Project Descriptions – Winter 2024

**This Page is Updated Regularly – Please check back often**

How To Select a Project: First, have a look at the projects listed here. Contact the supervisors for more information. When you have found a project, and the supervisor is willing/able to give you the project, you MUST complete a contract and both of you have to sign it. This space will have detailed information on how to submit your contract and proposals in the Fall term. ONLY when the contract is received is the project officially yours. You cannot register in the course prior to this. If you have conflicts or not meeting the eligibility criteria, please contact the undergraduate advisor or the course coordinator as soon as possible. You need to complete your contract in the Fall term.

Each project entry lists the maximum number of students who can participate in the project, and the number of students whom the supervisor has already agreed to supervise (conditional on receipt of the contract). If they are equal, no more slots remain for that project, and you should pursue a different one.
I strongly recommend that you try to find a project from the list below. However, if you can’t find a project here that fits your interests and abilities, you can propose a project yourself. In this case you will need to find a School of Computing faculty member who is willing to supervise your project. As a rough guideline to project design, the amount of work involved should be the same as a typical 400-level CISC course.

Note that CISC 495 is an alternative capstone course you can register in. For more information, please see the course website and the FAQ on capstones.

SupervisorProject TitleAvailability
Amber Simpson Risk Stratification for Managing Post-Partum Complications 4 / 4 spots taken
Burton Ma Illustrating data structures and algorithms using Penrose 3 / 4 spots taken
Catherine Stinson Evaluating AI Image Generation Models: Can they make art? 5 / 5 spots taken
Christian Muise Automated Agriculture Frontend 3 / 3 spots taken
Research Ready Smart Watch 2 / 2 spots taken
Campsite Scan 2 / 2 spots taken
Computational Art Installation 1 / 3 spots taken
Automated Agriculture Software Stack 3 / 3 spots taken
Erin Meger Probabilistic Simulations on Deterministic Social Network Models 0 / 2 spots taken
Teaching and Learning in STEM – Accessible Assessments 0 / 3 spots taken
Information Diffusion and Diversity in On-Line Social Networks 0 / 2 spots taken
Farhana Zulkernine Generating a Multi-Sensor Dataset for Human Activity Recognition 0 / 1 spots taken
Evaluating Open-Source Large Language Model for Medical Summarization Tasks 0 / 1 spots taken
Furkan Alaca and Juergen Dingel Analysis of security protocols with Alloy 1 / 2 spots taken
Hazem Abbas Predicting Neutropenia Diagnosis In Pediatric Patients Utilizing Artificial Intelligence Tools 0 / 2 spots taken
Predicting Hemophagocytic lymphohistiocytosis diagnosis in paediatric patients and building an individualized treatment algorithm 0 / 2 spots taken
Hesham ElSawy Explanatory AI (XAI) for Improving Deep Learning Performance 4 / 4 spots taken
Wireless Federated Visual Intelligence at the Network Edge 3 / 3 spots taken
James Stewart Fourier Transform Exploration Tool 0 / 1 spots taken
Video Enhancement for Lightboard Lectures 0 / 1 spots taken
Elevator Arrival Prediction 1 / 1 spots taken
Animations of Algorithms and Data Structures 0 / 1 spots taken
Race Car Telemetry Visualization 0 / 1 spots taken
Jana Dunfield Automatic bug-finding for TeX proofs 0 / 2 spots taken
Juergen Dingel Declarative Prompting 2 / 3 spots taken
Property-based Testing 2 / 3 spots taken
Kai Salomaa Converting finite-state machines to regular expressions 2 / 3 spots taken
Average-case complexity of the subset construction 0 / 3 spots taken
Kathrin Tyryshkin Genomic data preprocessing 1 / 1 spots taken
Transforming Bladder Cancer Diagnosis through Digital Pathology and Computing 1 / 1 spots taken
Randy E Ellis Discovering Fungal Metabolites from Mass Spectrometry Data 0 / 1 spots taken
Classifying Fungal Colonies from Mass Spectrometry Data 0 / 1 spots taken
Salimur Choudhury Deep Learning-Based Infrastructure Inspection with Drone Imagery: Focusing on Urban Elements 4 / 4 spots taken
Sara Nabil Exploring 3D-printing for Prototyping Wearable Technology 2 / 2 spots taken
Textile Interfaces and User Experience (UX) of Tactile Interaction 4 / 4 spots taken
Sara Nabil and Farhana Zulkernine Interactive Wearables for Building Biodata Predictive Models 0 / 3 spots taken
Selim Akl Unconventional Computation Problems 3 / 3 spots taken
Steven Ding Intrusion Detection System (IDS) for CAN Bus in Automobiles 0 / 3 spots taken
AI-Based Attack on Drone Swarm Algorithms 0 / 2 spots taken
AI for Swarm Counter-Swarms 0 / 3 spots taken
ERS0 - Firmware Software Bill of Material (SBOM) Analysis Platform Development 3 / 5 spots taken
Menu Item Recognition from Images 3 / 3 spots taken
AI for Video-Based Counter Uncrewed Aerial Systems Concept Development 2 / 3 spots taken
AI-Based Communication Protocol Reverse Engineering 0 / 3 spots taken
Adversarial Attacks Against Generative AIs such as ChatGPT 3 / 3 spots taken
Firmware Catalogue Development 2 / 3 spots taken
Vulnerability Detection in Firmware Images 0 / 3 spots taken
Timothy Wowk (Client) Using Monte Carlo simulation augmented by AI to Help Predict Revenues in the Charitable Sector 0 / 1 spots taken
Wendy Powley City of Kingston Housing Match 4 / 4 spots taken
Academic Integrity Form Filling Tool 2 / 2 spots taken
Yuan Tian Deep Learning-based Simulation of Texas Hold ‘em. 2 / 2 spots taken
Conversational Recommender System Leveraging Large Language Models 4 / 4 spots taken
Yuanzhu Chen IMDb Knowledge Graph 3 / 3 spots taken
Course-Curriculum Map 3 / 3 spots taken
Through the lens of LinkedIn 3 / 3 spots taken
Where do Queen’s and QSC stand? 3 / 3 spots taken

