Roman Bruch
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Roman Bruch has a total of 37 co-authors. Here are the top co-authors by number of shared publications:
- Rüdiger Rudolf
Institute of Molecular and Cell Biology, Mannheim University of Applied Sciences, Mannheim, Germany
7 shared publications - Markus Reischl
Institute for Automation and Applied Informatics (IAI), Karlsruhe Institute of Technology (KIT), Eggenstein-Leopoldshafen, Germany
5 shared publications - Mathias Hafner
Institute of Molecular and Cell Biology, Mannheim University of Applied Sciences, 68163 Mannheim, Germany
4 shared publications - Florian Keller
Institute of Molecular and Cell Biology, Mannheim University of Applied Sciences, 68163 Mannheim, Germany
3 shared publications Mario Vitacolonna
Institute of Molecular and Cell Biology, Mannheim University of Applied Sciences, Mannheim, Germany
3 shared publicationsMarkus Reischl
Institute for Automation and Applied Informatics, Karlsruhe Institute of Technology, Eggenstein-Leopoldshafen, Germany
3 shared publications
Latest Publications
- Open Access
Advanced 3D spheroid-based skin models and deep-learning based image analysis enable in-depth investigation of keratinocyte differentiation and barrier function
To date, a panel of different biological models have been used in skin research, ranging from in vivo testing to 2D and 3D cultures. Among these, organotypic skin models represent the current gold standard for preclinical dermatology and toxicology studies. However, they are variable in quality and require long maturation times and a lot of work and cells, the latter often of primary origin. Here, we propose dermal-epidermal spheroids as an alternative model that balances physiological relevance and throughput. Next to corresponding full thickness skin models, different fibroblast/keratinocyte coculture spheroids were generated. These used the commonly employed HaCaT cells as well as two recently immortalized keratinocyte cell lines, NHK-SV/TERT and NHK-E6/E7. To investigate their differentiation with detailed spatio-temporal resolution, a deep-learning segmentation-based pipeline, capable of revealing nuclear morphology and positioning as well as marker expression with single-cell precision, was developed and applied. Moreover, the formation of a functional barrier was assessed by live-imaging of Lucifer yellow diffusion. These experiments identified the NHK-E6/E7 cell line as the most and HaCaT cells as the least suitable alternative to primary keratinocytes in both spheroids and full thickness models. Furthermore, NHK-based coculture spheroids displayed functional maturation, including stratification, cornification, and barrier formation, closely recapitulating these features of corresponding full thickness models. Given the scalability and compatibility with automation, these micro-skin fibroblast/NHK-based 3D coculture spheroids might represent a promising new platform for pharmaceutical, cosmetic, and toxicological testing.
Open AccessAdvanced 3D spheroid-based skin models and deep-learning based image analysis enable in-depth investigation of keratinocyte differentiation and barrier function
To date, a panel of different biological models have been used in skin research, ranging from in vivo testing to 2D and 3D cultures. Among these, organotypic skin models represent the current gold standard for preclinical dermatology and toxicology studies. However, they are variable in quality and require long maturation times and a lot of work and cells, the latter often of primary origin. Here, we propose dermal-epidermal spheroids as an alternative model that balances physiological relevance and throughput. Next to corresponding full thickness skin models, different fibroblast/keratinocyte coculture spheroids were generated. These used the commonly employed HaCaT cells as well as two recently immortalized keratinocyte cell lines, NHK-SV/TERT and NHK-E6/E7. To investigate their differentiation with detailed spatio-temporal resolution, a deep-learning segmentation-based pipeline, capable of revealing nuclear morphology and positioning as well as marker expression with single-cell precision, was developed and applied. Moreover, the formation of a functional barrier was assessed by live-imaging of Lucifer yellow diffusion. These experiments identified the NHK-E6/E7 cell line as the most and HaCaT cells as the least suitable alternative to primary keratinocytes in both spheroids and full thickness models. Furthermore, NHK-based coculture spheroids displayed functional maturation, including stratification, cornification, and barrier formation, closely recapitulating these features of corresponding full thickness models. Given the scalability and compatibility with automation, these micro-skin fibroblast/NHK-based 3D coculture spheroids might represent a promising new platform for pharmaceutical, cosmetic, and toxicological testing.
