Biomedical Engineering - Theses

Permanent URI for this collection

Search Results

Now showing 1 - 10 of 87
  • Item
    Thumbnail Image
    Positron emission tomography image reconstruction and analysis in the study of Alzheimer's disease
    Ruwanpathirana, Gihan Pramuditha ( 2023-07)
    Alzheimer’s disease (AD) is a progressive neurological disorder that eventually leads to the onset of dementia. Neuropathology of AD is characterised by two stereotypical protein accumulations: extracellular amyloid-Beta (ABeta) plaques and intracellular tau neurofibrillary tangles. With the development of Positron Emission Tomography (PET) radiotracer technologies, tracers were developed to image these proteins in vivo. The extent of the ABeta-PET accumulation in the brain is commonly used and quantified using standardized uptake value ratio (SUVR) and Centiloid scale (CL), which is a scaled variation of SUVR values. The PET image reconstruction algorithms and associated parameters alter the PET image quality measures such as spatial resolution, that will impact the quantitative metrics derived from these PET images. However, the impact of PET image reconstruction on both cross-sectional and longitudinal ABeta-PET quantitation hasn’t been studied. Therefore, this thesis focuses on investigating the impact of the spatial resolution as determined by the PET reconstruction configurations on ABeta-PET quantitation. Clinical PET scanners exhibit variations in PET scanner technologies and reconstruction methods. Consequently, imaging the same patient on different PET scanners introduces variability in PET images and the quantitative measures of ABeta-PET. This is a challenge in ABeta-PET studies as they are normally carried out across multiple centers. Therefore, it is essential to perform ABeta-PET harmonisation to ensure consistent quantitation measures for ABeta-PET imaging across different PET scanners. One objective of this thesis is to develop an inter-scanner ABeta-PET harmonisation procedure by utilizing a barrel phantom to match the measured spatial resolution. This research extends beyond phantom data and harmonisation scheme is also validated using subjects scanned on multiple scanners. Despite the presence of ABeta plaques and tau fibrillary tangles in AD brains, the interaction between these two proteins in the pathogenesis of AD remains incompletely understood. Previous cross-sectional studies that explored the relationship between ABeta-PET and tau-PET using linear, voxel-wise, or region-of-interest (ROI) techniques, omitting the spatial dependencies between voxels or brain regions. However, it is widely recognized that tau accumulation in the brain exhibits strong spatial dependencies. Therefore, it is important to employ analysis methods that can explore multivariate, nonlinear relationships between ABeta and tau across spatially distant brain regions. This approach has the potential to provide novel insights into the interactions between these two molecular species. Therefore, this thesis also investigates the use of deep learning, a Convolutional Neural Network, to examine the relationship between tau-PET images and ABeta-PET quantification throughout the AD continuum.
