Patients with neuromuscular diseases that are unable to speak, but whose cognitive ability has been maintained, can be benefited from Brain Computer Interfaces (BCIs). The decoding of inner (covert) speech from EEGs consists of one of the state of the art methods that aim to tackle this issue. High variability between subjects, as well as low signal to noise ratio (SNR) undermine the methods used, and introduce the need for computer assisted solutions. Thus, machine learning models as well as large amounts of recorded data are required to design effective algorithms and produce substantial results. In this study, covert vowel classification was performed in a systematic way, by making use of two openly shared databases from literature; the Coretto database, that contains EEG recordings of native Spanish speakers, and the DAIS dataset, which includes EEG recordings of native Dutch speakers. Six classifiers were initially selected to perform 5-class classification: a Random Forest (RF), a k Nearest Neighbours (kNN), a Gaussian Naive Bayers (GNB), a Deep Convolutional Neural Network (DCNN), a Shallow Convolutional Neural Network (SCNN) and a Long Short Term Memory Recurrent Neural Network (LSTM). The DCNN outperformed the other methods, with average intra-subject accuracies of 35% for Coretto and 39% for DAIS (chance level 20%). Afterwards, an Overt versus Covert trials experiment was implemented, to test the limits of overt speech decoding from EEGs. The overt result was slightly higher than covert, with an intra-subject average value of 37.8% for Coretto and 40.5% for DAIS (chance level 20%). Finally, binary classification was performed to identify those pairs of vowels that can be classified more efficiently. Vowels /a/ and /u/ seemed to perform better in average in both datasets (average of 64.8% for Coretto and 64.4% for DAIS with a chance accuracy of 50%). Future work should focus on identifying the useful parts of the EEG recordings, increasing the SNR and the resolution of the electrodes, and defining the most appropriate dictionaries of words/vowels for a BCI. Also, more studies should follow systematic ways of comparisons between datasets, to obtain less ambiguous insights and lead this field to improvements.

Shoulder exoskeleton is a popular solution to work-related shoulder disorders and muscle fatigue. With a wide range of exoskeletons designed, a comprehensive report on how the use of shoulder exoskeletons changes shoulder biomechanics is still missing. In this project, the impact of exoskeletons on shoulder biomechanics was investigated with the musculoskeletal simulation OpenSim. This study proposed a "human-in-the-loop" optimization-based design tool for shoulder exoskeletons. This design tool incorporates the predicted biomechanical effects of a shoulder exoskeleton from musculoskeletal simulations into design considerations. This design tool was validated with a case study designing a shoulder exoskeleton based on a compliant beam and testing the design in the musculoskeletal simulation and experiments. The exoskeleton design tool is a coupling of finite element analysis and OpenSim. OpenSim calculates the deformation of the exoskeleton with human motion, and the finite element analysis calculates the force exerted from the exoskeleton upon deformation. Then OpenSim computes muscle activities under the external force from the exoskeleton. By merging muscle activities and the resultant glenohumeral joint reaction force to an objective function, the optimization-based design loop is closed by looking for the best objective value iteratively. Several exoskeletons were designed by the new design tool to assist different types of tasks. The design tool exhibited good ability in finding optimal solutions for a range of design choices and design requirements. Simulated tests of designed exoskeletons showed significant effects on reducing muscle activities and good robustness in resisting the influence of perturbed motions in arm-elevated tasks. An exoskeleton was selected to be tested with an experiment set up in the same way as the simulated test. Experiment results supported the performance of the exoskeleton predicted in the simulated test. This project established a method to comprehensively predict the effect of an exoskeleton on shoulder biomechanics and provided a more comprehensive understanding of biomechanical effects of shoulder exoskeletons. This facilitated the “human-in-the-loop” design process of shoulder exoskeletons which could greatly save money and time investments into prototyping, testing, and validation.