Details

  • Genomic data preprocessing
    Supervisor: Kathrin Tyryshkin
    Description:
    Analysis of a genomic data often involves pre-processing, quality control, normalization, feature selection and classification and differential expression analysis. Many methods exist, however, the best technique depends on the dataset. Therefore, it is often required to try different techniques to select the one that works best for a given dataset.
    This project involves incorporating the preprocessing pipeline that was developed in our lab into the MATLAB Biopipeline Designer. Through this project students will learn the methods used in gene expression analysis and apply their computational technique to genomic and cancer research.
    To apply for this project please send your recent transcript and a brief note on how you think you fit to this project.
  • Automated Agriculture Software Stack
    Supervisor: Christian Muise
    Description:
    The MuLab and Machine Intelligence & Biocomputing (MIB) Laboratory are building a lab-scale platform for the exploration of autonomous agriculture. This will include sensors such as video feeds, nutrient detection, etc., and actuators such as lighting, watering, etc. This project will involve working with the MIB lab and MuLab to significantly expand the initial software for the first iteration of the system being built in 2023/24 academic year. It will be primarily based on the Home Assistant platform.

    Vertical Farm
  • Automated Agriculture Frontend
    Supervisor: Christian Muise
    Description:
    The MuLab and Machine Intelligence & Biocomputing (MIB) Laboratory are building a lab-scale platform for the exploration of autonomous agriculture. This will include sensors such as video feeds, nutrient detection, etc., and actuators such as lighting, watering, etc. This project will involve working with the MIB lab and MuLab to build a frontend to both analyze the health and status of the plants, and control the system as a whole. The development will be focused on a web implementation that takes advantage of touch and dial-based inputs of the Microsoft Surface Studio.

    Vertical Farm
  • Research Ready Smart Watch
    Supervisor: Christian Muise
    Description:
    To conduct ongoing research in both continuous authentication and smart office analytics, this project aims to build a custom Android Wear application that will let us make use of all the sensors in modern smartwatches. Students located in Kingston will optionally have access to modern wearables in order to test the developed application, and the resulting prototype will contribute directly to research in both the MuLab and CSRL research labs.

    smart watch
  • Campsite Scan
    Supervisor: Christian Muise
    Description:
    This research project aims to reduce the number of hazardous sites left, by requiring campers to
    complete an automated checkout-safety scan. Following the leave-no-trace policy, the proposed
    verification system will leverage computer vision techniques to ensure that fires are completely
    put-out, and garbage is not left behind before campers are allowed to leave. The project will investigate
    the use of pre-trained YOLO models for object detection to identify key items such as the fire pit,
    wrappers, cans, and other belongings that may have been left behind. The system will also
    further employ a model to detect if the identified firepit poses a risk (fire hazard vs fully
    extinguished).

    campsite
  • Computational Art Installation
    Supervisor: Christian Muise
    Description:
    This project will involve working with some number of members from the MuLab to visualize, artistically, some of the ongoing research. The final artefact will be an installation in Goodwin 627 (the MuLab), and this open-ended project is ideal for students who are in COCA or are generally interested in the field of computational art.

