White Blood Cell (WBC) Leukaemia is caused by excessive production of leukocytes in the bone marrow, and image-based detection of malignant WBCs is important for its detection. Convolutional Neural ...Networks (CNNs) present the current state-of-the-art for this type of image classification, but their computational cost for training and deployment can be high. We here present an improved hybrid approach for efficient classification of WBC Leukemia. We first extract features from WBC images using VGGNet, a powerful CNN architecture, pre-trained on ImageNet. The extracted features are then filtered using a statistically enhanced Salp Swarm Algorithm (SESSA). This bio-inspired optimization algorithm selects the most relevant features and removes highly correlated and noisy features. We applied the proposed approach to two public WBC Leukemia reference datasets and achieve both high accuracy and reduced computational complexity. The SESSA optimization selected only 1 K out of 25 K features extracted with VGGNet, while improving accuracy at the same time. The results are among the best achieved on these datasets and outperform several convolutional network models. We expect that the combination of CNN feature extraction and SESSA feature optimization could be useful for many other image classification tasks.
Osteocytes and their cell processes reside in a large, interconnected network of voids pervading the mineralized bone matrix of most vertebrates. This osteocyte lacuno-canalicular network (OLCN) is ...believed to play important roles in mechanosensing, mineral homeostasis, and for the mechanical properties of bone. While the extracellular matrix structure of bone is extensively studied on ultrastructural and macroscopic scales, there is a lack of quantitative knowledge on how the cellular network is organized. Using a recently introduced imaging and quantification approach, we analyze the OLCN in different bone types from mouse and sheep that exhibit different degrees of structural organization not only of the cell network but also of the fibrous matrix deposited by the cells. We define a number of robust, quantitative measures that are derived from the theory of complex networks. These measures enable us to gain insights into how efficient the network is organized with regard to intercellular transport and communication. Our analysis shows that the cell network in regularly organized, slow-growing bone tissue from sheep is less connected, but more efficiently organized compared to irregular and fast-growing bone tissue from mice. On the level of statistical topological properties (edges per node, edge length and degree distribution), both network types are indistinguishable, highlighting that despite pronounced differences at the tissue level, the topological architecture of the osteocyte canalicular network at the subcellular level may be independent of species and bone type. Our results suggest a universal mechanism underlying the self-organization of individual cells into a large, interconnected network during bone formation and mineralization.
Spatial biological networks are abundant on all scales of life, from single cells to ecosystems, and perform various important functions including signal transmission and nutrient transport. These ...biological functions depend on the architecture of the network, which emerges as the result of a dynamic, feedback-driven developmental process. While cell behavior during growth can be genetically encoded, the resulting network structure depends on spatial constraints and tissue architecture. Since network growth is often difficult to observe experimentally, computer simulations can help to understand how local cell behavior determines the resulting network architecture. We present here a computational framework based on directional statistics to model network formation in space and time under arbitrary spatial constraints. Growth is described as a biased correlated random walk where direction and branching depend on the local environmental conditions and constraints, which are presented as 3D multilayer grid. To demonstrate the application of our tool, we perform growth simulations of a dense network between cells and compare the results to experimental data from osteocyte networks in bone. Our generic framework might help to better understand how network patterns depend on spatial constraints, or to identify the biological cause of deviations from healthy network function.
Display omitted
Despite of the progress made to engineer structured microtissues such as BioMEMS and 3D bioprinting, little control exists how microtissues transform as they mature, as the misbalance ...between cell-generated forces and the strength of cell-cell and cell-substrate contacts can result in unintended tissue deformations and ruptures. To develop a quantitative perspective on how cellular contractility, scaffold curvature and cell-substrate adhesion control such rupture processes, human aortic smooth muscle cells were grown on glass substrates with submillimeter semichannels. We quantified cell sheet detachment from 3D confocal image stacks as a function of channel curvature and cell sheet tension by adding different amounts of Blebbistatin and TGF-β to inhibit or enhance cell contractility, respectively. We found that both higher curvature and higher contractility increased the detachment probability. Variations of the adhesive strength of the protein coating on the substrate revealed that the rupture plane was localized along the substrate-extracellular matrix interface for non-covalently adsorbed adhesion proteins, while the collagen-integrin interface ruptured when collagen I was covalently crosslinked to the substrate. Finally, a simple mechanical model is introduced that quantitatively explains how the tuning of substrate curvature, cell sheet contractility and adhesive strength can be used as tunable parameters as summarized in a first semi-quantitative phase diagram. These parameters can thus be exploited to either inhibit or purposefully induce a collective detachment of sheet-like microtissues for the use in tissue engineering and regenerative therapies.