Introduction: Amputees may be eligible to use a socket or osseointegrated (OI) prosthesis after losing a limb. A lot is known about the stump-socket interface in transtibial (TT) amputees. However, there are hardly any articles that explain the load transfers between the tibia and an OI prosthesis. For this reason, this study will focus on the load pattern at the bone-implant interface. The aim is to investigate how the osseointegrated loading for TT amputees is compared to able-bodied subjects. Method: Two generic musculoskeletal models were created in OpenSim: a transtibial osseointegrated (TTO) model and an able-bodied model. The able-bodied model includes all muscles in the lower limb, where the TTO model had no muscles around the tibia. Two datasets, consisting of five able-bodied participants (healthy control (HC)) and five transtibial amputees (TTA) with a socket prosthesis, were used. The internal reaction forces and moments were analysed for the affected side during the stance phase of the gait cycle. These normalized forces and moments were expressed in bodyweight (BW) and bodyweight x height (BWH), respectively. Results: The largest forces at the bone-implant interface were measured in the longitudinal direction (-6.92 BW) for the able-bodied model, the lowest in the mediolateral direction (-0.09 BW) for the TTO model. The loading differences were the highest between the able-bodied model and the TTO model with TTA data. Around the antero-posterior axis, the largest moments were found. The able-bodied model measured 0.027 BWH, where the TTO model with HC data found 0.023 BWH and 0.017 BWH for the TTA data. Conclusion: Transtibial amputees have decreased loading at the bone-implant interface compared to able-bodied subjects. The muscles play the most prominent role in comparison with the gait pattern.

Inertial Measurement Units (IMUs) have become increasingly popular human motion estimation due to their portability, self-contained features, and cost-effectiveness compared to marker-based sensing systems which rely on external cameras to observe the position of the markers. Over the years, many studies have proposed various models and algorithms based on IMUs and the kinematics of the human body for motion estimation, neglecting the fact that IMUs measure acceleration, which can be directly related to joint torque with known inertial parameters. In turn, by estimating joint torque and incorporating kinetics into the model, it becomes possible to address a long-standing problem in the field of biomechanics: the marker-based sensing system’s inability to provide a reliable estimation of kinetics due to the need for numerical differentiation. In a previous study, a kinetics model was proposed for estimating the motion but validated only on a robotic arm. In our study, we further explore the performance of using IMUs to estimate wheelchair user motion data based on Extended Kalman Filter (EKF) and the kinetics model, with marker-based Inverse Kinematics (IK)/Inverse Dynamics (ID) as the benchmark. Compared to Marker-based IK, the method leveraging kinetics achieves a Root Mean Squared Difference (RMSD) below 16◦ for three out of four tasks across all joints throughout the trial. After analyzing the only task with degradation in estimation, we conclude that the erroneous IMUs measurements results from Soft Tissue Artefacts (STA) is the most likely reason. For joint torque estimation, the RMSD for joint torque estimation is below 3.05Nm for the tasks less affected STA. Through fine-tuning the EKF, we can achieve fast and responsive estimation results without being affected by numerical differentiation, enabling us to capture both sudden and subtle changes in joint torque estimation. The kinetics model performs better than the kinematics-based model in estimating both kinematics and kinetics and also reduced drifting behavior. Compared to OpenSense, which depends on magnetometer measurements, kinetics model estimation shows comparable kinematics estimation accuracy while excluding the use of heading information. The results show that including the kinetics model for human motion estimation can improve estimation accuracy and robustness encouraging further studies to include kinetics for human motion estimation.

Total Hip Arthroplasty (THA) is a procedure where a defective hip joint with pain symptoms or functional impairment is replaced by a hip prosthesis. During this procedure, the optimal hip prosthesis should be chosen and placed in the correct orientation to achieve a stable hip joint. Failure to do so can lead to dislocation, mechanical failure or infection of the hip joint. To improve the outcome of the THA, range of motion (ROM) of the hip joint can be evaluated during the surgery after placement of the hip prosthesis. Current methods of measuring hip ROM have some limitations. For example, there is a low intra- and inter-test reliability between identification of the bony landmarks and goniometric alignment. The goal of this project was to develop an instrument that can measure the ROM during THA without the limitations of current instruments. An instrument was developed that determines its orientation via an inertial measurement unit (IMU). The IMU is connected on the upper leg via a strap and aligned with the bony landmarks via laser beams. A test has been performed to validate the usage of the instrument. The ROM of the hip for flexion and abduction motion was measured with the instrument and compared to a reference measurement made with two video cameras. A significant difference was found: the instrument did not stay aligned with the hip joint. For further development of the instrument, the drift should be minimised and multiple straps should be tested to allow a better fixation. Additionally, the added-value of the laser requires further research.