    Robot Art
  • Evaluating AI Image Generation Models: Can they make art?
    Supervisor: Catherine Stinson
    Description:
    Tools like DALL-E, Midjourney and Stable Diffusion can create images based on text prompts. While the outputs seem impressive, how can we evaluate whether they produced the right image? How can we evaluate whether they are able to produce images with artistic merit? These models are trained using unsupervised adversarial methods, with no role for human judgment of these questions.
    A previous 499 project took initial steps toward developing evaluation metrics for AI image generators. This project will extend that work by expanding to new datasets, and validating the metrics using other kinds of image+text pairs. This project will also develop new methods for evaluating the success of AI image generation tools.
    Experience with art history or artistic practice would be beneficial.
  • Conversational Recommender System Leveraging Large Language Models
    Supervisor: Yuan Tian
    Description:
    Note: the recommender target may be updated to another domain rather than Programming, initially, we want to recommend learning resources for novice programmers.

    Novice programmers are different from experienced programmers. When they're starting out, they often face a variety of unique needs and questions as they navigate the world of coding—for instance, choosing the right programming language and tools for their tasks, setting up a development environment, debugging, and understanding basic programming concepts. Therefore, a recommender system that can recommend learning resources for novices would be greatly helpful.

    With the recent advanced in large language models, including chatgpt, gpt-4, and llama, this project aims to create a web application that implements a novel recommendation system based on conversations between novice developers and a chatbot. Such web applications would also help Queen's undergraduate students learn computing more effectively.

    This project prefers someone who has a machine learning background or web application development background.
    If you are interested, please send me an email and attach your unofficial transcript as a reference.
  • Unconventional Computation Problems
    Supervisor: Selim Akl
    Description:
    This project is in the area of theoretical computer science. The objective is to design and analyze computational problems that disprove the principle of universality in computing. There is no programming involved in this project.

    The task is akin to designing and analyzing an algorithm for solving a computational problem. The difference here is that, in this case, the problem is not given. The purpose is for you to design a problem to be solved. You are then to show that the problem is eminently solvable, but not solvable by a universal computer, that is, a computer that is finite and fixed once and for all.

    A student taking on this project is required to submit three original unconventional computation problems, along with a proof that each of these problems demonstrates non-universality in computation.

    References
    • S.G. Akl, Unconventional computational problems, in: Encyclopedia of Complexity and Systems Science, Springer, New York, 2018, pp. 631 - 639.
    • S.G. Akl, Evolving computational systems, Chapter One in: Handbook of Parallel Computing: Models, Algorithms, and Applications, Rajasekaran, S. and Reif, J.H., Eds., Taylor & Francis, CRC Press, Boca Raton, Florida, 2008, pp. 1 - 22.
    • S.G. Akl, Computational nonuniversality: Philosophical and artistic perspectives, in: Thoughts on Unconventional Computing, Adamatzky, A. and Lestocart, L.-J., Eds., Luniver Press, Beckington, U.K., 2021, pp. 10 - 17. Also in: Unconventional Computing, Arts, Philosophy, World Scientific Publishers, London, 2022, pp. 41 – 55.
  • Wireless Federated Visual Intelligence at the Network Edge
    Supervisor: Hesham ElSawy
    Description:
    Federated learning (often referred to as collaborative learning) is a decentralized approach to training machine learning models without the need to exchange data from clients, and hence, preserve the data privacy. With the ubiquity of wireless connectivity, it is important operate federated learning over wireless networks. In this context, this project integrates federated learning and wireless communications to create a distributed wireless visual intelligence system. The primary focus of the project will be to leverage the power of federated learning to collaboratively train machine learning models using data from multiple cameras while utilizing wireless communication to transmit model updates. This project aims to address the challenges of privacy, scalability, and wireless communication impairments in the context of real-time distributed visual data analysis.

    Objectives:
    1. Use Python machine learning packages (e.g., TensorFlow) to implement real-world systems. In particular, the students will design and implement a federated learning infrastructure that allows multiple wireless cameras (e.g., mobile phone camaras or computer webcams) to collaboratively train a common (i.e., global) machine learning model while keeping data decentralized and preserving individual camera privacy.
    2. Understand and account for wireless communications impairments. In particular, the students will learn how to establish wireless communication channels to exchange of model parameters and updates among the cameras while considering factors like latency, reliability, and security.
    3. Optimize the global model. The students will learn how to tune the system parameters to ensure efficient aggregation algorithms that can effectively merge local model updates from different cameras into a global model, ensuring convergence and accuracy.
    4. Real-time visual analysis. The students will demonstrate the application of the federated learning-enabled wireless camera network for real-time visual analysis tasks such as object detection, activity recognition, and anomaly detection.

  • Where do Queen’s and QSC stand?
    Supervisor: Yuanzhu Chen
    Description:
    In this project, you will look into the QS World University Rankings and make sense of where Queen’s University, QSC and possibly other disciplines stand world-wide, how our ranks evolved over the years, and where the top universities and Computing departments are located.
  • Through the lens of LinkedIn
    Supervisor: Yuanzhu Chen
    Description:
    In this project, you will explore data gathering of where Queen's Computing alumni are by crawling LinkedIn user files, make sense of their career path, and potentially recommend changes to the School's undergraduate education.
  • Illustrating data structures and algorithms using Penrose
    Supervisor: Burton Ma
    Description:
    Penrose is a software framework for generating visual diagrams of mathematical statements. Penrose uses programming languages to describe mathematical objects and their visual representation. Broadly speaking, the goal of this project is to use Penrose to illustrate concepts in data structures and/or algorithms by developing what Penrose refers to as Domain and Style packages so that end-users can easily generate Substance.