  • Item
    Thumbnail Image
    Biomechanical function of the native and prosthetic temporomandibular joints
    Woodford, Sarah Claire ( 2023-08)
    The temporomandibular joint (TMJ) is a coupled synovial joint connecting the skull and the mandible; and is essential for biting, swallowing and speech, and required for maintaining quality of life. Disorders of the TMJ are common, with 5% to 60% of the adult population showing clinical signs of a TMJ disorder, including muscle and joint pain and tenderness, and restricted mandibular movements. Total TMJ replacement (TMJR) surgery is the established treatment for end-stage TMJ disorders such as degenerative joint disease and ankylosis and is used in mandibular reconstruction following condylar fractures and tumor resection. This procedure has been shown to increase mouth opening capacity, reduce pain and improve quality of life; however, biomechanical TMJ function following total TMJR surgery is poorly understood. As a result, the ability for total TMJR surgery to restore jaw function remains limited. The objective of this thesis was to evaluate the jaw and TMJ kinematics, maximum voluntary bite force, muscle functionality and TMJ loading for unilateral total TMJR patients and compare them to healthy control subjects. A systematic literature review was conducted to critically investigate published literature on measurement modalities used to quantify mandibular and TMJ kinematics in six degrees-of-freedom. Measurement techniques fell into four main categories, namely mechanical linkage systems, magnetic tracking systems, video motion analysis and radiographic tracking. Magnetic tracking was identified as a technique with high motion measurement accuracy which circumvents skin motion artefact by attaching sensors directly to the teeth. Following this, magnetic tracking was used to develop a modelling framework to evaluate subject-specific three-dimensional (3D), dynamic bite forces during biting and chewing. Measurement of jaw motion was captured on one healthy adult using a Dental Motion Decoder (DMD) system (Ignident, Ludwigshafen, Germany); this samples the position and orientation of two low-profile electromagnetic field sensors attached to the teeth. The subjects occlusal anatomy was quantified using an intra-oral scanner. Following this, the subject was instructed to chew and maximally compress a rubber sample between their teeth. The occlusal anatomy, rubber geometry and experimentally measured rubber material properties were combined in a finite element model. The measured mandibular motion was used to kinematically drive model simulations of chewing and biting on the rubber sample, and 3D dynamic bite forces were calculated. A validation study found excellent correlation between experimentally measured force during tooth-rubber penetration and that predicted from the modelling framework. Following this, an optoelectronic tracking system was used to measure jaw kinematics during mandibular border movements and biting of a rubber sample for 6 total unilateral TMJR patients and 10 age- and gender-matched healthy controls. Optoelectronic tracking was employed for its ease of use in a research setting, as the DMD system previously used is restricted to clinical environments. Finite element analysis (FEA) was employed to calculate the maximum voluntary bite force. Musculoskeletal models were used to investigate muscle and TMJ loading. Unilateral total TMJR patients had restricted anterior mobility of their prosthetic condyles, this resulted in limited opening, lateral and protrusive movement capacity. During chewing, unilateral total TMJR patient’s prosthetic condyles were positioned significantly inferiorly to those of healthy controls, suggesting a weakness of the elevator muscles on the operated side. There was a non-significant trend that healthy controls had higher maximum voluntary bite force than unilateral total TMJR patients. Furthermore, total unilateral TMJR patents had higher maximum voluntary bite force when biting on their contralateral molars than their ipsilateral molars. Unilateral total TMJR patients had high medial pterygoid forces on the ipsilateral side and significantly lower temporalis forces on the ipsilateral side compared to the contralateral side. In addition, total unilateral TMJR patients had lower TMJ loading than control subjects, particularly on their native contralateral joint. This abnormal loading may be a mechanism associated with degenerative changes that have been observed in the contralateral joint following unilateral total TMJR surgery. This thesis presents new data describing jaw kinematics and kinetics following TMJR surgery, and proposes new techniques to support rigorous biomechanical analysis of the mandible, dental structures and TMJ both in clinical and research settings. The results of this thesis may guide the design and evaluation of TMJ prosthetics, prosthodontic devices, dental restorations and maxillofacial orthopaedic implants; furthermore, these results may influence physiotherapy prescription following unilateral total TMJR surgery to enhance long-term patient outcomes.
