21090 research outputs found
Sort by
Validating Internal Density Calibration in The Proximal Humerus to Estimate Bone Stiffness Using Finite Element Analysis for Stemless Shoulder Arthroplasty
No full textStemless humeral head components are a popular choice for patients undergoing shoulder arthroplasty for end-stage osteoarthritis (OA). OA is known to alter bone density in the humeral head, which can compromise implant stability and increase the need for surgical revisions. Current pre-operative clinical assessments are limited in evaluating bone mineral density (BMD) and fail to consider the mechanical properties of bone in the region directly supporting the stemless component, leaving a critical gap in understanding the structural integrity of the bone supporting the stemless component. Traditionally, in-scan phantom calibration determines volumetric bone mineral density (vBMD) from greyscale intensity in computed tomography (CT) images, but this method is rarely used in clinical practice due to limited time and resources. As a result, alternative density measures for determining accurate vBMD from clinical CT images are needed. Internal density calibration using internal tissues as references has been validated in the spine and hip, however, it has yet to be validated in the proximal humerus. Additionally, vBMD derived from internal density calibrated images has yet to be linked to finite element model (FEM) apparent stiffness in the context of stemless shoulder arthroplasty. Establishing stiffness as a measure of bone mechanical properties is a first step in accurately predicting bone strength in clinical CT images.
The purpose of this thesis was to 1) determine the correlation between phantom and internal density calibration in the proximal humerus using three different tissue combinations, 2) compare vBMD in an end-stage OA patient group to a non-pathologic group, and 3) determine the correlation between vBMD and apparent stiffness. Non-pathologic cadaveric single-energy CT images containing a dipotassium phosphate (K2HPO4) phantom were used to analyze a 10 mm thick volume of interest (VOI) directly below the anatomic neck. Phantom and internal density calibration was performed on each phantom-containing cadaveric specimen. vBMD was extracted and FEMs were generated from the VOI. The internal calibration with the lowest bias was used to calibrate images for all end-stage OA patient specimens. VOIs were created for cortical, trabecular, and combined (integral) bone compartments across specimens and vBMD was extracted for each compartment. FEMs were generated using the integral VOI to estimate apparent stiffness.
Statistical analysis revealed a strong correlation between internal and phantom density calibration, establishing internal calibration as a valid metric for determining vBMD (AAdC R2 = 0.80; AAdCM R2 = 0.88; ACM R2 = 0.90). The ACM (Air, Cortical Bone, Skeletal Muscle) tissue combination had the lowest error (Mean: 13.08 mgK2HPO4/cm3). The end-stage OA patient group had significantly lower integral (Patient: 119 mg K₂HPO₄/cm³; Cadaver: 159 mg K₂HPO₄/cm³), cortical (Patient: 518 mg K₂HPO₄/cm³; Cadaver: 643 mg K₂HPO₄/cm³), and trabecular (Patient: 79.8 mg K₂HPO₄/cm³; Cadaver: 110 mg K₂HPO₄/cm³) vBMD than the non-pathologic cadaveric group (p<0.001), highlighting the biological relevance of vBMD. Mean apparent stiffness was found to be significantly lower in the end-stage OA group (672 MPa) relative to the non-pathologic cadaveric group (1261 MPa) (p < 0.001), however stiffness was not correlated with cortical vBMD in either group (Patient: R² = -0.018, p = 0.73; Cadaver: R² = -0.018, p = 0.71), suggesting the need for a multi-factorial approach when quantifying mechanical properties using FEMs
Predicting Cardiovascular Events or Death in People with Dysglycemia Using Machine Learning Methods
No full textCox regression is commonly used to analyze time-to-event for patients in tabular medical data. Hazard ratio can then be calculated to show how much riskier an event may occur for a patient in one group versus the other. Nonetheless, the hazard ratio and Cox regression rely on the proportional hazards assumption, which is not guaranteed. In this paper, we investigate the use of machine learning models to predict patients’ outcomes and identify key factors that may influence the outcomes. We focused on the ORIGIN Trial dataset because it has undergone extensive analysis using Cox regression, allowing us to compare the machine learning model results with previous findings. Three outcomes, major adverse cardiovascular events (MACE), expanded composite outcome (COPRIM2), and all-cause death (ALLDTH), were analyzed in this thesis. The