    Interested students should do the following before contacting the supervisor:

    1. read the academic paper that describes Penrose
    2. visit the Penrose website
    3. try running some Penrose examples

    Suggested skills and interests:

    - programming
    - graphic design
    - data structures
    - algorithms

    Suitable for up to four groups of one or two students.
  • Discovering Fungal Metabolites from Mass Spectrometry Data
    Supervisor: Randy E Ellis
    Description:
    I have agreed to supervise Zoey Drassinower in this project
  • Deep Learning-Based Infrastructure Inspection with Drone Imagery: Focusing on Urban Elements
    Supervisor: Salimur Choudhury
    Description:
    Inspecting infrastructure elements using drone imagery is a precious and widely applied use case. Drones equipped with high-resolution cameras and specialized sensors can inspect various types of infrastructure in urban areas, offering numerous benefits, including cost-efficiency, safety, and access to hard-to-reach areas. The idea is to utilize drone imagery to inspect one of the selected infrastructure elements by analyzing drone imagery using deep learning methods. The infrastructure elements that can be inspected include (but are not limited to) bridges, roads, utility infrastructure, buildings, etc. The students can focus on one of the infrastructure elements to limit the project's scope.
  • Academic Integrity Form Filling Tool
    Supervisor: Wendy Powley
    Description:
    Academic Integrity cases require instructors to gather information regarding students involved and input this information into a standard word document provided by the Faculty of Arts and Science. In this project, you will design an automated workflow process to reduce the time required for instructors to complete the paperwork required for academic integrity cases. This work will involve reading data from spreadsheets, and working with word documents and/or pdfs.
  • Property-based Testing
    Supervisor: Juergen Dingel
    Description:
    Property-based testing (PBT) generalizes traditional, 'example-based' testing through the specification of properties that the output has to satisfy under, perhaps, the assumption that the inputs satisfy some other properties. PBT tools allow for the automatic generation of inputs that satisfy the input properties and then check that the output produced indeed has the desired output properties. Moreover, once an input that results in incorrect output has been found, the tool will automatically try to 'shrink' the input, i.e., look for the smallest input that makes the software fail. PBT thus not only allows for the specification of more general and expressive tests that avoid unnecessary or even misleading specialization, but also features a smart test case generation. More detailed descriptions of PBT can be found online (e.g., here). PBT can be used in any language, and PBT tools exist for, e.g, Pyton, Haskell, Java, JavaScript, C++, Kotlin, and Go.

    The goal of the project is to explore and illustrate the use of PBT in a context of a case study such as a suitable implementation for decision tree construction or optimization. The project is most suitable for students with interests in testing and software development. Completion of CISC 422 will be beneficial, but is not a requirement.
  • Declarative Prompting
    Supervisor: Juergen Dingel
    Description:
    Specifications are instrumental to describe properties of artifacts in engineering in general, including software engineering. The use of generative AI tools such as ChatGPT to help create and evolve software artifacts is a topic of active research. The goal of the project is to explore the potential utility of different kinds of specifications for the use generative AI tools to facilitate software engineering tasks. Topics of interest include the use of declarative specifications in prompts, but also as a result to be returned by the AI tool. The project is most suitable for students with interests in formal specification (as in, e.g., CISC 223 and CISC 422), logic, and software engineering. Completion of CISC 422 will be beneficial, but is not a requirement.

    Xi Ye, Qiaochu Chen, Isil Dillig, Greg Durrett. Satisfiability-Aided Language Models Using Declarative Prompting. Arxiv. 2023. Access it here
  • Average-case complexity of the subset construction
    Supervisor: Kai Salomaa
    Description:
    It is well known that in the worst case the minimal
    deterministic finite automaton (DFA) equivalent to a given
    $n$ state nondeterministic finite automaton (NFA)
    needs $2^n$ states. On the other hand, the determinization
    algorithm based on the subset construction typically works
    well in practice - which seems to indicate that the
    exponential size blow-up is a relatively rare occurrence.

    The goal of this project is to study the {\em average case complexity
    of the NFA-to-DFA transformation\/} (strictly speaking,
    the NFA-to-minimized DFA
    transformation)
    by running the subset construction algorithm
    on a large number of randomly generated NFAs and minimizing
    the resulting DFAs. Determinization
    of NFAs and DFA minimization are included in software
    libraries such as {\em Fado, } or {\em Vaucanson}:

    \begin{itemize}
    \item Fado {\tt http://fado.dcc.fc.up.pt/}, or,
    \item Vaucanson {\tt http://vaucanson-project.org/?eng}.
    \end{itemize}
    The software libraries provide a collection of operations
    to convert an NFA to a DFA and for minimizing DFAs.

