Real-time polarization microscopy of fibrillar collagen in histopathology.

Over the past two decades, fibrillar collagen reorganization parameters such as the amount of collagen deposition, fiber angle and alignment have been widely explored in numerous studies. These parameters are now widely accepted as stromal biomarkers and linked to disease progression and survival time in several cancer types. Despite all these advances, there has not been a significant effort to make it possible for clinicians to explore these biomarkers without adding steps to the clinical workflow or by requiring high-cost imaging systems. In this paper, we evaluate previously described polychromatic polarization microscope (PPM) to visualize collagen fibers with an optically generated color representation of fiber orientation and alignment when inspecting the sample by a regular microscope with minor modifications. This system does not require stained slides, but is compatible with histological stains such as H&E. Consequently, it can be easily accommodated as part of regular pathology review of tissue slides, while providing clinically useful insight into stromal composition.

Collagen Organization in Relation to Ductal Carcinoma In Situ Pathology and Outcomes.

BACKGROUND: There is widespread interest in discriminating indolent from aggressive ductal carcinoma in situ (DCIS). We sought to evaluate collagen organization in the DCIS tumor microenvironment in relation to pathologic characteristics and patient outcomes. METHODS: We retrieved fixed tissue specimens for 90 DCIS cases within the population-based Vermont DCIS Cohort. We imaged collagen fibers within 75 µm of the tumor/stromal boundary on hematoxylin and eosin-stained slides using multiphoton microscopy with second-harmonic generation. Automated software quantified collagen fiber length, width, straightness, density, alignment, and angle to the tumor/stroma boundary. Factor analysis identified linear combinations of collagen fiber features representing composite attributes of collagen organization. RESULTS: Multiple collagen features were associated with DCIS grade, necrosis pattern, or periductal fibrosis (P < 0.05). After adjusting for treatments and nuclear grade, risk of recurrence (defined as any second breast cancer diagnosis) was lower among cases with greater collagen fiber width [hazard ratio (HR), 0.57 per one standard deviation increase; 95% confidence interval (CI), 0.39-0.84] and fiber density (HR, 0.60; 95% CI, 0.42-0.85), whereas risk was elevated among DCIS cases with higher fiber straightness (HR, 1.47; 95% CI, 1.05-2.06) and distance to the nearest two fibers (HR, 1.47; 95% CI, 1.06-2.02). Fiber length, alignment, and fiber angle were not associated with recurrence (P > 0.05). Five composite factors were identified, accounting for 72.4% of the total variability among fibers; three were inversely associated with recurrence (HRs ranging from 0.60 to 0.67; P ≤ 0.01). CONCLUSIONS: Multiple aspects of collagen organization around DCIS lesions are associated with recurrence risk. IMPACT: Collagen organization should be considered in the development of prognostic DCIS biomarker signatures.

Microstructure and resident cell-types of the feline optic nerve head resemble that of humans.

The lamina cribrosa (LC) region of the optic nerve head (ONH) is considered a primary site for glaucomatous damage. In humans, biology of this region reflects complex interactions between retinal ganglion cell (RGC) axons and other resident ONH cell-types including astrocytes, lamina cribrosa cells, microglia and oligodendrocytes, as well as ONH microvasculature and collagenous LC beams. However, species differences in the microanatomy of this region could profoundly impact efforts to model glaucoma pathobiology in a research setting. In this study, we characterized resident cell-types, ECM composition and ultrastructure in relation to microanatomy of the ONH in adult domestic cats (Felis catus). Longitudinal and transverse cryosections of ONH tissues were immunolabeled with astrocyte, microglia/macrophage, oligodendrocyte, LC cell and vascular endothelial cell markers. Collagen fiber structure of the LC was visualized by second harmonic generation (SHG) with multiphoton microscopy. Fibrous astrocytes form glial fibrillary acidic protein (GFAP)-positive glial columns in the pre-laminar region, and cover the collagenous plates of the LC region in lamellae oriented perpendicular to the axons. GFAP-negative and alpha-smooth muscle actin-positive LC cells were identified in the feline ONH. IBA-1 positive immune cells and von Willebrand factor-positive blood vessel endothelial cells are also identifiable throughout the feline ONH. As in humans, myelination commences with a population of oligodendrocytes in the retro-laminar region of the feline ONH. Transmission electron microscopy confirmed the presence of capillaries and LC cells that extend thin processes in the core of the collagenous LC beams. In conclusion, the feline ONH closely recapitulates the complexity of the ONH of humans and non-human primates, with diverse ONH cell-types and a robust collagenous LC, within the beams of which, LC cells and capillaries reside. Thus, studies in a feline inherited glaucoma model have the potential to play a key role in enhancing our understanding of ONH cellular and molecular processes in glaucomatous optic neuropathy.