machine learning models we used are Neural Network (NN), Random Forest (RF) and Gradient Boosted Trees (GBT), which were trained with nested cross-validation to tune their hyperparameters. When testing the trained models for all three outcomes, we found that machine learning models had higher Area-Under-the-Curve scores (AUCs) than Cox regression (0.91-0.95 vs 0.63-0.65), and Random Forest and Gradient Boosted Trees had excellent recall scores (0.80 - 0.88). Subsequently, we used SHAP values, mean decrease in AUC, and partial dependency plots (PDPs) to further examine variable importance for RF and GBT. For MACE and COPRIM2, prior cardiovascular events (priorcv), cancer, and blood lipid measures are the most important variables, while for ALLDTH, cancer and kidney functions related measures are the most important variables. The PDPs are harder to analyze than hazard ratio due to having no assumptions and fewer restrictions, but it is useful to estimate the non-linear relation between an explanatory variable and the average probability of the outcome occurring to patients in the dataset
Manifold-Aware Regularization for Self-Supervised Representation Learning
No full textSelf-supervised learning (SSL) has emerged as a dominant paradigm for representation learning, yet much of its recent progress has been guided by empirical heuristics rather than unifying theoretical principles. This thesis advances the understanding of
SSL by framing representation learning as a problem of geometry preservation on the data manifold, where the objective is to shape embedding spaces that respect intrinsic structure while remaining discriminative for downstream tasks. We develop a suite
of methods—ranging from optimal transport–regularized contrastive learning (SinSim) to kernelized variance–invariance–covariance regularization (Kernel VICReg)—that systematically move beyond the Euclidean metric paradigm toward geometry-adaptive distances and statistical dependency measures, such as maximum mean discrepancy (MMD) and Hilbert–Schmidt independence criterion (HSIC). Our contributions span both theory and practice. Theoretically, we unify contrastive and non-contrastive SSL objectives under a manifold-aware regularization framework, revealing deep connections between dependency reduction, spectral geometry, and invariance principles. We also challenge the pervasive assumption that Euclidean distance is the canonical measure for alignment, showing that embedding metrics are themselves learnable design choices whose compatibility with the manifold geometry critically affects representation quality. Practically, we validate our framework across diverse domains—including
natural images and structured scientific data—demonstrating improvements in downstream generalization, robustness to distribution shift, and stability under limited augmentations. By integrating geometric priors, kernel methods, and distributional alignment into SSL, this work reframes representation learning as a principled interaction between statistical dependence control and manifold geometry. The thesis concludes by identifying open theoretical questions at the intersection of Riemannian geometry, kernel theory, and self-supervised objectives, outlining a research agenda for the next generation of geometry-aware foundation models
Proactive Characterization of Wildfire Impacts on Drinking Water Treatability
No full textForested catchments are important sources of drinking water globally. They are increasingly threatened by disturbances, prominently climate shocks, including large wildfires. Wildfires alter watershed hydrology and biogeochemistry, leading to reduced infiltration, increased overland flow, and enhanced delivery of sediments, burned vegetation, and pyrogenic material into aquatic systems. Such inputs can alter drinking water source quality and challenge treatability.
Ash is the residual material from wildland fuel combustion, composed of mineral particles and organic matter that can leach into water. While inorganic dissolved compounds from ash can impact water quality by, for example, increasing ionic strength and alkalinity, water-extractable organic matter (WEOM) from wildfire ash contributes to increased post-fire dissolved organic carbon (DOC) concentrations. During drinking water treatment, higher DOC concentrations increase chemical demand (e.g., coagulant, disinfectant), enhance the formation of potentially harmful disinfection by-products (DBPs), cause taste and odor issues, and promote bacterial regrowth in distribution systems. These impacts may also necessitate new infrastructure to manage changes in source water quality, ultimately increasing overall treatment costs.
