Department of Computer Science

Permanent link for this collection

https://hdl.handle.net/20.500.14078/7

Browse

Now showing 1 - 20 of 79

Virtual reality training simulators as a learning method for medical equipment: The IV infusion pump
(2025) Aizon, Jehdidae; Neumeier, Melanie; Qorbani, Sam
The integration of virtual reality (VR) simulators in medical education shows promise, particularly for training with equipment like the IV infusion pump. This study explores VR as an innovative tool to enhance learning, engagement, and information retention. Research supports VR’s effectiveness in visualizing complex concepts and reinforcing higher-order learning. To evaluate its impact, we tested a VR training simulator with five nursing student volunteers. Participants found the experience engaging after adjusting to the controls, though some reported discomfort with the Meta Quest 3 headset and blurry visuals due to lens distance. While the study was limited to a student capstone timeframe and in scale, findings suggest VR simulations have strong potential as medical training tools.
Developing virtual simulation of optic lab in physics education using virtual reality
(2025) Kajtazovic, Haris; Bakridi, Nadir; Qorbani, Sam
As virtual reality (VR) technology improves, the impact VR has on other areas, such as education, also improves. Educational VR applications provide support to students with an immersive environment in which such an environment may not be available. This paper as a student project focuses on the development of an educational VR application focused on the topic of optics. Ideas like collaboration between computer science students and physics educators, accessibility, and learning theories are explored. A pilot study with six participants was conducted, offering insight into the effectiveness of the educational VR application. The results gathered from the pilot study, limitations faced, and future research directions discussed offer a basis for researchers interested in educational VR work.
The motivational appeal of persuasive strategies in a healthy eating behaviour change game
(2025) Ndulue, Chinenye; Oyebode, Oladapo; Orji , Rita
Persuasive game designers employ persuasive strategies to improve the effectiveness of behaviour change games. Since persuasive strategies are intended to motivate the players toward the desired behaviours, the motivational appeal of these persuasive strategies can play an important role in the effectiveness of these behaviour change games. Therefore, it is important to understand the effectiveness of persuasive strategies and their motivational appeal. To advance research in this direction, this paper explores the relationship between the effectiveness of four popular persuasive strategies (reward, competition, praise, suggestion) and their motivational appeal in a persuasive game for healthy eating. In a study of 124 participants, our results showed that all the persuasive strategies were perceived to be effective in promoting behaviour change. We also discovered that the reward, competition and suggestion strategies showed a completely consistent relationship with all the motivational appeal dimensions. We also observed the strongest motivational appeal dimension for rewards was attention, competition and praise predominantly impacted satisfaction, while relevance stood out as the most significant motivational appeal dimension for suggestions. We conclude by offering some insights on how to implement persuasive strategies that amplify the four motivational appeal dimensions, in order to design games with better persuasive appeal.
Wideband corrugated horn design based on machine learning technique
(2025) Ibrahim, Aminah; Mohamed, Saphia; Gadelrab, Mahmoud; Elsaadany, Mahmoud; Shams, Shoukry I.
Corrugated horn antennas are essential in satellite and space communication systems due to their wide bandwidth, low cross-polarization, low side lobes, and excellent return loss. In this paper, several machine learning algorithms are used and trained on CST Microwave Studio data to predict antenna design parameters. The result values achieve a wideband response below 10 dB and a gain error within ±2dBi. These methods offer an efficient initial point for antenna design and reduce development time.
Fast power allocation technique for downlink transmission in cell-free massive MIMO systems
(2025) Gamal, Fatma; AbdelRaheem, Mohamed; Abdellatif, Mohammad M.; Elsaadany, Mahmoud; Aly, Omar A. M.
In this paper, we study power allocation techniques for the downlink in cell-free (CF) massive multiple-input multiple-output (mMIMO) systems. In such systems, choosing the optimal power allocation plays an important role in maximizing the achievable data rate and guaranteeing fairness among the users. However, this optimization problem is typically non-convex with an iterative computationally-expensive solution. This paper presents two new formalizations of the power allocation problem using duality. The dual problems have smaller solution spaces that would yield a fast, less-complex power allocation without any performance compromise. Simulation results show the solution to the dual problems produces the same per-user achievable capacity of the upper bound obtained using the classical bi-section method.
Pilot assignment in cell-free massive MIMO: a low-complexity clusterization technique
(2024) Gamal, Fatma; Elsaadany, Mahmoud; AbdelRaheem, Mohamed; Aly, Omar A.M.
