Stroke is a cerebrovascular disorder with 15 million cases every year worldwide. The most common symptom is motor deficits. In order to overcome such symptoms, the motor brain either repairs the damaged tissue or reorganises to compensate for the injured brain region. To stimulate this reorganisation transcranial Direct Current Stimulation (tDCS) is considered to be a promising thera- peutic intervention. Simulations of electric field distributions generated by tDCS currently entail individualised volume conduction models to improve tDCS. A volume conduction model includes geometry and conductivity properties of tissue types in healthy subjects. When applying existing models to chronic stroke subjects, electric field distribution patterns differ substantially compared to healthy subject distribution patterns. In current models, the lesion is not identified and acknowledged as a distinctive tissue type, as it is yet unclear what the lesion influence is. However, the lesion is a potential source of variability in desired electric field distribution which could result in different motor recovery. A volume conduction model is designed by combining the software SimNIBS, which can segment the head of healthy subjects and LINDA, able to distinguish lesion tissue of chronic stroke subjects. The location and the conductivity value of the lesion seem to influence the electric field distribution of tDCS where this individualised model is preferred. Including the lesion is an important advance towards the use of volume conduction models for chronic stroke subjects to prospectively find optimal electrode configurations, keep the safety margins and to prospectively analyse the results of tDCS.

The central nervous system (CNS) exhibits remarkable plasticity throughout life. The physiological changes in the CNS that occur due to plasticity allow us to perform new skills and old ones more effectively and efficiently over time. Recently, it has been demonstrated that plasticity can be used to help people recover motor function after spinal cord injury (SCI), stroke, or other neurodegenerative diseases. Following injury or illness, neuronal pathways are disrupted, leading to exaggerated reflexes and motor impairments. Rehabilitation methods can help restore motor function by triggering beneficial plasticity (i.e., neuronal and/or synaptic changes that improve motor functions). H-reflex operant conditioning that triggers beneficial plasticity is one promising new therapeutic approach to motor rehabilitation. In this paradigm, participants are operantly conditioned to change the size of abnormal reflexes associated with motor deficiencies (either increased or decreased as needed), which consequently improves movement. H-reflex operant conditioning has no known adverse side effects and it can complement other therapies. Two present limitations of H-reflex operant conditioning are its success rate and the length of time required to complete the conditioning. Given that the beneficial plasticity induced by this paradigm is modeled to start in the sensorimotor cortex, we designed an enhanced H-reflex operant conditioning system that provides people with brain-computer interface (BCI)-based feedback on activity from this region of the brain. We hypothesize that by guiding this critical first stage of plasticity, it should be possible to enhance the efficacy and efficiency of this paradigm. This thesis is organized as follows. Chapter 1 introduces the H-reflex operant conditioning and the logic for our enhanced H-reflex operant conditioning system. Chapters 2 and 3 describe experiments conducted to identify and train participants to use our BCI-based feedback system. Five participants completed the training; four of these participants learned to use the BCI with better than 70% accuracy and three of these four participants significantly improved their accuracy with training. Chapter 4 lays out the design of the enhanced H-reflex conditioning system. Finally, Chapter 5 presents plans for experiments to test the system when human-based research is able to safely resume following the global COVID-19 pandemic and potential directions of future work.

Dravet Syndrome is a rare epileptic disorder, which is caused in more than 70% of all patients by a loss-of-function mutation in the SCN1A gene encoding sodium channel Nav1.1. One of the main problems for treatment of patients with Dravet Syndrome is drug resistance and there are no biomarkers to monitor the efficacy of a particular treatment. The Dravet Syndrome phenotype can be recapitulated in mice by knockout of the Scn1a gene. Previous studies revealed a reduced sodium current in inhibitory GABAergic interneurons in a mouse model of Dravet Syndrome which leads to reduced excitability of inhibitory GABAergic inter-neurons. Inhibition of inhibitory GABAergic interneurons can lead to decreased phase-amplitude coupling between theta (5-10 Hz) and gamma (50-70 Hz) or theta and high gamma (100-140 Hz) oscillations in the brain. To investigate whether impaired phase-amplitude coupling could be a potential biomarker to monitor the efficacy of treatment for Dravet Syndrome, we recorded electrocorticogram signals from the primary visual cortex in two mice models of Dravet Syndrome and wild type mice. We found a decrease in theta-high gamma coupling for mice with Dravet Syndrome in the early epileptic stage and a decrease in theta-gamma and theta-high gamma coupling in mice with Dravet Syndrome in which seizure onset had already started, during rapid-eye-movement sleep. Reduced excitability of inhibitory GABAergic interneurons can also cause hyper-excitability that leads to seizures. These results suggest that phase amplitude coupling could be a potential biomarker to monitor the efficacy of treatment for Dravet Syndrome. Efficient treatment decreases the number of seizures and therefore we expect to see an increase in theta-gamma and theta-high gamma coupling.