Although they cannot reflect all watershed processes, bench-scale evaluations provide valuable insights into wildfire impacts on drinking water treatability by isolating treatment-relevant mechanisms at controlled laboratory conditions. However, different approaches used to prepare wildfire ash-impacted waters (WAIWs) limit the inferences that can be drawn from them. Here, key factors (e.g., mixing duration and condition, ash-to-water ratio, and source water quality) that can impact organic matter leaching from wildfire ash to water were investigated. WEOM concentration increased within the first 24 hours of mixing before plateauing or declining as mixing progressed, regardless of ash type and background water source. Continuous mixing yielded higher WEOM concentrations than stagnant conditions, indicating that particle-particle interactions and surface exposure enhanced leaching. WEOM yield also decreased as ash-to-water ratios increased. Despite anecdotal suggestions, a relationship between wildfire ash color and WEOM concentration was not observed (Chapter 2).
Wildfire ash collection methods may also impact inferences drawn from bench-scale drinking water treatability assessments. Unburned vegetation, rocks, or other debris may have physico-chemical properties different from those of ash deposits; thus, increasing uncertainty in treatability assessments. Dry ash homogenization methods (i.e., manual separation, sieving, and pulverization) were investigated because they may mitigate these impacts. Sieving was shown to be the most practical and reliable method for ensuring ash homogeneity. Pulverization enhanced organic matter release from large particles by increasing surface area, but it also generated aerosolized ash, complicating sample handling. In addition, pulverization altered WEOM character, potentially by increasing the availability of smaller organic matter compounds previously encapsulated within ash particles or by mechanically fragmenting larger organic molecules into smaller compounds (Chapter 3).
Subsequent investigations examined the role of settleable ash solids (SAS), a previously overlooked fraction of wildfire ash. SAS substantially increased water alkalinity and make pH control for coagulation extremely difficult. Although pH adjustment enhances DOC removal from WAIW, SAS increased acid demand substantially. The removal of SAS reduced both alkalinity and acid demand; however, as ionic strength was concurrently reduced, floc formation and turbidity reduction for a given coagulant dose decreased somewhat. A limited complementary analysis was conducted to evaluate whether atmospheric ash deposition could also act as a significant driver of source water quality and treatability change. While the impact of atmospheric deposition of ash on water alkalinity depends on the surface area of water body, only exceptionally high atmospheric ash loading could meaningfully alter source water alkalinity in reservoirs that hold large volume of water (Chapter 4).
Wildfire ash alters multiple aspects of water quality concurrently, including turbidity, DOC concentration and character, and alkalinity, so its overall implications for water treatment cannot be adequately assessed by examining individual mechanisms in isolation. Coagulation experiments with WAIWs demonstrated these interacting impacts. At low coagulant (i.e., alum) doses, turbidity was effectively reduced, yet DOC removal remained limited, despite pH adjustment to coagulant-specific optima. Enhanced coagulation combined with higher alum doses improved DOC removal but introduced trade-offs, as turbidity reduction declined somewhat because of reduced ionic strength associated with decreased alkalinity. The results indicated, while wildfire ash can severely deteriorate water quality by increasing turbidity, alkalinity, DOC concentration, and aromaticity, which may increase coagulant demand or necessitate more advanced treatment methods, the underlying coagulation mechanisms for WAIW remain consistent with those in natural waters. Thus, wildfire ash does not present fundamentally new challenges to coagulation; rather, the magnitude of water quality changes following wildfire can pose risk to treatment performance and operational resilience (Chapter 5).
