This paper investigates the effect of intrinsic muscle stiffness on neural control parameters in biological musculoskeletal control of stabilisation or reaching tasks. Current model implementations of intrinsic muscle properties are highly simplified, limiting their accuracy in replicating experimental short-range stiffness (SRS) behaviour, which appears to be important for stabilisation tasks. The Hill model, often used in musculoskeletal simulations, cannot account for SRS, while the Huxley model, which can account for non-linear muscle phenomena such as SRS , has a higher computational burden. The study compares a simplified Huxley-type model to two Hill-type models and determines the effect of intrinsic SRS on the control parameters of stabilizing 1- and 2-Degree of Freedom musculoskeletal models over various positive and negative stiffness positions in the force-length curve. Furthermore, the effect of the intrinsic muscle stiffness on the robustness of the feedback parameters of simple individual muscle feedback systems is determined in reaching experiments similar to classic experiments. The study finds that the Huxley model shows positive SRS in the negative flank of the force-length curve, achieves stabilisation through only co-contraction using a lower level of required muscle excitation than both Hill-type models and stabilises both musculoskeletal systems at a larger muscle range than the Hill-type models, including in the negative stiffness flank. The feedback parameters dominantly responsible for muscle activation patterns are also more robust to change in the Huxley model. These findings suggest that intrinsic muscle stiffness impacts neural control parameters in stabilisation and reaching tasks, and further musculoskeletal modelling should consider using more complex muscle stiffness calculations for improved accuracy.

Biologically inspired neural networks are a promising approach to understand the causes and improve the treatments of brain damage. Parkinson's disease is a progressive nervous system disorder that affects mainly movements, speech and cognitive problems. It symptoms cannot be cured, though medications can significantly improve the condition. Among the symptoms, tremor is the only one which remains unaffected by medications and is only responsive to deep-brain stimulation. A simplified, cortico-thalamo-cerebellar model will be simulated with spiking neural networks to evaluate the disease effects under dopamine depletion and connectivity weight changes. Confirming previous findings, striatal dopamine depletion was not found to cause tremor, nor its injection to affect tremor severity. The model showed evidence that parkinsonian weight changes in the pallidal inner feedback loop (GPi-GPe) are responsible of creating a suitable environment for the PD tremor oscillations to rise in the thalamus. Furthermore, both the GPi and the GPe present enhanced maximal activity coherent with muscular co-contraction onsets showing evidence of abnormal basal ganglia firing during re-emergent tremor. These findings may connect abnormal basal ganglia activity to the main parkinsonian motor impairments and may help explaining the beneficial effects of deep-brain stimulation on tremor severity.

Neuropathic pain (NP) affects approximately seven to ten percent of the general population. Seventeen percent of NP patients scored their life as “worse than death”. A myriad of causes may underlie NP, such as stroke or spinal cord injury. Also, damage or disease of the peripheral nervous system (PNS) may result in NP. One of the main issues of NP caused by a peripheral nerve injury (PNI) is the development of a neuroma, which is a tumor-like mass at the proximal end of a severed nerve that can become very painful. Neuromas show unique neurophysiological characteristics. Cell membrane alternations lead to different ion channel distributions, which in turn result in subthreshold oscillations (SO) and ectopic discharges (ED). It is assumed that this behavior could lead to NP generation. Electrical neurostimulation (ENS) is used to treat patients, thereby applying pre-programmed stimulation patterns to the affected nerves. However, the pain-provoking signals which run through the nerves are not detected and analyzed before ENS is provided. Furthermore, it is questionable whether (the currently applied) pre-programmed ENS defuses these signals anyway. In addition, pre-programmed ENS is not effective at all moments of the pain experience caused by fluctuations in signal intensity. As the clinical results are discouraging, and in view of the high costs, the popularity of this technology is currently waning. Optimization of this potentially powerful technique is needed to improve the outcome and make this technology useful to implement in the treatment strategy of patients with intractable otherwise difficult to treat pain syndromes. Theoretically, optimization of stimulation technology is possible by actually neutralizing SO and ED, which should lead to mitigating the generation of NP. We propose an approach to neutralize SO and ED consisting of several steps. Firstly, the nerve activity is real-time monitored. Secondly, a decision mechanism (called a ‘controller’) is developed that constructs electrical neurostimulation (ENS) patterns to neutralize SO and ED. Finally, these patterns are actually applied to the nerve by an electrical stimulator...

