Clinicopathological and immunohistochemical study of 29 cases of solid-pseudopapillary neoplasms of the pancreas in patients under 20 years of age along with detailed review of literature

Background

Multiplex immunohistochemistry (mIHC) permits the labeling of six or more distinct cell types within a single histologic tissue section. The classification of each cell type requires detection of uniquely colored chromogens localized to cells expressing biomarkers of interest. The most comprehensive and reproducible method to evaluate such slides is to employ digital pathology and image analysis pipelines to whole-slide images (WSIs). Our suite of deep learning tools quantitatively evaluates the expression of six biomarkers in mIHC WSIs. These methods address the current lack of readily available methods to evaluate more than four biomarkers and circumvent the need for specialized instrumentation to spectrally separate different colors. The use case application for our methods is a study that investigates tumor immune interactions in pancreatic ductal adenocarcinoma (PDAC) with a customized mIHC panel.

Methods

Six different colored chromogens were utilized to label T-cells (CD3, CD4, CD8), B-cells (CD20), macrophages (CD16), and tumor cells (K17) in formalin-fixed paraffin-embedded (FFPE) PDAC tissue sections. We leveraged pathologist annotations to develop complementary deep learning-based methods: (1) ColorAE is a deep autoencoder which segments stained objects based on color; (2) U-Net is a convolutional neural network (CNN) trained to segment cells based on color, texture and shape; and (3) ensemble methods that employ both ColorAE and U-Net, collectively referred to as ColorAE:U-Net. We assessed the performance of our methods using: structural similarity and DICE score to evaluate segmentation results of ColorAE against traditional color deconvolution; F1 score, sensitivity, positive predictive value, and DICE score to evaluate the predictions from ColorAE, U-Net, and ColorAE:U-Net ensemble methods against pathologist-generated ground truth. We then used prediction results for spatial analysis (nearest neighbor).

Results

We observed that (1) the performance of ColorAE is comparable to traditional color deconvolution for single-stain IHC images (note: traditional color deconvolution cannot be used for mIHC); (2) ColorAE and U-Net are complementary methods that detect six different classes of cells with comparable performance; (3) combinations of ColorAE and U-Net in ensemble methods outperform ColorAE and U-Net alone; and (4) ColorAE:U-Net ensemble methods can be employed for detailed analysis of the tumor microenvironment (TME).

Summary

We developed a suite of scalable deep learning methods to analyze 6 distinctly labeled cell populations in mIHC WSIs. We evaluated our methods and found that they reliably detected and classified cells in the PDAC tumor microenvironment. We also utilized the ColorAE:U-Net ensemble method to analyze 3 mIHC WSIs with nearest neighbor spatial analysis. We demonstrate a proof of concept that these methods can be employed to quantitatively describe the spatial distribution of immune cells within the tumor microenvironment. These complementary deep learning methods are readily deployable for use in clinical research studies.

     

Limitations on cell volume determination using three-dimensional optical microscopy

One important application for three-dimensional microscopy is the determination of the volumes of biological cells. In this paper we investigate the application of three-dimensional deconvolution techniques to the problem of cell volume determination. The theoretical discussion is based on a Fourier sampling condition for reliable image reconstruction and it is shown, using this condition and simulated images, that the cell volume determination is reliable only for cells with a shape and orientation such that their length along the optic axis is much greater than their width.     

A novel voltage clamp technique for mapping ionic currents from cultured skeletal myotubes.

The biophysical properties and cellular distribution of ion channels largely determine the input/output relationships of electrically excitable cells. A variety of patch pipette voltage clamp techniques are available to characterize ionic currents. However, when used by themselves, such techniques are not well suited to the task of mapping low-density channel distributions. We describe here a new voltage clamp method (the whole cell loose patch (WCLP) method) that combines whole-cell recording through a tight-seal pipette with focal extracellular stimulation through a loose-seal pipette. By moving the stimulation pipette across the cell surface and using a stationary whole-cell pipette to record the evoked patch currents, this method should be suitable for mapping channel distributions, even on large cells possessing low channel densities. When we applied this method to the study of currents in cultured chick myotubes, we found that the cell cable properties and the series resistance of the recording pipette caused significant filtering of the membrane currents, and that the filter characteristics depended in part upon the distance between the stimulating and recording pipettes. We describe here how we determined the filter impulse response for each loose-seal pipette placement and subsequently recovered accurate estimates of patch membrane current through deconvolution.     