Collectively, this research demonstrates that while bench-scale studies cannot fully replicate the complexity of post-fire watershed processes and wildfire impacts on water quality, they remain essential for isolating and investigating the specific effects of wildfire ash on drinking water treatment processes. Accordingly, it is practical to adopt methods that maximize the extraction of organic matter from wildfire ash and represent worst-case treatment scenarios. These methodological insights help ensure the comparability of bench-scale investigations. This work also shows that wildfire impacts coagulation primarily by complicating pH control and deteriorating drinking water source quality, increasing the need for more intensive treatment processes. Overall, this research establishes a robust methodological foundation for reliably assessing wildfire ash impacts on water quality and for informing the development of strategies to mitigate wildfire impacts on drinking water treatability
Preservation Under Siege: Adaptive Conservation Approaches for Yemen’s Heritage
No full textYemen is no stranger to conflict. In fact, Yemen’s urban heritage, marked by tower houses and fortifications, reflects enduring architecture and craftsmanship developed over centuries. Towards the end of the 20th century, international interest in the preservation of Islamic cities placed Yemen center stage, leading to the award of three UNESCO World Heritage titles within a decade and to major publications and conservation projects.
However, this momentum stalled following the 2011 Arab Spring revolution and the devastating coalition-led aerial raids in 2015, which caused the loss of many historical sites and disrupted the passing down of traditional construction techniques from one generation to the next. The thesis explores heritage preservation amidst Yemen’s ongoing conflict. Unlike existing scholarship that anticipates a stable government and resumption of foreign aid, this thesis proposes adaptive and local solutions for preserving Yemen’s heritage in its current, war-torn reality.
It explores three proposals: re-purposing abandoned settlements of Yemenite Jews as shelters for displaced refugees, leveraging the Yemeni diaspora as potential donors, and utilizing the highly debated waqf system as a self-sustaining mechanism. The village of Bayt Baws in Sana’a is introduced as a case study, providing a grounded, culturally resonant, and logistically feasible model for heritage preservation in Yemen today
Hypothesis Testing of Multivariate Biomechanical Responses using Statistical Parametric Mapping and Arc-Length Re-Parameterization
No full textThis is a post-peer-review, pre-copyedit version of an article published in Annals of Biomedical Engineering. The final authenticated version is available online at: https://doi.org/10.1007/s10439-025-03788-xDetection of differences between experimental biomechanical data sets is critical to quantify effects and their significance. Many forms of biomechanical data are continuous and multivariate in nature, yet contemporary statistical analysis and hypothesis testing most often utilize single-value scalar metrics. However, reducing continuous responses to single-value scalar metrics can introduce bias and eliminate much of the physical context of a response. This study proposes a methodology to perform hypothesis testing directly on continuous multivariate experimental data sets. The methodology couples arc-length re-parameterization with statistical parametric mapping (SPM) to provide a general framework that can be applied to many of the response types found in biomechanics, including sets of responses that do not terminate at a common coordinate or are hysteretic, such as load-unload data. The arc-length-based SPM methodology was applied to three literature data sets representing a cross-section of the types of responses encountered in biomechanics. In each case, the arc-length-based SPM methodology produced results that agreed with contemporary statistical techniques while providing quantification and identification of statistically significant differences between the data sets. The proposed method provided important contextual information and a deeper understanding of the underlying behaviour of a dataset that would otherwise be missed using contemporary single-value scalar metric statistical techniques, such as highlighting specific response features that drive differences between datasets
Dry Extraction of Nickel from Mixed-Hydroxide Precipitates via Reduction and Carbonylation