BACKGROUND: Knee problems are the most common complaints of the lower extremity in the Netherlands. Osteoarthritis has the highest prevalence of all knee complaints. Knee osteoarthritis is likely to start at the patellofemoral joint and is associated with the aggravation of pain. Mechanical overloading is hypothesized to contribute to the development and progression of patellofemoral osteoarthritis (PFOA). Several studies have explored the mechanical loading pattern of the patellofemoral joint. Because different biomechanical models were used, vastly different estimations of the patellofemoral joint contact force (PFJCF) were concluded. This diversity prevents a clear understanding of the role of mechanical overloading in PFOA, since it is unknown how different biomechanical models affect the PFJCF estimation. OBJECTIVES: This study will explore how different biomechanical models of the patellofemoral joint affect the estimated PFJCF for common weight-bearing activities. METHODS: Ten healthy participants were included in this study. Common weight-bearing activities (walking, stair ascending, stair descending, sit-to-stand and stand-to-sit) were performed in the motion lab. Marker trajectory data and force plate data were collected. The data were input to biomechanical models used to estimate the quadriceps muscle force and subsequently the PFJCF. The quadriceps muscle force was estimated using the inverse dynamics and static optimization method. From there on, the PFJCF was estimated using three PFJCF to quadriceps muscle force ratios (P2QFRs), each based on a different patellofemoral joint model (i.e. van Eijden’s model, Yamaguchi’s model and Gill’s model). For each weight-bearing activity, the peak PFJCF was obtained and the magnitude of the difference among the biomechanical models was explored. RESULTS: The static optimization method resulted in a significantly higher peak PFJCF compared to the inverse dynamics method in walking (largest effect size was 0.10 BW). However, for stair descending, sit-to-stand and stand-to-sit, the inverse dynamics method resulted in a significantly higher peak PFJCF compared to the static optimization method (largest effect size was 0.45 BW, 1.17 BW, 1.25 BW, respectively). No significantly difference was found for stair ascending. For walking, Yamaguchi’s model resulted in a significantly higher peak PFJCF compared to van Eijden’s model and Gill’s model, and van Eijden’s model resulted in a significantly higher peak PFJCF compared to Gill’s model (largest effect size was 0.06 BW). For stair ascending, stair descending, sit-to-stand and stand-to-sit, Gill’s model resulted in a significantly higher peak PFJCF compared to van Eijden’s model and Yamaguchi’s model. For stair descending the van Eijden’s model resulted in a significantly higher peak PFJCF compared to Yamaguchi’s model. The largest effect size was 0.15 BW, 0.32 BW, 0.72 BW and 0.72 BW, for respectively stair ascending, stair descending, sit-to-stand and stand-to-sit. CONCLUSION: The choice of a biomechanical model has a critical effect on the estimation of the magnitude of the PFJCF. Its differences might reach half the clinical size effects when comparing control to symptomatic PF pain patients.

Humans and other biological bipedal walkers are extraordinarily agile and robust. This is especially apparent when certain features of the environment are unknown such as unexpected downsteps (for example, suddenly walking off the side-walk or not expecting the last stair when descending). Humans can negotiate these downsteps ---both expected and unexpected--- with remarkable agility and ease. The contributions from active compensation, passive dynamics, and reflexes that humans employ to overcome these downsteps are however not inherently present in bipedal robots. This motivates us to assess whether externally observed behavior can be used to achieve the same capabilities in bipedal robots. One of the challenges with this proposal is the morphological differences between humans and robots. Taking the bipedal robot Cassie as an example, these differences predominantly constitute to overall- and segment mass, leg morphology, and a lack of an upper body. The observed behavior from the human should subsequently be scaled to represent an appropriate nominal and compensatory behavior of the robot. This thesis aims to systematically study the translation of this human behavior to bipedal walking robots, regardless of the morphology of the walkers. We start from human experimental data where nominal walking, expected downstep, and unexpected downstep trials were conducted. The human data is analyzed from the perspective of a reduced-order representation of the human. The reduced-order representation encodes the center of mass dynamics and contact forces. An equivalent reduced-order model is used to represent the bipedal robot, which allows for the translation of the nominal walking and downstep behaviors between human and robot. The morphological differences between the human and the robot are therefore resolved by realizing dynamically equivalent behavior which is embedded into the full-order dynamics of a bipedal robot via optimization-based controllers. The results demonstrate traversing expected and unexpected downsteps in simulation on the underactuated 3D walking robot Cassie.