Despite of the significant progress in 3D tissue fabrication technologies at the microscale, there is still no quantitative model that can predict if cells seeded on a 3D structure maintain the imposed geometry while they form a continuous microtissue. Especially, detachment or loss of shape control of growing tissue is a major concern when designing 3D-structured scaffolds. Utilizing semi-cylindrical channels and vascular smooth muscle cells, we characterized how geometrical and mechanical parameters such as curvature of the substrate, cellular contractility, or protein-substrate adhesion strength tune the catastrophic detachment of microtissue. Observed results were rationalized by a theoretical model. The phase diagram showing how unintended tissue detachment progresses would help in designing of mechanically-balanced 3D scaffolds in future tissue engineering applications.
This study investigated how substrate geometry influences in-vitro tissue formation at length scales much larger than a single cell. Two-millimetre thick hydroxyapatite plates containing circular ...pores and semi-circular channels of 0.5 mm radius, mimicking osteons and hemi-osteons respectively, were incubated with MC3T3-E1 cells for 4 weeks. The amount and shape of the tissue formed in the pores, as measured using phase contrast microscopy, depended on the substrate geometry. It was further demonstrated, using a simple geometric model, that the observed curvature-controlled growth can be derived from the assembly of tensile elements on a curved substrate. These tensile elements are cells anchored on distant points of the curved surface, thus creating an actin "chord" by generating tension between the adhesion sites. Such a chord model was used to link the shape of the substrate to cell organisation and tissue patterning. In a pore with a circular cross-section, tissue growth increases the average curvature of the surface, whereas a semi-circular channel tends to be flattened out. Thereby, a single mechanism could describe new tissue growth in both cortical and trabecular bone after resorption due to remodelling. These similarities between in-vitro and in-vivo patterns suggest geometry as an important signal for bone remodelling.
Localization-based super-resolution microscopy resolves macromolecular structures down to a few nanometers by computationally reconstructing fluorescent emitter coordinates from diffraction-limited ...spots. The most commonly used algorithms are based on fitting parametric models of the point spread function (PSF) to a measured photon distribution. These algorithms make assumptions about the symmetry of the PSF and thus, do not work well with irregular, non-linear PSFs that occur for example in confocal lifetime imaging, where a laser is scanned across the sample. An alternative method for reconstructing sparse emitter sets from noisy, diffraction-limited images is compressed sensing, but due to its high computational cost it has not yet been widely adopted. Deep neural network fitters have recently emerged as a new competitive method for localization microscopy. They can learn to fit arbitrary PSFs, but require extensive simulated training data and do not generalize well. A method to efficiently fit the irregular PSFs from confocal lifetime localization microscopy combining the advantages of deep learning and compressed sensing would greatly improve the acquisition speed and throughput of this method.
Here we introduce ReCSAI, a compressed sensing neural network to reconstruct localizations for confocal dSTORM, together with a simulation tool to generate training data. We implemented and compared different artificial network architectures, aiming to combine the advantages of compressed sensing and deep learning. We found that a U-Net with a recursive structure inspired by iterative compressed sensing showed the best results on realistic simulated datasets with noise, as well as on real experimentally measured confocal lifetime scanning data. Adding a trainable wavelet denoising layer as prior step further improved the reconstruction quality.
Our deep learning approach can reach a similar reconstruction accuracy for confocal dSTORM as frame binning with traditional fitting without requiring the acquisition of multiple frames. In addition, our work offers generic insights on the reconstruction of sparse measurements from noisy experimental data by combining compressed sensing and deep learning. We provide the trained networks, the code for network training and inference as well as the simulation tool as python code and Jupyter notebooks for easy reproducibility.