  • Item
    Thumbnail Image
    A predictive neural model of visual information processing
    Zhang, Yu ( 2023-08)
    Recent understanding of how the brain processes sensory input has moved away from understanding sensory processing as just being the passive processing of input to a framework that employs a more active role of higher-level expectations in sensory processing. One such theory is predictive coding in which the brain generates a prediction of the sensory input that it will receive and compares this prediction with the actual sensory input. Because the transmission of visual information from the eyes to the brain takes time, for the brain to accurately respond to the real-time location of a moving object, the prediction mechanism has to take into account the change in object’s location during the period of transmission latency. Understanding such temporal prediction mechanism will extend our understanding of the way by which the brain actively interacts with the environment we live in. This research project investigated the predictive signal transmission pathways of the mammalian visual system and focused on the early stages of the visual pathway, including the retina, the Lateral Geniculate Nucleus (LGN) and the primary visual cortex (V1). These structures play critical roles in visual signal gathering and integration. Mathematical and computational models were constructed based on predictive coding strategies and spike-based neural coding principles, where neurons with specific firing timings are arranged into hierarchical areas, and upper areas predict the neuronal behaviours of lower areas that receive sensory stimuli. The first goal of the project is to investigate the encoding of visual information in precise neuron spike timings and neuronal interactions, because the temporal prediction mechanism involves small time scales and detailed object motions. We intend to show that results obtained via spike-based neural principles, which involves cumulative computations in small time scales, do not contradict with the results from classical rate-based neural networks that operate based on longer time scales, and results from physiological recordings. The second goal is to investigate the mechanism by which temporal prediction can be achieved using the spike-based neural network, given moving input stimuli. Through the project, we validated that a predictive coding network can be built based on spike-based neural principles, and it has the potential to encode moving stimuli with less error compared with rate-based approaches. Based on the model developed, the next step is to study the specific mechanism by which the alignment between the real and the perceived locations of a moving object can be achieved, i.e., a mechanism that compensates for the signal transmission delay from the eyes to the brain. Outcomes of the research are expected to advance our understanding about human visual system and provide new insights into the development of neural implants, prostheses and machine learning algorithms. The principles investigated are hypothesised to apply throughout the cerebral cortex. Consequently, the results are anticipated to have application to the processing of other modes of sensory stimulus, such as auditory and olfactory inputs, applications can also be expanded to the research areas of memory, motor control, cognition and decision making.
  • Item
    Thumbnail Image
    Artifact Reduction Methods for Ultra-high Field Magnetic Resonance Imaging
    Yaghmaie, Negin ( 2023-04)
    Ultra-high field Magnetic Resonance Imaging (MRI) provides high spatial resolution and increased signal-to-noise ratio for brain imaging. While certain applications particularly benefit from the increased signal and contrast offered by imaging at ultrahigh field, challenges remain, such as Radio Frequency (RF) field inhomogeneity, increased energy deposition and associated artifacts. This thesis introduces three artifact reduction methods for different stages in the MRI acquisition and analysis pipeline: Firstly, an artifact reduction method for Quantitative Susceptibility Mapping (QSM) is introduced as part of a post-processing pipeline to suppress the streaking artifacts near high susceptibility sources in the brain. In the second part, an eddy current artifact reduction method for diffusion-weighted MRI is proposed, in which phase data from an additional readout during the acquisition is used to estimate the eddy current fields and retrospectively correct the reconstruction. Lastly, a new acquisition and reconstruction scheme using spatially selective RF pulses is introduced to enhance the performance of slice-accelerated imaging.
  • Item
    Thumbnail Image
    Cycles and Seizure Forecasting in Epilepsy
    Stirling, Rachel Elizabeth ( 2023-08)
    People with epilepsy were once termed “lunatics” because their seizures appeared to occur in synchrony with the lunar cycle. The cause was sinning at the wrong time of the lunar phase: “If they [the planets] should scrutinize while the Moon is putting an end to a certain phase, they produce maniacs, ecstatics, epileptics, those who chant”. This conviction was consistent with medicine at the time, which believed that the moon caused an unnaturally moist brain, leading to epileptic seizures. Nowadays, we know that most people with epilepsy have at least one seizure cycle, the periods of which are unique to the individual (7 days, 10 days, 20 days, 28 days, etc.). Yet we have a very limited understanding of where they derive from and how to best understand them for seizure forecasting and other applications. This thesis aims to expand our scientific understanding of the co-existing factors and mechanistic drivers of multiday cyclical patterns in epilepsy while investigating their utility in improving the performance of seizure forecasting algorithms. The presented research addresses this overarching aim by answering three questions: (1) What physiological signals correlate with cyclical patterns in epileptic seizures? (2) Can physiological signals correlating with cyclical patterns in epileptic seizures be tracked and utilised to improve seizure forecasting algorithms? (3) How can we begin investigating the systemic mechanisms of cycles in epilepsy and beyond? By answering these questions, this thesis aspires to give future researchers the foundational knowledge and resources needed to gain greater insight into cycles in epilepsy and mammalian biology.