Inertial Measurement Unit (IMU)-based motion capture has gained interest over the years due to its potential to measure human movement in the clinic and on the sports field at low cost. Still, IMU-based motion reconstruction remains a challenging task as these IMU measurements are corrupted by noise and bias. There have been many filtering and sensor fusion algorithms developed to address this problem, but limited attention has been paid to including the system dynamics of the subject to estimate kinematics and kinetics from multiple IMUs. I propose a novel motion reconstruction algorithm for capturing 3D motions in a markerless and unconstrained environment using gyroscope and accelerometer measurements. The algorithm is based upon an Iterated Extended Kalman Filter for state estimation and the 3D dynamical model of the subject created in OpenSim (IEKF-OS). The IEKF-OS algorithm consists of two stages. The first stage incorporates the dynamics of the subject to predict the motion. The second stage tracks measurements from multiple IMUs to improve model estimates of joint angles and speeds. The goal of this thesis is threefold. Firstly, the IEKF-OS is derived and verified using simulations performed on a six-link robot manipulator including sensor placement errors. Secondly, experiments are performed to validate the algorithm’s estimations against the robot’s ground truth joint encoder values. Thirdly, the algorithm is compared against an inverse kinematics method to track IMU estimated orientations (OpenSense) provided by OpenSim. The IEKF-OS algorithm showed lower motion tracking errors compared to OpenSense for various motions performed by the robot manipulator. The joint angle estimations as computed by both methods are compared against the gold-standard ground truth robot encoder values. OpenSense joint angle values were in the range of 0.6-6.4 [deg] RMSE, whereas the IEKF-OS algorithm estimated joint angles and speeds in the range of 0.4-2.8 [deg] and 0.3-2 [deg/s] RMSE, respectively. These results highlight accurate 3D motion reconstruction on a six-link robot manipulator. Contrary to OpenSense, the IEKF-OS is able to accurately reconstruct motions regardless of magnetic disturbances because it does not rely on magnetometer data.

To better understand and predict osteoarthritis, researchers are developing so-called coupled modelling workflows. Coupled workflows convert data from gait analysis studies to subject-specific tissue mechanical response estimations through the use of musculoskeletal and finite element models. The tissue mechanical response inside the joint is thought to play an integral role in the onset and progression of osteoarthritis. To study this, the design of coupled workflow was proposed. This design contained subject-specific gait data which was processed by a musculoskeletal model with a single degree of freedom knee joint. Musculoskeletal output was transferred through an adjusted generic finite element model of the knee to calculate maximum principal stress and shear strain values in the tibial cartilage of the knee. Proper marker placement is crucial for making an accurate assessment of the patient’s function in gait analysis studies. It has been claimed that marker misplacement is the main cause of measurement variability in gait analysis studies. It however remains unclear how potential marker misplacement propagates to the coupled modelling workflow results. To investigate this, in addition to the design of a coupled workflow, a sensitivity analysis was performed. With this sensitivity analysis we tried to answer the following question: How does marker placement of knee joint markers in gait analysis influence the tissue mechanical response calculated by a coupled workflow? For the sensitivity analysis, knee joint marker placements were virtually perturbed along anterior-posterior, proximal-distal and medial-lateral direction, to mimic marker misplacement. Corresponding knee biomechanics were estimated from the perturbed input data in the coupled workflow. Peak maximum principal stress values varied by up to 0.60MPa and peak shear strain varied by up to 0.08% as a result of perturbed knee marker placement. For cumulative stress levels, broader relative ranges were found. Moreover, the results showed that marker placement along the anterior-posterior direction had the greatest influence on corresponding tissue mechanical response estimations. In future studies a standard error of measurement margin is proposed in the assessment of coupled modelling worfklow results. In conclusion, the proposed workfow was relatively easy to build and provided similar tissue mechanical response result to those as reported by more complex models which were more computationally intensive. This implies that in the future, coupled modelling processes may very well be incorporated into the clinical decision-making process for musculoskeletal disorders like osteoarthritis.

