PRISM | Institutional Repository
Communities in PRISM
Select a community to browse its collections.
Recent Submissions
Item type: Item , Access status: Open Access , Graduate Student Partnership Plan: Working with students in educational development(2026-06) Grant, Kimberley A.; Arshad, M. Adil; Dyjur, Patti; Smith, Erika E.This form for working with graduate students in academic settings can be used to discover the needs and expectations of both partners. Consider providing it to the student prior to discussing to allow time for reflection. It is helpful to revisit the plan over time as your needs change and evolve.Item type: Item , Access status: Open Access , Exploring Parents’ Perceptions of Autistic Youths’ Experiences in an Adaptive Basketball Program(2026-06-05) Dickner, Sydney; McCrimmon, Adam; Makarenko, Erica; Ellard, JohnSports programming has been widely shown to support children and youths’ wellbeing and overall development. However, autistic individuals are often excluded from or face significant barriers to participation in traditional sport settings. In response, adaptive sport programs have emerged as a more accessible alternative, with research demonstrating promising benefits in terms of skill development. Despite these findings, much of the existing literature has focused on quantitative outcomes, leaving limited understanding of how autistic youth and their families experience participation in these programs. The present study explores parent perspectives on their autistic children’s experiences in an adaptive basketball program in Western Canada. An autistic advisory panel informed the study, helping to ensure the research aligned with community priorities. Four parents participated in semi-structured interviews, which were analyzed using Reflexive Thematic Analysis. The analysis generated five themes and 14 subthemes, capturing both shared and unique experiences across participants. Overall, the findings highlight largely positive experiences within the adaptive basketball program, particularly in relation to perceived improvements in skill development, inclusion, and wellbeing. These results contribute to the growing body of literature on adaptive sport and offer important clinical implications, as well as directions for future research.Item type: Item , Access status: Open Access , Cost-Effective Mapping Platform to Generate HD-Maps for Autonomous Vehicles and Digital Twins(2026-06-04) Bharadwaj, Akshay Shankar; Lichti, Derek; El-Sheimy, Naser M; Yang, Hongzhou; O'Keefe, Kyle; Armenakis, CostasScalable and cost-effective generation and updating of high-definition maps (HD-maps) remains a founda-tional challenge for the widespread deployment of autonomous vehicles, contingent on the accuracy of the underlying sensor calibration and localization pipelines. To enable precise mapping, a rigorous full camera system modeling approach is adopted. Checkerboard-based calibration methods, despite not being intended for mission-critical applications, have become widely used in practice. In this work, intrinsic and extrinsic camera calibration are revisited within the framework of PSC. It is demonstrated that PSC consistently outperform checkerboard-based methods in both accuracy and precision across diverse vision tasks, includ-ing 3D reconstruction, SfM, visual SLAM, and novel view synthesis. These results advocate a shift toward calibration methods that more accurately capture the physical and projective properties of real-world camera systems. Accurate fusion of camera imagery with onboard sensors such as LiDAR and IMU is critical for reli-able perception and mapping. While camera–inertial calibration is well studied, LiDAR–camera extrinsic calibration remains challenging, particularly outside controlled environments. To address this limitation, a two-step learning-based calibration framework is proposed. Instance-wise SCs derived from everyday objects in the scene are first used to estimate coarse extrinsic parameters, which are subsequently refined using an end-to-end neural network to obtain fine calibration estimates. This approach enables continuous adaptation to sensor pose variations in real-world settings. To construct and maintain geometrically consistent maps from LiDAR scans, neural implicit methods have emerged as an effective alternative to classical approaches, as they achieve comparable accuracy while requiring significantly lower storage for map representation. However, although strong performance is ob-served under low-dynamic motion, their accuracy degrades in highly dynamic scenarios such as pedestrian motion. To address this limitation, a neural implicit inertial odometry framework is introduced that jointly integrates LiDAR and inertial measurements to produce geometrically consistent 3D PCs. Fine geometric detail is preserved through adaptive LiDAR downsampling and the use of neural point correspondences for accurate inter-scan odometry estimation. Furthermore, an attention-based loop closure mechanism is incorporated to ensure global consistency in large-scale mapping scenarios. The proposed software tools and algorithms are evaluated on benchmark datasets, including Oxford-Spires, KITTI, and EuRoC-MAV, and further validated on a newly collected dataset acquired using a cost-effective prototype mobile mapping system developed as part of this work, in both driving and backpack configurations.Item type: Item , Access status: Open Access , Alpha-1 Adrenoceptor Activation