  • Item
    Thumbnail Image
    Techniques for Time Series Modelling in low Signal-to-Noise Conditions
    Haderlein, Jonas Felix ( 2023-06)
    This thesis revolves around the problem of analysing low signal-to-noise data streams in domains such as computational neuroscience and brain research. Our current understanding of the brain is not unlike the pre-Copernican question for a good model for planetary motion. We are given a complex system, i.e., the brain, but do not know the underlying system state and its dynamics. This state is opaque for the same reasons planetary motion used to be: measurements from a particular brain are data streams taken, for example, through brain-computer interfaces capturing only a low-dimensional ‘telescope image’ of neuronal activity. These observations are incomplete and noisy as our devices cannot easily resolve the activity of every single neuron, but measure only a tiny fraction of the whole system through a ‘primitive lens’ such as voltage measurements from the surface of the brain. The brain is further in constant interaction with its environment. External sensory input makes the system exhibit constantly and unpredictably changing dynamics. The complexity of the system thus stands in stark contrast to our data, which makes the ‘true laws of motion’ inaccessible at the current stage. Luckily, from an engineering perspective, we aim to derive models not for the laws governing the brain as a whole, but for specific applied tasks. One example is detecting or forecasting events like seizures in the case of epileptic brain behaviour. Information towards such an anomalous state may be hidden in the complexity of measurements. Throughout this work, we use different model states of the brain that (a) are consistent with the current observed data and (b) are low-dimensional projections of the brain dynamics capturing an essential part or behaviour. This thesis presents methods to map measurements onto model states. To this end, we develop novel techniques for time series modelling under uncertainty about the ‘true’ dynamics that gen- erated the data and the model-specific distinction between signal and noise. Our investigations lead us to a paradigm that draws from modern machine learning and optimisation theory. The contributions to the classical domains of ‘systems identification’ generalise to similar contexts where one needs to retrieve models for the unknown dynamics of complex systems from data only.
  • Item
    Thumbnail Image
    Acoustic metasurfaces for micromanipulation
    Xu, Mingxin ( 2023-07)
    Acoustic methods are ideal tools for micromanipulation due to their biocompatibility and ability to generate acoustic fields with micro/nano-scale resolution. These methods utilize the force generated by acoustic waves to pattern, manipulate, and sort cells and microparticles without physical contact. These approaches have several advantages including the ability to handle fragile or sensitive cells without damaging them, and the potential for high-throughput manipulation of multiple cells/micro-particles simultaneously. While acoustic methods have shown great potential in micromanipulation, several challenges need to be addressed. One of the primary challenges is to generate flexible and complex acoustic fields to achieve desired manipulations, such as the trapping of cells or particles with complex patterns. Another challenge is the ability to generate programmable acoustic fields to generate instantly reconfigurable acoustic fields. Acoustic methods for micromanipulation also face challenges related to planarization and integration with other techniques. Therefore, this work introduces the following acoustic approaches based on metasurfaces for micromanipulation: (1) sawtooth acoustic metasurfaces for generating flexible acoustic fields in microfluidic channels using only a single travelling acoustic wave, (2) micropillar-based acoustic metasurfaces for applications in generating complex acoustic holographic patterns, (3) the creation of reconfigurable acoustic holograms through the modification of the sound velocity of the medium, and (4) the integration of acoustic holography with microfluidic systems to generate complex acoustic fields in microfluidic channels, where acoustic metasurfaces with subwavelength structures can achieve unique acoustic properties that do not normally exist in nature. These metasurface methods can generate flexible, complex, and programmable acoustic fields through subwavelength structures, holding great significance for various applications including acoustic tweezers, microfluidics, and biomedical sensing.