Non-disruptive collagen characterization in clinical histopathology using cross-modality image synthesis.

The importance of fibrillar collagen topology and organization in disease progression and prognostication in different types of cancer has been characterized extensively in many research studies. These explorations have either used specialized imaging approaches, such as specific stains (e.g., picrosirius red), or advanced and costly imaging modalities (e.g., second harmonic generation imaging (SHG)) that are not currently in the clinical workflow. To facilitate the analysis of stromal biomarkers in clinical workflows, it would be ideal to have technical approaches that can characterize fibrillar collagen on standard H&E stained slides produced during routine diagnostic work. Here, we present a machine learning-based stromal collagen image synthesis algorithm that can be incorporated into existing H&E-based histopathology workflow. Specifically, this solution applies a convolutional neural network (CNN) directly onto clinically standard H&E bright field images to extract information about collagen fiber arrangement and alignment, without requiring additional specialized imaging stains, systems or equipment.

Fibrillar Collagen Quantification With Curvelet Transform Based Computational Methods.

Quantification of fibrillar collagen organization has given new insight into the possible role of collagen topology in many diseases and has also identified candidate image-based bio-markers in breast cancer and pancreatic cancer. We have been developing collagen quantification tools based on the curvelet transform (CT) algorithm and have demonstrated this to be a powerful multiscale image representation method due to its unique features in collagen image denoising and fiber edge enhancement. In this paper, we present our CT-based collagen quantification software platform with a focus on new features and also giving a detailed description of curvelet-based fiber representation. These new features include C++-based code optimization for fast individual fiber tracking, Java-based synthetic fiber generator module for method validation, automatic tumor boundary generation for fiber relative quantification, parallel computing for large-scale batch mode processing, region-of-interest analysis for user-specified quantification, and pre- and post-processing modules for individual fiber visualization. We present a validation of the tracking of individual fibers and fiber orientations by using synthesized fibers generated by the synthetic fiber generator. In addition, we provide a comparison of the fiber orientation calculation on pancreatic tissue images between our tool and three other quantitative approaches. Lastly, we demonstrate the use of our software tool for the automatic tumor boundary creation and the relative alignment quantification of collagen fibers in human breast cancer pathology images, as well as the alignment quantification of in vivo mouse xenograft breast cancer images.

Quantitative phase imaging of stromal prognostic markers in pancreatic ductal adenocarcinoma.

New quantitative prognostic markers are needed for improved pancreatic ductal adenocarcinoma (PDAC) prognosis. Second harmonic generation microscopy has been used to show that collagen fiber alignment in PDAC is a negative prognostic factor. In this work, a series of PDAC and normal adjacent tissue (NAT) biopsies were imaged with spatial light interference microscopy (SLIM). Quantitative analysis performed on the biopsy SLIM images show that PDAC fiber structures have lower alignment per unit length, narrower width, and are longer than NAT controls. Importantly, fibrillar collagen in PDAC shows an inverse relationship between survival data and fiber width and length (p < 0.05).

Intensity-based registration of bright-field and second-harmonic generation images of histopathology tissue sections.

The use of second-harmonic generation (SHG) microscopy in biomedical research is rapidly increasing. This is due in large part to the wide spread interest of using this imaging technique to examine the role of fibrillar collagen organization in diseases such as cancer. The co-examination of SHG images and traditional bright-field (BF) images of hematoxylin and eosin (H&E) stained tissue as a gold standard clinical validation is usually required. However, image registration of these two modalities has been mostly done by manually selecting corresponding landmarks which is labor intensive and error prone. We designed, implemented, and validated the first image intensity-based registration method capable of automatically aligning SHG images and BF images. In our algorithmic approach, a feature extractor is used to pre-process the BF image to block the content features not visible in SHG images and the output image is then aligned with the SHG image by maximizing the common image features. An alignment matrix maximizing the image mutual information is found by evolutionary optimization and the optimization is facilitated using a hierarchical multiresolution framework. The automatic registration results were compared to traditional manual registration to assess the performance of the algorithm. The proposed algorithm has been successfully used in several biomedical studies such as pancreatic and kidney cancer studies and shown great efficacy.

Super-resolution recurrent convolutional neural networks for learning with multi-resolution whole slide images.

