Up to now, modern Machine Learning is mainly based on fitting high dimensional functions to enorm... more Up to now, modern Machine Learning is mainly based on fitting high dimensional functions to enormous data sets, taking advantage of huge hardware resources. We show that biologically inspired neuron models such as the Leaky-Integrate-and-Fire (LIF) neurons provide novel and efficient ways of information encoding. They can be integrated in Machine Learning models, and are a potential target to improve Machine Learning performance. Thus, we derived simple update-rules for the LIF units from the differential equations, which are easy to numerically integrate. We apply a novel approach to train the LIF units supervisedly via backpropagation, by assigning a constant value to the derivative of the neuron activation function exclusively for the backpropagation step. This simple mathematical trick helps to distribute the error between the neurons of the pre-connected layer. We apply our method to the IRIS blossoms image data set and show that the training technique can be used to train LIF neurons on image classification tasks. Furthermore, we show how to integrate our method in the KERAS (tensorflow) framework and efficiently run it on GPUs. To generate a deeper understanding of the mechanisms during training we developed interactive illustrations, which we provide online. With this study we want to contribute to the current efforts to enhance Machine Intelligence by integrating principles from biology.
Fiber alignment in 3D collagen networks as a biophysical marker for cell contractility
Cells cultured in 3D fibrous biopolymer matrices exert traction forces on their environment that ... more Cells cultured in 3D fibrous biopolymer matrices exert traction forces on their environment that induce deformations and remodeling of the fiber network. By measuring these deformations, the traction forces can be reconstructed if the mechanical properties of the matrix and the force-free matrix configuration are known. These requirements severely limit the applicability of traction force reconstruction in practice. In this study, we test whether force-induced matrix remodeling can instead be used as a proxy for cellular traction forces. We measure the traction forces of hepatic stellate cells and different glioblastoma cell lines and quantify matrix remodeling by measuring the fiber orientation and fiber density around these cells. In agreement with simulated fiber networks, we demonstrate that changes in local fiber orientation and density are directly related to cell forces. By resolving Rho-kinase (ROCK) Inhibitor-induced changes of traction forces and fiber alignment and densit...
Immune cells such as natural killer (NK) cells migrate with high speeds of several μm/min through... more Immune cells such as natural killer (NK) cells migrate with high speeds of several μm/min through dense tissue, but the traction forces are unknown. We present a method to measure dynamic traction forces of fast migrating cells in non-linear biopolymer matrices. We find that NK cells display bursts of large traction forces that increase with matrix stiffness and facilitate migration through tight constrictions.
How is information processed in the cerebral cortex? To answer this question a lot of effort has ... more How is information processed in the cerebral cortex? To answer this question a lot of effort has been undertaken to create novel and to further develop existing neuroimaging techniques. Thus, a high spatial resolution of fMRI devices was the key to exactly localize cognitive processes. Furthermore, an increase in time-resolution and number of recording channels of electro-physiological setups has opened the door to investigate the exact timing of neural activity. However, in most cases the recorded signal is averaged over many (stimulus) repetitions, which erases the fine-structure of the neural signal. Here, we show that an unsupervised machine learning approach can be used to extract meaningful information from electro-physiological recordings on a single-trial base. We use an auto-encoder network to reduce the dimensions of single local field potential (LFP) events to create interpretable clusters of different neural activity patterns. Strikingly, certain LFP shapes correspond to...
Cognitive computational neuroscience (CCN) suggests that to gain a mechanistic understanding of b... more Cognitive computational neuroscience (CCN) suggests that to gain a mechanistic understanding of brain function, hypothesis driven experiments should be accompanied by biologically plausible computational models. This novel research paradigm offers a way from alchemy to chemistry, in auditory neuroscience.Withaspecial focus on tinnitus – as the prime example of auditory phantom perception – we review recent work at the intersection of artificial intelligence, psychology, and neuroscience, foregrounding the idea that experiments will yield mechanistic insight only when employed to test formal or computational models. This view challenges the popular notion that tinnitus research is primarily data limited, and that producing large, multi-modal, and complex data-sets, analyzed with advanced data analysis algorithms, will lead to
Calcium supplementation of bioinks reduces shear stress-induced cell damage during bioprinting
Biofabrication
During bioprinting, cells are suspended in a viscous bioink and extruded under pressure through s... more During bioprinting, cells are suspended in a viscous bioink and extruded under pressure through small diameter printing needles. The combination of high pressure and small needle diameter exposes cells to considerable shear stress, which can lead to cell damage and death. Approaches to monitor and control shear stress-induced cell damage are currently not well established. To visualize the effects of printing-induced shear stress on plasma membrane integrity, we add FM 1-43 to the bioink, a styryl dye that becomes fluorescent when bound to lipid membranes, such as the cellular plasma membrane. Upon plasma membrane disruption, the dye enters the cell and also stains intracellular membranes. Extrusion of alginate-suspended NIH/3T3 cells through a 200 µm printing needle led to an increased FM 1-43 incorporation at high pressure, demonstrating that typical shear stresses during bioprinting can transiently damage the plasma membrane. Cell imaging in a microfluidic channel confirmed that ...
Evaluates the fiber orientation around cells in biological tissue as measure of contractile stren... more Evaluates the fiber orientation around cells in biological tissue as measure of contractile strength.
Up to now, modern machine learning (ML) has been based on approximating big data sets with high-d... more Up to now, modern machine learning (ML) has been based on approximating big data sets with high-dimensional functions, taking advantage of huge computational resources. We show that biologically inspired neuron models such as the leaky-integrate-and-fire (LIF) neuron provide novel and efficient ways of information processing. They can be integrated in machine learning models and are a potential target to improve ML performance. Thus, we have derived simple update rules for LIF units to numerically integrate the differential equations. We apply a surrogate gradient approach to train the LIF units via backpropagation. We demonstrate that tuning the leak term of the LIF neurons can be used to run the neurons in different operating modes, such as simple signal integrators or coincidence detectors. Furthermore, we show that the constant surrogate gradient, in combination with tuning the leak term of the LIF units, can be used to achieve the learning dynamics of more complex surrogate gra...
