Jakob Nikolas Kather

Multi-class texture analysis in colorectal cancer histology.

Automatic recognition of different tissue types in histological images is an essential part in the digital pathology toolbox. Texture analysis is commonly used to address this problem; mainly in the context of estimating the tumour/stroma ratio on histological samples. However, although histological images typically contain more than two tissue types, only few studies have addressed the multi-class problem. For colorectal cancer, one of the most prevalent tumour types, there are in fact no published results on multiclass texture separation. In this paper we present a new dataset of 5,000 histological images of human colorectal cancer including eight different types of tissue. We used this set to assess the classification performance of a wide range of texture descriptors and classifiers. As a result, we found an optimal classification strategy that markedly outperformed traditional methods, improving the state of the art for tumour-stroma separation from 96.9% to 98.6% accuracy and setting a new standard for multiclass tissue separation (87.4% accuracy for eight classes). We make our dataset of histological images publicly available under a Creative Commons license and encourage other researchers to use it as a benchmark for their studies.

Angiopoietin-1 Is Regulated by miR-204 and Contributes to Corneal Neovascularization in KLEIP-Deficient Mice

PURPOSE Corneal neovascularization can cause loss of vision. The introduction of anti-VEGF therapy has been a major improvement in therapeutic options. Recently, we established Kelch-like Ect2-interacting protein (KLEIP/KLHL20) knockout mice as a model of spontaneous corneal neovascular dystrophy. The aim of the present study was to characterize corneal neovascularization in progressive corneal dystrophy in KLEIP(-/-) mice, to evaluate the efficacy of anti-VEGF therapy, and to identify novel molecular regulators in this experimental model. METHODS Corneal neovascularization was assessed by immunohistochemistry. Vascular endothelial growth factor signaling was inhibited by injection of a blocking antibody. Microarrays were used to measure expression of mRNA and microRNA (miRNA) in dystrophic corneae. Results were validated by immunohistochemistry and Western blotting. RESULTS Blood vessels and lymphatics grew from the limbus toward the dystrophic epithelium in corneae of KLEIP(-/-) mice. Blocking VEGF signaling did not reduce phenotype progression. Correspondingly, microarray analysis revealed no upregulation of canonical vascular growth factors in late dystrophy. During phenotype progression, angiopoietin-1 expression increased while miR-204 expression decreased. Bioinformatic analysis identified a binding site for miR-204 in the angiopoietin-1 gene. Validation by in vitro experiments confirmed regulation of angiopoietin-1 by miR-204. CONCLUSIONS Vascular endothelial growth factor does not act as a major player in corneal neovascularization in KLEIP(-/-) mice. However, the proangiogenic factor angiopoietin-1 was strongly upregulated in late-stage phenotype, correlating with loss of miR-204 expression. Correspondingly, we identified miR-204 as a novel regulator of angiopoietin-1 in vitro. These findings may explain the incomplete efficacy of anti-VEGF therapy in the clinic and may provide new candidates for pharmaceutical intervention.

Transgenic Mouse Models of Corneal Neovascularization: New Perspectives for Angiogenesis Research

Corneal neovascularization (NV) refers to the growth of blood vessels and/or lymphatics into the physiologically avascular cornea, which occurs in several pathological processes. In mouse models, corneal NV can be artificially induced to investigate mechanisms of corneal pathologies. However, mouse models of corneal NV are not restricted to cornea-specific research, but also are widely used to investigate general mechanisms of angiogenesis. Because the cornea is transparent and easily accessible, corneal NV models are among the most useful in vivo models in angiogenesis research. The three different approaches that are used to study corneal NV in mice are based on direct application of proangiogenic or antiangiogenic transmitters, external injury to the cornea, or genetically engineered mice, which spontaneously develop corneal NV. The aim of this review is to compare the scope and limitations of the different approaches for corneal NV in mice. Our main focus is to highlight the potential of transgenic spontaneous models of corneal NV. Transgenic models do not require any experimental interference and make it possible to investigate different interconnected proangiogenic signaling cascades. As a result, transgenic models are highly useful for disease-centered angiogenesis research. In summary, transgenic models of corneal NV will complement and advance existing ocular NV assays, and help to discover new angiogenesis-related treatment strategies for ocular and extraocular diseases.

