Organotypic brain slice cultures are well-established and valuable three-dimensional (3D) models in neuroscience, particularly for cancer research. However, many existing 3D models rely on adult animals or are restricted to specific brain regions, such as the hippocampus. To address these limitations, we established a whole, large-format brain slice model that preserves the complex cellular architecture of the brain and its diverse cell populations. In addition, the use of 6-9-day-old postnatal mice allows a reduction in animal numbers by sectioning each brain into multiple contiguous slices of 300 µm thickness. The slices were maintained in vitro for 7–14 days under stable conditions and served as carrier tissue for brain tumor cell spheroids, thereby creating an ex vivo glioblastoma (GBM) model. To simulate tumor growth and invasion, spheroids of rat 9L/lacZ GBM cells were seeded onto the murine slices. Two different fluorescence-based methods were used to visualize and quantify the viability of the model. The AlamarBlue assay was used to identify the optimal culture medium and to longitudinally monitor whole-slice viability over a 10-day period. In parallel, the structural integrity of the brain slices and tumor invasion patterns were visualized using live/dead staining for up to 14 days post-slicing. Both methods demonstrated that the slices remained viable and stable for one week before a gradual decrease in viability. When co-cultured with GBM spheroids, the slices’ viability slightly decreased, likely due to the tumor’s invasive nature, which compromised the complex brain structures and cell networks. In summary, our ex vivo whole-brain slice GBM model preserves the complex organotypic brain and tumor structures essential for a robust in vitro GBM 3D model. The model stability for a minimum of 7 days, with the potential for extended culture duration, makes it a valuable tool in brain cancer research.

Wound healing is an intricate physiological process, with acute and chronic wounds imposing significant burdens on the healthcare systems worldwide. This study reports the development fully biodegradable silica gel fiber (SGF) scaffolds for enhanced skin tissue regeneration. Three types of wound dressings, differing in their structure, are fabricated: pressure-spun silica gel µ-fibers (pSGF) allowing cell penetration, electrospun sub-µ silica gel fibers (eSGF) mimicking an extracellular matrix (ECM)-like sub-µ-structure with narrow mesh sizes allowing no cell ingrowth, and a hybrid scaffold combining both fiber types (peSGF) that combines the advantages of both structures. Comprehensive characterization with different techniques and in-vitro degradation confirmed the complete biodegradation of scaffolds into water-soluble ortho-silicic acid (oSA). In-vitro studies demonstrated excellent biocompatibility, with peSGF showing good fibroblast attachment, proliferation, and cytoskeletal organization compared to other scaffolds. In-vivo studies further corroborated these findings, with peSGF exhibiting accelerated wound contraction, advanced re-epithelialization, and improved vascularization compared to controls. Histopathological analysis highlighted significant tissue regeneration, with peSGF scaffolds promoting collagen deposition and muscle fiber formation while minimizing scar tissue. These results emphasize the potential of peSGF combined with µ-fibers and ECM-like sub-µ-fibers as a promising solution for skin tissue regeneration and wound healing.

Advanced tissue-engineered respiratory models are essential for studying drug or cosmetic toxicity, infection biology and xenobiotic metabolism. Here, we investigated a polyamide 6 (PA6)-based electrospun stromal scaffold as a substitute for porcine-derived small intestinal submucosa (SIS) to build human airway mucosa tissue models at the air-liquid interface. We demonstrate that the porous PA6 scaffold supports extracellular matrix production by human nasal fibroblasts and facilitates the complete differentiation of respiratory epithelial cells to the mucociliary phenotype. These models reduce reliance on animal-derived materials, improve reproducibility, and minimize potential interference from animal-derived antigens and pathogens. Both PA6- and SIS-based models promote fibroblast migration, epithelial differentiation, and the expression of key xenobiotic metabolizing enzymes. They exhibit comparable epithelial barrier integrity and susceptibility to influenza A virus infections. These findings establish PA6 scaffolds as a suitable, animal-free alternative to the SIS to build human airway mucosa tissue models.