In this paper, we present a novel pilot assignment algorithm leveraging a low-complexity clustering approach for cell-free massive MIMO networks. The proposed algorithm utilizes the geographical positions of users to refine user clustering and pilot allocation. A low-complexity iterative migration of users across neighbouring clusters is employed to enhance the average data rate and achieve an optimal user count per cluster, in alignment with the finite orthogonal pilot signals used. Comparative simulation results show the superior performance of our proposed method in data rate enhancement while maintaining a reduced complexity relative to existing pilot assignment techniques in the literature.
Crossguide Coupler Design using Deep-Learning model
(2025) Issa, Nadine; Mohamed, Sahra; Gadelrab, Mahmoud; Elsaadany, Mahmoud; Shams, Shoukry I.
Crossguide couplers are essential components in high power applications, where a sample is collected from the forward and reverse path to ensure operation. The design of the coupling section was intensively investigated but no accurate model exists. Accordingly, most of the literature models are used as starting points followed by lengthy numerical optimization. Here, we introduced a deep learningbased design for the first time. The proposed design is used to generate several design, where the recorded coupling value error is below 2 % for the validation cases. The generated designs satisfied a coupling flatness within ±1.5dB and the directivity beyond 15 dB.
Design of corrugated horn antenna for LEO satellite feeding system: efficient hybrid optimization approach
(2025) Gadelrab, Mahmoud; Shams, Shoukry I.; Elsaadany, Mahmoud; Gagnon, Ghyslain
Low Earth Orbit (LEO) satellites are receiving fresh interest from the wireless communications society. The transceivers on these satellites require high-gain antennas with consistent performance over the operational spectrum. Corrugated horn antennas are the most typical feed used in such systems. Furthermore, several satellite systems use dual-polarized antennas to provide polarisation variety. However, simulating the full satellite feed transceiver is a time-consuming task. This paper presents a revolutionary hybrid technique that combines real-measured data with simulation tools to reduce the optimization time for the entire system while providing more accurate and robust results. This optimized design is fabricated using a CNC lathe machine with ±0.002 inches accuracy. The fabricated prototype is measured in an anechoic chamber, where the measured results are in good agreement with the simulated ones.
Ridge gap waveguide low pass filters: a systematic design approach
(2024) Gadelrab, Mahmoud; Shams, Shoukry I.; Elsaadany, Mahmoud; Sebak, Abdelrazik
In satellite communication systems, Low Pass Filters (LPFs) are used to remove harmonics generated from the power source, minimize interference, and enhance the signal-to-noise ratio. Through the filter design process, various aspects must be considered such as the insertion loss, bandwidth, weight, and size. A stepped Impedance filter is a typical topology to realize the LPFs. The stepped impedance filter design process goes through multiple design steps starting from the normalized prototype design and ending with the realization of the filter sections. The filter realization significantly depends on the host guiding structure. This paper, presents, for the first time, a systematic design approach for the stepped impedance filter based on ridge gap waveguide technology. An accurate mathematical model for calculating a virtual cutoff for the ridge gap waveguide is introduced, which is deployed in the proposed design methodology. Moreover, a prototype of the stepped impedance filter is fabricated and measured, with measured results closely aligning with simulations.
Charging optimization in multi-app wireless sensor networks through reinforcement learning
(2025) Hamacher, Neal; Lawrence, Benjamin; Elmorsy, Mohammed
Wireless sensor networks are becoming increasingly prevalent in modern systems. These networks can be outfitted with a mobile charger that travels the network and replenishes the energy of the nodes within. This paper introduces a novel resource management problem for controlling mobile chargers in rechargeable wireless sensor networks shared among multiple applications. A reinforcement learning approach is developed to optimize the charger's actions, increasing the network's lifetime while ensuring that each application's throughput and coverage requirements are met to the best of the charger's ability. The resultant algorithm optimizes mobile charger network traversal and energy usage to maximize the network's lifespan while meeting application Quality of Service (QoS) requirements. It can also adjust the mobile charger behaviour when some applications are assigned higher priority than others, ensuring critical network operations are maintained more effectively. Numerical results show that the proposed approach ensures minimum QoS requirements are met through network node energy level maintenance and prolonged network up-time.