- Open Access
- 11 January 2025
- Published by Springer Nature in Communications Biology
- Vol. 8 (1) , 1-13
Improving 3D deep learning segmentation with biophysically motivated cell synthesis
Biomedical research increasingly relies on three-dimensional (3D) cell culture models and artificial-intelligence-based analysis can potentially facilitate a detailed and accurate feature extraction on a single-cell level. However, this requires for a precise segmentation of 3D cell datasets, which in turn demands high-quality ground truth for training. Manual annotation, the gold standard for ground truth data, is too time-consuming and thus not feasible for the generation of large 3D training datasets. To address this, we present a framework for generating 3D training data, which integrates biophysical modeling for realistic cell shape and alignment. Our approach allows the in silico generation of coherent membrane and nuclei signals, that enable the training of segmentation models utilizing both channels for improved performance. Furthermore, we present a generative adversarial network (GAN) training scheme that generates not only image data but also matching labels. Quantitative evaluation shows superior performance of biophysical motivated synthetic training data, even outperforming manual annotation and pretrained models. This underscores the potential of incorporating biophysical modeling for enhancing synthetic training data quality.
- 11 January 2025
- Published by Springer Nature in Communications Biology
- Vol. 8 (1) , 1-13
Open AccessImproving 3D deep learning segmentation with biophysically motivated cell synthesis
Biomedical research increasingly relies on three-dimensional (3D) cell culture models and artificial-intelligence-based analysis can potentially facilitate a detailed and accurate feature extraction on a single-cell level. However, this requires for a precise segmentation of 3D cell datasets, which in turn demands high-quality ground truth for training. Manual annotation, the gold standard for ground truth data, is too time-consuming and thus not feasible for the generation of large 3D training datasets. To address this, we present a framework for generating 3D training data, which integrates biophysical modeling for realistic cell shape and alignment. Our approach allows the in silico generation of coherent membrane and nuclei signals, that enable the training of segmentation models utilizing both channels for improved performance. Furthermore, we present a generative adversarial network (GAN) training scheme that generates not only image data but also matching labels. Quantitative evaluation shows superior performance of biophysical motivated synthetic training data, even outperforming manual annotation and pretrained models. This underscores the potential of incorporating biophysical modeling for enhancing synthetic training data quality.
- Open Access
Improving 3D deep learning segmentation with biophysically motivated cell synthesis
Biomedical research increasingly relies on 3D cell culture models and AI-based analysis can potentially facilitate a detailed and accurate feature extraction on a single-cell level. However, this requires for a precise segmentation of 3D cell datasets, which in turn demands high-quality ground truth for training. Manual annotation, the gold standard for ground truth data, is too time-consuming and thus not feasible for the generation of large 3D training datasets. To address this, we present a novel framework for generating 3D training data, which integrates biophysical modeling for realistic cell shape and alignment. Our approach allows the in silico generation of coherent membrane and nuclei signals, that enable the training of segmentation models utilizing both channels for improved performance. Furthermore, we present a new GAN training scheme that generates not only image data but also matching labels. Quantitative evaluation shows superior performance of biophysical motivated synthetic training data, even outperforming manual annotation and pretrained models. This underscores the potential of incorporating biophysical modeling for enhancing synthetic training data quality.