A major cause for voiding dysfunction is the inability to relax the urethral sphincter. KiloHertz Frequency Alternating Current (KHFAC) stimulation can block signals that are travelling through the body; applying this type of stimulation at the pudendal nerve could inhibit the pulses that lead to contraction of the urethral sphincter, and restore voiding ability. In order to design a successful KHFAC block therapy for the pudendal nerve, it is necessary to understand what impact different stimulation parameters have on efficacy, safety and power-efficiency. This thesis will therefore test earlier researched KHFAC stimulation parameters against a new quality measure, study the impact of new waveform alterations, and study how bipolar electrode design can improve KHFAC therapy. By utilizing the theory behind the mechanism of the KHFAC nerve block, a new block-determination model was developed that is over thirty times faster than the classic model. The McIntyre-Richardson-Grill model was chosen as the implementation of the axon model, and the bipolar electrode was modelled as an electric dipole. The simulation experiments revealed that the charge per phase of the KHFAC signal at block threshold could be reduced, without increasing the amplitude of the signal, by introducing interphase delays to the waveforms and by creating asymmetric charge-balanced waveforms. Triangular waveforms were shown to also require less charge per phase than a regular square wave to block, albeit with a higher amplitude. A correctly aligned bipolar electrode set-up with an interpolar distance that was about the same as the electrode-to-axon distance was shown to result in reduced block thresholds. Overall, this thesis has shown how stimulation parameters can be chosen to develop an effective KHFAC block therapy for the pudendal nerve.

The human brain is arguably the most complex system we know of. For centuries, understanding the brain and the way it works has been of great concern to scientists. In 1952, the first mathematical model of a neuron was developed by Hodgkin and Huxley. This work pioneered the field of computational neuroscience, which studies the organization of the brain and the way it processes information. Modeling and simulation of neurons and small brain regions at the cellular level offer great insights into the details of brain mechanics, however, they are far away from clinical applications. The Virtual Brain (TVB) is a simulation framework that takes a different approach to brain modeling. TVB reduces complexity at the micro level to achieve the macro level of brain modeling and simulation. This approach in brain modeling is referred to as Large-scale Brain Network or Neural-Mass Modeling. Additionally, TVB incorporates individual brain-imaging data with the models to achieve personalized, patient-specific digital brain twins. The digital brain twins can be used in clinical settings such as neurosurgeries and real-time brain-interface systems. However, building and utilizing such large-scale brain models requires high performance and low latency in terms of computation. Today, heterogeneous architectures, due to their enhanced performance, energy efficiency, and better flexibility are being utilized increasingly in computing systems ranging from low-power edge devices to high-performance cloud infrastructures. Versal Adaptive Compute Acceleration Platform (ACAP) is a high-performance, heterogeneous computing platform developed by Xilinx/AMD. Versal ACAP offers an array of vector processors, specialized in artificial intelligence (AI) and signal processing workloads, next to the traditional programmable logic with extensive memory resources. The large-scale neural-mass simulation problem at hand has heterogeneous computational needs, which makes it interesting to see how it maps to the Versal ACAP. In this work, we explore the benefits and challenges of using heterogeneous systems (specifically, the Versal ACAP) in the high-performance workload of large-scale neural-mass simulation. We designed and implemented a dataflow-style system for large-scale brain network simulation on the Versal ACAP. Two different versions of this system, namely AIE-Only and Heterogeneous, were developed in order to explore the capabilities of the Versal ACAP when accelerating our application. Compared to the current GPU version of TVB, the Heterogeneous implementation performed on average around 4× slower in terms of throughput. However, the Versal ACAP Heterogeneous implementation delivered around 13× lower latency, two orders of magnitude better energy efficiency, and around 2× better power efficiency compared to the GPU version of TVB. Although the Heterogeneous implementation falls short in terms of throughput compared to the GPU version, its lower latency and energy efficiency make it suitable for real-time applications that might use large-scale neural-mass modeling such as virtual surgery or brain interface devices.

The olivocerebellar system plays a central role in motor learning, crucially contributing to the coordination, precision and accurate timing of movements. The system is formed by Purkinje cells (PC), the Deep Cerebellar Nucleus (DCN) and Inferior Olive (IO). Its input activate the PC which produces simple (SS) and complex spikes (CS). The latter induced by the IO. The oscillatory nature of the production of CS seems to play a role in motor control and motor timing. In addition, CS modulate the parallel fiber-Purkinje cell (PF-PC) synaptic plasticity. IO synchrony can help the system learn timing with the PF input as the timing context. Furthermore, the level of synchronization of the coupled IO cells could determine the function of complex spikes, implying that the olivocerebellum is capable of switching between modes of learning by changing the level of synchronization. In this paper we introduced a novel computational model to analyze the role of coupling of the IO and long-term plasticity at the PF-PC synapse in the response of the olivocerebellar loop. It is a resonant system formed by a detailed IO model and Integrate-and-Fire PC and DCN models, with physiologically observed firing frequencies. The IO cells are modeled as coupled oscillators and plasticity is incorporated through a timing dependent specialization of Hebbian learning (Spike-Timing Dependent Plasticity or STDP). Two different simulations are performed both for the coupled and uncoupled scenarios. The STDP is used only in the second type of simulation. Both types of simulations use the same noisy input, which is applied twice during the second type of simulation. The second half of the latter is interpreted as the response of a trained loop. Results show that in the presence of coupling the correlation of the firing rates distribution decreases. This indicates that for the coupled scenario PCs are separating the patterns, while for the uncoupled scenarios the noise is encoded more robustly. Furthermore, a drop in the noise current (inhibiting the PC) leads to an IO spike at about 100 milliseconds later. After training, however, the loop recognizes a drop in the noise and depresses the synapses avoiding an increase in the firing rate of the PC. This effect is more noticeable in the coupled scenario. In conclusion, the model shows that plasticity can lead to learning, and that this process is more efficient for a coupled system.