A comparison of deconvolution techniques for stress relaxation

Stress relaxation (or equivalently creep) allows a large range of the relaxation (retardation) spectrum of materials to be examined, particularly at lower frequencies. However, higher frequency components of the relaxation curves (typically of the order of Hertz) are attenuated due to the finite time taken to strain the specimen. This higher frequency information can be recovered by deconvolution of the stress and strain during the loading period. This paper examines the use of three separate deconvolution techniques: numerical (Fourier) deconvolution, semi-analytical deconvolution using a theoretical form of the strain, and deconvolution by a linear approximation method. Both theoretical data (where the exact form of the relaxation function is known) and experimental data were used to assess the accuracy and applicability of the deconvolution methods. All of the deconvolution techniques produced a consistent improvement in the higher frequency data up to the frequencies of the order of Hertz, with the linear approximation method showing better resolution in high-frequency analysis of the theoretical data. When the different deconvolution techniques were applied to experimental data, similar results were found for all three deconvolution techniques. Deconvolution of the stress and strain during loading is a simple and practical method for the recovery of higher frequency data from stress-relaxation experiments.     

Image restoration for confocal microscopy: improving the limits of deconvolution, with application to the visualization of the mammalian hearing organ

Deconvolution algorithms have proven very effective in conventional (wide-field) fluorescence microscopy. Their application to confocal microscopy is hampered, in biological experiments, by the presence of important levels of noise in the images and by the lack of a precise knowledge of the point spread function (PSF) of the system. We investigate the application of wavelet-based processing tools to deal with these problems, in particular wavelet denoising methods, which turn out to be very effective in application to three-dimensional confocal images. When used in combination with more classical deconvolution algorithms, these methods provide a robust and efficient restoration scheme allowing one to deal with difficult imaging conditions. To make our approach applicable in practical situations, we measured the PSF of a Biorad-MRC1024 confocal microscope under a large set of imaging conditions, including in situ acquisitions. As a specific biological application, we present several examples of restorations of three-dimensional confocal images acquired inside an intact preparation of the hearing organ. We also provide a quantitative assessment of the gain in quality achieved by wavelet-aided restorations over classical deconvolution schemes, based on a set of numerical experiments that we performed with test images.     

PERT: A Method for Expression Deconvolution of Human Blood Samples from Varied Microenvironmental and Developmental Conditions

The cellular composition of heterogeneous samples can be predicted using an expression deconvolution algorithm to decompose their gene expression profiles based on pre-defined, reference gene expression profiles of the constituent populations in these samples. However, the expression profiles of the actual constituent populations are often perturbed from those of the reference profiles due to gene expression changes in cells associated with microenvironmental or developmental effects. Existing deconvolution algorithms do not account for these changes and give incorrect results when benchmarked against those measured by well-established flow cytometry, even after batch correction was applied. We introduce PERT, a new probabilistic expression deconvolution method that detects and accounts for a shared, multiplicative perturbation in the reference profiles when performing expression deconvolution. We applied PERT and three other state-of-the-art expression deconvolution methods to predict cell frequencies within heterogeneous human blood samples that were collected under several conditions (uncultured mono-nucleated and lineage-depleted cells, and culture-derived lineage-depleted cells). Only PERT''s predicted proportions of the constituent populations matched those assigned by flow cytometry. Genes associated with cell cycle processes were highly enriched among those with the largest predicted expression changes between the cultured and uncultured conditions. We anticipate that PERT will be widely applicable to expression deconvolution strategies that use profiles from reference populations that vary from the corresponding constituent populations in cellular state but not cellular phenotypic identity.     

Optimizing deconvolution techniques by the application of the Münchhausen meta algorithm.