Advanced in vitro models are crucial for studying human airway biology. Our objective was the development and optimization of 3D in vitro models representing diverse airway regions, including deep lung alveolar region. This initiative was aimed at assessing the influence of selective scaffold materials on distinct airway co-culture models. While PET membranes (30 µm thickness) were unsuitable for alveolar models due to their stiffness and relatively high Young’s modulus, a combination of collagenous scaffolds seeded with Calu-3 cells and fibroblasts, showed increased mucus production going from week 1 to week 4 of air lift culture. Meanwhile standard electrospun polymer membrane (50–60 µm thick), which possesses a considerably low modulus of elasticity, offered higher flexibility and supported co-cultures of primary alveolar epithelial (huAEC) and endothelial cells (hEC) in concert with lung biopsy-derived fibroblasts which enhanced maturation of the tissue model. As published, designing human alveolar in vitro models require thin scaffold to mimic the required ultra-thin ECM, in addition to assuring right balanced AT1/AT2 ratio for biomimetic representation. We concluded that co-cultivation of primary/stem cells or cell lines has a higher influence on the function of the airway tissue models than the applied scaffolds.

The development of bone metastases poses a clinical challenge for patients with malignancies originating from diverse organs, including lungs, prostate, and breast. These metastatic events correlate with a poorer prognosis, and treatments are predominantly palliative. Furthermore, traditional in-vivo models for bone metastasis research have limitations, primarily due to species-specific differences. Therefore, we sought an alternative in-vitro model that is human cell based, robust and capable of a relatively quick establishment. Initially, we studied the potential of a decellularized porcine jejunum scaffold named Small Intestinal Submucosa with mucosa (SISmuc) and compared it to decellularized human bone and two synthetic scaffolds. We assessed various parameters related to osteoblast differentiation such as calcification, collagen I protein expression, and further differentiation to mature osteocytes indicated by sclerostin protein expression. The synthetic electrospun matrix demonstrated to be a good alternative to SISmuc, particularly when combined with bioactive glass 45S5, which enhanced calcification. We also incorporated osteoclasts into our model and demonstrated that monocytes could be more effectively differentiated into osteoclasts in a 3D environment compared to traditional 2D cultures. Within seven weeks, we successfully generate an in-vitro model incorporating osteoblasts, osteoclasts, and tumor cells. As a potential strategy for treating bone metastases, we found Chimeric Antigen Receptor T cells within lung and pancreatic bone metastasis models to be effective in reducing the number of tumor cells and inducing apoptosis. This innovative research establishes a foundation for more sophisticated 3D in-vitro models of bone metastasis, holding significant promise for advancing preclinical cancer research and therapeutic strategies.

Sinkholes can cause significant damage to infrastructures, agriculture, and endanger lives in active karst regions like the Dead Sea’s eastern shore at Ghor Al-Haditha. The common sinkhole mapping methods often require costly high-resolution data and manual, time-consuming expert analysis. This study introduces an efficient deep learning model designed to improve sinkhole mapping using accessible satellite imagery, which could enhance management practices related to sinkholes and other geohazards in evaporite karst regions. The developed AI system is centered around the U-Net architecture. The model was initially trained on a high-resolution drone dataset (0.1 m GSD, phase I), covering 250 sinkhole instances. Subsequently, it was additionally fine-tuned on a larger dataset from a Pleiades Neo satellite image (0.3 m GSD, phase II) with 1038 instances. The training process involved an automated image-processing workflow and strategic layer freezing and unfreezing to adapt the model to different input scales and resolutions. We show the usefulness of initial layer features learned on drone data, for the coarser, more readily-available satellite inputs. The validation revealed high detection accuracy for sinkholes, with phase I achieving a recall of 96.79% and an F1 score of 97.08%, and phase II reaching a recall of 92.06% and an F1 score of 91.23%. These results confirm the model’s accuracy and its capability to maintain high performance across varying resolutions. Our findings highlight the potential of using RGB visual bands for sinkhole detection across different karst environments. This approach provides a scalable, cost-effective solution for continuous mapping, monitoring, and risk mitigation related to sinkhole hazards. The developed system is not limited only to sinkholes however, and can be naturally extended to other geohazards as well. Moreover, since it currently uses U-Net as a backbone, the system can be extended to incorporate super-resolution techniques, leveraging U-Net based latent diffusion models to address the smaller-scale, ambiguous geo-structures that are often found in geoscientific data.