Open AccessImproving 3D deep learning segmentation with biophysically motivated cell synthesis
Biomedical research increasingly relies on 3D cell culture models and AI-based analysis can potentially facilitate a detailed and accurate feature extraction on a single-cell level. However, this requires for a precise segmentation of 3D cell datasets, which in turn demands high-quality ground truth for training. Manual annotation, the gold standard for ground truth data, is too time-consuming and thus not feasible for the generation of large 3D training datasets. To address this, we present a novel framework for generating 3D training data, which integrates biophysical modeling for realistic cell shape and alignment. Our approach allows the in silico generation of coherent membrane and nuclei signals, that enable the training of segmentation models utilizing both channels for improved performance. Furthermore, we present a new GAN training scheme that generates not only image data but also matching labels. Quantitative evaluation shows superior performance of biophysical motivated synthetic training data, even outperforming manual annotation and pretrained models. This underscores the potential of incorporating biophysical modeling for enhancing synthetic training data quality.
- Open Access
A Multiparametric Analysis Reveals Differential Behavior of Spheroid Cultures on Distinct Ultra-Low Attachment Plates Types
Spheroids have become principal three-dimensional biological models to study cancer, developmental processes, and drug efficacy. For spheroid generation, ultra-low attachment plates are noteworthy due to their simplicity, compatibility with automation, and experimental and commercial accessibility. Nonetheless, it is unknown whether and to what degree the plate type impacts spheroid formation and biology. This study employed automated brightfield microscopy to systematically compare the size and eccentricity of spheroids formed in six different plate types using four distinct human cell lines, i.e., CCD-1137Sk fibroblasts, HaCaT keratinocytes, and MDA-MB-231 and HT-29 cancer cells. Results showed that all plate types exhibited similar sphe-roid-forming capabilities, and the gross patterns of growth or shrinkage during four days after seeding were comparable. Yet, size and eccentricity varied systematically among specific cell lines and plate types. A confocal wholemount analysis by a novel pipeline of AI-based 3D-image analysis procedures revealed changes in cell proliferation, cell number, nuclear volume, and keratino-cyte differentiation, which were accompanied by altered YAP1-signals. The findings show that the plate type may influence the outcome of experimental campaigns. It is advisable to scan different plate types for the optimal configuration for a specific investigation instead of using one standard plate for all kinds of applications.
Open AccessA Multiparametric Analysis Reveals Differential Behavior of Spheroid Cultures on Distinct Ultra-Low Attachment Plates Types
Spheroids have become principal three-dimensional biological models to study cancer, developmental processes, and drug efficacy. For spheroid generation, ultra-low attachment plates are noteworthy due to their simplicity, compatibility with automation, and experimental and commercial accessibility. Nonetheless, it is unknown whether and to what degree the plate type impacts spheroid formation and biology. This study employed automated brightfield microscopy to systematically compare the size and eccentricity of spheroids formed in six different plate types using four distinct human cell lines, i.e., CCD-1137Sk fibroblasts, HaCaT keratinocytes, and MDA-MB-231 and HT-29 cancer cells. Results showed that all plate types exhibited similar sphe-roid-forming capabilities, and the gross patterns of growth or shrinkage during four days after seeding were comparable. Yet, size and eccentricity varied systematically among specific cell lines and plate types. A confocal wholemount analysis by a novel pipeline of AI-based 3D-image analysis procedures revealed changes in cell proliferation, cell number, nuclear volume, and keratino-cyte differentiation, which were accompanied by altered YAP1-signals. The findings show that the plate type may influence the outcome of experimental campaigns. It is advisable to scan different plate types for the optimal configuration for a specific investigation instead of using one standard plate for all kinds of applications.
- Open Access
- 9 November 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 11, 127895-127906
Improving Generative Adversarial Networks for Patch-Based Unpaired Image-to-Image Translation
Deep learning models for image segmentation achieve high-quality results, but need large amounts of training data. Training data is primarily annotated manually, which is time-consuming and often not feasible for large-scale 2D and 3D images. Manual annotation can be reduced using synthetic training data generated by generative adversarial networks that perform unpaired image-to-image translation. As of now, large images need to be processed patch-wise during inference, resulting in local artifacts in border regions after merging the individual patches. To reduce these artifacts, we propose a new method that integrates overlapping patches into the training process. We incorporated our method into CycleGAN and tested it on our new 2D tiling strategy benchmark dataset. The results show that the artifacts are reduced by 85% compared to state-of-the-art weighted tiling. Additionally, we demonstrate transferability to real-world 3D biological image data, receiving a high-quality synthetic dataset. Increasing the quality of synthetic training datasets can reduce manual annotation, increase the quality of model output, and can help develop and evaluate deep learning models.