No full textThe global transition towards electric vehicles (EVs) has prompted significant research into the sustainable and efficient production of battery-grade materials. Among the critical components of rechargeable batteries, nickel (Ni) is of particular importance due to its central role in cathode materials, specifically for Nickel Manganese Cobalt (NMC) and Nickel Cobalt Aluminum (NCA) batteries. Ni is conventionally extracted from primary sources such as laterite ores (containing 2-3% Ni by mass) through hydrometallurgy (with acid-intensive processing) or pyrometallurgy (with high-temperature, energy-intensive processing). Hydrometallurgical extraction produces an intermediate product called mixed-hydroxide precipitate (MHP), which can contain up to 50% Ni by mass on a dry basis, but still requires further processing to obtain high-purity nickel. This study explores an alternative, sustainable and selective extraction pathway for nickel from MHPs derived from laterite ores and spent battery materials (black mass). The explored vapour metallurgical approach is a two-step, dry process: 1) hydrogen reduction of nickel hydroxides with the MHP to metallic nickel at temperatures between 400°C to 500°C, and 2) selective nickel extraction via carbonylation and conditions of 100°C to 120°C and 150 psig to 450 psig. The carbonylation of metallic Ni using carbon monoxide (CO) produces a volatile molecule called nickel tetracarbonyl (Ni(CO)4), which selectively extracts Ni into the vapour phase. Rigorous safety protocols were employed in this research study to handle the toxic nature of the produced Ni(CO)4 molecules, including CO detectors to identify leaks, and an in-situ decomposition furnace downstream of the reactor to thermally decompose the carbonyls. Reduction and subsequent carbonylation experiments were conducted in a pressurized thermogravimetric analyzer (PTGA), allowing for real-time monitoring of mass changes associated with the reactions. Characterization techniques, including Fourier Transform Infrared (FTIR) spectroscopy, inductively coupled plasma–optical emission spectroscopy (ICP-OES), and Brunauer-Emmett-Teller (BET) analysis, were used to quantify Ni extraction, evaluate morphological changes from fresh samples to reaction residue, and confirm the formation of Ni(CO)4. Significant results demonstrated that the Ni extraction via carbonylation is strongly dependent on the precursor’s structural properties, specifically requiring high surface areas, adequate pore sizes, and minimal cobalt content to enhance transport of CO and Ni(CO)4. Optimal reduction conditions were identified at 450°C, producing residues with a balanced surface area and average pore size, favourable for the carbonylation reaction. Increased carbonylation pressure, at 450 psig, improved Ni extraction efficiency to 95% for a black mass-based MHP
Beneath the Tracks
No full textAs of 2025, Vancouver is facing a worsening housing crisis, with homelessness
rising sharply each year and vacancy rates remaining critically low. Given the city’s
limited available land, it is essential to explore alternative spatial strategies. One such
looked opportunity, lies beneath Vancouver’s elevated SkyTrain tracks - a space in
which hundreds of commuters pass over every day yet remains underused. This thesis
explores the untapped potential of these areas, reimagining them as prime locations for
affordable housing that is targeted at young urban residents such as students, recent
graduates and young professionals.
To understand the viability of this housing strategy, four categories of international
case studies relating to under-SkyTrain communities are explored: [1] Reclaiming
Elevated Spaces, [2] Prefabricated Units, [3] Tiny and Flexible Spaces, [4] Communal
and Landscape Engagement. Many of these categories overlap, with the central case
study encompassing all four categories being the Chūō Line in Tokyo. It is a project
that most closely parallels the thesis vision, which successfully integrates housing with
community and commercial life beneath elevated rail infrastructure. As a fully realized
and active project, it offers an example of what an under-SkyTrain development in
Vancouver could become. Other examples from dense cities like Hong Kong and Paris
show innovative responses to limited land, many of which are becoming increasingly
relevant in a rapidly growing city like Vancouver. The thesis also examines local cases
to see how these strategies might actually play out in reality. By examining the range of
both global and local projects, the thesis identifies key design and planning strategies
that may be applicable to Vancouver’s own spatial context and housing challenges.
The following section considers how these spaces could be integrated in
Vancouver, and what it means to build so close to transit infrastructure. It explores topics
such as the Transit-Oriented Development (TOD) strategy, historical context of the
SkyTrain and sound mitigation.
The research, case studies, and context studies are ultimately synthesized into
two design ideas that test how prefabricated housing and community can be integrated
beneath the SkyTrain. The first explores co-living and retail near Metrotown Station
(a high-density area), while the second looks at live-work housing around Royal Oak
Station (a medium-density area). A lower-density site isn’t proposed, since those areas
still have room to grow without needing to build under infrastructure like the SkyTrain.
This thesis challenges the idea that dense cities like Vancouver have run out of space.
