Daniel Heim

Detection and Segmentation of Cell Nuclei in Virtual Microscopy Images: A Minimum-Model Approach

Automated image analysis of cells and tissues has been an active research field in medical informatics for decades but has recently attracted increased attention due to developments in computer and microscopy hardware and the awareness that scientific and diagnostic pathology require novel approaches to perform objective quantitative analyses of cellular and tissue specimens. Model-based approaches use a priori information on cell shape features to obtain the segmentation, which may introduce a bias favouring the detection of cell nuclei only with certain properties. In this study we present a novel contour-based “minimum-model” cell detection and segmentation approach that uses minimal a priori information and detects contours independent of their shape. This approach avoids a segmentation bias with respect to shape features and allows for an accurate segmentation (precision = 0.908; recall = 0.859; validation based on ∼8000 manually-labeled cells) of a broad spectrum of normal and disease-related morphological features without the requirement of prior training.

CognitionMaster: an object-based image analysis framework

BackgroundAutomated image analysis methods are becoming more and more important to extract and quantify image features in microscopy-based biomedical studies and several commercial or open-source tools are available. However, most of the approaches rely on pixel-wise operations, a concept that has limitations when high-level object features and relationships between objects are studied and if user-interactivity on the object-level is desired.ResultsIn this paper we present an open-source software that facilitates the analysis of content features and object relationships by using objects as basic processing unit instead of individual pixels. Our approach enables also users without programming knowledge to compose “analysis pipelines“ that exploit the object-level approach. We demonstrate the design and use of example pipelines for the immunohistochemistry-based cell proliferation quantification in breast cancer and two-photon fluorescence microscopy data about bone-osteoclast interaction, which underline the advantages of the object-based concept.ConclusionsWe introduce an open source software system that offers object-based image analysis. The object-based concept allows for a straight-forward development of object-related interactive or fully automated image analysis solutions. The presented software may therefore serve as a basis for various applications in the field of digital image analysis.

Cancer beyond organ and tissue specificity: next-generation-sequencing gene mutation data reveal complex genetic similarities across major cancers.

Cancer medicine relies on the paradigm that cancer is an organ‐ and tissue‐specific disease, which is the basis for classifying tumors. With the extensive genomic information now available on tumors it is possible to conduct analyses to reveal common genetic features across cancer types and to explore whether the established anatomy‐based tumor classification is actually reflected on the genetic level, which might provide important guides to new therapeutic directions. Here, we have conducted an extensive analysis of the genetic similarity of tumors from 14 major cancer entities using somatic mutation data from 4,796 cases available through The Cancer Genome Atlas (TCGA) based on all available genes as well as different cancer‐related gene sets. Our analysis provides a systematic account of the genetic similarity network for major cancer types and shows that in about 43% of the cases on average, tumors of a particular anatomic site are genetically more similar to tumors from different organs and tissues (trans‐similarity) than to tumors of the same origin (self‐similarity). The observed similarities exist not only for carcinomas from different sites but are also present among neoplasms from different tissue origin, such as melanoma, acute myeloid leukemia, and glioblastoma. The current WHO cancer classification is therefore reflected on the genetic level by only about 57% of the tumors. These results provide a rationale to reconsider organ‐ and tissue‐specificity in cancer and contribute to the discussion about whether personalized therapies targeting specific genetic alterations may be transferred to cancers from other anatomic sites with similar genetic properties.

Copy Number Changes of Clinically Actionable Genes in Melanoma, Non‐Small Cell Lung Cancer and Colorectal Cancer – A Survey Across 822 Routine Diagnostic Cases

Targeted deep massive parallel sequencing has been implemented in routine molecular diagnostics for high‐throughput genetic profiling of formalin‐fixed paraffin‐embedded (FFPE) cancer samples. This approach is widely used to interrogate simple somatic mutations but experience with the analysis of copy number variations (CNV) is limited. Here, we retrospectively analyzed CNV in 822 cancer cases (135 melanoma, 468 non‐small cell lung cancers (NSCLC), 219 colorectal cancers (CRC)). We observed a decreasing frequency of CNV in clinically actionable genes from melanoma to NSCLC to CRC. The overall cohort displayed 168 (20%) amplifications in 17 druggable targets. The majority of BRAF mutant melanomas (54%) showed co‐occurring CNV in other genes, mainly affecting CDKN2A. Subsets showed clustered deletions in ABL1, NOTCH1, RET or STK11, GNA11, and JAK3. Most NRAS mutant melanomas (49%) harbored CNVs in other genes with CDKN2A and FGFR3 being most frequently affected. Five BRAF/NRASwt tumors had co‐amplifications of KDR, KIT, PDGFRA and another six mutated KIT. Among all NSCLC, we identified 14 EGFRamp (with ten EGFRmut) and eight KRASamp (with seven KRASmut). KRASmut tumors displayed frequent amplifications of MYC (nu2009=u200910) and MDM2 (nu2009=u20095). Fifteen KRAS/EGFR/BRAFwt tumors had MET mutations/amplifications. In CRC, amplified IGF2 was most prevalent (nu2009=u200913) followed by MYC (nu2009=u20099). Two cases showed amplified KRAS wildtype alleles. Two of the KRASmut cases harbored amplifications of NRAS and three KRASwt cases amplification of EGFR. In conclusion, we demonstrate that our approach i) facilitates detection of CNV, ii) enables detection of known CNV patterns, and iii) uncovers new CNV of clinically actionable genes in FFPE tissue samples across cancers.