New Colors for Histology: Optimized Bivariate Color Maps Increase Perceptual Contrast in Histological Images.

Background Accurate evaluation of immunostained histological images is required for reproducible research in many different areas and forms the basis of many clinical decisions. The quality and efficiency of histopathological evaluation is limited by the information content of a histological image, which is primarily encoded as perceivable contrast differences between objects in the image. However, the colors of chromogen and counterstain used for histological samples are not always optimally distinguishable, even under optimal conditions. Methods and Results In this study, we present a method to extract the bivariate color map inherent in a given histological image and to retrospectively optimize this color map. We use a novel, unsupervised approach based on color deconvolution and principal component analysis to show that the commonly used blue and brown color hues in Hematoxylin—3,3’-Diaminobenzidine (DAB) images are poorly suited for human observers. We then demonstrate that it is possible to construct improved color maps according to objective criteria and that these color maps can be used to digitally re-stain histological images. Validation To validate whether this procedure improves distinguishability of objects and background in histological images, we re-stain phantom images and N = 596 large histological images of immunostained samples of human solid tumors. We show that perceptual contrast is improved by a factor of 2.56 in phantom images and up to a factor of 2.17 in sets of histological tumor images. Context Thus, we provide an objective and reliable approach to measure object distinguishability in a given histological image and to maximize visual information available to a human observer. This method could easily be incorporated in digital pathology image viewing systems to improve accuracy and efficiency in research and diagnostics.

Rho guanine exchange factors in blood vessels: fine-tuners of angiogenesis and vascular function.

The angiogenic cascade is a multi-step process essential for embryogenesis and other physiological and pathological processes. Rho family GTPases are binary molecular switches and serve as master regulators of various basic cellular processes. Rho GTPases are known to exert important functions in angiogenesis and vascular physiology. These functions demand a tight and context-specific control of cellular processes requiring superordinate control by a multitude of guanine nucleotide exchange factors (GEFs). GEFs display various features enabling them to fine-tune the actions of Rho GTPases in the vasculature: (1) GEFs regulate specific steps of the angiogenic cascade; (2) GEFs show a spatio-temporally specific expression pattern; (3) GEFs differentially regulate endothelial function depending on their subcellular location; (4) GEFs mediate crosstalk between complex signaling cascades and (5) GEFs themselves are regulated by another layer of interacting proteins. The aim of this review is to provide an overview about the role of GEFs in regulating angiogenesis and vascular function and to point out current limitations as well as clinical perspectives.

Identification of a characteristic vascular belt zone in human colorectal cancer

Blood vessels in cancer Intra-tumoral blood vessels are of supreme importance for tumor growth, metastasis and therapy. Yet, little is known about spatial distribution patterns of these vessels. Most experimental or theoretical tumor models implicitly assume that blood vessels are equally abundant in different parts of the tumor, which has far-reaching implications for chemotherapy and tumor metabolism. In contrast, based on histological observations, we hypothesized that blood vessels follow specific spatial distribution patterns in colorectal cancer tissue. We developed and applied a novel computational approach to identify spatial patterns of angiogenesis in histological whole-slide images of human colorectal cancer. A characteristic spatial pattern of blood vessels in colorectal cancer In 33 of 34 (97%) colorectal cancer primary tumors blood vessels were significantly aggregated in a sharply limited belt-like zone at the interface of tumor tissue to the intestinal lumen. In contrast, in 11 of 11 (100%) colorectal cancer liver metastases, a similar hypervascularized zone could be found at the boundary to surrounding liver tissue. Also, in an independent validation cohort, we found this vascular belt zone: 22 of 23 (96%) samples of primary tumors and 15 of 16 (94%) samples of liver metastases exhibited the above-mentioned spatial distribution. Summary and implications We report consistent spatial patterns of tumor vascularization that may have far-reaching implications for models of drug distribution, tumor metabolism and tumor growth: luminal hypervascularization in colorectal cancer primary tumors is a previously overlooked feature of cancer tissue. In colorectal cancer liver metastases, we describe a corresponding pattern at the invasive margin. These findings add another puzzle piece to the complex concept of tumor heterogeneity.