- 9 November 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 11, 127895-127906
Open AccessImproving Generative Adversarial Networks for Patch-Based Unpaired Image-to-Image Translation
Deep learning models for image segmentation achieve high-quality results, but need large amounts of training data. Training data is primarily annotated manually, which is time-consuming and often not feasible for large-scale 2D and 3D images. Manual annotation can be reduced using synthetic training data generated by generative adversarial networks that perform unpaired image-to-image translation. As of now, large images need to be processed patch-wise during inference, resulting in local artifacts in border regions after merging the individual patches. To reduce these artifacts, we propose a new method that integrates overlapping patches into the training process. We incorporated our method into CycleGAN and tested it on our new 2D tiling strategy benchmark dataset. The results show that the artifacts are reduced by 85% compared to state-of-the-art weighted tiling. Additionally, we demonstrate transferability to real-world 3D biological image data, receiving a high-quality synthetic dataset. Increasing the quality of synthetic training datasets can reduce manual annotation, increase the quality of model output, and can help develop and evaluate deep learning models.
- Open Access
- 31 March 2023
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 18 (3) , e0283828
Synthesis of large scale 3D microscopic images of 3D cell cultures for training and benchmarking
The analysis of 3D microscopic cell culture images plays a vital role in the development of new therapeutics. While 3D cell cultures offer a greater similarity to the human organism than adherent cell cultures, they introduce new challenges for automatic evaluation, like increased heterogeneity. Deep learning algorithms are able to outperform conventional analysis methods in such conditions but require a large amount of training data. Due to data size and complexity, the manual annotation of 3D images to generate large datasets is a nearly impossible task. We therefore propose a pipeline that combines conventional simulation methods with deep-learning-based optimization to generate large 3D synthetic images of 3D cell cultures where the labels are known by design. The hybrid procedure helps to keep the generated image structures consistent with the underlying labels. A new approach and an additional measure are introduced to model and evaluate the reduced brightness and quality in deeper image regions. Our analyses show that the deep learning optimization step consistently improves the quality of the generated images. We could also demonstrate that a deep learning segmentation model trained with our synthetic data outperforms a classical segmentation method on real image data. The presented synthesis method allows selecting a segmentation model most suitable for the user’s data, providing an ideal basis for further data analysis.
- 31 March 2023
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 18 (3) , e0283828
Open AccessSynthesis of large scale 3D microscopic images of 3D cell cultures for training and benchmarking
The analysis of 3D microscopic cell culture images plays a vital role in the development of new therapeutics. While 3D cell cultures offer a greater similarity to the human organism than adherent cell cultures, they introduce new challenges for automatic evaluation, like increased heterogeneity. Deep learning algorithms are able to outperform conventional analysis methods in such conditions but require a large amount of training data. Due to data size and complexity, the manual annotation of 3D images to generate large datasets is a nearly impossible task. We therefore propose a pipeline that combines conventional simulation methods with deep-learning-based optimization to generate large 3D synthetic images of 3D cell cultures where the labels are known by design. The hybrid procedure helps to keep the generated image structures consistent with the underlying labels. A new approach and an additional measure are introduced to model and evaluate the reduced brightness and quality in deeper image regions. Our analyses show that the deep learning optimization step consistently improves the quality of the generated images. We could also demonstrate that a deep learning segmentation model trained with our synthetic data outperforms a classical segmentation method on real image data. The presented synthesis method allows selecting a segmentation model most suitable for the user’s data, providing an ideal basis for further data analysis.
