Using dropout based active learning and surrogate models in the inverse viscoelastic parameter identification of human brain tissue.

Inverse mechanical parameter identification enables the characterization of ultrasoft materials, for which it is difficult to achieve homogeneous deformation states. However, this usually involves high computational costs that are mainly determined by the complexity of the forward model. While simulation methods like finite element models can capture nearly arbitrary geometries and implement involved constitutive equations, they are also computationally expensive. Machine learning models, such as neural networks, can help mitigate this problem when they are used as surrogate models replacing the complex high fidelity models. Thereby, they serve as a reduced order model after an initial training phase, where they learn the relation of in- and outputs of the high fidelity model. The generation of the required training data is computationally expensive due to the necessary simulation runs. Here, active learning techniques enable the selection of the "most rewarding" training points in terms of estimated gained accuracy for the trained model. In this work, we present a recurrent neural network that can well approximate the output of a viscoelastic finite element simulation while significantly speeding up the evaluation times. Additionally, we use Monte-Carlo dropout based active learning to identify highly informative training data. Finally, we showcase the potential of the developed pipeline by identifying viscoelastic material parameters for human brain tissue.

Inverse identification of region-specific hyperelastic material parameters for human brain tissue.

The identification of material parameters accurately describing the region-dependent mechanical behavior of human brain tissue is crucial for computational models used to assist, e.g., the development of safety equipment like helmets or the planning and execution of brain surgery. While the division of the human brain into different anatomical regions is well established, knowledge about regions with distinct mechanical properties remains limited. Here, we establish an inverse parameter identification scheme using a hyperelastic Ogden model and experimental data from multi-modal testing of tissue from 19 anatomical human brain regions to identify mechanically distinct regions and provide the corresponding material parameters. We assign the 19 anatomical regions to nine governing regions based on similar parameters and microstructures. Statistical analyses confirm differences between the regions and indicate that at least the corpus callosum and the corona radiata should be assigned different material parameters in computational models of the human brain. We provide a total of four parameter sets based on the two initial Poisson's ratios of 0.45 and 0.49 as well as the pre- and unconditioned experimental responses, respectively. Our results highlight the close interrelation between the Poisson's ratio and the remaining model parameters. The identified parameters will contribute to more precise computational models enabling spatially resolved predictions of the stress and strain states in human brains under complex mechanical loading conditions.

Mechanisms of mechanical load transfer through brain tissue.

Brain injuries are often characterized by diffusely distributed axonal and vascular damage invisible to medical imaging techniques. The spatial distribution of mechanical stresses and strains plays an important role, but is not sufficient to explain the diffuse distribution of brain lesions. It remains unclear how forces are transferred from the organ to the cell scale and why some cells are damaged while neighboring cells remain unaffected. To address this knowledge gap, we subjected histologically stained fresh human and porcine brain tissue specimens to compressive loading and simultaneously tracked cell and blood vessel displacements. Our experiments reveal different mechanisms of load transfer from the organ or tissue scale to single cells, axons, and blood vessels. Our results show that cell displacement fields are inhomogeneous at the interface between gray and white matter and in the vicinity of blood vessels-locally inducing significant deformations of individual cells. These insights have important implications to better understand injury mechanisms and highlight the importance of blood vessels for the local deformation of the brain's cellular structure during loading.

Tissue-Scale Biomechanical Testing of Brain Tissue for the Calibration of Nonlinear Material Models.

Brain tissue is one of the most complex and softest tissues in the human body. Due to its ultrasoft and biphasic nature, it is difficult to control the deformation state during biomechanical testing and to quantify the highly nonlinear, time-dependent tissue response. In numerous experimental studies that have investigated the mechanical properties of brain tissue over the last decades, stiffness values have varied significantly. One reason for the observed discrepancies is the lack of standardized testing protocols and corresponding data analyses. The tissue properties have been tested on different length and time scales depending on the testing technique, and the corresponding data have been analyzed based on simplifying assumptions. In this review, we highlight the advantage of using nonlinear continuum mechanics based modeling and finite element simulations to carefully design experimental setups and protocols as well as to comprehensively analyze the corresponding experimental data. We review testing techniques and protocols that have been used to calibrate material model parameters and discuss artifacts that might falsify the measured properties. The aim of this work is to provide standardized procedures to reliably quantify the mechanical properties of brain tissue and to more accurately calibrate appropriate constitutive models for computational simulations of brain development, injury and disease. Computational models can not only be used to predictively understand brain tissue behavior, but can also serve as valuable tools to assist diagnosis and treatment of diseases or to plan neurosurgical procedures. © 2022 The Authors. Current Protocols published by Wiley Periodicals LLC.