Polyphonic sonification of electrocardiography signals for diagnosis of cardiac pathologies

Electrocardiography (ECG) data are multidimensional temporal data with ubiquitous applications in the clinic. Conventionally, these data are presented visually. It is presently unclear to what degree data sonification (auditory display), can enable the detection of clinically relevant cardiac pathologies in ECG data. In this study, we introduce a method for polyphonic sonification of ECG data, whereby different ECG channels are simultaneously represented by sound of different pitch. We retrospectively applied this method to 12 samples from a publicly available ECG database. We and colleagues from our professional environment then analyzed these data in a blinded way. Based on these analyses, we found that the sonification technique can be intuitively understood after a short training session. On average, the correct classification rate for observers trained in cardiology was 78%, compared to 68% and 50% for observers not trained in cardiology or not trained in medicine at all, respectively. These values compare to an expected random guessing performance of 25%. Strikingly, 27% of all observers had a classification accuracy over 90%, indicating that sonification can be very successfully used by talented individuals. These findings can serve as a baseline for potential clinical applications of ECG sonification.

Dimensionality Reduction Strategies for CNN-Based Classification of Histopathological Images

Features from pre-trained Convolutional Neural Newtorks (CNN) have proved to be effective for many tasks such as object, scene and face recognition. Compared with traditional, hand-designed image descriptors, CNN-based features produce higher-dimensional feature vectors. In specific applications where the number of samples may be limited – as in the case of histopatological images – high dimensionality could potentially cause overfitting and redundancy in the information to be processed and stored. To overcome these potential problems feature reduction methods can be applied, at the cost of a moderate reduction in the discrimination accuracy. In this paper we investigate dimensionality reduction schemes for CNN-based features applied to computer-assisted classification of histopathological images. The purpose of this study is to find the best trade-off between accuracy and dimensionality. Specifically, we test two well-known techniques (i.e.: Principal Component Analysis and Gaussian Random Projection) and propose a novel reduction strategy based on the cross-correlation between the components of the feature vector. The results show that it is possible to reduce CNN-based features by a high ratio with a moderate decrease in accuracy with respect to the original values.

Topography of cancer-associated immune cells in human solid tumors

Lymphoid and myeloid cells are abundant in the tumor microenvironment, can be quantified by immunohistochemistry and shape the disease course of human solid tumors. Yet, there is no comprehensive understanding of spatial immune infiltration patterns (‘topography’) across cancer entities and across various immune cell types. In this study, we systematically measure the topography of multiple immune cell types in 965 histological tissue slides from N = 177 patients in a pan-cancer cohort. We provide a definition of inflamed (‘hot’), non-inflamed (‘cold’) and immune excluded patterns and investigate how these patterns differ between immune cell types and between cancer types. In an independent cohort of N = 287 colorectal cancer patients, we show that hot, cold and excluded topographies for effector lymphocytes (CD8) and tumor-associated macrophages (CD163) alone are not prognostic, but that a bivariate classification system can stratify patients. Our study adds evidence to consider immune topographies as biomarkers for patients with solid tumors.

Color-coded visualization of magnetic resonance imaging multiparametric maps

Multiparametric magnetic resonance imaging (mpMRI) data are emergingly used in the clinic e.g. for the diagnosis of prostate cancer. In contrast to conventional MR imaging data, multiparametric data typically include functional measurements such as diffusion and perfusion imaging sequences. Conventionally, these measurements are visualized with a one-dimensional color scale, allowing only for one-dimensional information to be encoded. Yet, human perception places visual information in a three-dimensional color space. In theory, each dimension of this space can be utilized to encode visual information. We addressed this issue and developed a new method for tri-variate color-coded visualization of mpMRI data sets. We showed the usefulness of our method in a preclinical and in a clinical setting: In imaging data of a rat model of acute kidney injury, the method yielded characteristic visual patterns. In a clinical data set of N = 13 prostate cancer mpMRI data, we assessed diagnostic performance in a blinded study with N = 5 observers. Compared to conventional radiological evaluation, color-coded visualization was comparable in terms of positive and negative predictive values. Thus, we showed that human observers can successfully make use of the novel method. This method can be broadly applied to visualize different types of multivariate MRI data.