UWSpace

UWSpace is the University of Waterloo’s institutional repository for the free, secure, and long-term home of research produced by faculty, students, and staff.

Depositing Theses/Dissertations or Research to UWSpace

Are you a Graduate Student depositing your thesis to UWSpace? See our Thesis Deposit Help and UWSpace Thesis FAQ pages to learn more.

Are you a Faculty or Staff member depositing research to UWSpace? See our Waterloo Research Deposit Help and Self-Archiving pages to learn more.

Photo by Waterloo staff

Communities in UWSpace

Select a community to browse its collections.

Recent Submissions

  • Thumbnail Image
    Item type: Item ,
    Neuromodulatory Enhancement of BCI Critical EEG Features: Site Specific iTBS Effects on Movement Related Cortical Activity
    (University of Waterloo, 2026-05-29) Wolfe, Paul James
    Self-paced voluntary movement generation induces a complex cascade of activity in the brain to coordinate and execute the desired movement. This activity can be observed as the movement-related cortical potential (MRCP) in the time domain, and the event-related spectral perturbation (ERSP) in the time-frequency domain. The MRCP is generated primarily by the supplementary motor area (SMA) and primary motor cortex (M1) during self-paced movement. Similar networks drive the ERSP leading to event-related desynchronization (ERD) and synchronization (ERS) during movement preparation and execution. Brain computer interfaces (BCIs) leveraging the MRCP and ERSP have emerged as a therapeutic option for supporting neurorehabilitation. These devices rely on high detection accuracy to function effectively, a trait hampered by the inherently noisy data. While research has considered various methods to improve detection accuracy, limited work has investigated methods which improve the saliency of the signal itself. Transcranial magnetic stimulation offers non-invasive methods which transiently modulate the brain. Intermittent theta burst stimulation (iTBS) is one such method which increases cortical excitability in a targeted region of the cortex. This thesis proposes the application of iTBS to enhance the neural correlates of movement preparation and execution to ultimately improve the signal-to-noise ratio within the MRCP and ERSP. Different locations of iTBS stimulation were investigated targeting the primary generators of these signals to maximize the induced change. Three experiments were conducted differentiated by the location of iTBS stimulation, over M1, SMA, and both M1 and SMA sequentially. In each experiment, participants performed a self-paced ballistic gripping task before and after receiving the intervention. The MRCP was analyzed as mean amplitudes of four components: the readiness potential (RP), negative slope (NS′), motor potential (MP), and peak negativity (PN). In the frequency domain, ERSP was investigated through mean amplitudes of ERD/ERS in alpha and beta bands both before and after movement onset. Time windows of interest in the pre-movement phase were selected to align with MRCP components. Post-movement ERS was quantified over two windows of time to reflect early and late ERS. Results indicated that iTBS significantly modulated the MRCP and ERSP across the three experiments. Within the MRCP, the NS′ and MP were enhanced after sequential M1 + SMA iTBS and M1 iTBS, respectively. The difference between these outcomes suggests that sequential iTBS preferentially modulated intracortical connections between SMA and M1. Surprisingly, SMA iTBS suppressed the PN while not significantly modulating any other component. While the RP was not modulated across any of the experiments, beta ERD was amplified following SMA iTBS in this pre-movement time window. However, this effect was inverted after sequential iTBS which induced a reduction in beta ERD during the pre-movement period. Both alpha and beta bands following movement onset were facilitated by sequential iTBS while M1 iTBS trended towards facilitation of only beta band ERS. SNR of the MRCP was significantly enhanced by sequential iTBS but not following single site iTBS at either site. Conversely, alpha and beta SNR were not significantly adjusted by any of the interventions. Lastly, time to maximal effects differed between the experiments such that iTBS protocols involving SMA exhibited their strongest effects thirty minutes after stimulation compared to M1 iTBS which was most prominent ten minutes after stimulation. The findings underscore the influence iTBS has upon the neural correlates of movement preparation. Single- and multi-site iTBS significantly increased the size of the various components of the MRCP and the ERSP. Importantly, the interventions involving SMA iTBS are capable of modulating early movement preparation processes, a necessary trait to benefit many neurorehabilitation approaches. The ability of iTBS to modulate neural signatures of motor planning positions itself as a promising avenue for improving signal salience and ultimately BCI performance. Interestingly, the findings suggest potential advantages to each iTBS protocol rather than a generalized “one-size-fits-all” protocol. For example, BCIs operating in the frequency domain may find superior performance with the early beta ERD facilitation induced by SMA iTBS. Conversely, sequential iTBS may benefit MRCP-based BCI due to robust facilitation of the NS′. By examining three novel applications of iTBS, this thesis demonstrates that iTBS can modulate neural mechanisms underlying movement preparation, thereby advancing understanding of its potential for BCI and, by extension, neurorehabilitation.