Reinforcement learning for self driving racing car games
(2025) Beaunoyer, Adam; Beaunoyer, Cory; Elmorsy, Mohammed; Saleh, Hanan
This research aims to create a reinforcement learning agent capable of racing in challenging simulated environments with a low collision count. We present a reinforcement learning agent that can navigate challenging tracks using both a Deep Q-Network (DQN) and a Soft Actor-Critic (SAC) method. A challenging track includes curves, jumps, and varying road widths throughout. Using open-source code on Github, the environment used in this research is based on the 1995 racing game WipeOut. The proposed reinforcement learning agent can navigate challenging tracks rapidly while maintaining low racing completion time and collision count. The results show that the SAC model outperforms the DQN model by a large margin. We also propose an alternative multiple-car model that can navigate the track without colliding with other vehicles on the track. The SAC model is the basis for the multiple-car model where it can complete the laps quicker than the single-car model but has a higher collision rate with the track wall.
Leveraging machine learning to predict factors that drive successful basketball team formation
(2025) El-Hajj, Mohamad; Kwon, Benjamin; Jethro Infante, Craeg; Steed, Jackson; Gore, Victor; Phan, Nhi; Elmorsy, Mohammed; Pang, Xiaodan
This study delves deep into the key factors affecting the likelihood of NCAA basketball players getting drafted into the NBA. The study highlights the importance of offensive metrics such as points scored and offensive ratings in predicting an NCAA player’s chances of being drafted into the NBA by utilizing an unsupervised learning clustering model and a supervised decision tree model. This underscores the significance of offensive statistics in a player’s skill set and suggests that players and coaches should prioritize improving these metrics to enhance a player’s draft potential. The study found that defensive metrics like defensive ratings and blocks have less impact on overall draft potential than offensive metrics. A crucial point to note is that a team’s success often relies on having its top players actively participating on the court. This research enhances our understanding of the factors influencing the draft prospects of NCAA basketball players. It underscores the advancement of basketball analytics and paves the way for further research on player performance metrics and their influence on the scouting and selection of professional athletes.
Analyzing factors that lead to NBA regular season success
(2024) El-Hajj, Mohamad; Steed, Jackson; Gore, Victor; Infante, Craeg; Flores, Raniel; Wakista, Danindu; Elmorsy, Mohammed
The National Basketball Association (NBA) values regular-season success and acknowledges the crucial role of a team’s roster composition in determining overall performance. This study uses machine learning techniques, specifically unsupervised learning clustering and decision tree models, to predict the composition of a winning roster. Our research identified three distinct clusters based on win percentage and the distribution of players across different skill levels. Successful teams typically have more top-tier players and a significant representation of players in the lowest skill level. In contrast, teams that spread their talent across the entire roster are less successful. We have noticed that players with average to above-average skills are notably affected by excessive playing time in the previous game, which leads to decreased performance and potential losses for the team in the next game. Considering the time of year and the gap between games, we recommend prioritizing the rest and recovery of top players, especially in the latter half of the season. It’s crucial to ensure that players who are not as skilled as the top players but still make significant contributions to the team maintain consistent performance, especially during the first half of the season. Analyzing height’s impact on basketball player performance has revealed practical insights that can empower coaches and management. We found that the shortest and tallest players often perform less than those of average height. Most top performers in the NBA tend to have heights closer to the average. However, for players who frequently operate near the net and encounter numerous rebound opportunities, it is generally preferable to have an average or taller player for slightly enhanced overall performance compared to below-average height players. Teams can use these insights to improve their roster construction and maximize player utilization by coaches from one game to the next. This research provides practical strategies that can be immediately implemented to enhance team performance.
Fitting and filtering functional data for use in video data analysis
(2024) Smith, Iain; El-Hajj, Mohamad
Our research focuses on advancing the capabilities of machine learning applications that involve analyzing video data. To achieve this, we have created a novel method for integrating functional data into video. Our approach entails the direct application of convolutional filters to functional data, as well as the introduction of new filters that make use of derivatives, which represent an exciting avenue for further exploration. In order to validate the effectiveness of our approach, we conducted experiments using both synthetic and real-world datasets. These experiments helped us establish our method’s potential in practical scenarios. We propose a specific parameter ratio for incorporating functional data into the original input frames. This parameter ratio has been shown to require less information while offering substantial potential for exploration within the realm of machine-learning applications for video data. Furthermore, we found that additional operations applicable to functions, such as derivatives, yield valuable information that can be harnessed to enhance machine learning applications involving video data. This opens up exciting possibilities for leveraging the richness of functional data in video analysis.