Oxidized Hyaluronic Acid-Gelatin-Based Hydrogels for Tissue Engineering and Soft Tissue Mimicking.

Hydrogels are ideal materials for mimicking and engineering soft tissue. Hyaluronic acid is a linear polysaccharide native to the human extracellular matrix. In this study, we first develop and characterize two hydrogel compositions built from oxidized HA and gelatin with and without alginate-di-aldehyde (ADA) crosslinked by ionic and enzymatic agents with potential applications in soft tissue engineering and tissue mimicking structures. The stability under incubation conditions was improved by adjusting crosslinking times. Through large-strain mechanical measurements, the hydrogels' properties were compared to human brain tissue and the samples containing ADA revealed similar mechanical properties to the native tissue specimens in cyclic compression-tension. In vitro characterization demonstrated a high viability of encapsulated mouse embryonic fibroblasts and a spreading of the cells in case of ADA-free samples. Impact statement Brain mimicking materials are required in several medical and industrial fields for the development of safety gear, testing of medical imaging techniques, surgical training, tissue engineering, and modeling of the mechanical behavior of tissues. The materials must resemble the microstructure, chemistry, and mechanical properties of the native tissue extracellular matrix while being adjustable in degradation to suit the various applications. In this article, different methods are used to evaluate a novel hydrogel material and its suitability as brain mimicking matrix.

Dendritic Cell-Triggered Immune Activation Goes along with Provision of (Leukemia-Specific) Integrin Beta 7-Expressing Immune Cells and Improved Antileukemic Processes.

Integrin beta 7 (ß7), a subunit of the integrin receptor, is expressed on the surface of immune cells and mediates cell-cell adhesions and interactions, e.g., antitumor or autoimmune reactions. Here, we analyzed, whether the stimulation of immune cells by dendritic cells (of leukemic derivation in AML patients or of monocyte derivation in healthy donors) leads to increased/leukemia-specific ß7 expression in immune cells after T-cell-enriched mixed lymphocyte culture-finally leading to improved antileukemic cytotoxicity. Healthy, as well as AML and MDS patients' whole blood (WB) was treated with Kit-M (granulocyte-macrophage colony-stimulating factor (GM-CSF) + prostaglandin E1 (PGE1)) or Kit-I (GM-CSF + Picibanil) in order to generate DCs (DCleu or monocyte-derived DC), which were then used as stimulator cells in MLC. To quantify antigen/leukemia-specific/antileukemic functionality, a degranulation assay (DEG), an intracellular cytokine assay (INTCYT) and a cytotoxicity fluorolysis assay (CTX) were used. (Leukemia-specific) cell subtypes were quantified via flow cytometry. The Kit treatment of WB (compared to the control) resulted in the generation of DC/DCleu, which induced increased activation of innate and adaptive cells after MLC. Kit-pretreated WB (vs. the control) led to significantly increased frequencies of ß7-expressing T-cells, degranulating and intracellular cytokine-producing ß7-expressing immune cells and, in patients' samples, increased blast lysis. Positive correlations were found between the Kit-M-mediated improvement of blast lysis (vs. the control) and frequencies of ß7-expressing T-cells. Our findings indicate that DC-based immune therapies might be able to specifically activate the immune system against blasts going along with increased frequencies of (leukemia-specific) ß7-expressing immune cells. Furthermore, ß7 might qualify as a predictor for the efficiency and the success of AML and/or MDS therapies.

Unraveling the Local Relation Between Tissue Composition and Human Brain Mechanics Through Machine Learning.