Machine learning (ML) algorithms can be used in Earth system models (ESMs) to emulate sub-grid-scale processes. Due to the statistical nature of ML algorithms and the high complexity of ESMs, these hybrid ML ESMs require careful validation. Simulation stability needs to be monitored in fully coupled simulations, and the plausibility of results needs to be evaluated in suitable experiments.We present the coupling of SuperdropNet, a machine learning model for emulating warm-rain processes in cloud microphysics, with ICON (Icosahedral Nonhydrostatic) model v2.6.5. SuperdropNet is trained on computationally expensive droplet-based simulations and can serve as an inexpensive proxy within weather prediction models. SuperdropNet emulates the collision-coalescence of rain and cloud droplets in a warm-rain scenario and replaces the collision-coalescence process in the two-moment cloud microphysics scheme.We address the technical challenge of integrating SuperdropNet, developed in Python and PyTorch, into ICON, written in Fortran, by implementing three different coupling strategies: embedded Python via the C foreign function interface (CFFI), pipes, and coupling of program components via Yet Another Coupler (YAC).We validate the emulator in the warm-bubble scenario and find that SuperdropNet runs stably within the experiment. By comparing experiment outcomes of the two-moment bulk scheme with SuperdropNet, we find that the results are physically consistent and discuss differences that are observed in several diagnostic variables.In addition, we provide a quantitative and qualitative computational benchmark for three different coupling strategies - embedded Python, coupler YAC, and pipes - and find that embedded Python is a useful software tool for validating hybrid ML ESMs.

Earthquake prediction relies on identification of distinctive patterns of precursory parameters that might precede a large earthquake. However, these patterns are typically not reliably observed in the field. Laboratory stick-slip experiments provide an analog of seismic cycle observed in nature in fully controlled conditions with the associated Acoustic Emission (AE) activity reproducing basic characteristics of seismicity preceding and following the large lab earthquake. Recent laboratory studies showed that the deployment of Machine Learning/Artificial Intelligence techniques has lead to new state of the art results in lab earthquake prediction on smooth faults while using simple statistical features derived from raw AE signals and AE-derived catalogs. However, not enough work has been done on explainability of earthquakes preparatory process on rough faults by leveraging deep learning techniques. In this work we attempt to mitigate this gap and analyze/grade a pool of explainable seismo-mechanical features through the eyes of neural networks.We used AE data from three laboratory stick-slip experiments performed in triaxial pressure vessel on Westerly Granite samples. Samples were first fractured at 75MPa confining pressure creating rough fault surfaces. The following stick-slip experiments were performed at constant displacement rate. The experimental procedure led to an extremely complex slip pattern composed of large and small slips of the whole surface, as well as the confined slips highlighted only with AE data and no externally measured slip. The AE catalog was used to extract temporal evolution of 16 seismo-mechanical and statistical features characterizing evolution of stress and damage in response to the axial stress change. The feature pool included clearly physically interpretable parameters such as AE rates, b-value, fractal dimension, AE localization, clustering and triggering properties, and features characterizing the variability of local stress field. We apply explainable AI techniques to identify what features are more important to forecast axial stress and stress drop. Our feature ranking and importance evaluation with the help of neural networks can serve as an indicator as to what research directions are more promising to take for further feature engineering efforts with an emphasis on explainability of earthquake phenomena.