Aims: Desminopathies comprise hereditary myopathies and cardiomyopathies caused by mutations in t... more Aims: Desminopathies comprise hereditary myopathies and cardiomyopathies caused by mutations in the intermediate filament protein desmin that lead to severe and often lethal degeneration of striated muscle tissue. Animal and single cell studies hinted that this degeneration process is associated with massive ultrastructural defects correlating with increased susceptibility of the muscle to acute mechanical stress. The underlying mechanism of mechanical susceptibility, and how muscle degeneration develops over time, however, has remained elusive. Methods: Here, we investigated the effect of a desmin mutation on the formation, differentiation, and contractile function of in vitro-engineered three-dimensional microtissues grown from muscle stem cells (satellite cells) isolated from heterozygous R349P desmin knock-in mice. Results: Micro-tissues grown from desmin-mutated cells exhibited spontaneous unsynchronised contractions, higher contractile forces in response to electrical stimulation, and faster force recovery compared with tissues grown from wild-type cells. Within 1 week of culture, the majority of R349P desmin-mutated tissues disintegrated, whereas wild-type tissues remained intact over at least three weeks. Moreover, under tetanic stimulation lasting less than 5 s, desmin-mutated tissues partially or completely ruptured, whereas wild-type tissues did not display signs of damage. Conclusions: Our results demonstrate that the progressive degeneration of desminmutated micro-tissues is closely linked to extracellular matrix fibre breakage associated with increased contractile forces and unevenly distributed tensile stress. This suggests that the age-related degeneration of skeletal and cardiac muscle in patients suffering from desminopathies may be similarly exacerbated by mechanical damage from highintensity muscle contractions. We conclude that micro-tissues may provide a valuable Marina Spörrer and Delf Kah contributed equally to this work.
Pylustrator: An Interactive Interface To Style Matplotlib Plots
Visualisations of data are at the core of every publication of scientific research results. They ... more Visualisations of data are at the core of every publication of scientific research results. They have to be as clear as possible to facilitate the communication of research. As data can have different formats and shapes, the visualisations often have to be adapted to reflect the data as well as possible. We developed Pylustrator, an interface to directly edit python generated matplotlibgraphs to finalize them for publication. Therefore, subplots can be resized and dragged around by the mouse, text and annotations can be added. The changes can be saved to the initial plot file as python code.
High resolution panoramic images for 04/01/2014 to 04/21/2014 used to evaluate emperor penguin nu... more High resolution panoramic images for 04/01/2014 to 04/21/2014 used to evaluate emperor penguin numbers and arrival patter
Video (4008x2672, 5 fps, 60s) recorded on 07/22/2013 at 11:40:47 UTC used to evaluate the movemen... more Video (4008x2672, 5 fps, 60s) recorded on 07/22/2013 at 11:40:47 UTC used to evaluate the movement characteristics of single emperor penguins at the huddle boundarie
Meteorological data and time lapse image recordings (04/02/2013 to 04/07/2013) used to evaluate t... more Meteorological data and time lapse image recordings (04/02/2013 to 04/07/2013) used to evaluate the influence of wind speed and wind direction on the position of the Atka Bay emperor penguin colon
Scientific applications often require an exact reconstruction of object positions and distances f... more Scientific applications often require an exact reconstruction of object positions and distances from digital images. Therefore, the images need to be corrected for perspective distortions. We present CameraTransform, a python package that performs a perspective image correction whereby the height, tilt/roll angle and heading of the camera can be automatically obtained from the images if additional information such as GPS coordinates or object sizes are provided. We present examples of images of penguin colonies that are recorded with stationary cameras and from a helicopter.
ABSTRACTNumerous cell functions are accompanied by phenotypic changes in viscoelastic properties,... more ABSTRACTNumerous cell functions are accompanied by phenotypic changes in viscoelastic properties, and measuring them can help elucidate higher-level cellular functions in health and disease. We present a high-throughput, simple and low-cost microfluidic method for quantitatively measuring the elastic (storage) and viscous (loss) modulus of individual cells. Cells are suspended in a high-viscosity fluid and are pumped with high pressure through a 5.8 cm long and 200 µm wide microfluidic channel. The fluid shear stress induces large, near ellipsoidal cell deformations. In addition, the flow profile in the channel causes the cells to rotate in a tank-treading manner. From the cell deformation and tank treading frequency, we extract the frequency-dependent viscoelastic cell properties based on a theoretical framework developed by R. Roscoe1 that describes the deformation of a viscoelastic sphere in a viscous fluid under steady laminar flow. We confirm the accuracy of the method using at...
Deep neural networks typically outperform more traditional machine learning models in their abili... more Deep neural networks typically outperform more traditional machine learning models in their ability to classify complex data, and yet is not clear how the individual hidden layers of a deep network contribute to the overall classification performance. We thus introduce a Generalized Discrimination Value (GDV) that measures, in a non-invasive manner, how well different data classes separate in each given network layer. The GDV can be used for the automatic tuning of hyper-parameters, such as the width profile and the total depth of a network. Moreover, the layer-dependent GDV(L) provides new insights into the data transformations that self-organize during training: In the case of multi-layer perceptrons trained with error backpropagation, we find that classification of highly complex data sets requires a temporal {\em reduction} of class separability, marked by a characteristic 'energy barrier' in the initial part of the GDV(L) curve. Even more surprisingly, for a given data ...
Uploads
Papers by Richard Gerum