  • Thumbnail Image
    Item type: Item ,
    Categorical Limits of Quantum Graphs and Possibilities Induced by Quantum Pseudometrics
    (University of Waterloo, 2026-05-29) Zhu, Jennifer
    Two monographs [Wea12] and [KW12] introduced new notions of quantum relations and quantum pseudometric spaces incorporating inspiration and techniques from a broad array of fields related to quantum theory. We begin by investigating quantum relations; namely, we find a new formulation of a morphism of quantum relations. Under the general principle that classical functions should be dualized to contravariant maps between associated algebras in quantum theory, we use some operator space theory to analogously dualize the complement of a subset of vertices. This framework yields a representation independent expression of a morphism of quantum relations that aligns with previously representation dependent ones under the appropriate assumptions. Under these morphisms, the categorical (co)limit of a subclass of quantum relations has an obvious candidate. We also define these morphisms on the level of bimodules We next investigate quantum pseudometrics with an emphasis on the quantum mechanical interpretation of the background and results. The motivational theorem (due to unpublished work by Farah and Weaver) is that pure states of a von Neumann algebra 𝓜 are in bijection with maximal filters in the projection lattice of 𝓜. Under the observation that the neighborhood filter of a point in a topological space is also a maximal filter and armed with a notion of distance 𝜌 between projections given by quantum pseudometrics, we investigate whether 𝜌 induces a notion of distance between pure states.
  • Thumbnail Image
    Item type: Item ,
    Improving OOD Detection, Recognition, and Understanding via Multi-Modal Feature Alignment
    (University of Waterloo, 2026-05-29) Wang, Yimu
    The deployment of artificial intelligence (AI) systems in real-world scenarios, such as autonomous vehicles encountering novel road conditions and medical AI analyzing rare pathologies, requires robust handling of out-of-distribution (OOD) data—inputs that differ from the training distribution due to semantic shift (e.g., novel object categories) or covariate shift (e.g., changes in lighting or sensor noise). Achieving this robustness requires that AI systems progress from detecting OOD data, to recognizing novel categories within OOD data, and ultimately understanding OOD scenarios to answer user queries—an essential capability for safety-critical applications where understanding unfamiliar situations is required. This motivates the use of increasingly advanced vision-language capabilities, yet current models face technical barriers in multi-modal feature alignment that limit practical deployment in OOD detection, recognition, and understanding. The challenges and applications of multi-modal feature alignment for these tasks have not been fully explored. This thesis makes three key contributions to advance the understanding and application of multi-modal feature alignment in vision-language models (VLMs). First, we address the limitations of VLMs in OOD detection. We observe that the modality gap between image and text features causes high false positive rates, as OOD samples can exhibit high similarity to in-distribution (ID) text prototypes. To overcome this limitation, we propose a novel few-shot OOD detection method that incorporates ID image prototypes alongside ID text prototypes. Our method introduces the Bias Prompt Generation module to enhance image-text fusion and the Image-Text Contrastive module to reduce the modality gap. This multi-modal prototype approach significantly improves OOD detection accuracy across multiple benchmarks. Next, we tackle OOD recognition through 3D open-vocabulary semantic segmentation, which leverages VLMs to generate point-wise recognition results for novel object categories. Due to the lack of large-scale 3D-language data, current methods distill knowledge from pre-trained 2D VLMs into 3D models. However, this distillation is supervised by misaligned 3D-scene-image-to-text data pairs, leading to suboptimal performance. To address this issue, we propose an aligned 3D open-vocabulary semantic segmentation framework with two novel modules: a CLIP-Rewarded Alignment Module that generates high-quality, well-aligned 3D-scene-image-to-text pairs through temperature-based generation and CLIP-rewarded sampling, and an Adaptive Segmentation Module that introduces trainable tokens within the text encoder to adapt it to 3D contexts. This approach significantly outperforms previous methods on representative benchmarks. Finally, we explore efficient multi-modal feature alignment for OOD understanding. In real-world applications such as autonomous driving and medical diagnosis, AI systems must not only detect and recognize OOD data but also generate appropriate responses by understanding the scenario—for instance, determining safe actions when encountering an unexpected obstacle or providing diagnostic insights for rare pathologies. Multi-modal large language models (MLLMs, as a class of generative VLMs) offer strong generalization capabilities for such understanding tasks, but their substantial computational requirements limit practical deployment. To address this, we propose an efficient MLLM that incorporates a novel conditional token reduction module to consolidate visual tokens based on their similarity to text tokens and learnable queries, and a novel mixture of multi-modal experts module with a router that takes both text and visual tokens as input for better switching between different low-rank adaptation (LoRA) experts. The proposed method achieves competitive performance while using significantly fewer visual tokens, enabling efficient OOD understanding without sacrificing effectiveness. This thesis demonstrates that systematic improvements in multi-modal feature alignment can address multiple complex OOD challenges, from detection through recognition to understanding. These contributions establish a foundation for the deployment of AI systems in open-world environments, enabling more reliable and scalable AI systems that can robustly handle novel scenarios.