A deconvolution applied to disturbed data often gives poor results, due to fundamental difficulties associated with ill-posed problems. Many numerical and theoretical methods have been invented to circumvent this phenomenon. Their performance varies, depending on the given problem and data. The main aim of this paper is to provide a decision rule for choosing a method for deconvolution and application of this method to the same data. We have called this meta-algorithm Münchhausen. In this paper we introduce and describe for the first time the basic principle of artificial disturbance of the data in the set-up of deconvolution. We demonstrate some interesting features of the random procedure Münchhausen, such as the non parametric set-up, robustness to disturbance of the data and last but not least good performance.     

Generation of Subcutaneous and Intrahepatic Human Hepatocellular Carcinoma Xenografts in Immunodeficient Mice

In vivo experimental models of hepatocellular carcinoma (HCC) that recapitulate the human disease provide a valuable platform for research into disease pathophysiology and for the preclinical evaluation of novel therapies. We present a variety of methods to generate subcutaneous or orthotopic human HCC xenografts in immunodeficient mice that could be utilized in a variety of research applications. With a focus on the use of primary tumor tissue from patients undergoing surgical resection as a starting point, we describe the preparation of cell suspensions or tumor fragments for xenografting. We describe specific techniques to xenograft these tissues i) subcutaneously; or ii) intrahepatically, either by direct implantation of tumor cells or fragments into the liver, or indirectly by injection of cells into the mouse spleen. We also describe the use of partial resection of the native mouse liver at the time of xenografting as a strategy to induce a state of active liver regeneration in the recipient mouse that may facilitate the intrahepatic engraftment of primary human tumor cells. The expected results of these techniques are illustrated. The protocols described have been validated using primary human HCC samples and xenografts, which typically perform less robustly than the well-established human HCC cell lines that are widely used and frequently cited in the literature. In comparison with cell lines, we discuss factors which may contribute to the relatively low chance of primary HCC engraftment in xenotransplantation models and comment on technical issues that may influence the kinetics of xenograft growth. We also suggest methods that should be applied to ensure that xenografts obtained accurately resemble parent HCC tissues.     

Estimation of growth hormone secretion rate: impact of kinetic assumptions intrinsic to the analytical approach

We compared four common mathematical techniques to determine daily endogenous growth hormone (GH) secretion rates from diurnal plasma GH concentration profiles in 24 women (16 upper- or lower-body obese and 8 normal-weight individuals). Two forms of deconvolution analysis and two techniques based on a priori determined GH clearance estimates were employed. Deconvolution analyses revealed significant differences in the 24-h GH secretion rate between normal-weight and upper-body obese women, whereas the other two techniques did not. Moreover, deconvolution analyses predicted that the reduction in mean plasma GH concentrations in upper-body obese women was accounted for by impaired GH secretion, whereas the other methods suggested that obesity increases GH metabolic clearance. Thus we infer that disparate conclusions concerning GH secretion can be drawn from the same primary data set. The different inferences likely reflect dissimilar kinetic assumptions and the particular limitations intrinsic to each analytical approach. Accordingly, we urge caution in the facile comparison of calculated GH secretion data in humans, especially when kinetic and secretion measurements are performed under different conditions. The most appropriate way to determine the GH secretion rate in humans must be balanced by the exact intent of the experiment and the acceptability of different assumptions in that context.     

Perinatal stem cells: A promising cell resource for tissue engineering of craniofacial bone

In facing the mounting clinical challenge and suboptimal techniques of craniofacial bone defects resulting from various conditions, such as congenital malformations, osteomyelitis, trauma and tumor resection, the ongoing research of regenerative medicine using stem cells and concurrent advancement in biotechnology have shifted the focus from surgical reconstruction to a novel stem cell-based tissue engineering strategy for customized and functional craniofacial bone regeneration. Given the unique ontogenetical and cell biological properties of perinatal stem cells, emerging evidence has suggested these extraembryonic tissue-derived stem cells to be a promising cell source for extensive use in regenerative medicine and tissue engineering. In this review, we summarize the current achievements and obstacles in stem cell-based craniofacial bone regeneration and subsequently we address the characteristics of various types of perinatal stem cells and their novel application in tissue engineering of craniofacial bone. We propose the promising feasibility and scope of perinatal stem cell-based craniofacial bone tissue engineering for future clinical application.     