We study a problem scenario of super-resolution (SR) algorithms in the context of whole slide imaging (WSI), a popular imaging modality in digital pathology. Instead of just one pair of high- and low-resolution images, which is typically the setup in which SR algorithms are designed, we are given multiple intermediate resolutions of the same image as well. The question remains how to best utilize such data to make the transformation learning problem inherent to SR more tractable and address the unique challenges that arises in this biomedical application. We propose a recurrent convolutional neural network model, to generate SR images from such multi-resolution WSI datasets. Specifically, we show that having such intermediate resolutions is highly effective in making the learning problem easily trainable and address large resolution difference in the low and high-resolution images common in WSI, even without the availability of a large size training data. Experimental results show state-of-the-art performance on three WSI histopathology cancer datasets, across a number of metrics.

Quantitative Histopathology of Stained Tissues using Color Spatial Light Interference Microscopy (cSLIM).

Tissue biopsy evaluation in the clinic is in need of quantitative disease markers for diagnosis and, most importantly, prognosis. Among the new technologies, quantitative phase imaging (QPI) has demonstrated promise for histopathology because it reveals intrinsic tissue nanoarchitecture through the refractive index. However, a vast majority of past QPI investigations have relied on imaging unstained tissues, which disrupts the established specimen processing. Here we present color spatial light interference microscopy (cSLIM) as a new whole-slide imaging modality that performs interferometric imaging on stained tissue, with a color detector array. As a result, cSLIM yields in a single scan both the intrinsic tissue phase map and the standard color bright-field image, familiar to the pathologist. Our results on 196 breast cancer patients indicate that cSLIM can provide stain-independent prognostic information from the alignment of collagen fibers in the tumor microenvironment. The effects of staining on the tissue phase maps were corrected by a mathematical normalization. These characteristics are likely to reduce barriers to clinical translation for the new cSLIM technology.

A multiscale Mueller polarimetry module for a stereo zoom microscope.

Mueller polarimetry is a quantitative polarized light imaging modality that is capable of label-free visualization of tissue pathology, does not require extensive sample preparation, and is suitable for wide-field tissue analysis. It holds promise for selected applications in biomedicine, but polarimetry systems are often constrained by limited end-user accessibility and/or long-imaging times. In order to address these needs, we designed a multiscale-polarimetry module that easily couples to a commercially available stereo zoom microscope. This paper describes the module design and provides initial polarimetry imaging results from a murine preclinical breast cancer model and human breast cancer samples. The resultant polarimetry module has variable resolution and field of view, is low-cost, and is simple to switch in or out of a commercial microscope. The module can reduce long imaging times by adopting the main imaging approach used in pathology: scanning at low resolution to identify regions of interest, then at high resolution to inspect the regions in detail. Preliminary results show how the system can aid in region of interest identification for pathology, but also highlight that more work is needed to understand how tissue structures of pathological interest appear in Mueller polarimetry images across varying spatial zoom scales.

Collagen organization of renal cell carcinoma differs between low and high grade tumors.

BACKGROUND: The traditional pathologic grading for human renal cell carcinoma (RCC) has low concordance between biopsy and surgical specimen. There is a need to investigate adjunctive pathology technique that does not rely on the nuclear morphology that defines the traditional grading. Changes in collagen organization in the extracellular matrix have been linked to prognosis or grade in breast, ovarian, and pancreatic cancers, but collagen organization has never been correlated with RCC grade. In this study, we used Second Harmonic Generation (SHG) based imaging to quantify possible differences in collagen organization between high and low grades of human RCC. METHODS: A tissue microarray (TMA) was constructed from RCC tumor specimens. Each TMA core represents an individual patient. A 5 µm section from the TMA tissue was stained with standard hematoxylin and eosin (H&E). Bright field images of the H&E stained TMA were used to annotate representative RCC regions. In this study, 70 grade 1 cores and 51 grade 4 cores were imaged on a custom-built forward SHG microscope, and images were analyzed using established software tools to automatically extract and quantify collagen fibers for alignment and density assessment. A linear mixed-effects model with random intercepts to account for the within-patient correlation was created to compare grade 1 vs. grade 4 measurements and the statistical tests were two-sided. RESULTS: Both collagen density and alignment differed significantly between RCC grade 1 and RCC grade 4. Specifically, collagen fiber density was greater in grade 4 than in grade 1 RCC (p < 0.001). Collagen fibers were also more aligned in grade 4 compared to grade 1 (p < 0.001). CONCLUSIONS: Collagen density and alignment were shown to be significantly higher in RCC grade 4 vs. grade 1. This technique of biopsy sampling by SHG could complement classical tumor grading approaches. Furthermore it might allow biopsies to be more clinically relevant by informing diagnostics. Future studies are required to investigate the functional role of collagen organization in RCC.