    In addition to running experiments on randomly generated
    NFAs the main goals of the project include:
    \begin{itemize}
    \item identifying a reasonable method to
    generate ``random'' NFAs,
    \item identifying different types of NFAs where the
    determinization cost is close to the worst case,
    \item searching in the literature for theoretical work
    dealing with average case complexity of
    the subset construction algorithm.
    \end{itemize}

    The project involves both a theoretical and an implementation
    component.
    This project is suitable for two or three students.
    (If a third student does not register, two students can complete
    the project.)

  • Converting finite-state machines to regular expressions
    Supervisor: Kai Salomaa
    Description:
    The well known state
    elimination algorithm converts a (nondeterministic) finite automaton
    (NFA) to a regular expression. This algorithm tends to generate very
    large and redundant regular expressions, partly because it
    does no simplification. In this project you will develop
    and implement heuristics for simplifying regular expressions and
    apply these heuristics to the problem of converting finite
    state machines to regular expressions.

    The tasks can include
    the following:
    \begin{itemize}
    \item study the expressions that tend to be generated in conversion
    of finite state machines, and characterize the types of
    simplification that would be useful
    \item develop and implement a simplification heuristics for
    regular expressions
    \item test the efficacy of your algorithms by using them in the conversion
    of finite state machines to regular expressions
    \item investigate some of the
    current literature on simplification of regular expressions
    \item (optional, if time permits) evaluate the cost of the
    simplification heuristics and develop a metric for deciding
    when to use them
    \end{itemize}

    The algorithm converting NFAs to regular expressions
    has been implemented in various libraries such as
    \begin{itemize}
    \item Fado {\tt http://fado.dcc.fc.up.pt/}, or,
    \item Vaucanson {\tt http://vaucanson-project.org/?eng}
    \end{itemize}
    The software libraries provide a collection of operations
    to manipulate
    finite-state machines and convert them to regular
    expressions or vice versa.

    The project requires a solid understanding of the basics of finite automata
    and regular expressions. Additionally a fair amount of programming is
    required and you should expect to run a significant number of simulations and
    other experiments.

    The project involves both a theoretical and an implementation
    component.
    This project is suitable for two or three students.
    (If a third student does not register, two students can complete
    the project.)
  • Using Monte Carlo simulation augmented by AI to Help Predict Revenues in the Charitable Sector
    Supervisor: Timothy Wowk (Client)
    Description:
    This project is based on work completed and published in 2021 where MC simulation was used to forecast charitable revenue. The work showed promise and would potentially be of interest to the broader philanthropic sector if the interface with client data could be simplified and the program (or app potentiall) had an integrated AI component that would improve its predictive ability - over time - based on new or refined data. I have attached a copy of the article in which the original work was published.

    Rolling the Dice on Revenue Forecasting: A Case Study

    One of the most significant challenges facing college and university advancement operations is revenue forecasting. The ability to accurately forecast future philanthropic revenues is vitally important to charities as it impacts on the organization’s ability to plan, to meet financial obligations and to identify any needed changes in business practice. The challenge is a systemic one that affects all charities regardless of size, structure or focus. At the major gifts level the forecasting challenge is particularly acute. Consider this. Major gifts programs typically account for 90 percent or more of philanthropic revenues at post-secondary institutions in Canada and the United States. The ability to articulate and defend a major gifts revenue forecast is an essential skill for fundraising executives. The standard forecasting approach – up till now – typically involves a line-by-line (or aggregate) analysis of the MG pipeline and may include (but often does not) the application of a “discount” factor based on the level of confidence in the proposal. There a several problems with this methodology – it is time consuming, susceptible to personal bias and forecasts are difficult to defend in an analytics-driven environment. But…there may be a better way.
    Placing a Bet…on Monte Carlo
    One solution to the intractable problem of forecasting fundraising revenue is simulation. Simply stated, simulation, or probability simulation, is a way to model random events so that the outcomes closely match real-world events. This technique has been used for decades by industry to measure risk and possible outcomes in such diverse fields as energy, manufacturing, mining and insurance. While there are many probability tools the one that is arguably best suited to major gifts forecasting is Monte Carlo simulation as it allows the fundraiser to factor in a range of values for variables like gift amount, days in stage etc.
    Monte Carlo simulation works by building a model of possible results —a probability distribution—for any factor (like a gift amount) that has inherent uncertainty. It then calculates the possible results over and over, each time using a different set of random values from the probability functions. The concept itself is a bit esoteric and a frequently-cited gambling analogy may help show the true value of this approach. A novice player walks up to a “craps” table for the first time and makes a practice roll of 8, consisting of two 4’s. Based on the practice throw the player may assume – incorrectly – that 8 is likely to turn up regularly in the game. To determine the true probability of “8” turning up the player would need throw the dice several thousand times. In doing this a probability distribution for rolling “8” would be produced (Fig1.). This same logic can be applied to forecasting the value of a major gifts pipeline. For example, the result of any major gift solicitation is a range of outcomes from 0 to the full amount of the “ask”. By simulating the “ask” thousands of times – and making a few fact-based assumptions on the likelihood of receiving a “yes” or “no” answer - we can come up with an evidence-based and reasonably accurate forecasting tool for any major gifts pipeline. Best of all – it’s not that hard to do.
    How It Works
    Basic revenue projection using Monte Carlo techniques can be accomplished with just a few basic operational statistics from the unit. Proposal terminology and processes differ from institution to institution but, in a nutshell, you will need the following to get started:
    • “Ready” and “Ask Made” (or similar) Proposal Data
    • Proposal “Ask Made” Date
    • Proposal Ask Amount
    • Proposal Response(s) (Accepted vs. Declined)
    • Proposal Ask Amount vs. Received Amount