It doesn’t claim to solve homelessness overnight, but it argues that under-bridge spaces
shouldn’t be dismissed as leftover gaps. With a shift in perspective, they can become
seeds for community
Nanocellulose from Hemp: Characterization for Molded Pulp Applications
No full textThe global shift to sustainable packaging solutions has generated growing attention towards biodegradable substitutes for traditional plastic materials. Molded pulp products, originating from lignocellulosic fibers, are both biodegradable and recyclable; however, they frequently demonstrate inadequate mechanical strength and moisture resistance, which restricts their use in high-performance packaging applications. This study explores the capabilities of cellulose nanofibers (CNF) as a reinforcing additive to enhance the characteristics of molded pulp.
CNF was generated from multiple hemp-derived sources through Masuko grinding at varying pass levels and characterized using centrifugation-based techniques, such as water retention value (WRV) and settling volume, to assess their degree of fibrillation and dispersion. TEM analysis validated the findings from the centrifugation-based techniques, confirming that the trends noted in settling volume and WRV correspond to fibrillation quality at the nanoscale. Among the samples evaluated, Dry Anka Bast processed at 12 passes exhibited exceptional dispersion characteristics and was chosen for further application. CNF was integrated into molded pulp of four types: softwood, hardwood, thermo-mechanical eucalyptus pulp (TMP), and kraft eucalyptus pulp. Mechanical testing was performed to evaluate the impact of CNF incorporation on tensile strength and structural integrity.
The findings indicated that CNF markedly improved the mechanical properties of molded pulp, especially in both softwood and hardwood samples, where there was a notable increase in tensile strength. Tensile strength increased from 4 MPa to 13 MPa in hardwood pulp, from 4 MPa to 18 Mpa in softwood pulp, from 3 MPa to 14 MPa in Kraft Eucalyptus pulp, and from 1 MPa to 3 MPa in TMP Eucalypts. The findings validate the enhancing capabilities of CNF and emphasize the significance of the CNF source and processing conditions in maximizing the performance of molded pulp. The results of this study contribute to the development of efficient, bio-based packaging solutions and support wider initiatives aimed at minimizing plastic waste via sustainable material innovation
Accumulation and Recovery of Prolonged Low-Frequency Force Depression at Different Intensities of Repetitive Isometric Contractions
No full textProlonged Low Frequency Force Depression (PLFFD) may impact performance in the workplace by influencing musculoskeletal disorder (MSD) risk or reducing force stability. PLFFD is a reduction in low-frequency stimulated force with little change in high-frequency stimulated force for long period after contraction. The objective of this research was to measure changes in PLFFD over exposure and recovery from a half-shift of an isometric contraction task at two intensities on separate visits, and to determine whether there is a relationship between PLFFD and force stability.
Participants repetitively supported a weight with their elbow flexors for 4 consecutive 1-hour work-segments. Participants performed a low-force high-duty cycle and a high-force low-duty cycle workload-matched protocol on two different days. PLFFD and force stability were measured in the biceps brachii muscle throughout task exposure and recovery. PLFFD was measured as the ratio of elbow flexion force produced at low (10 Hz) and high (100 Hz) frequency transcutaneous stimulations on both the left (non-intervention) and right (intervention) arms. A repeated measures ANOVA detected a progression of PLFFD in the intervention arm through intervention and recovery, but no protocol-effects were detected. Force stability metrics of variability (normalized standard deviation) and unsteadiness (average rate of change) of force during an isometric elbow flexion force matching task were poorly predicted by PLFFD and better predicted by changes in Maximum Voluntary Force (MVF).
This research expanded on incidental findings of PLFFD from past research, and reinforced relationships between muscle fatigue and force stability. PLFFD did not recover at different rates depending on exertion intensity, nor impact force stability metrics, therefore likely not affecting worker performance. Muscle fatigue was shown to impact force unsteadiness to a greater extent than force variability, leading to the suggestion that force stability changes caused by muscle fatigue should be considered when designing a workplace