Enhancing patient care: machine learning’s role in reducing wait times for medical procedures
(2025) El-Hajj, Mohamad; Collins, Liam; Steed, Jackson; Heß, Claudia; Kunz, Sibylle
The healthcare system faces a critical challenge with extended wait times for medical procedures, significantly impacting both patients and healthcare professionals. While increasing funding and hiring more doctors may seem like effective solutions, these approaches are often impractical due to various constraints. This research examines the factors driving medical procedure wait times in Canada, specifically in British Columbia, Nova Scotia, and Quebec, highlighting the urgent need to address delays caused by resource limitations. By leveraging machine learning techniques—including random forest methods, k-means clustering, and linear regression—alongside statistical models such as bar graphs, correlation matrices, and z-score normalization, the study, conducted in both Python and R Studio, identifies key contributors to these delays. Based on the findings, a strategic approach to physician hiring is proposed, emphasizing the optimization of seniority levels. Specifically, the study recommends capping the hiring of entry-level doctors at 18% and senior-level doctors at 5%, while increasing the absolute population of entry-level physicians by 27% and reducing the physician-to-100,000 population ratio by 2%, which could lead to a 15% reduction in wait times. By addressing the complexities of medical procedure delays, this research aims to enhance the efficiency and fairness of surgical care delivery.
Understanding AI in cybersecurity and secure AI: challenges, strategies and trends
(2025) Sharma, Dilli Prasad; Lashkari, Arash Habibi; Daghmehchi Firoozjaei, Mahdi; Mahdavifar, Samaneh; Xiong, Pulei
This book presents an overview of the emerging topics in Artificial Intelligence (AI) and cybersecurity and addresses the latest AI models that could be potentially applied to a range of cybersecurity areas. Furthermore, it provides different techniques of how to make the AI algorithms secure from adversarial attacks. The book presents the cyber threat landscape and explains the various spectrums of AI and the applications and limitations of AI in cybersecurity. Moreover, it explores the applications and limitations of secure AI. The authors discuss the three categories of machine learning (ML) models and reviews cutting-edge recent Deep Learning (DL) models. Furthermore, the book provides a general AI framework in security as well as different modules of the framework; similarly, chapter four proposes a general framework for secure AI. It explains different aspects of network security including malware and attacks. The book also includes a comprehensive study of various scopes of application security; categorised into three groups of smartphone, web application, and desktop application and delves into the concepts of cloud security. The authors discuss state-of-the-art Internet of Things (IoT) security and describe various challenges of AI for cybersecurity, such as data diversity, model customising, explainability, and time complexity and includes some future work. They provide a comprehensive understanding of adversarial machine learning including the up-to-date adversarial attacks and defences. The book finishes off with a discussion of the challenges and future work in secure AI. Overall, this book covers applications of AI models to various fields of cybersecurity and appeals not only to an scholarly audience but also to professionals wanting to learn more about the new developments in these areas
Disentangling hippocampal shape variations: A study of neurological disorders using mesh variational autoencoder with contrastive learning
(2025) Rabbi, Jakaria; Kiechle, Johannes; Beaulieu, Christian; Ray, Nilanjan; Cobzas, Dana
This paper presents a comprehensive study focused on disentangling hippocampal shape variations from diffusion tensor imaging (DTI) datasets within the context of neurological disorders. Leveraging a Mesh Variational Autoencoder (VAE) enhanced with Supervised Contrastive Learning, our approach aims to improve interpretability by disentangling two distinct latent variables corresponding to age and the presence of diseases. In our ablation study, we investigate a range of VAE architectures and contrastive loss functions, showcasing the enhanced disentanglement capabilities of our approach. This evaluation uses synthetic 3D torus mesh data and real 3D hippocampal mesh datasets derived from the DTI hippocampal dataset. Our supervised disentanglement model outperforms several state-of-the-art (SOTA) methods like attribute and guided VAEs in terms of disentanglement scores. Our model distinguishes between age groups and disease status in patients with Multiple Sclerosis (MS) using the hippocampus data. Our Mesh VAE with Supervised Contrastive Learning shows the volume changes of the hippocampus of MS populations at different ages, and the result is consistent with the current neuroimaging literature. This research provides valuable insights into the relationship between neurological disorder and hippocampal shape changes in different age groups of MS populations using a Mesh VAE with Supervised Contrastive loss.
