Institutes
Refine
Document Type
- Article (3)
Language
- English (3)
Has Fulltext
- yes (3)
Is part of the Bibliography
- no (3)
Keywords
- DNA methylation (1)
- Deep learning (1)
- Digital pathology (1)
- EMT (1)
- Glioma (1)
- IDH mutation (1)
- SLUG (1)
- TGF-β (1)
- Transfer learning (1)
- Whole slide images (1)
Institute
- Medizin (3) (remove)
We provide in this paper a comprehensive comparison of various transfer learning strategies and deep learning architectures for computer-aided classification of adult-type diffuse gliomas. We evaluate the generalizability of out-of-domain ImageNet representations for a target domain of histopathological images, and study the impact of in-domain adaptation using self-supervised and multi-task learning approaches for pretraining the models using the medium-to-large scale datasets of histopathological images. A semi-supervised learning approach is furthermore proposed, where the fine-tuned models are utilized to predict the labels of unannotated regions of the whole slide images (WSI). The models are subsequently retrained using the ground-truth labels and weak labels determined in the previous step, providing superior performance in comparison to standard in-domain transfer learning with balanced accuracy of 96.91% and F1-score 97.07%, and minimizing the pathologist's efforts for annotation. Finally, we provide a visualization tool working at WSI level which generates heatmaps that highlight tumor areas; thus, providing insights to pathologists concerning the most informative parts of the WSI.
Linking epigenetic signature and metabolic phenotype in IDH mutant and IDH wildtype diffuse glioma
(2020)
Aims: Changes in metabolism are known to contribute to tumour phenotypes. If and how metabolic alterations in brain tumours contribute to patient outcome is still poorly understood. Epigenetics impact metabolism and mitochondrial function. The aim of this study is a characterisation of metabolic features in molecular subgroups of isocitrate dehydrogenase mutant (IDHmut) and isocitrate dehydrogenase wildtype (IDHwt) gliomas. Methods: We employed DNA methylation pattern analyses with a special focus on metabolic genes, large-scale metabolism panel immunohistochemistry (IHC), qPCR-based determination of mitochondrial DNA copy number and immune cell content using IHC and deconvolution of DNA methylation data. We analysed molecularly characterised gliomas (n = 57) for in depth DNA methylation, a cohort of primary and recurrent gliomas (n = 22) for mitochondrial copy number and validated these results in a large glioma cohort (n = 293). Finally, we investigated the potential of metabolic markers in Bevacizumab (Bev)-treated gliomas (n = 29). Results: DNA methylation patterns of metabolic genes successfully distinguished the molecular subtypes of IDHmut and IDHwt gliomas. Promoter methylation of lactate dehydrogenase A negatively correlated with protein expression and was associated with IDHmut gliomas. Mitochondrial DNA copy number was increased in IDHmut tumours and did not change in recurrent tumours. Hierarchical clustering based on metabolism panel IHC revealed distinct subclasses of IDHmut and IDHwt gliomas with an impact on patient outcome. Further quantification of these markers allowed for the prediction of survival under anti-angiogenic therapy. Conclusion: A mitochondrial signature was associated with increased survival in all analyses, which could indicate tumour subgroups with specific metabolic vulnerabilities.
Aims: In primary central nervous system tumours, epithelial-to-mesenchymal transition (EMT) gene expression is associated with increased malignancy. However, it has also been shown that EMT factors in gliomas are almost exclusively expressed by glioma vessel-associated pericytes (GA-Peris). In this study, we aimed to identify the mechanism of EMT in GA-Peris and its impact on angiogenic processes.
Methods; In glioma patients, vascular density and the expression of the pericytic markers platelet derived growth factor receptor (PDGFR)-β and smooth muscle actin (αSMA) were examined in relation to the expression of the EMT transcription factor SLUG and were correlated with survival of patients with glioblastoma (GBM). Functional mechanisms of SLUG regulation and the effects on primary human brain vascular pericytes (HBVP) were studied in vitro by measuring proliferation, cell motility and growth characteristics.
Results: The number of PDGFR-β- and αSMA-positive pericytes did not change with increased malignancy nor showed an association with the survival of GBM patients. However, SLUG-expressing pericytes displayed considerable morphological changes in GBM-associated vessels, and TGF-β induced SLUG upregulation led to enhanced proliferation, motility and altered growth patterns in HBVP. Downregulation of SLUG or addition of a TGF-β antagonising antibody abolished these effects.
Conclusions: We provide evidence that in GA-Peris, elevated SLUG expression is mediated by TGF-β, a cytokine secreted by most glioma cells, indicating that the latter actively modulate neovascularisation not only by modulating endothelial cells, but also by influencing pericytes. This process might be responsible for the formation of an unstructured tumour vasculature as well as for the breakdown of the blood–brain barrier in GBM.