The Delft Hand and Wrist model is a recently created musculoskeletal model in the OpenSim environment. The current model lacks a validation of the thumb muscles. Therefore, the main goal of this work is to perform a quantitative trend validation by analyzing the correlation between experimentally measured muscle activity data and muscle forces estimated by the model from markerless motion capture data. The original model is reduced to a minimal model containing only the thumb muscles. A second model is created by changing optimal muscle fiber length, maximum isometric force and tendon slack length parameter values in the minimal model to values reported in literature to evaluate the effects of these adjustments on the estimated muscle force. An experiment is conducted in which participants are instructed to perform repetitive thumb motions while muscle activity and 2D kinematic data is captured. 3D kinematics are obtained through the machine learning toolboxes DeepLabCut and Anipose. Muscle forces are estimated through inverse dynamic static optimization in OpenSim. The results showed that the correlation between the estimated muscle forces and experimentally measured muscle activity data for both the minimal model and the adjusted model was very low to moderate, meaning both models yield unrealistic muscle force estimations. In contrast, the measured muscle activity and kinematic data show expected results thus are captured correctly, which is useful for reference in future work. There is room for improvement of the Delft Hand and Wrist model. Nevertheless, this work provides suggestions for the optimization of the current model and paves the way towards muscle force estimation and quantitative validation using experimentally measured muscle activity data for the Delft Hand and Wrist model in OpenSim.

Electroencephalography (EEG) source localization has been applied in the development of braincomputer interfaces to control hand prostheses. When performing fine movements, our brain uses sensory feedback regarding position, velocity, and force to improve performance. Understanding the cortical mechanisms underlying individual finger movements can lead to a higher number of degrees of freedom (DoF) when developing BCI-controlled hand prostheses. Our goal was to test the efficacy of separating the activity of two individual fingers during a pinch-and-hold motor task using EEG source localization. EEG data from three healthy participants performing the motor task using different fingers were collected and analyzed using two parametric and two non-parametric source localization methods. A statistical analysis was performed on the source space to test whether it is possible to distinguish between the two fingers. We were able to measure the cortical response to the perturbations on the channel level during the hold phase of the motor task. However, source power in the primary motor (M1) and somatosensory (S1) cortices was low for all conditions. The most active sources were found in the frontal cortex over Brodmann area 8. A cluster-based permutation test performed on the source space results did not reveal differences between the two fingers on the cortical area. Statistically significant (p < 0:05) source differences are reported in one case, however, the locations of the sources indicate this effect is irrelevant to the motor task. Our findings indicate that there are no measurable source-level differences regarding the motor activity of individual fingers during the hold phase of the motor task, independently of the source localization method used.

Advanced anatomical knowledge and understanding of the muscles involved in various movements are crucial for medical practitioners to reach the correct diagnostic and successfully predict surgery outcomes. To acquire this knowledge, 3D graphical anatomical models which are displayed stereoscopically can effectively supplement cadaveric dissections. Nevertheless, the movements implemented in the available graphical models do not accurately reproduce the intricate dynamics of the human body, which is especially relevant in the case of the hand. Biomechanical models, on the other hand, provide accurate movement simulations from experimental data, while lacking a detailed graphical representation. Thus, the current paper focuses on the incorporation of the biomechanical model of the hand developed by Mirakhorlo et al. (2018) into a comprehensive graphical anatomical model (Zygote Media Group Inc), to be used for educational purposes. Motion capture data of a pinch task was acquired to validate the combinational approach, and an inverse kinematics simulation was performed in OpenSim using the musculoskeletal model. A reference value based on the fingertip distance difference at the pinch pose was calculated from the experimental data and the simulated motion by the musculoskeletal model. This value was used for validation of the musculoskeletal model reproducibility by the graphical model. Comparison shows that the graphical model reproduced the simulated motion with satisfactory visual effects and within an acceptable range from the reference metric. The presented approach is considered a successful first step towards a biomechanically and anatomically accurate graphical model of the human hand. This lays the foundation for further work on minimising the effect of the anatomical differences between the two models in order to achieve a better match.