- Open Access
- 1 August 2022
- Published by Walter de Gruyter GmbH in Current Directions in Biomedical Engineering
- Vol. 8 (2) , 305-308
Prediction of Fluorescent Ki67 Staining in 3D Tumor Spheroids
3D cell culture models are important tools for the development and testing of new therapeutics. In combination with immunoassays and 3D confocal microscopy, crucial information like morphological or metabolic changes can be examined during drug testing. However, a common limitation of immunostainings is the number of dyes that can be imaged simultaneously, as overlaps in the spectral profiles of the different dyes may result in cross talk. We therefore present a 3D deep learning method, able to predict fluorescent stainings of specific antigens on the basis of a nuclei staining. Using the proliferation marker Ki67, we showed that the presented model was able to predict the Ki67 staining with a strong correlation to the real signal. Additional analysis showed, that the model was not relying on signal cross talk. This approach, based on staining of the cell nuclei and subsequent prediction of the target antigen, could reduce the number of parallel antigen stains to a minimum and incompatible staining panels could be circumvented in the future.
- 1 August 2022
- Published by Walter de Gruyter GmbH in Current Directions in Biomedical Engineering
- Vol. 8 (2) , 305-308
Open AccessPrediction of Fluorescent Ki67 Staining in 3D Tumor Spheroids
3D cell culture models are important tools for the development and testing of new therapeutics. In combination with immunoassays and 3D confocal microscopy, crucial information like morphological or metabolic changes can be examined during drug testing. However, a common limitation of immunostainings is the number of dyes that can be imaged simultaneously, as overlaps in the spectral profiles of the different dyes may result in cross talk. We therefore present a 3D deep learning method, able to predict fluorescent stainings of specific antigens on the basis of a nuclei staining. Using the proliferation marker Ki67, we showed that the presented model was able to predict the Ki67 staining with a strong correlation to the real signal. Additional analysis showed, that the model was not relying on signal cross talk. This approach, based on staining of the cell nuclei and subsequent prediction of the target antigen, could reduce the number of parallel antigen stains to a minimum and incompatible staining panels could be circumvented in the future.
- Open Access
hiPSC-Derived Schwann Cells Influence Myogenic Differentiation in Neuromuscular Cocultures
Motoneurons, skeletal muscle fibers, and Schwann cells form synapses, termed neuromuscular junctions (NMJs). These control voluntary body movement and are affected in numerous neuromuscular diseases. Therefore, a variety of NMJ in vitro models have been explored to enable mechanistic and pharmacological studies. So far, selective integration of Schwann cells in these models has been hampered, due to technical limitations. Here we present robust protocols for derivation of Schwann cells from human induced pluripotent stem cells (hiPSC) and their coculture with hiPSC-derived motoneurons and C2C12 muscle cells. Upon differentiation with tuned BMP signaling, Schwann cells expressed marker proteins, S100b, Gap43, vimentin, and myelin protein zero. Furthermore, they displayed typical spindle-shaped morphologies with long processes, which often aligned with motoneuron axons. Inclusion of Schwann cells in coculture experiments with hiPSC-derived motoneurons and C2C12 myoblasts enhanced myotube growth and affected size and number of acetylcholine receptor plaques on myotubes. Altogether, these data argue for the availability of a consistent differentiation protocol for Schwann cells and their amenability for functional integration into neuromuscular in vitro models, fostering future studies of neuromuscular mechanisms and disease.