    Having several fiscal years of data on hand is helpful as this will allow for more extensive back-testing of the model. With program/pipeline information in hand the next step is to build a model of the business activity – in this case fundraising revenue projection. There are a number of commercially available programs to help with this step (Queen’s used @Risk by Palisade) and all of them provide good online support and tutorial options. Regardless of the software selected the steps to building the forecasting simulation follow the same basic pattern:

    • Build a quantitative model of the business activity
    • Define the Input Parameters
    • Create Random Data (Calculation Model)
    • Simulate Output (Forecast)

    It is important to note that the process of developing and testing a forecasting model at Queen’s was iterative and characterized by many cycles of trial and error. While our current model appears to be reasonably accurate, it will take several more quarters of testing before we know for sure.



    This process is repeated as many times as necessary (usually 10,000 times) in order to consider enough scenarios to create a 90% confidence interval for Total Revenues, as it is shown in Figure 1. For instance, if a prospect has a good level of confidence (with a LC=1) the process simulates the event of a donation by generating a random value of “zero” or “one” (according to a 20% probability of having a “one”). If a donation is confirmed (OC=1), the process computes a random amount around the potential donation for the prospect +/- 1.5%. This deviation was obtained as a result of calibrating historical data of past donations.


    A Question of Confidence & Variability

    The proof, as they say, is in the pudding. With this in mind Queen’s University decided to back-test the accuracy of our preliminary Monte Carlo simulation against major gift receipts for the fiscal years 2015-19. The initial results were encouraging but lacked precision. For example, when we applied our simulation against known values in both the Ask Ready and Ask Made pipeline (Fig.1) the range between the low and high values >$100M over the three year fiscal period. While our real-world major gift revenues for the period fell within this range, the model had far too much variance to be a useful forecasting tool. A quick caveat before digging into the numbers. For confidentiality reasons the authors chose to replace actual major gift totals at our institution with re-scaled results. While the re-scaled numbers may look “off” they accurately reflect scaled major gift revenues and probabilities for the 2015-19 period. The forecast, presented in Figure 1, suggests that we had a 90 percent chance (or probability) of receiving between $ 17.450 and $ 19.894 in major gift revenues during the 2015-19 period. Actual major gift receipts for the period were $ 18.794 – roughly the mid-point of the re-scaled forecast value. This provides strong validation of the Monte Carlo technique as a forecasting tool.


    Fig. 1
    The ability to forecast aggregate major gift revenues over a multi-year period is important as it allows Advancement leadership to develop (and influence) strategic policy and planning for the unit. Of equal, if not greater importance, is the ability to accurately forecast philanthropic revenue on a fiscal year basis. As anyone who has worked in Advancement (or fundraising more broadly) can attest – dollars-raised “this year” is a very important determinant of program effectiveness. Having said that, how well did the Queen’s simulation perform on a year-over-year basis? Let’s take a look.




    Actual Fundraising Major Gift (MG) Revenues 2015-19 (Re-scaled)
    Year Simulated 90% CI MG Revenue Actual MG Revenue Coefficient of Variation
    2015-16 8.876 – 10.295 10.728 5.35%
    2016-17 1.452 – 2.709 1.682 26.65%
    2017-18 2.073 – 2.659 2.130 8.59%
    2018-19 4.629 – 5.528 4.294 5.43%

    As shown in Table 1, when we compare fiscal year results against forecast values the results are promising. For example, even though in 2015 the forecast range did not capture the Actual MG Revenue value, it was close and only 4.2 percent away from the upper limit. Simulated results for 2016 were also very good and captured Actual MG Revenue in a 90% Confidence interval. However, the variance of results was significantly greater than in the previous year possibly owing to the comparatively small number of major gift received for the year. Overall the results of the simulation approach are impressive.