  • Item
    Thumbnail Image
    Understanding cognition-related neuronal dynamics and criticality in brain networks: From modeling to in-vitro and in-vivo cortical networks
    Habibollahi Saatlou, Forough ( 2023-04)
    Investigating the neural dynamics associated with cognition in the brain’s cortical networks can contribute significantly to understanding the way in which the brain processes information and adapts to changes in the environment, as well as to developing treatments for cognitive disorders. Electrical oscillations in the brain are rhythmic patterns of neural activity that occur at different frequencies. These oscillations are thought to play an important role in coordinating the activity of different brain regions and allowing for the transfer of information between them. The generation of this intrinsic network rhythmicity is governed largely by the synaptic conductances in the network, but few studies have previously examined the effects of voltage-gated ion channels (VGICs) on these rhythms. In the first study, a pyramidal-interneuron-gamma (PING) network consisting of excitatory pyramidal cells and two types of inhibitory interneurons is used to investigate the effects of several synaptic conductances upon network theta and gamma frequency oscillations. A conductance-based neural network is constructed incorporating a persistent sodium current (INaP), a delayed rectifier potassium current (IKDR), an inactivating potassium current (IA) and a hyperpolarization-activated current (IH). The results show that theta power is altered by all conductances tested. Gamma rhythmogenesis is dependent on IA and IH. The IKDR currents in excitatory pyramidal cells, as well as both types of inhibitory interneurons, were essential for theta rhythmogenesis and altered gamma rhythm properties. Increasing INaP suppresses both gamma and theta rhythms. The addition of noise does not alter these patterns. These findings indicate that VGICs strongly affect brain network rhythms and, therefore, the cognition-related functions associated with these rhythms. A dynamical system is described as “critical” when it is at the borderline between ordered and disordered states, at which the input is neither strongly damped nor excessively amplified. In the second study, the role of near-critical dynamics in a neuronal network’s response to an increased load of input information was investigated. An in vitro neural network of cortical neurons was utilized that were trained to play a simplified game of ‘Pong’. The results of this study demonstrate that critical dynamics emerge when neural networks receive task-related structured sensory input, reorganizing the system to a near-critical state. Additionally, better task performance was found to correlate with proximity to critical dynamics. However, criticality alone is insufficient for a neuronal network to demonstrate learning in the absence of additional information regarding the consequences of previous actions. This provides compelling data to better understand the role of neural criticality in how brains process information. In the third study, criticality in awake behaving animals was investigated. Miniscope recordings were undertaken from several hundred hippocampal CA1 neurons in freely-behaving mice in three states: rest, the cognitive task of novel object recognition, and novel object recognition following scopolamine administration, which greatly impairs spatial memory encoding. The results of this study demonstrate that, while the hippocampus neural network exhibits some characteristics of a near-critical system at rest, the network activity shifts significantly closer to a critical state when the mice engage in novel object recognition. The dynamics are shifted away from criticality again when the animal’s performance in the novel object recognition test deteriorates due to scopolamine-induced memory impairment. In contrast to previous theoretical predictions, the results of this study indicate that hippocampus neural networks move closer to criticality under cognitive load, and that critical dynamical regimes produce a near-optimal state for cognitive operations.