Convolutional neural networks for whole slide image superresolution.

We present a computational approach for improving the quality of the resolution of images acquired from commonly available low magnification commercial slide scanners. Images from such scanners can be acquired cheaply and are efficient in terms of storage and data transfer. However, they are generally of poorer quality than images from high-resolution scanners and microscopes and do not have the necessary resolution needed in diagnostic or clinical environments, and hence are not used in such settings. The driving question of this presented research is whether the resolution of these images could be enhanced such that it would serve the same diagnostic purpose as high-resolution images from expensive scanners or microscopes. This need is generally known as the image super-resolution (SR) problem in image processing, and it has been studied extensively. Even so, none of the existing methods directly work for the slide scanner images, due to the unique challenges posed by this modality. Here, we propose a convolutional neural network (CNN) based approach, which is specifically trained to take low-resolution slide scanner images of cancer data and convert it into a high-resolution image. We validate these resolution improvements with computational analysis to show the enhanced images offer the same quantitative results. In summary, our extensive experiments demonstrate that this method indeed produces images that are similar to images from high-resolution scanners, both in quality and quantitative measures. This approach opens up new application possibilities for using low-resolution scanners, not only in terms of cost but also in access and speed of scanning for both research and possible clinical use.

Methods for Quantifying Fibrillar Collagen Alignment.

Recent evidence has implicated collagen, particularly fibrillar collagen, in a number of diseases ranging from osteogenesis imperfecta and asthma to breast and ovarian cancer. A key property of collagen that has been correlated with disease has been the alignment of collagen fibers. Collagen can be visualized using a variety of imaging techniques including second-harmonic generation (SHG) microscopy, polarized light microscopy, and staining with dyes or antibodies. However, there exists a great need to easily and robustly quantify images from these modalities for individual fibers in specified regions of interest and with respect to relevant boundaries. Most currently available computational tools rely on calculation of pixel-wise orientation or global window-wise orientation that do not directly calculate or give visible fiber-wise information and do not provide relative orientation against boundaries. We describe and detail how to use a freely available, open-source MATLAB software framework that includes two separate but linked packages "CurveAlign" and "CT-FIRE" that can address this need by either directly extracting individual fibers using an improved fiber tracking algorithm or directly finding optimal representation of fiber edges using the curvelet transform. This curvelet-based framework allows the user to measure fiber alignment on a global, region of interest, and fiber basis. Additionally, users can measure fiber angle relative to manually or automatically segmented boundaries. This tool does not require prior experience of programming or image processing and can handle multiple files, enabling efficient quantification of collagen organization from biological datasets.

Fluorescence of Picrosirius Red Multiplexed With Immunohistochemistry for the Quantitative Assessment of Collagen in Tissue Sections.

The low cost and simplicity of picrosirius red (PSR) staining have driven its popularity for collagen detection in tissue sections. We extended the versatility of this method by using fluorescent imaging to detect the PSR signal and applying automated quantification tools. We also developed the first PSR protocol that is fully compatible with multiplex immunostaining, making it possible to test whether collagen structure differs across immunohistochemically labeled regions of the tissue landscape. We compared our imaging method with two gold standards in collagen imaging, linear polarized light microscopy and second harmonic generation imaging, and found that it is at least as sensitive and robust to changes in sample orientation. As proof of principle, we used a genetic approach to overexpress beta catenin in a patchy subset of mouse prostate epithelial cells distinguished only by immunolabeling. We showed that collagen fiber length is significantly greater near beta catenin overexpressing cells than near control cells. Our fluorescent PSR imaging method is sensitive, reproducible, and offers a new way to guide region of interest selection for quantifying collagen in tissue sections.

Quantification of Collagen Organization after Nerve Repair.