The regional mechanical properties of brain tissue are not only key in the context of brain injury and its vulnerability towards mechanical loads, but also affect the behavior and functionality of brain cells. Due to the extremely soft nature of brain tissue, its mechanical characterization is challenging. The response to loading depends on length and time scales and is characterized by nonlinearity, compression-tension asymmetry, conditioning, and stress relaxation. In addition, the regional heterogeneity-both in mechanics and microstructure-complicates the comprehensive understanding of local tissue properties and its relation to the underlying microstructure. Here, we combine large-strain biomechanical tests with enzyme-linked immunosorbent assays (ELISA) and develop an extended type of constitutive artificial neural networks (CANNs) that can account for viscoelastic effects. We show that our viscoelastic constitutive artificial neural network is able to describe the tissue response in different brain regions and quantify the relevance of different cellular and extracellular components for time-independent (nonlinearity, compression-tension-asymmetry) and time-dependent (hysteresis, conditioning, stress relaxation) tissue mechanics, respectively. Our results suggest that the content of the extracellular matrix protein fibronectin is highly relevant for both the quasi-elastic behavior and viscoelastic effects of brain tissue. While the quasi-elastic response seems to be largely controlled by extracellular matrix proteins from the basement membrane, cellular components have a higher relevance for the viscoelastic response. Our findings advance our understanding of microstructure - mechanics relations in human brain tissue and are valuable to further advance predictive material models for finite element simulations or to design biomaterials for tissue engineering and 3D printing applications.

High-shear- and-thrombin-inducible platelet adhesion and aggregation in patients undergoing percutaneous coronary intervention. Effects of unfractionated heparin versus bivalirudin.

Thrombin-generation and activation of platelets during percutaneous coronary intervention (PCI) play a key role for early thrombotic events. Heparin and bivalirudin are approved anticoagulants for PCI. We examined the specific effects of these anticoagulants on platelet adhesion and aggregation under high shear conditions, and the presence of excess thrombin. To simulate in vivo conditions that may precipitate a bleeding/thrombotic event, we added thrombin in vitro to blood samples from 89 stable patients who had been randomly assigned to receive heparin or bivalirudin for elective PCI and examined thrombin-inducible platelet adhesion and aggregation under high shear conditions. Platelet adhesion increased by 10% of baseline with heparin, but decreased by 20% with bivalirudin (p=0.0047). Thrombin-inducible platelet adhesion and size of aggregates was equally inhibited by heparin and bivalirudin. Thus, under high shear conditions and excessive thrombin generation as they occur in atherosclerotic vascular compartments and acute vascular syndromes, heparin and bivalirudin inhibit thrombin-induced platelet adhesion and aggregation to a similar extent, while they have opposite effects on platelet adhesion in the absence of thrombin.

Thrombin-inducible platelet adhesion and regulation of the platelet seven-transmembrane thrombin receptor-1 (PAR-1): effects of unfractionated heparin and lepirudin.

Heparin may induce platelet activation and even heparin-induced thrombocytopenia. Lepirudin has been approved for HIT treatment. We speculated that lepirudin inhibits platelet function under high shear and the platelet thrombin receptor PAR-1 better than heparin. Thrombin-inducible platelet adherence under high shear conditions and the expression of PAR-1 were studied after samples from healthy donors were exposed in vitro to increasing concentrations of unfractionated heparin or lepirudin. Compared to baseline and to lepirudin, heparin induced platelet P-selectin expression (p = 0.04). Platelet adherence increased slightly in the presence of lepirudin, but not heparin (p = 0.04). Thrombin-inducible platelet aggregate formation and consecutive adherence under high shear conditions was inhibited by both anticoagulants (p = 0.004). Further, heparin and lepirudin inhibited thrombin-inducible cleavage and internalization of PAR-1 at a dosage of 1.0 U/ml and 1.6 microg/ml, respectively (p = 0.004). Thus, heparin and lepirudin inhibit thrombin-inducible platelet activation in vitro to a similar extent.

Regulation of PAR-1 in patients undergoing percutaneous coronary intervention: effects of unfractionated heparin and bivalirudin.

Aims We examined the specific effects of unfractionated heparin and bivalirudin on thrombin-inducible platelet PAR-1 in patients undergoing percutaneous coronary intervention (PCI). Methods and results To simulate in vivo conditions that may precipitate a bleeding event, we added thrombin in vitro to blood samples from 89 patients who had been randomly assigned to receive heparin or bivalirudin for elective PCI and examined thrombin-inducible PAR-1 expression. Thrombin-inducible cleavage of PAR-1 was inhibited by heparin, but not affected by bivalirudin (P = 0.0001). Further, PAR-1 internalization was more effectively inhibited by heparin than bivalirudin (P = 0.002). Conclusion Heparin has stronger inhibitory effects on thrombin-dependent PAR-1 cleavage and internalization, thus providing a biological explanation for lower clinical bleeding rates with bivalirudin.