Lameness is characterized by abnormal gait and is an indicator of various hoof diseases in cattle. Not only does this raise animal welfare issues, it also causes significant economic loss from reduced milk yield and fertility. Despite that, prevalence of lameness in dairy farms is high because farmers are unable to dedicate time and labor to identify lame cattle. Many automatic lameness detection solutions have been proposed in literature. Machine vision solutions using cameras are especially attractive because cameras do not require much space and is relatively low cost. However, none of the machine vision solutions so far have been robust enough to be useful to dairy farmers. This thesis attempts to remedy that by applying deep learning methods to the pose estimation of cattle to analyze their gait and detect the presence of lameness. 313 videos of cattle walking were recorded at a dairy farm. Images were randomly extracted from those videos and 17 body parts were manually annotated on the images to fine-tune a deep neural network pretrained on ImageNet. The fine-tuned network is then used to automatically find the trajectories of the 17 body parts in all 313 videos. 84 gait features were extracted from each video based on these trajectories. Each video was also manually given a locomotion score between 1-5 by 2 experts. Due to the small number of locomotion score 5 cows, locomotion score 4 and 5 were merged into one group for analysis. Two experiments were conducted with these gait features and locomotion scores: a) Data analysis to test significant differences between locomotion score groups and b) Automatic locomotion score classification. Data analysis was done using ANOVA followed by Bonferroni correction. Stance time related features were the best at differentiating locomotion score groups, but were unable to differentiate between locomotion score pair 2<>3. Step length related features were also relatively good at differentiating different locomotion score groups, but have trouble differentiating between locomotion score pairs 1<>2 and 2<>3. For the automatic classification, the 84 gait features were first reduced to 3 features using LDA. Then, various classifiers were trained with these 3 features and locomotion score as labels. The linear discriminant classifier achieved the highest classification rate at 85.6%, but this number was heavily skewed by the high classification rate of locomotion score 1 group (95.3%), which also makes up the largest portion of the dataset. To correct this imbalance in the dataset, the prior probabilities were set equal for each locomotion score group and the classifiers were trained again. This resulted in much better classification rates with the linear discriminant classifier for locomotion score 2 and 3 (71% and 84% respectively) at the expense of lowering the classification rate of locomotion score 1 (83.4%).

A distal radius fracture (DRF) is one of the most common fractures, especially in older women. Previous research shows that a DRF disturbs sensorimotor functions even eight weeks after finished treatment, which could influence neuromuscular control. The neuromuscular system exists of a sensory, motor and integration part, which all interact with each other as a closed-loop system. This study researches if a history of DRF disturbs neuromuscular control and if so, which part of the neuromuscular control system is the origin of this impairment. Nine healthy participants and eleven participants with a DRF history (who finished their treatment 0.2 – 4 years ago) executed posture tasks and reproduction tasks with a wrist perturbator. A posture task with force perturbations was done to test neuromuscular control. Changed environmental dynamics were applied to test the adaptation of the participants during the posture task. A position and force reproduction task were executed to test sensory position and force feedback. To test the motor part of the neuromuscular system, muscle activity was measured during the tasks with electromyography. The responses to the posture task did not differ between the groups. The position reproduction task was found to be significantly different between the two groups. Moreover, people with a DRF did not adapt to changed environmental dynamics while control participants did. This implies that processing of sensory position feedback does not work properly in people with a DRF history while neuromuscular control during a posture task with small deviations is still intact. A possible explanation for these results is that different neural networks are used during reproduction tasks and posture tasks. It is concluded that sensory feedback which is used in cortical processes is disturbed in people with a history of DRF while peripheral reflexes are still intact.

Patients suffering from diseases affecting verbal communication can make use of assistive devices to improve communication. Some severely disabled patients can only produce yes-or-no responses to communicate. These responses can be created through a physical switch, eye blink, or a 'mental' click created by brain activity. The yes-or-no responses can be used to communicate by making multiple selections between two groups of letters. Through deduction paradigms the target letter can be determined. Current assistive devices use a Row Column or Huffman paradigm to communicate. Communication rates achieved with these paradigms are slow compared to regular conversation rates and to assistive devices using eye-tracking. Furthermore, these deduction paradigms have only been tested in assistive devices with no or little noise. Noise affects the yes-or-no responses and leads to the selection of incorrect letters. There are assistive devices that suffer from high noise levels which affect communication rates. This work evaluates four communication paradigms for a range of noise conditions to improve communication rates. Row Column, Huffman, and two novel paradigms are evaluated. The two novel paradigms, based on Variable-Length Error-Correcting code and Weighted Huffman encoding, are designed for environments with noise. Evaluation of these paradigms is done through simulation and human experiments. A mathematical model is developed for simulation, and an emulator emulating an assistive device is used for human experiments. Spelling speed and cognitive effort are used as performance measures. The simulations were shown to be useful as a tool for predicting the relative performance of the paradigms in real use situations. Results from the simulation and experiments found that the effect of noise is paradigm dependent and should be taken into consideration. A novel paradigm was shown to be optimal for selective noise conditions. Row Column scanning scored the best on cognitive effort, while Huffman encoding resulted in the fastest typing speed in almost all noise conditions. Through emulation and simulation, Huffman encoding is validated as the optimal paradigm to increase communication rates. The mathematical model set a basis on which more research into optimal communication paradigms for discrete control assistive devices can be conducted.