Automatic deep learning segmentation of the hippocampus on high resolution diffusion MRI and its application to the healthy lifespan
(2024) Efird, Cory; Neumann, Samuel; Solar, Kevin; Beaulieu, Christian; Cobzas, Dana; Miller, Dylan
Diffusion tensor imaging (DTI) can provide unique contrast and insight into microstructural changes with age or disease of the hippocampus, although it is difficult to measure the hippocampus because of its comparatively small size, location, and shape. This has been markedly improved by the advent of a clinically feasible 1-mm isotropic resolution 6-min DTI protocol at 3 T of the hippocampus with limited brain coverage of 20 axial-oblique slices aligned along its long axis. However, manual segmentation is too laborious for large population studies, and it cannot be automatically segmented directly on the diffusion images using traditional T1 or T2 image-based methods because of the limited brain coverage and different contrast. An automatic method is proposed here that segments the hippocampus directly on high-resolution diffusion images based on an extension of well-known deep learning architectures like UNet and UNet++ by including additional dense residual connections. The method was trained on 100 healthy participants with previously performed manual segmentation on the 1-mm DTI, then evaluated on typical healthy participants (n = 53), yielding an excellent voxel overlap with a Dice score of ~ 0.90 with manual segmentation; notably, this was comparable with the inter-rater reliability of manually delineating the hippocampus on diffusion magnetic resonance imaging (MRI) (Dice score of 0.86). This method also generalized to a different DTI protocol with 36% fewer acquisitions. It was further validated by showing similar age trajectories of volumes, fractional anisotropy, and mean diffusivity from manual segmentations in one cohort (n = 153, age 5–74 years) with automatic segmentations from a second cohort without manual segmentations (n = 354, age 5–90 years). Automated high-resolution diffusion MRI segmentation of the hippocampus will facilitate large cohort analyses and, in future research, needs to be evaluated on patient groups.
Explaining anatomical shape variability: supervised disentangling with a variational graph autoencoder
(2023) Kiechle, Johannes; Miller, Dylan; Slessor, Jordan; Pietrosanu, Matthew; Kong, Linglong; Beaulieu, Christian; Cobzas, Dana
This work proposes a modular geometric deep learning framework that isolates shape variability associated with a given scalar factor (e.g., age) within a population (e.g., healthy individuals). Our approach leverages a novel graph convolution operator in a variational autoencoder to process 3D mesh data and learn a meaningful, low-dimensional shape descriptor. A supervised disentanglement strategy aligns a single component of this descriptor with the factor of interest during training. On a toy synthetic dataset and a high-resolution diffusion tensor imaging (DTI) dataset, the proposed model is better able to disentangle the learned latent space with a simulated factor and patient age, respectively, relative to other state-of-the-art methods. The relationship between age and shape estimated in the DTI analysis is consistent with existing neuroimaging literature.
Scrum package for Software Engineering education
(2025) Pang, Candy
Nowadays, software is embedded in almost all devices (e.g., car, refrigerator, kettle). Creating high-quality software is extremely important now and in the future. The process of creating high-quality software is called Software Engineering (SE). As far as I know, every undergraduate Computer Science (CS) curriculum includes SE education. Agile is the most used SE methodology in the IT industry, and Scrum is the most popular Agile framework. Therefore, most SE courses in the CS undergraduate curriculum teach Scrum through software-developing course projects to prepare students for their careers. One challenge in teaching Scrum is its lack of a definitive definition. In the IT industry, organizations customize the Scrum framework according to their needs without restriction. Therefore, teachers must have in-depth Scrum experience to teach students how to employ Scrum when dealing with various software development challenges. Through these challenges, teachers guide students to understand the pros and cons of the Scrum framework and learn how to apply Scrum in different organizations under different situations. Teachers having in-depth Scrum experience are essential in undergraduate SE education. However, many teachers, including instructors and teaching assistants (TAs), lack industrial Scrum experience. As a technical architect consultant in the IT industry for over 15 years, who participated in over 20 projects in more than 10 organizations, mainly using Scrum, I created a Scrum teaching package for my single-term Introduction to Software Engineering (Intro-SE) course. The package includes Scrum process templates, template instructions, common mistake descriptions, evaluation schemas, and feedback suggestions. With the package, teachers can teach Scrum professionally without industrial experience. I used the teaching package in Intro-SE for six terms through three years and enhanced the package to improve students' learning experience. I describe the package's content, usage, and benefits in this paper. A Scrum project starts with defining a product backlog. Then, the project is split into sprints of a few weeks each. For the project, the package included a product backlog template. For the sprints, the package consists of sprint backlog template, sprint planning template, sprint task board template, sprint burndown chart template, sprint tracking template, sprint retrospective template, and sprint demo template. The package also includes detailed template evaluation schemas with feedback suggestions that enforce positive behaviors and recommend improvements. With the package, the teachers may mark students' submissions according to the industrial practices with meaningful feedback. This paper also points out significant shortcomings and downfalls of the students so that the teachers may be more aware of the learning opportunities, guide the students to learn from their mistakes, and apply remediations.

Browse

Recent Submissions