Scaffolds for tissue engineering are usually designed to support cell viability with large adhesion surfaces and high permeability to nutrients and oxygen. Recent experiments support the idea that, ...in addition to surface roughness, elasticity and chemistry, the macroscopic geometry of the substrate also contributes to control the kinetics of tissue deposition. In this study, a previously proposed model for the behavior of osteoblasts on curved surfaces is used to predict the growth of bone matrix tissue in pores of different shapes. These predictions are compared to in vitro experiments with MC3T3-E1 pre-osteoblast cells cultivated in two-millimeter thick hydroxyapatite plates containing prismatic pores with square- or cross-shaped sections. The amount and shape of the tissue formed in the pores measured by phase contrast microscopy confirms the predictions of the model. In cross-shaped pores, the initial overall tissue deposition is twice as fast as in square-shaped pores. These results suggest that the optimization of pore shapes may improve the speed of ingrowth of bone tissue into porous scaffolds.
Myofibroblasts orchestrate wound healing processes, and if they remain activated, they drive disease progression such as fibrosis and cancer. Besides growth factor signaling, the local extracellular ...matrix (ECM) and its mechanical properties are central regulators of these processes. It remains unknown whether transforming growth factor-β (TGF-β) and tensile forces work synergistically in up-regulating the transition of fibroblasts into myofibroblasts and whether myofibroblasts undergo apoptosis or become deactivated by other means once tissue homeostasis is reached. We used three-dimensional microtissues grown in vitro from fibroblasts in macroscopically engineered clefts for several weeks and found that fibroblasts transitioned into myofibroblasts at the highly tensed growth front as the microtissue progressively closed the cleft, in analogy to closing a wound site. Proliferation was up-regulated at the growth front, and new highly stretched fibronectin fibers were deposited, as revealed by fibronectin fluorescence resonance energy transfer probes. As the tissue was growing, the ECM underneath matured into a collagen-rich tissue containing mostly fibroblasts instead of myofibroblasts, and the fibronectin fibers were under reduced tension. This correlated with a progressive rounding of cells from the growth front inward, with decreased α-smooth muscle actin expression, YAP nuclear translocation, and cell proliferation. Together, this suggests that the myofibroblast phenotype is stabilized at the growth front by tensile forces, even in the absence of endogenously supplemented TGF-β, and reverts into a quiescent fibroblast phenotype already 10 μm behind the growth front, thus giving rise to a myofibroblast-to-fibroblast transition. This is the hallmark of reaching prohealing homeostasis.
Neurotransmitter release is stabilized by homeostatic plasticity. Presynaptic homeostatic potentiation (PHP) operates on timescales ranging from minute- to life-long adaptations and likely involves ...reorganization of presynaptic active zones (AZs). At Drosophila melanogaster neuromuscular junctions, earlier work ascribed AZ enlargement by incorporating more Bruchpilot (Brp) scaffold protein a role in PHP. We use localization microscopy (direct stochastic optical reconstruction microscopy dSTORM) and hierarchical density-based spatial clustering of applications with noise (HDBSCAN) to study AZ plasticity during PHP at the synaptic mesoscale. We find compaction of individual AZs in acute philanthotoxin-induced and chronic genetically induced PHP but unchanged copy numbers of AZ proteins. Compaction even occurs at the level of Brp subclusters, which move toward AZ centers, and in Rab3 interacting molecule (RIM)-binding protein (RBP) subclusters. Furthermore, correlative confocal and dSTORM imaging reveals how AZ compaction in PHP translates into apparent increases in AZ area and Brp protein content, as implied earlier.
Display omitted
•HDBSCAN analysis of dSTORM data allows precise determination of AZ nanotopology•AZ compaction correlates with acute and chronic PHP•AZ compaction comprises reduced AZ area and enhanced protein density•Higher protein density equals apparent enlargement in confocal microscopy
Applying localization microscopy, Mrestani et al. introduce compaction of presynaptic active zone scaffolds and individual Bruchpilot and RIM-binding protein subclusters correlating with presynaptic homeostatic potentiation. Active zone compaction without protein recruitment enhances local molecule concentrations, which may be misjudged as enlargement by intensity-based imaging approaches.
In correlative light and electron microscopy (CLEM), the fluorescent images must be registered to the EM images with high precision. Due to the different contrast of EM and fluorescence images, ...automated correlation-based alignment is not directly possible, and registration is often done by hand using a fluorescent stain, or semi-automatically with fiducial markers. We introduce "DeepCLEM", a fully automated CLEM registration workflow. A convolutional neural network predicts the fluorescent signal from the EM images, which is then automatically registered to the experimentally measured chromatin signal from the sample using correlation-based alignment. The complete workflow is available as a Fiji plugin and could in principle be adapted for other imaging modalities as well as for 3D stacks.