Bacterial infection is a crucial complication in implant restoration, in particular in permanent skin-penetrating implants. Therein, the resulting gap between transcutaneous implant and skin represents a permanent infection risk, limiting the field of application and the duration of application. To overcome this limitation, a tight physiological connection is required to achieve a biological and mechanical welding for a long-term stable closure including self-healing probabilities. This study describes a new approach, wherein the implant is connected covalently to a highly porous electrospun fleece featuring physiological dermal integration potential. The integrative potential of the scaffold is shown in vitro and confirmed in vivo, further demonstrating tissue integration by neovascularization, extracellular matrix formation, and prevention of encapsulation. To achieve a covalent connection between fleece and implant surface, self-initiated photografting and photopolymerization of hydroxyethylmethacrylate is combined with a new crosslinker (methacrylic acid coordinated titanium-oxo clusters) on proton-abstractable implant surfaces. For implant modification, the attached fleece is directed perpendicular from the implant surface into the surrounding dermal tissue. First in vitro skin implantations demonstrate the implants' dermal integration capability as well as wound closure potential on top of the fleece by epithelialization, establishing a bacteria-proof and self-healing connection of skin and transcutaneous implant.

In recent years, machine learning (ML) based parameterizations have become increasingly common in Earth System Models (ESM). Sub-grid scale physical processes that would be computationally too expensive, e.g., atmospheric chemistry and cloud microphysics, can be emulated by ML algorithms such as neural networks.Neural networks are trained first on simulations of the sub-grid scale process that is to be emulated. They are then used in so-called inference mode to make predictions during the ESM run, replacing the original parameterization. Training usually requires GPUs, while inference may be done on CPU architectures.At first, neural networks are evaluated offline, i.e., independently of the ESM on appropriate datasets. However, their performance can ultimately only be evaluated in an online setting, where the ML algorithm is coupled to the ESM, including nonlinear interactions.We want to shorten the time spent in neural network development and offline testing and move quickly to online evaluation of ML components in our ESM of choice, ICON (Icosahedral Nonhydrostatic Weather and Climate Model). Since ICON is written in Fortran, and modern ML algorithms are developed in the Python ecosystem, this requires efficient bridges between the two programming languages. The Fortran-Python bridge must be flexible to allow for iterative development of the neural network. Changes to the ESM codebase should be as few as possible, and the runtime overhead should not limit development.In our contribution we explore three strategies to call the neural network inference from within Fortran using (i) embedded Python code compiled in a dynamic library, (ii) pipes, and (iii) MPI using the ICON coupler YAC. We provide quantitative benchmarks for the proposed Fortran-Python bridges and assess their overall suitability in a qualitative way to derive best practices. The Fortran-Python bridge enables scientists and developers to evaluate ML components in an online setting, and can be extended to other parameterizations and ESMs.

This dataset provides the experiment results for the publication " Efficient and Stable Coupling of the SuperdropNet Deep Learning-based Cloud Microphysics (v0.1.0) to the ICON Climate and Weather Model (v2.6.5) ", to be submitted.

Artificial intelligence (AI) is proliferating and developing faster than any domain scientist can adapt. To support the scientific enterprise in the Helmholtz association, a network of AI specialists has been set up to disseminate AI expertise as an infrastructure among domain scientists. As this effort exposes an evolutionary step in science organization in Germany, this article aspires to describe our setup, goals, and motivations. We comment on past experiences, current developments, and future ideas as we bring our expertise as an infrastructure closer to scientists across our organization. We hope that this offers a brief yet insightful view of our activities as well as inspiration for other science organizations.

