Lenz, E. Rijcken, M. Vieth, N. Ding, S. Ketelhut, F. Rieder, B. Kemper, and D. Bettenworth, “Quantitative Phase Imaging Using Digital Holographic Microscopy Reliably Assesses Morphology and Reflects Elastic Properties of Fibrotic Intestinal Tissue,” Sci. Rep. 9(1), 19388 (2019). [CrossRef]
23. C. Hu, M. Santi, O. Adelaja, A. Kajdacsy-Balla, G. Popescu, and W. Kobak, “Imaging Collagen Properties in the Uterosacral Ligaments of Women With Pelvic Organ Prolapse Using Spatial Light Interference Microscopy (SLIM),” Front. Phys. 7, 72 (2019). [CrossRef]
24. W. Luo, A. Greenbaum, Y. Zhang, and A. Ozcan, “Synthetic aperture-based on-chip microscopy,” Light: Sci. Appl. 4(3), e261 (2015). [CrossRef]
25. A. Greenbaum, Y. Zhang, A. Feizi, P.-L. Chung, W. Luo, S. R. Kandukuri, and A. Ozcan, “Wide-field computational imaging of pathology slides using lens-free on-chip microscopy,” Sci. Transl. Med. 6(267), 267ra175 (2014). [CrossRef]
26. H. Hugonnet, Y. W. Kim, M. Lee, S. Shin, R. Hruban, S.-M. Hong, and Y. Park, “Multiscale label-free volumetric holographic histopathology of thick-tissue slides with subcellular resolution,” Adv. Photonics 3(2), 026004 (2021). [CrossRef]
27. C. Zuo, J. Li, J. Sun, Y. Fan, J. Zhang, L. Lu, R. Zhang, B. Wang, L. Huang, and Q. Chen, “Transport of intensity equation: a tutorial,” Opt. Lasers Eng. 135, 106187 (2020). [CrossRef]
28. M. Chen, L. Tian, and L. Waller, “3D differential phase contrast microscopy,” Biomed. Opt. Express 7(10), 3940–3950 (2016). [CrossRef]
29. C. Zuo, J. Sun, J. Li, A. Asundi, and Q. Chen, “Wide-field high-resolution 3D microscopy with Fourier ptychographic diffraction tomography,” Opt. Lasers Eng. 128, 106003 (2020). [CrossRef]
30. H. Hugonnet, M. Lee, and Y. Park, “Optimizing illumination in three-dimensional deconvolution microscopy for accurate refractive index tomography,” Opt. Express 29(5), 6293–6301 (2021). [CrossRef]
31. S. Ban, E. Min, Y. Ahn, G. Popescu, and W. Jung, “Effect of tissue staining in quantitative phase imaging,” Journal of Biophotonics 11, e201700402 (2018).
32. V. Lauer, “New approach to optical diffraction tomography yielding a vector equation of diffraction tomography and a novel tomographic microscope,” J. Microsc. 205(2), 165–176 (2002). [CrossRef]
33. S. Preibisch, S. Saalfeld, and P. Tomancak, “Globally optimal stitching of tiled 3D microscopic image acquisitions,” Bioinformatics 25(11), 1463–1465 (2009). [CrossRef]
34. S. Berg, D. Kutra, T. Kroeger, C. N. Straehle, B. X. Kausler, C. Haubold, M. Schiegg, J. Ales, T. Beier, M. Rudy, K. Eren, J. I. Cervantes, B. Xu, F. Beuttenmueller, A. Wolny, C. Zhang, U. Koethe, F. A. Hamprecht, and A. Kreshuk, “ilastik: interactive machine learning for (bio)image analysis,” Nat. Methods 16(12), 1226–1232 (2019). [CrossRef]
35. S. Graham, Q. D. Vu, S. E. A. Raza, A. Azam, Y. W. Tsang, J. T. Kwak, and N. Rajpoot, “Hover-Net: Simultaneous segmentation and classification of nuclei in multi-tissue histology images,” Med. Image Anal. 58, 101563 (2019). [CrossRef]
36. G. Sauter, F. Algaba, M. Amin, C. Busch, J. Cheville, T. Gasser, and D. Grignon, “Tumours of the urinary system,” World Health Organization classification of tumours pathology and genetics: tumours of the urinary system and male genital organs 89–157 (2004).
37. I. Orsolic, D. Jurada, N. Pullen, M. Oren, A. G. Eliopoulos, and S. Volarevic, “The relationship between the nucleolus and cancer: Current evidence and emerging paradigms,” Semin. Cancer Biol. 37-38, 36–50 (2016). [CrossRef]
38. D. Stępiński, “The nucleolus, an ally, and an enemy of cancer cells,” Histochem. Cell Biol. 150(6), 607–629 (2018). [CrossRef]
39. K. Kim, W. S. Park, S. Na, S. Kim, T. Kim, W. Do Heo, and Y. Park, “Correlative three-dimensional fluorescence and refractive index tomography: bridging the gap between molecular specificity and quantitative bioimaging,” Biomed. Opt. Express 8(12), 5688–5697 (2017). [CrossRef]
40. P. Popławski, J. Bogusławska, K. Hanusek, and A. Piekiełko-Witkowska, “Nucleolar Proteins and Non-Coding RNAs: Roles in Renal Cancer,” Int. J. Mol. Sci. 22(23), 13126 (2021). [CrossRef]
41. M. Penzo, L. Montanaro, D. Treré, and M. Derenzini, “The Ribosome Biogenesis—Cancer Connection,” Cells 8(1), 55 (2019). [CrossRef]
42. H. Hugonnet, M. J. Lee, and Y. K. Park, “Quantitative phase and refractive index imaging of 3D objects via optical transfer function reshaping,” Opt. Express 30(8), 13802–13809 (2022). [CrossRef]
43. A. S.-Y. Leong and J. Wright, “The contribution of immunohistochemical staining in tumour diagnosis,” Histopathology 11, 1295–1305 (1987). [CrossRef]