    What Next
    Getting major gift revenue projection right – or at least getting in the right ballpark – is vitally important for any charity. Important staff, program and policy decisions and made on the basis of major gift revenue projections and Monte Carlo simulation appears to have real potential in this area. When combined with other, more traditional techniques, Monte Carlo simulation can provide a sophisticated and complimentary approach to forecasting that is evidence-based and timely. The intersection of simulation, machine-learning and artificial intelligence is happening all around us and it makes perfect sense to harness this technology in our work as Advancement professionals. It will help us do our jobs better. You can bet on it.





  • Classifying Fungal Colonies from Mass Spectrometry Data
    Supervisor: Randy E Ellis
    Description:
    I have agreed to supervise Mercy Doan in this project.
  • Interactive Wearables for Building Biodata Predictive Models
    Supervisor: Sara Nabil and Farhana Zulkernine
    Description:
    In computing, smart textiles and wearables technology provides the unique opportunity of embedding computers in everyday garments and on-body accessories in ways that blend with our existing environment. Off-the-shelf sensors (https://mbientlab.com/store/) provide capabilities for collecting massive data sets for machine learning predictive models but lack being neatly integrated in what people actually want to wear and hang on to with undesirable and unfashionable black Velcro wrist and sleeve bands. Can you help us redesign them in better ways?

    In this project, you will have the opportunity to blend these two fields together to create interactive garments that can collect sensor data. This project will have two main components. The first (HCI element) will involve designing and prototyping, with support and training from iStudio lab researchers, where you will aim to understand the digital fabrication process of smart textiles (i.e. technical embroidery and digital weaving: https://youtu.be/wIVCDS8fYmk?si=IFxskJA9o4Le4Z0Q). Using your designs, you will then build a simple computational model for data collection (ML element) using people sensor data. For example, imagine designing a garment that collects data for a model to predict unstable walking of an elder and thus prevent near-fall event before they even happen by notifying them to take a seat, hold on to something, or at least fall gracefully.

    This project is well suited for teams who would like to pursue User Experience Research (UXR) careers or work with biomedical data in roles in industry or academia after their degree. As part of this research project, on top of prototyping with physical computing, you will gain valuable and everlasting industry skillsets such as how to conduct research interviews, conduct prototype evaluations with users, facilitate design workshops, and write up your analysis. You will also get the opportunity to leverage the lab’s tools, such as our professional PhotoBox studio and equipment, to document and create a portfolio-worthy project.

    No previous experience with wearable fabrication required, but an interest in design, physical computing and any previous experience with Arduino will help you be successful in this project. While the later will be working at the iStudio, other group members will be working closely with the Big-data Analytics and Management Laboratory (BAMLab).
  • Exploring 3D-printing for Prototyping Wearable Technology
    Supervisor: Sara Nabil
    Description:
    Have you ever thought you can 3D-print your clothes? here is your chance!

    Prototyping with 3D-printing can produce materials that morph, sense, or actuate in various effects. As a material, 3D-printing filament or resin survives different fabrication operations, such as sawing, screwing, and nailing, among other additive and subtractive methods, aesthetic and functional properties.

    In this project, you'll get familiar with the state-of-the-art 3D-printing techniques, digital fabrication methods, and smart materials that can be used to create smart furniture and artifacts. You will be mainly playing and experimenting with both our 3D printing systems:
    1. FormLab Form 3: Resin Printing (SLA)
    https://formlabs.com/3d-printers/form-3/
    2. Ultimaker S5 Studio: Filament Printing (Nylon, CPE, PLA, or PVA) with dual extrusion
    https://shop3d.ca/collections/ultimaker-3d-printers/products/ultimaker-s5-studio
    Applications include designing prototypes for/with people with physical disabilities, prosthetics, wearables, and smart textiles.
    This project is well suited for individuals who would like to pursue careers in design, User Interfaces (UI), and User Experience (UX) roles in industry or academia after their degree. As part of this research project, on top of prototyping with physical computing, you will gain valuable and everlasting industry skillsets such as how to conduct research interviews, conduct prototype evaluations with users, and write up your analysis. You will also get the opportunity to leverage the lab’s tools, such as our desktop vacuum former, 3D-scanner, digital fabrication machines, and professional PhotoBox studio and equipment, to document and create a portfolio-worthy project.
  • Textile Interfaces and User Experience (UX) of Tactile Interaction
    Supervisor: Sara Nabil
    Description:
    Electronic textiles (e-textiles) is an emerging industry in Canada that combines increasingly small computers and embeds them in textiles. This results in user interfaces that are fuzzy, squishy, and stretchy, and that act differently than traditional screen-based devices. These types of devices also expand where computers belong, such as in clothing, soft furnishings, and soft objects.