  • Thumbnail Image
    Item type: Item ,
    Multi-scale computational modeling towards efficacy in radiopharmaceutical therapies while minimizing side effects: Modeling of amino acid infusion
    (Public Library of Science, 2025-07-16) Golzaryan, Aryan; Souri, Mohammad; Kashkooli, Farshad M.; Rahmim, Arman; Soltani, M.
    Amino acid infusion (AAI) is a technique used in radiopharmaceutical therapy (RPT) to reduce toxicity in kidney and increase clearance rate of radiopharmaceuticals from body. In this study our aim is to evaluate its effect in personalized RPT considering kidney and salivary glands as dose limiting organs using a multiscale modeling framework. We developed a Physiologically-Based Pharmacokinetic (PBPK) model consisting of 19 compartments, personalized it for four prostate cancer patients using data derived from gamma camera imaging. This model was used to investigate the influence of AAI on the absorbed dose to tumors and organs at risk. We then computed the maximum safe injected activity based on the PBPK model. To address the effects of interstitial fluid pressure (IFP) and tumor heterogeneity, we coupled the PBPK model with convection-diffusion-reaction (CDR) equations. To compare the effectiveness of our modeling approaches, we calculated absorbed doses to the tumors with and without AAI, using both the standalone PBPK model and the coupled PBPK-CDR model. Our findings revealed a relative error (RE) of 9.6% ± 2.2% (mean ± SD) in total tumor absorbed dose calculation between PBPK and CDR equations, attributable to the consideration of IFP. Moreover, AAI proved beneficial for RPT when the kidney was designated as the organ-at-risk. It enabled an increase in radiopharmaceutical injection from 12.3 ± 6.32 MBq (mean ± SD) to 15.45 ± 6.95 MBq (RE: 28.5% ± 15.7%), resulting in a corresponding increase in tumor absorbed dose from 67.8 ± 47.45 Gy to 72.43 ± 51.03 Gy (RE: 8.6% ± 5.4%), while maintaining critical kidney absorbed dose limits. However, this was not observed when the salivary gland was considered the dose-limiting organ. Although, AAI allowed for increased therapeutic injection ranging from 4.22 ± 2.23 MBq to 5.25 ± 3.14 MBq (RE: 19.2% ± 9.9%), it results in a minimal increase in tumor absorbed dose of 0.22 ± 0.04 (RE: 1.4% ± 1.3%). Statistical analysis using the Wilcoxon Signed-Rank Test revealed significant effects of AAI on administered activity and tumor absorbed dose (p-value = 0.007 < 0.05). Finally, a local sensitivity analysis was performed on selected radiation and tumor transportation parameters individually to evaluate their impact on the tumor absorbed dose. In conclusion, selection of organ-at-risk in personalized RPT is critical, as it determines the injected activity amount and the efficacy of delivery-enhancing techniques.
  • Thumbnail Image
    Item type: Item ,
    The structure of informal water markets: Insights from spatial monitoring in Lodwar, Kenya
    (Public Library of Science, 2025-08-05) Zhao, Jane; Garrick, Dustin; Ekwar, Paul; Hoque, Sonia Ferdous; Hope, Robert; Whittington, Dale
    Public water utilities have struggled to keep pace with rapid urbanization, particularly in towns and small to medium-sized cities of low-income regions. Informal water markets have proliferated to fill gaps in piped water coverage and service delivery through a wide range of water vending activities (from private water sources to tanker trucks and handcart operators that distribute water). Despite the prevalence and persistence of water vending, the structure, impacts, and evolution of informal water markets in these settings remain poorly understood, especially the interaction between private vendors and public utilities. We seek to improve our understanding of mobile, distributing vendors (tankers, motorcycles) by advancing high-frequency, spatially explicit monitoring of water vendor transactions in Lodwar, Kenya. We examine both the market and spatial structure of the informal water supply system and then draw inferences about their impacts and evolution. We find that vendors that use motorcycles are not making profits from transporting water. We also identify many linkages between the formal and informal systems. For example, purchases of bulk water by water vendors account for 28% of the public water utility’s revenue. We also find that while most consumers of vended water are located outside of the piped water service area, many households and institutions inside the service area still purchase from private water vendors due to concerns about reliability and quality. These results highlight the complementarities between public utilities and private water vending and the corresponding importance of mapping water vending networks to support planning, policy, and investment and to protect consumers.