We studied the impact of two different thawing mechanisms on the microstructure of defrosted cryopreserved 3D tissue models using transmission Mueller microscopy and a statistical analysis of polarimetric images of thin histological sections of defrosted tissue models. The cryopreserved 3D tissue models were thawed by using either a 37 °C water bath or radio-frequency inductive heating with the magnetic nanoparticles embedded into the 3D tissue model during the preparation process. Polarimetric measurements were conducted at 700 nm and the acquired Mueller matrices of the samples were post-processed using the differential decomposition and the statistical analysis of the maps of the azimuth of the optic axis. Our results indicate the sensitivity of polarimetry to the changes in thawed tissue morphology compared to that of reference non-frozen tissue. Thus, Mueller microscopy can be used as a fast complementary technique to the currently accepted gold standard methods for the assessment of the cryopreserved tissue microstructure after thawing.

Recently, Machine learning (ML) has been widely utilized for laboratory earthquake (labquake) prediction using various types of data. This study pioneers in time to failure (TTF) prediction based on ML using acoustic emission (AE) records from three laboratory stick-slip experiments performed on Westerly granite samples with naturally fractured rough faults, more similar to the heterogeneous fault structures in the nature. 47 catalog-driven seismo-mechanical and statistical features are extracted introducing some new features based on focal mechanism. A regression voting ensemble of Long-Short Term Memory (LSTM) networks predicts TTF with a coefficient of determination (R2) of 70% on the test dataset. Feature importance analysis revealed that AE rate, correlation integral, event proximity, and focal mechanism-based features are the most important features for TTF prediction. Results reveal that the network uses all information among the features for prediction, including general trends in high correlated features as well as fine details about local variations and fault evolution involved in low correlated features. Therefore, some highly correlated and physically meaningful features may be considered less important for TTF prediction due to their correlation with other important features. Our study provides a ground for applying catalog-driven to constrain TTF of complex heterogeneous rough faults, which is capable to be developed for real application.

The development of novel fibrous biomaterials and further processing of medical devices is still challenging. For instance, titanium(IV) oxide is a well-established biocompatible material, and the synthesis of TiOx particles and coatings via the sol-gel process has frequently been published. However, synthesis protocols of sol-gel-derived TiOx fibers are hardly known. In this publication, the authors present a synthesis and fabrication of purely sol-gel-derived TiOx fiber fleeces starting from the liquid sol-gel precursor titanium ethylate (TEOT). Here, the α-hydroxy-carboxylic acid lactic acid (LA) was used as a chelating ligand to reduce the reactivity towards hydrolysis of TEOT enabling a spinnable sol. The resulting fibers were processed into a non-woven fleece, characterized with FTIR, 13C-MAS-NMR, XRD, and screened with regard to their stability in physiological solution. They revealed an unexpected dependency between the LA content and the dissolution behavior. Finally, in vitro cell culture experiments proved their potential suitability as an open-mesh structured scaffold material, even for challenging applications such as therapeutic medicinal products (ATMPs).

Porous Fibrous Scaffolds The dimensions of electrospun fibers are predestined for synthetic replacement of extracellular matrix structures. In article number 2106780, Tobias Weigel and co-workers generate highly porous electrospun scaffolds, featuring an ensemble of loose fibers and pores of appropriate size to enable applications in engineering animal-free 3D biomedical tissues. The mechanical properties and structural composition describe a tradeoff between current synthetic porous scaffolds and animal-derived hydrogels, which are deployed as electrospun-based multilayered skin equivalents.

GNSS Reflectometry (GNSS-R) is a novel remote sensing technique for the monitoring of geophysical parameters using reflected GNSS signals from the Earth's surface. Ocean wind speed monitoring is the main objective of the recently launched Cyclone GNSS (CyGNSS), a GNSS-R constellation of eight microsatellites, launched in late 2016. In this study, the capability of deep learning, especially, for an operational wind speed data derivation from the measured Delay-Doppler Maps (DDMs) is characterized. CyGNSSnet is based on convolutional layers for the feature extraction from bistatic radar cross section (BRCS) DDMs, along with fully connected layers for processing ancillary technical and higher-level input parameters. The best architecture is determined on a validation set and is evaluated over a completely blind dataset from a different time span than that of the training data to validate the generality of the model for operational usage. After a data quality control, CyGNSSnet results in an RMSE of 1.36 m/s leading to a significant improvement by 28% in comparison to the officially operational retrieval algorithm. The RMSE is the lowest among those seen in the literature for any conventional or machine learning-based algorithm. The benefits of the convolutional layers, the advantages and weaknesses of the model are discussed. CyGNSSnet offers efficient processing of GNSS-R measurements for high-quality global ocean winds.