  • Item
    Thumbnail Image
    Perturbation-based biomarkers outperform passive ones as indicators for changes in cortical excitability and seizure transitions
    Qin, Wei ( 2023-06)
    Epilepsy is a neurological disorder that affects patients differently and manifests as spontaneous recurrent seizures. It is a prevalent condition that affects about 50 million people globally, but its exact aetiology and pathophysiology often remain elusive. Functional imaging techniques allow us to observe alterations in brain activity during seizures. For instance, EEG data reveals hyper-excitable and hyper-synchronised neuronal firing in the brain. This thesis explores the potential of using biomarkers to track cortical excitability and detect state transitions in epileptic models. Nine biomarkers were developed based on theoretical concepts such as Critical Slowing Down (CSD) and signal processing methods. The performances of these biomarkers were evaluated in neural mass models, animal models of epilepsy and human data. In this thesis, we first employed two mesoscopic neural mass models to study state transitions in a controlled way. Through simulation and perturbation, we found that biomarkers can anticipate state transitions before they occur. They are effective when a system undergoes a critical transition, but less so when the system jumps between multiple stable states due to stochastic noise. Overall, active biomarkers with perturbations outperform passive biomarkers in terms of accuracy and robustness. We also investigated how biomarkers and perturbations can be used to identify state transitions using experimental data. We contrasted active and passive biomarkers on various time scales: pre-seizure scale, circadian cycle and lifespan scale in animal data. We found that active biomarkers with perturbations are superior to passive biomarkers in monitoring state transitions and giving early warnings before seizures. Moreover, by examining two datasets of human epileptic transitions, we further assess the possibility of adopting biomarkers in clinical studies. It is observed that changes in biomarkers differ depending on both patients and seizures. We propose that seizure transitions are not only patient-dependent but also seizure-dependent. Through this thesis, we have shown that biomarkers can detect the underlying changes that precede a seizure by measuring cortical excitability with perturbations. By applying a small perturbation, it is possible to probe changes in brain states by measuring the response to perturbation, or cortical excitability. Understanding changes in cortical response to perturbation during brain state transitions may yield important insights into brain disorders. The methods employed in this project are anticipated to be applicable to clinical settings for seizure forecasting and epilepsy management.
  • Item
    No Preview Available
    Novel machine-learning approaches to create a structurally accurate virtual model of the heart cell
    Khadankishandi, Afshin ( 2023-06)
    Cardiomyocytes are densely packed with parallel columns of myofibrils and mitochondria. Research has shown a strong correlation between changes in ultrastructure and changes in the heart’s function. For example, heart pumping is often compromised in heart diseases such as cardiomyopathy, hypertension and diabetes. Hence, understanding the 3D architecture of cardiac cells will underpin breakthroughs in cardiovascular disease treatment and prevention. With the advent of high-throughput microscopy image datasets resulting from modalities such as serial block-face and focused ion-beam scanning electron microscopy, we can acquire large datasets of cardiac muscle cells in 3D. However, segmenting these datasets is challenging due to low contrast and high noise ratio. The community often relies on manual segmentation and image tracing, a laborious and cost-inefficient approach that hinders novel breakthroughs. This thesis proposes state-of-the-art deep neural networks to segment ultrastructures of cardiac cells in EM datasets and obtains 3D statistical architecture of the cardiomyocyte. EM-net and EM-stellar are cloud-based software proposed to segment EM image datasets and benchmark a wide range of segmentation performance measures. EM-net is a scalable convolutional neural network offering fast convergence during optimisation and can be trained with minimal ground-truth information due to its novel architecture. EM-stellar is hosted on Google Colab, and it can be used to benchmark the performance of state-of-the-art deep neural networks on a user-specified dataset. Together these pipelines offer the research community more efficient ways to segment and analyse cardiac muscle ultrastructure from electron microscopy datasets. Finally, we propose CardioVinci, a workflow utilising generative adversarial networks (GANs) to obtain a statistical 3D model of cardiomyocyte architecture. CardioVinci addresses a significant challenge with large EM datasets: the time taken to collect tissue samples, acquire the data, extract key characteristics and statistically analyse 3D changes in the ultrastructures. It encodes the 2D and 3D variations in the ultrastructures across the image volume into a generative model. As a result, the community will be able to statistically quantify the morphologies and spatial assembly of mitochondria, myofibrils, and Z-disks with minimal manual annotation.