BACKGROUND: Clinical outcomes after nerve injury and repair remain suboptimal. Patients may be plagued by poor functional recovery and painful neuroma at the repair site, characterized by disorganized collagen and sprouting axons. Collagen deposition during wound healing can be intrinsically imaged using second harmonic generation (SHG) microscopy. The purpose of this study was to develop a protocol for SHG imaging of nerves and to assess whether collagen alignment can be quantified after nerve repair. METHODS: Sciatic nerve transection and epineural repair was performed in male rats. The contralateral nerves were used as intra-animal controls. Ten-millimeter nerve segments were harvested and fixed onto slides. SHG images were collected using a 20× objective on a multiphoton microscope. Collagen fiber alignment was calculated using CurveAlign software. Alignment was calculated on a scale from 0 to 1, where 1 represents perfect alignment. Statistical analysis was performed using a linear mixed-effects model. RESULTS: Eight male rats underwent right sciatic nerve repair using 9-0 Nylon suture. There were gross variations in collagen fiber organization in the repaired nerves compared with the controls. Quantitatively, collagen fibers were more aligned in the control nerves (mean alignment 0.754, SE 0.055) than in the repairs (mean alignment 0.413, SE 0.047; P < 0.001). CONCLUSIONS: SHG microscopy can be used to quantitate collagen after nerve repair via fiber alignment. Given that the development of neuroma likely reflects aberrant wound healing, ex vivo and/or in vivo SHG imaging may be useful for further investigation of the variables predisposing to neuroma.

Highly aligned stromal collagen is a negative prognostic factor following pancreatic ductal adenocarcinoma resection.

Risk factors for pancreatic ductal adenocarcinoma (PDAC) progression after surgery are unclear, and additional prognostic factors are needed to inform treatment regimens and therapeutic targets. PDAC is characterized by advanced sclerosis of the extracellular matrix, and interactions between cancer cells, fibrillar collagen, and other stromal components play an integral role in progression. Changes in stromal collagen alignment have been shown to modulate cancer cell behavior and have important clinical value in other cancer types, but little is known about its role in PDAC and prognostic value. We hypothesized that the alignment of collagen is associated with PDAC patient survival. To address this, pathology-confirmed tissues from 114 PDAC patients that underwent curative-intent surgery were retrospectively imaged with Second Harmonic Generation (SHG) microscopy, quantified with fiber segmentation algorithms, and correlated to patient survival. The same tissue regions were analyzed for epithelial-to-mesenchymal (EMT), α-SMA, and syndecan-1 using complimentary immunohistostaining and visualization techniques. Significant inter-tumoral variation in collagen alignment was found, and notably high collagen alignment was observed in 12% of the patient cohort. Stratification of patients according to collagen alignment revealed that high alignment is an independent negative factor following PDAC resection (p = 0.0153, multivariate). We also found that epithelial expression of EMT and the stromal expression of α-SMA and syndecan-1 were positively correlated with collagen alignment. In summary, stromal collagen alignment may provide additional, clinically-relevant information about PDAC tumors and underscores the importance of stroma-cancer interactions.

Comparison of Picrosirius Red Staining With Second Harmonic Generation Imaging for the Quantification of Clinically Relevant Collagen Fiber Features in Histopathology Samples.

Stromal collagen alignment has been shown to have clinical significance in a variety of cancers and in other diseases accompanied by fibrosis. While much of the biological and clinical importance of collagen changes has been demonstrated using second harmonic generation (SHG) imaging in experimental settings, implementation into routine clinical pathology practice is currently prohibitive. To translate the assessment of collagen organization into routine pathology workflow, a surrogate visualization method needs to be examined. The objective of the present study was to quantitatively compare collagen metrics generated from SHG microscopy and commonly available picrosirius red stain with standard polarization microscopy (PSR-POL). Each technique was quantitatively compared with established image segmentation and fiber tracking algorithms using human pancreatic cancer as a model, which is characterized by a pronounced stroma with reorganized collagen fibers. Importantly, PSR-POL produced similar quantitative trends for most collagen metrics in benign and cancerous tissues as measured by SHG. We found it notable that PSR-POL detects higher fiber counts, alignment, length, straightness, and width compared with SHG imaging but still correlates well with SHG results. PSR-POL may provide sufficient and additional information in a conventional clinical pathology laboratory for certain types of collagen quantification.

Second-harmonic generation imaging of cancer.

The last 30 years has seen great advances in optical microscopy with the introduction of sophisticated fluorescence-based imaging methods such as confocal and multiphoton laser scanning microscopy. There is increasing interest in using these methods to quantitatively examine sources of intrinsic biological contrast including autofluorescent endogenous proteins and light interactions such as second-harmonic generation (SHG) in collagen. In particular, SHG-based microscopy has become a widely used quantitative modality for imaging noncentrosymmetric proteins such as collagen in a diverse range of tissues. Due to the underlying physical origin of the SHG signal, it is highly sensitive to collagen fibril/fiber structure and, importantly, to collagen-associated changes that occur in diseases such as cancer, fibrosis, and connective tissue disorders. An overview of SHG physics background and technologies is presented with a focused review on applications of SHG primarily as applied to cancer.