Open AccesshiPSC-Derived Schwann Cells Influence Myogenic Differentiation in Neuromuscular Cocultures
Motoneurons, skeletal muscle fibers, and Schwann cells form synapses, termed neuromuscular junctions (NMJs). These control voluntary body movement and are affected in numerous neuromuscular diseases. Therefore, a variety of NMJ in vitro models have been explored to enable mechanistic and pharmacological studies. So far, selective integration of Schwann cells in these models has been hampered, due to technical limitations. Here we present robust protocols for derivation of Schwann cells from human induced pluripotent stem cells (hiPSC) and their coculture with hiPSC-derived motoneurons and C2C12 muscle cells. Upon differentiation with tuned BMP signaling, Schwann cells expressed marker proteins, S100b, Gap43, vimentin, and myelin protein zero. Furthermore, they displayed typical spindle-shaped morphologies with long processes, which often aligned with motoneuron axons. Inclusion of Schwann cells in coculture experiments with hiPSC-derived motoneurons and C2C12 myoblasts enhanced myotube growth and affected size and number of acetylcholine receptor plaques on myotubes. Altogether, these data argue for the availability of a consistent differentiation protocol for Schwann cells and their amenability for functional integration into neuromuscular in vitro models, fostering future studies of neuromuscular mechanisms and disease.
- Open Access
- 1 August 2021
- Published by Elsevier in SLAS Technology
- Vol. 26 (4) , 367-376
epiTracker: A Framework for Highly Reliable Particle Tracking for the Quantitative Analysis of Fish Movements in Tanks
- 1 August 2021
- Published by Elsevier in SLAS Technology
- Vol. 26 (4) , 367-376
Open AccessepiTracker: A Framework for Highly Reliable Particle Tracking for the Quantitative Analysis of Fish Movements in Tanks
- Open Access
Extracellular Matrix Components Regulate Bone Sialoprotein Expression in MDA-MB-231 Breast Cancer Cells
Bone sialoprotein (BSP) has become a target in breast cancer research as it is associated with tumor progression and metastasis. The mechanisms underlying the regulation of BSP expression have been largely elusive. Given that BSP is involved in the homing of cancer cells in bone metastatic niches, we addressed regulatory effects of proteolytic cleavage and extracellular matrix components on BSP expression and distribution in cell culture models. Therefore, MDA-MB-231 human breast cancer cells were kept in 2D and 3D spheroid cultures and exposed to basement membrane extract in the presence or absence of matrix metalloproteinase 9 or the non-polar protease, dispase. Confocal imaging of immunofluorescence samples stained with different antibodies against human BSP demonstrated a strong inducing effect of basement membrane extract on anti-BSP immunofluorescence. Similarly, protease incubation led to acute upregulation of anti-BSP immunofluorescence signals, which was blocked by cycloheximide, suggesting de novo formation of BSP. In summary, our data show that extracellular matrix components play an important function in regulating BSP expression and hint at mechanisms for the formation of bone-associated metastasis in breast cancer that might involve local control of BSP levels by extracellular matrix degradation and release of growth factors.
Open AccessExtracellular Matrix Components Regulate Bone Sialoprotein Expression in MDA-MB-231 Breast Cancer Cells
Bone sialoprotein (BSP) has become a target in breast cancer research as it is associated with tumor progression and metastasis. The mechanisms underlying the regulation of BSP expression have been largely elusive. Given that BSP is involved in the homing of cancer cells in bone metastatic niches, we addressed regulatory effects of proteolytic cleavage and extracellular matrix components on BSP expression and distribution in cell culture models. Therefore, MDA-MB-231 human breast cancer cells were kept in 2D and 3D spheroid cultures and exposed to basement membrane extract in the presence or absence of matrix metalloproteinase 9 or the non-polar protease, dispase. Confocal imaging of immunofluorescence samples stained with different antibodies against human BSP demonstrated a strong inducing effect of basement membrane extract on anti-BSP immunofluorescence. Similarly, protease incubation led to acute upregulation of anti-BSP immunofluorescence signals, which was blocked by cycloheximide, suggesting de novo formation of BSP. In summary, our data show that extracellular matrix components play an important function in regulating BSP expression and hint at mechanisms for the formation of bone-associated metastasis in breast cancer that might involve local control of BSP levels by extracellular matrix degradation and release of growth factors.