Exploring cancer stem cell niche directed tumor growth

The finding that only a sub-fraction of tumor cells, so called Cancer Stem Cells (CSC), is endowed with the capacity to initiate new tumors has important consequences for fundamental as well as clinical cancer research. Previously we established by computational modeling techniques that CSC driven tumor growth instigates infiltrative behavior, and perhaps most interesting, stimulates tumor cell heterogeneity. An important question that remains is to what extend CSC functions are intrinsically regulated or whether this capacity is orchestrated by the microenvironment, i.e. a putative CSC niche. Here we investigate how extrinsic regulation of CSC properties affects the characteristics of malignancies. We find that highly invasive growth in tumors dependent on a small subset of cells is not restricted to CSC-driven tumors, but is also observed in tumors where the CSC capacity of tumor cells is completely defined by the microenvironment. Importantly, also the high level of heterogeneity that was observed for CSC-driven tumors is preserved and partially even increased in malignancies with a microenvironmentally orchestrated CSC population. This indicates that invasive growth and high heterogeneity are fundamental properties of tumors fueled by a small population of tumor cells.     

High-definition mapping of neural activity using voltage-sensitive dyes

The distribution of patterns of activity in different brain structures has been related to the encoding and processing of sensory information. Consequently, it is important to be able to image the distribution of these patterns to understand basic brain functions. The spatial resolution of voltage-sensitive dye (VSD) methods has recently been enhanced considerably by the use of video imaging techniques. The main factor that now hampers the resolution of VSD patterns is the inherent limitation of the optical systems. Unfortunately, the intrinsic characteristics of VSD images impose important limitations that restrict the use of general deconvolution techniques. To overcomes this problem, in this study an image restoration procedure has been implemented that takes into consideration the limiting characteristics of VSD signals. This technique is based on applying a set of imaging processing steps. First, the signal-to-noise (S/N) ratio of the images was improved to avoid an increase in the noise levels during the deconvolution procedures. For this purpose, a new filter technique was implemented that yielded better results than other methods currently used in optical imaging. Second, focal plane images were deconvolved using a modification of the well-known nearest-neighbor deconvolution algorithm. But to reduce the light exposure of the preparation and simplify image acquisition procedures, adjacent image planes were modeled according to the in-focus image planes and the empirical point spread function (PSF) profiles. Third, resulting focal plane responses were processed to reduce the contribution of optical responses that originate in distant image planes. This method was found to be satisfactory under simulated and real experimental conditions. By comparing the restored and unprocessed images, it was clearly demonstrated that this method can effectively remove the out-of-focus artifacts and produce focal plane images of better quality. Evaluations of the tissue optical properties allowed assessment of the maximum practical optical section thickness using this deconvolution technique in the optical system tested. Determination of the three-dimensional PSF permitted the correct application of deconvolution algorithms and the removal of the contaminating light arising from adjacent as well as distant optical planes. The implementation of this deconvolution approach in salamander olfactory bulb allowed the detailed study of the laminar distribution of voltage-sensitive changes across the bulb layer. It is concluded that (1) this deconvolution procedure is well suited to deconvolved low-contrast images and offers important advantages over other alternatives; (2) this method can be properly used only when the tissue optical properties are first determined; (3) high levels of light scattering in the tissue reduce the optical section capabilities of this technique as well as other deconvolution procedures; and (4) use of the highest numerical aperture in the objectives is advisable because this improves not only the light-collecting efficiency to detect poor-contrast images, but also the spatial frequency differences between adjacent image planes. Under this condition it is possible to overcome some of the limitations imposed by the light scattering/birefringence of the tissue.     