The olivocerebellar system plays a crucial role in control of movements of the human body in terms of coordination, precision and timing. Long-term plasticity is directly linked to motor learning and control. In this research, we developed a phenomenological model of the olivocerebellar system with balancing of long-term potentiation (LTP) and long-term depression (LTD) at the parallel fiber-Purkinje cell (PF-PC) synapse. By ranging the PF input over frequencies, we found that PCs can select frequencies in a highly non-linear manner. There is a sharp contrast in synaptic weight change between neighbouring frequencies, which is caused by the temporal spiking property of the inferior olive (IO) cell. This research found a novel signal processing capability of the PC.

Complex mathematical models are used in computational neuroscience to stimulate brain activity to understand the biological processes involved. The simulation of such models is computationally costly, and thus highperformance computing systems are selected as a potential solution to increase performance. This thesis aims to implement a new versatile, multi-GPU eHH simulator (mgpuHH), explore its performance and make general observations on performance scalability over different modeling and cluster configuration properties. This work offers a multinode multi-GPU solution that offers excellent scalability performance due to how the simulator is constructed, with the use of OpenMPI and CUDA. The simulator is configured with JSON configuration files, containing the neural descriptions and simulatorspecific settings. Consequently, enabling a userfriendly environment, for the neuroscientists, without the need of recompiling or understanding the source code. The gap junction calculations are identified as the critical function bottlenecking performance of the simulator. Therefore, an algorithm tailored to utilize GPU performance is implemented to decrease wallclock time for these specific calculations. For internode communication, OpenMPI can be configured in two ways. Eiter share all possible compartments potentials with every node in the network or only share the compartments potentials to nodes that need them. These methods rely internally on MPI Allgather and Alltoallv respectively. When available, GPUDirect, NVlink, and RDMA are supported. The implementation hides communication overhead, when possible, by concurrently executable compute kernels. A neuron model from the Inferior Olivary Nucleus is selected for benchmarking. Reported results go up to 32 Nodes with a total of 64 GPU cards. The design shows linear weak and strong scaling within the experimental setups for intranode and internode scalability. With this simulator, networks over 10 million cells become available to model on largescale GPU clusters, setting a new standard for eHH simulations. Comparisons against related work on CPU and FPGAs have been conducted, a 100x speedup is achieved versus a single cpu threaded solution. Furthermore, a 2x speedup is achieved over an FPGA solution (flexHH) and 10 fold over a multithreaded CPU (GenEHH, with 128 threads) solution, both reported speedups are for a fully connected network with 7000 IO cells.

Background Electroencephalography (EEG) using a dry electrode cap is currently being investigated as a pre-hospital stroke triage instrument. Developing an algorithm for automatic interpretation of the EEG signals is challenging, considering the amount of artefacts often in the signal. Ideally, an algorithm should be capable of distinguishing between different artefact types to determine the appropriate action: whether to correct them, reject them, or consider their potential predictive value. Neural networks have demonstrated their value for these types of classifications in wet EEG data. However, this approach requires enormous amount of data and dry EEG data is sparse. Objective This study aims to develop a multi-class artefact classification model for dry EEG data using transfer learning. Methods First, a convolutional neural network (CNN) for multi-class (clean, eye movement, muscle activity and electrode artefact) classification was developed. Wet electrode EEG recordings from a publicly available dataset were used, containing data of 213 patients (Part I). Second, this model was implemented and transfer learned for multi-class (clean, pulse artefact, muscle activity and artefact) classification of dry EEG recordings using data of 13 subjects (Part II). The models were trained using annotated multi-channel input. Model performances were evaluated on unseen test data using accuracy, area under the receiver operating characteristic curve, F1-score, precision, and recall.  Results The pre-trained multi-class model achieved an overall accuracy of 74.8%. The fine-tuned model was able to correctly differentiate between the classes with an accuracy of 71.2%, with the best performance for the classes muscle activity (AUC 0.92, F1-score 0.80) and artefact (AUC 0.94, F1-score 0.80).  Conclusion Transfer learning enabled the development of a good performing multi-class artefact classification model specifically tailored for dry EEG data, even though available data was limited. The developed model could assist in assessing appropriate action for different artefact types in dry EEG data interpretation algorithms.