This research aims at developing and applying a machine learning based algorithm to detect geological structural features at the exposed Dead Sea shoreline. We focus on sinkhole, stream-channel and crack features that appear in different material at the coastlines, and post partly a threat for the local population and infrastructure. We use high resolution orthophotos and satellite images from the last years, as well as derived topographic models to reach an automatic identification and classification of these structures. The aim is to train a convolutional neural network that can identify these structures on recent datasets from satellite images in order to establish an automated detection of hazardous and active zones in the area. Furthermore, we use the algorithms to detect structures in the shallow waters of the lake.

The capability of Deep Learning (DL) for operational wind speed retrieval from the measured Delay-Doppler Maps (DDMs) is recently characterized. It is shown that such techniques can lead to a significant improvement in the derived atmospheric data products. A global ocean dataset is developed processing the measurements of NASA Cyclone GNSS (CYGNSS). The model is based on convolutional layers for direct feature extraction from bistatic radar cross-section (BRCS) DDMs and fully connected layers for processing ancillary technical and higher-level input parameters. This model leads to an RMSE of 1.36 m/s and a significant improvement of 28% in comparison to the officially operational retrieval algorithm. From the theoretical knowledge, several error sources are known, the modeling and correction of which is not easy due to their highly nonlinear interaction with other and the dependent parameters. DL is potentially able to learn such trends and correct the associated biases. For instance, rain splash on the ocean surface and swell waves alter the surface roughness, and consequently, the GNSS scattering patterns, which appear as a considerable bias in GNSS-R wind products. The magnitude of such biases is nonlinearly dependent on several technical and environmental parameters including the reflection geometry, and ocean surface state. After a brief introduction to the known physical mechanisms, we discuss how a DL-based fusion with data on bias-causing parameters, can improve the wind speed predictions.

This presentation reports on support done under the aegis of Helmholtz AI for a wide range of machine learning based solutions for research questions related to Earth and Environmental sciences. We will give insight into typical problem statements from Earth observation and Earth system modeling that are good candidates for experimentation with ML methods and report on our accumulated experience tackling such challenges with individual support projects. We address these projects in an agile, iterative manner and during the definition phase, we direct special attention towards assembling practically meaningful demonstrators within a couple of months. A recent focus of our work lies on tackling software engineering concerns for building ML-ESM hybrids. Our implementation workflow covers stages from data exploration to model tuning. A project may often start with evaluating available data and deciding on basic feasibility, apparent limitations such as biases or a lack of labels, and splitting into training and test data. Setting up a data processing workflow to subselect and compile training data is often the next step, followed by setting up a model architecture. We have made good experience with automatic tooling to tune hyperparameters and test and optimize network architectures. In typical implementation projects, these stages may repeat many times to improve results and cover aspects such as errors due to confusing samples, incorporating domain model knowledge, testing alternative architectures and ML approaches, and dealing with memory limitations and performance optimization. Over the past two years, we have supported Helmholtz-based researchers from many subdisciplines on making the best use of ML methods along with these steps. Example projects include wind speed regression on GNSS-R data, emulation of atmospheric chemistry modeling, Earth System model parameterizations with ML, marine litter detection, and rogue waves prediction. The poster presentation will highlight selected best practices across these projects. We are happy to share our experience as it may prove useful to applications in wider Earth System modeling. If you are interested in discussing your challenge with us, please feel free to chat with us.