Deconvolution can be used in electrooptic studies to correct for non-ideal electric excitation pulses only when the electric dipole moment of the studied molecules is predominantly induced

The electric field pulses used for most measurements of transient electrooptic properties such as birefringence and dichroism, are rectangular and assumed to be ideal, but in practice do all such pulses have non-zero rise and fall times. Claims have been made that this non-ideality may be taken into account by employing standard deconvolution techniques. We find that this approach yields exact results in the zero electric field limit when the electric dipole moment of the studied macromolecules is predominantly induced. However, for finite electric field strengths and/or macromolecules with partly or fully permanent electric dipole moments, we find that the deconvolution method yields erroneous estimates of the electrooptic relaxation times. When the decay time of the electric pulse and the electrooptic decay time are equal, and the system is operated in the Kerr domain, this systematic error for macromolecules with purely permanent electric dipole moment equals 37%. In a research field where the uncertainty of the reported relaxation times normally is assumed to be only a few percent this is an error that may seriously mislead unsuspecting users. We find that this systematic error can readily be avoided by employing standard numerical integration of a set of coupled first-order differential equations instead of the standard deconvolution techniques.     

Medoidshift clustering applied to genomic bulk tumor data

Despite the enormous medical impact of cancers and intensive study of their biology, detailed characterization of tumor growth and development remains elusive. This difficulty occurs in large part because of enormous heterogeneity in the molecular mechanisms of cancer progression, both tumor-to-tumor and cell-to-cell in single tumors. Advances in genomic technologies, especially at the single-cell level, are improving the situation, but these approaches are held back by limitations of the biotechnologies for gathering genomic data from heterogeneous cell populations and the computational methods for making sense of those data. One popular way to gain the advantages of whole-genome methods without the cost of single-cell genomics has been the use of computational deconvolution (unmixing) methods to reconstruct clonal heterogeneity from bulk genomic data. These methods, too, are limited by the difficulty of inferring genomic profiles of rare or subtly varying clonal subpopulations from bulk data, a problem that can be computationally reduced to that of reconstructing the geometry of point clouds of tumor samples in a genome space. Here, we present a new method to improve that reconstruction by better identifying subspaces corresponding to tumors produced from mixtures of distinct combinations of clonal subpopulations. We develop a nonparametric clustering method based on medoidshift clustering for identifying subgroups of tumors expected to correspond to distinct trajectories of evolutionary progression. We show on synthetic and real tumor copy-number data that this new method substantially improves our ability to resolve discrete tumor subgroups, a key step in the process of accurately deconvolving tumor genomic data and inferring clonal heterogeneity from bulk data.

     

A new deep learning method for image deblurring in optical microscopic systems

Deconvolution is the most commonly used image processing method in optical imaging systems to remove the blur caused by the point‐spread function (PSF). While this method has been successful in deblurring, it suffers from several disadvantages, such as slow processing time due to multiple iterations required to deblur and suboptimal in cases where the experimental operator chosen to represent PSF is not optimal. In this paper, we present a deep‐learning‐based deblurring method that is fast and applicable to optical microscopic imaging systems. We tested the robustness of proposed deblurring method on the publicly available data, simulated data and experimental data (including 2D optical microscopic data and 3D photoacoustic microscopic data), which all showed much improved deblurred results compared to deconvolution. We compared our results against several existing deconvolution methods. Our results are better than conventional techniques and do not require multiple iterations or pre‐determined experimental operator. Our method has several advantages including simple operation, short time to compute, good deblur results and wide application in all types of optical microscopic imaging systems. The deep learning approach opens up a new path for deblurring and can be applied in various biomedical imaging fields.     

Visualizing Escherichia coli sub-cellular structure using sparse deconvolution Spatial Light Interference Tomography

Studying the 3D sub-cellular structure of living cells is essential to our understanding of biological function. However, tomographic imaging of live cells is challenging mainly because they are transparent, i.e., weakly scattering structures. Therefore, this type of imaging has been implemented largely using fluorescence techniques. While confocal fluorescence imaging is a common approach to achieve sectioning, it requires fluorescence probes that are often harmful to the living specimen. On the other hand, by using the intrinsic contrast of the structures it is possible to study living cells in a non-invasive manner. One method that provides high-resolution quantitative information about nanoscale structures is a broadband interferometric technique known as Spatial Light Interference Microscopy (SLIM). In addition to rendering quantitative phase information, when combined with a high numerical aperture objective, SLIM also provides excellent depth sectioning capabilities. However, like in all linear optical systems, SLIM's resolution is limited by diffraction. Here we present a novel 3D field deconvolution algorithm that exploits the sparsity of phase images and renders images with resolution beyond the diffraction limit. We employ this label-free method, called deconvolution Spatial Light Interference Tomography (dSLIT), to visualize coiled sub-cellular structures in E. coli cells which are most likely the cytoskeletal MreB protein and the division site regulating MinCDE proteins. Previously these structures have only been observed using specialized strains and plasmids and fluorescence techniques. Our results indicate that dSLIT can be employed to study such structures in a practical and non-invasive manner.     