A Multi-resolution Approach for Combining Visual Information using Nuclei Segmentation and Classification in Histopathological Images

This paper describes a multi-resolution technique to combine diagnostically important visual information at different magnifications in H&E whole slide images (WSI) of gastric cancer. The primary goal is to improve the results of nuclei segmentation method for heterogeneous histopathological datasets with variations in stain intensity and malignancy levels. A minimum-model nuclei segmentation method is first applied to tissue images at multiple resolutions, and a comparative evaluation is performed. A comprehensive set of 31 nuclei features based on color, texture and morphology are derived from the nuclei segments. AdaBoost classification method is used to classify these segments into a set of pre-defined classes. Two classification approaches are evaluated for this purpose. A relevance score is assigned to each class and a combined segmentation result is obtained consisting of objects with higher visual significance from individual magnifications, thereby preserving both coarse and fine details in the image. Quantitative and visual assessment of combination results shows that they contain comprehensive and diagnostically more relevant information than in constituent

Histological tumor typing in the age of molecular profiling.

Clinical oncology and pathological diagnostics regard cancer as an organ- and tissue-specific disease. Comprehensive mutational profiling information from next-generation sequencing projects may be used to study to what extent the anatomic tumor classifications relate to the observed molecular profiles. Here, we review data that show substantial genetic similarities across major anatomic cancer types and that propose novel tumor classifications based on mutational profiling. Although these studies provide important insight into molecular tumor properties and some even propose novel tumor classification systems, current clinical evidence is lacking that genetic tumor profiling is sufficient to replace histological tumor typing. Recent studies rather show that targeted treatments efficaceous in one tumor type are not necessarily successful in another despite the presence of the same (actionable) mutations. We discuss the implications of the observed complex mutational tumor profiles for targeted therapy selection and future trial design in precision oncology.

Cell nuclei attributed relational graphs for efficient representation and classification of gastric cancer in digital histopathology

This paper describes a novel graph-based method for efficient representation and subsequent classification in histological whole slide images of gastric cancer. Her2/neu immunohistochemically stained and haematoxylin and eosin stained histological sections of gastric carcinoma are digitized. Immunohistochemical staining is used in practice by pathologists to determine extent of malignancy, however, it is laborious to visually discriminate the corresponding malignancy levels in the more commonly used haematoxylin and eosin stain, and this study attempts to solve this problem using a computer-based method. Cell nuclei are first isolated at high magnification using an automatic cell nuclei segmentation strategy, followed by construction of cell nuclei attributed relational graphs of the tissue regions. These graphs represent tissue architecture comprehensively, as they contain information about cell nuclei morphology as vertex attributes, along with knowledge of neighborhood in the form of edge linking and edge attributes. Global graph characteristics are derived and ensemble learning is used to discriminate between three types of malignancy levels, namely, non-tumor, Her2/neu positive tumor and Her2/neu negative tumor. Performance is compared with state of the art methods including four texture feature groups (Haralick, Gabor, Local Binary Patterns and Varma Zisserman features), color and intensity features, and Voronoi diagram and Delaunay triangulation. Texture, color and intensity information is also combined with graph-based knowledge, followed by correlation analysis. Quantitative assessment is performed using two cross validation strategies. On investigating the experimental results, it can be concluded that the proposed method provides a promising way for computer-based analysis of histopathological images of gastric cancer.

Somatic genome alterations in relation to age in lung squamous cell carcinoma

Lung squamous cell carcinoma (LUSC) is the most common cause of global cancer-related mortality and the major risk factors is smoking consumption. By analyzing ∼500 LUSC samples from The Cancer Genome Atlas, we detected a higher mutational burden as well as a higher level of methylation changes in younger patients. The SNPs mutational profiling showed enrichments of smoking-related signature 4 and defective DNA mismatch repair (MMR)-related signature 6 in younger patients, while the defective DNA MMR signature 26 was enriched among older patients. Furthermore, gene set enrichment analysis was performed in order to explore functional effect of somatic alterations in relation to patient age. Extracellular Matrix-Receptor Interaction, Nucleotide Excision Repair and Axon Guidance seem crucial disrupted pathways in younger patients. We hypothesize that a higher sensitivity to smoking-related damages and the enrichment of defective DNA MMR related mutations may contribute to the higher mutational burden of younger patients. The two distinct age-related defective DNA MMR signatures 6 and 26 might be crucial mutational patterns in LUSC tumorigenesis which may develop distinct phenotypes. Our study provides indications of age-dependent differences in mutational backgrounds (SNPs and CNVs) as well as epigenetic patterns that might be relevant for age adjusted treatment approaches.