The study of human motion consists of the analysis of kinematics, dealing with joint angles, velocities and accelerations, and kinetics, which deals with joint torques and interaction forces. Traditionally, kinematics and kinetics are estimated from optical marker data and sometimes measured interaction forces. Inertial measurement units (IMUs) were introduced as a low cost alternative that makes measurements outside the laboratory possible. Measurements are often used in a loosely coupled combination of marker-based inverse kinematics (IK) or orientation-based inverse kinematics (OBIK) and inverse dynamics (ID). The orientations for OBIK are first estimated from IMU data. Recently, tightly coupled estimation methods based on nonlinear state space models have been introduced for both sensor modalities to improve accuracy. Sensor fusion of marker and IMU data is a relatively unexplored area of research. Markers provide a measurement on the position level whereas IMUs measure on the velocity and acceleration levels. Fusing the two sensor modalities is hypothesized to result in the most accurate estimation of kinematics and kinetics. The objective of this thesis is to investigate the effects of the sensor modality and the estimation method on the accuracy of estimated kinematics and kinetics. An iterated extended Kalman filter (IEKF) was developed which estimates joint angles, velocities and torques using any combination of marker data, IMU data and measured interaction forces. To validate the system with respect to ground truth, two experiments (fast motion and interaction force) were carried out on a robot equipped with all three sensor types. For comparison, IK, OBIK and a marker/IMU-based kinematic extended Kalman filter (EKF) were used together with ID to estimate kinematics and kinetics as well. The IEKFs estimated joint angles with a root mean square error (RMSE) between 0.33◦ (markers and interaction forces only) and 1.58◦ (IMUs only, fast motion). The marker-based IEKF and the IMU-based IEKF outperformed IK and OBIK respectively in both experiments. The IMU-based IEKF improved joint velocity RMSE from 4.13 deg/s to 1.10 deg/s in the fast motion experiment. Sharp peaks in the joint torque signal were only estimated accurately using IMU data directly. The marker/IMU IEKF was the only method that estimated joint angles, velocities and torques with RMSEs below 1◦ , 1.5 deg/s and 1.5 Nm respectively in both experiments. It is concluded that using IMU data in a tightly coupled system improves joint velocity and torque estimation. Assuming accurate sensor registration and no soft tissue artifacts, a tightly coupled IMU-only method can be sufficiently accurate in practice.

Various neuromuscular disorders such as spinal cord injury, Charcot-Marie-Tooth disease and poliomyelitis lead to calf muscle weakness, which limits the patient's ability to propel their body forward during gait. Their abnormal gait pattern is characterized by increased ankle dorsiflexion, excessive knee flexion during stance and reduced ankle push-off power which leads to decreased walking speed and increased energy cost of gait. To improve walking ability, dorsal leaf spring (DLS) ankle-foot orthoses (AFOs) are often worn that provide stiffness around the ankle joint which is its most important characteristic. Selecting optimal AFO stiffness is important because it can substantially reduce the net metabolic cost of gait, while improper AFO stiffness can lead to discomfort, fatigue and overall reduced activity. Experimental data shows that the effect of AFOs is dependent on the individual characteristics of the patients, such as level of muscle weakness, spasticity and body mass. Despite the importance of AFO stiffness, empirically determining the stiffness of the applied AFO remains time-consuming and ad hoc. This master thesis aims to uncover the mechanism relating AFO stiffness to the metabolic cost of transport (CoT) as observed in experimental trends from individuals with calf muscle weakness. We used predictive forward musculoskeletal simulations to reproduce the experimental trends and analyzed our simulations to explain the relationship between AFO stiffness and metabolic CoT. The predicted optimal AFO stiffness was within 1 Nm/deg of the patient's experimentally measured one. From the experimentally observed effects, all kinematic and kinetic trends were predicted within 0.5-2.5 SD from the mean of the experimental slopes of all patients' results. Moreover, the differences between the slopes of the simulation results were mostly lower compared to the modelled patient's individual experimental slopes than compared to the experimental slope of the group average results. We identified a reduction in the vasti muscle metabolic cost due to larger external knee extension moments with increasing AFO stiffness as the main mechanism resulting in the net metabolic CoT trend. Limitations on translating our findings to patient behavior include: modelled calf muscle strength, where the model may be significantly stronger or weaker than the patient, and the relative importance of metabolic cost minimization to other goals during walking, such as minimizing the loading rate at heel strike.