in the Hypothalamus Regulates Neuron and Astrocyte-Derived Adenosine to Modulate Corticotropin-Releasing Hormone Neuronal Activity(2026-06-08) Grovue, Patrick Michael; Gordon, Grant; Borgland, Stephanie; Lohman, Alex; Faiz, MaryamThe paraventricular nucleus of the hypothalamus (PVN) is a key regulator of the neuroendocrine response to stress. Corticotropin-releasing hormone neurons within the PVN (CRHPVN neurons) initiate this response by releasing CRH, ultimately driving activation of the hypothalamic-pituitary-adrenal axis. During stress, noradrenaline (NA) released into the PVN acts on both CRH neurons and neighbouring astrocytes via G protein-coupled adrenoceptors, with α1-adrenoceptors (α1-ARs) being strongly linked to Gq signalling. Although astrocytes are known to contribute to noradrenergic modulation of synapses in other brain regions, their role in modulating CRHPVN neurons remains poorly understood. Adenosine is well known for its inhibitory role throughout the brain and is a plausible modulator of CRHPVN neuron activity downstream of noradrenergic and astrocytic signalling in the PVN. Using two-photon imaging of mouse brain slices expressing genetically encoded calcium indicators together with ATP or adenosine biosensors, I first identified a significantly greater functional importance of the α1A-AR subtype in PVN astrocytes compared to CRH neurons. Next, I demonstrated that phenylephrine (PE), an α1-AR agonist, increased extracellular ATP and adenosine concentrations, and showed that adenosine inhibits CRHPVN neurons via adenosine A1 receptors. Next, I examined the relative contributions of astrocytes and neurons to PE-induced adenosine release. Because Gq signalling in astrocytes can drive calcium-dependent gliotransmission, I tested the possibility that astrocytes contribute to adenosine generation through ATP release, which is readily degraded into adenosine in the extracellular space. Suppressing astrocyte calcium activity increased basal CRHPVN neuron activity but did not alter PE-induced adenosine release, whereas selectively driving astrocyte calcium activity with Gq-DREADDs decreased extracellular adenosine through an as-yet unidentified mechanism. Finally, the PE-induced adenosine signal was found to be strongly dependent on neuronal firing and mediated by efflux through equilibrative nucleoside transporters. Together, these findings illuminate mechanisms of noradrenergic and purinergic signalling in the PVN and suggest that, while neurons are the principal source of adenosine released in response to α1-AR activation, astrocytes may still play an important modulatory role in shaping CRH neuron activity and extracellular adenosine signalling in the PVN.Item type: Item , Access status: Embargo , A Generative Classifier for Trustworthy Ocular Condition Detection from Fundus Photographs(2026-06-04) Omar Ahsan, Ahmad; Wilms, Matthias; Forkert, Nils; Costello, FionaMillions of people worldwide are affected by ocular conditions that can severely impair vision and may lead to irreversible blindness if left untreated. Fortunately, many of these conditions, such as age-related macular degeneration (AMD) and diabetic retinopathy (DR), can be detected early using color fundus photographs (CFPs). However, manual interpretation of CFPs is time-consuming, and specialist availability is often limited. While existing discriminative deep learning (DL) models are accurate in ocular disease detection from CFPs, they operate as black boxes incapable of explaining their decisions and are usually limited to a single condition. Moreover, they often demonstrate performance disparities across potential bias attributes such as sex or camera. In combination, these models fail to build trust towards their use in clinical practice. Generative classifiers, a class of generative DL models repurposed for classification tasks, offer a promising alternative. These models are self-explanatory by design, as their generative capabilities enable them to produce counterfactual explanations that provide visual insights into their predictions. However, the use of generative classifiers in ophthalmology remains largely unexplored. Therefore, the overarching goal of this thesis is to develop and evaluate a self-explanatory generative classifier that can reliably detect multiple ocular conditions. This goal is achieved via two aims. First, a diffusion model-based generative classifier (RetCond) is proposed for multi-class classification of ocular conditions. RetCond achieved an accuracy of 96.99%, which is comparable to discriminative baseline models while providing transparent explanations via counterfactual images. Second, the impact of selected bias attributes (sex and camera) is investigated to identify potential subgroup performance disparities. Subsequently, the effectiveness of bias mitigation techniques on RetCond is explored, and counterfactual explanations are used to expose spurious correlations between image features and condition predictions. The results demonstrate that RetCond exhibits reduced aggregate subgroup performance disparity compared to a discriminative DL baseline. Although bias mitigation techniques can further reduce RetCond’s disparities, counterfactual analyses suggest that they may also reinforce spurious correlations. By enabling reliable diagnoses and meaningful explanations, this work aims to advance the development of trustworthy deep learning models for the early detection of ocular conditions