Abstract 3167: CNV patterns in 822 routine diagnostic cases of NSCLC, melanoma, and colorectal cancer

Targeted deep massive parallel sequencing (MPS) has been implemented in routine molecular diagnostics for high-throughput genetic profiling of formalin-fixed paraffin-embedded cancer samples. This approach is now widely used to interrogate simple somatic mutations but experience with the analysis of copy number variations (CNV) is still limited. Here, we retrospectively analyzed CNVs in 822 cancer cases (n = 135 melanoma, n = 468 non-small cell lung cancers (NSCLC), n = 219 colorectal cancers (CRC)) that were sent to our institution for routine molecular profiling using a semiconductor based sequencing platform. Amplifications and deletions inferred by MPS coverage data were independently validated by a qPCR assay. We observed a decreasing frequency of CNV in clinically actionable genes from melanoma to NSCLC to colorectal cancer. Of 56 melanomas with genetic aberrations in BRAF, 31 showed co-occurring CNV in other genes, mainly affecting CDKN2A. Some tumors (5 cases each) revealed clustered deletions affecting either ABL1, NOTCH1, and RET or STK11, GNA11, and JAK3. 8.1% of the cases had amplifications in clinically actionable genes. In the group of NRAS mutant tumors (n = 39), 26 showed co-occurring CNVs in other genes, such as CDKN2A and FGFR3, and 9 NRAS mutant cases were additionally mutated in BRAF. 19.1% had amplifications in clinically actionable genes. In contrast to BRAF mutant tumors, we did not see any specific CNV clusters. In the group of BRAF/NRASwt tumors (n = 11), we observed 5 cases with co-amplification of KDR, KIT, PDGFRA and another 6 cases with KIT mutations. While co-amplified cases had many gene deletions, KIT mutated tumors harbored only very few genetic aberrations in other genes. Across both NSCLC data sets, we identified 14 cases with amplified EGFR (10 of them harboring co-occurring EGFR mutations) and detected 8 NSCLC with KRAS amplifications (of which 7 had co-occurring mutations of KRAS). KRAS mutated tumors displayed frequent amplifications in MYC (n = 10) and MDM2 (n = 5). Of the 22 BRAF mutant tumors, two harbored mutated KRAS. In contrast to melanoma, we observed no clustering of CNVs in BRAFmut NSCLCs. Within the group of KRAS/EGFR/BRAFwt tumors, we identified 15 cases harboring genetic aberrations in MET (n = 8 mutations, n = 7 amplifications). Compared to melanoma and NSCLC, the number of CNV in CRC was rather low. IGF2 amplifications were most prevalent (n = 13) followed by MYC (n = 9). Two cases showed amplified wild-type alleles of KRAS. Two KRAS mutant tumors showed concomitant amplification of NRAS and three cases harbored amplified EGFR. In conclusion we demonstrate that i) detection of CNVs by targeted MPS data obtained from FFPE material is feasible and ii) could be validated independently, iii) this approach enables detection of known CNV patterns, and iv) uncovers new CNV patterns in clinically actionable targets across cancers. Citation Format: Albrecht Stenzinger, Roland Penzel, Frederick Klauschen, Arne Warth, Regine Brandt, Daniel Heim, Peter Schirmacher, Wilko Weichert, Volker Endris, Nicole Pfarr. CNV patterns in 822 routine diagnostic cases of NSCLC, melanoma, and colorectal cancer. [abstract]. In: Proceedings of the 107th Annual Meeting of the American Association for Cancer Research; 2016 Apr 16-20; New Orleans, LA. Philadelphia (PA): AACR; Cancer Res 2016;76(14 Suppl):Abstract nr 3167.

Mutational Similarities Across Cancers: Implications for Research, Diagnostics, and Personalized Therapy Design

Oncology and cancer research are based on the principle that cancers are regarded as organ- and tissue-specific diseases. One of the central aspects of histopathological tumor diagnostics is to determine the tumor’s anatomic origin and other morphological features that are the basis for selecting the appropriate therapy. Similarly, research programs are usually also focused on particular cancer entities. However, mutational tumor profiling performed with next-generation-sequencing techniques has made it possible to analyze whether this anatomical tumor classification is valid also on the genetic level. Here, we review recent evidence that substantial similarities exist among tumors across classical anatomic cancer entities on the mutational level. We furthermore discuss the implications of these complex mutational profiles and similarity patterns across cancers for diagnostics, research, and clinical study design and explain why the comprehensive genomic data should be complemented by functional proteomic analyses.