Observer-based techniques for the identification and analysis of avascular tumor growth

Cancer represents one of the most challenging issues for the biomedical research, due its large impact on the public health state. For this reason, many mathematical methods have been proposed to forecast the time evolution of cancer size and invasion. In this paper, we study how to apply the Gompertz’s model to describe the growth of an avascular tumor in a realistic setting. To this aim, we introduce mathematical techniques to discretize the model, an important requirement when discrete-time measurements are available. Additionally, we describe observed-based techniques, borrowed from the field of automation theory, as a tool to estimate the model unknown parameters. This identification approach is a promising alternative to traditional statistical methods, and it can be easily extended to other models of cancer growth as well as to the evaluation of not measurable variables, on the basis of the available measurements. We show an application of this method to the analysis of solid tumor growth and parameters estimation in presence of a chemotherapy agent.     

Compound analysis via graph kernels incorporating chirality

High accuracy is paramount when predicting biochemical characteristics using Quantitative Structural-Property Relationships (QSPRs). Although existing graph-theoretic kernel methods combined with machine learning techniques are efficient for QSPR model construction, they cannot distinguish topologically identical chiral compounds which often exhibit different biological characteristics. In this paper, we propose a new method that extends the recently developed tree pattern graph kernel to accommodate stereoisomers. We show that Support Vector Regression (SVR) with a chiral graph kernel is useful for target property prediction by demonstrating its application to a set of human vitamin D receptor ligands currently under consideration for their potential anti-cancer effects.     

Comparison of four computerized models to estimate 24-hour growth hormone secretion in girls with Turner's syndrome.

Daily pituitary growth hormone (GH) secretion can be estimated from a 24-hour GH profile by various methods. We have used four methods to assess GH secretion in 36 girls with Turner's syndrome: the method described by Thompson et al., the Pulsar algorithm combined with the method of Hellman et al. and two deconvolution techniques. The number of detected peaks varied considerably among the methods. The mean (+/- SD) total daily secretion per square meter body surface was 0.53 (0.19) U/m2.day by deconvolution, in contrast to 0.31 (0.17) with the Hellman method and 1.06 (0.37) according to Thompson. The differences are explained by different assumptions about the metabolic clearance rate and various methodological aspects. Assuming a degradation rate of 50%, the growth hormone substitution dosage would be 1-2 IU/m2.day in GH-deficient children. The usual dosage in girls with Turner's syndrome is expected to lead to serum GH levels approximately 4 times higher than in the untreated state.     

New algorithm to determine true colocalization in combination with image restoration and time-lapse confocal microscopy to MAP kinases in mitochondria

The subcellular localization and physiological functions of biomolecules are closely related and thus it is crucial to precisely determine the distribution of different molecules inside the intracellular structures. This is frequently accomplished by fluorescence microscopy with well-characterized markers and posterior evaluation of the signal colocalization. Rigorous study of colocalization requires statistical analysis of the data, albeit yet no single technique has been established as a standard method. Indeed, the few methods currently available are only accurate in images with particular characteristics. Here, we introduce a new algorithm to automatically obtain the true colocalization between images that is suitable for a wide variety of biological situations. To proceed, the algorithm contemplates the individual contribution of each pixel's fluorescence intensity in a pair of images to the overall Pearsońs correlation and Manders' overlap coefficients. The accuracy and reliability of the algorithm was validated on both simulated and real images that reflected the characteristics of a range of biological samples. We used this algorithm in combination with image restoration by deconvolution and time-lapse confocal microscopy to address the localization of MEK1 in the mitochondria of different cell lines. Appraising the previously described behavior of Akt1 corroborated the reliability of the combined use of these techniques. Together, the present work provides a novel statistical approach to accurately and reliably determine the colocalization in a variety of biological images.