    In this project you will learn how to prototype soft devices with the different textile fabrication machines available in the lab (such as embroidery, weaving, knitting, 3D printing, or laser cutting) with a focus on creating soft sensors with varying textures. After experimenting and developing several interactive textile samples you will get the opportunity to run design ideation workshops to better understand how users imagine interacting with these types of devices, and what applications they imagine using them for.

    This project is well suited for individuals who would like to pursue User Experience Design (UXD) or User Experience Research (UXR) roles in industry or academia after their degree. As part of this research project, on top of prototyping with physical computing, you will gain valuable and everlasting industry skillsets such as how to conduct research interviews, conduct prototype evaluations with users, facilitate design workshops, and write up your analysis. You will also get the opportunity to leverage the lab’s tools, such as our professional photo box studio and equipment, to document and create a portfolio-worthy project.

    No previous experience with textile fabrication required, but an interest in design, physical computing and any previous experience with Arduino will help you be successful in this project.
  • Information Diffusion and Diversity in On-Line Social Networks
    Supervisor: Erin Meger
    Description:
    On-Line Social Networks play a role in information diffusion across society. There are many theoretical models for online social networks. With the use of these models, we can measure the influence across diverse networks. In the biased preferential attachment model, Stoica et all examined the effect of diversity of node affiliations on the diffusion of information. This project will involve replicating known results, and supplementing the known models with increased diversity. The focus of the project is exploring the activation of information across a network.
  • Probabilistic Simulations on Deterministic Social Network Models
    Supervisor: Erin Meger
    Description:
    Models of complex networks allow researchers to explore the structure of real world phenomenon to better influence algorithm development and social patterns. The Iterated Independent Model for Social Networks (Meger and Raz) is a robust theoretical model that generalizes deterministic models of transitive and anti-transitive node interactions (such as the adage "the enemy of my enemy is my friend"). In this project, students will run simulations on this model. The focus of the project is to generate complex networks, determine the structure and properties of the models, and compare and contrast parameter values within the model.
  • Teaching and Learning in STEM – Accessible Assessments
    Supervisor: Erin Meger
    Description:
    Pedagogy surrounding deadlines and accommodations has evolved significantly since the influx of on-line learning after the Covid-19 Pandemic. Strategies surrounding best practices for assessment design and deadlines revolve mostly around arts education and small classes. This project will focus on data collection and analysis of secondary sources involving publicly available data on assessments, outcomes, and policies. By focusing our attention to STEM courses, we will seek to learn typical assessment styles, analyse learning outcomes, and create new and innovative best practices solutions to teaching and pedagogy in STEM.
  • Automatic bug-finding for TeX proofs
    Supervisor: Jana Dunfield
    Description:
    Proof assistants such as Agda are increasingly used to automate metatheory (proofs about programming languages). However, these tools require great effort to use; I used a proof assistant for just one of my research papers. And how do you do proofs about the proof assistants? Proofs in natural language—”paper” proofs—will always play a role.

    The goal of this project is to build a tool that looks for certain kinds of bugs in proofs written using TeX. Unlike a proof assistant, the goal is not to build a tool that checks that the proof is completely correct, but to find (some) probable bugs in the proof.

    The kinds of bugs detected might include:

    • certain wrong uses of the induction hypothesis;
    • circular references;
    • mismatched lemmas (the proof “uses” Lemma 2 but the result derived has the wrong shape);
    • “obvious” missing cases.

    To avoid, or at least reduce, the need to understand natural language, the tool you create would be restricted to proofs written in TeX using a specific set of macros.

    Basic familiarity with TeX is strongly recommended. Knowledge of a functional programming language, such as thorough understanding of the Haskell material from CISC 360, is required. Knowledge of material from CISC 465 and/or CISC 458 would be ideal, though since most 499 students will have not yet taken those courses, concurrent enrolment in 465 and/or 458 in the Winter term is helpful. Knowledge of natural language processing may be useful.
  • Analysis of security protocols with Alloy
    Supervisor: Furkan Alaca and Juergen Dingel
    Description:
    Security is an important quality attribute in many applications. Unfortunately, it can also be very challenging to establish. Formal methods offer techniques and tools that can allow more rigorous analysis of software artifacts than traditional techniques such as inspection and testing. The use of formal methods to verify security-sensitive applications is of great interest for researchers and practitioners [1]. Alloy is an open source language and analyzer for software modeling [2]. Specifications are expressed in first-order predicate logic enriched with support for relations which have proven very useful to capture relationships between collections of objects. The analysis is based on bounded satisfiability checking.

    The goal of the project is to explore and possibly extend the state-of-the-art in the use of Alloy to analyze web security protocols. The work described in [3] and [4] will provide a starting point.

    The project is most suited for students with interests in formal modeling and security. Completion of CISC 422 and 447 would be ideal.

    [1] S. Chong et al, "Report on the NSF workshop on formal methods for security", August 1, 2016, available here
    [2] D. Jackson, "Alloy: A Language and Tool for Exploring Software Designs", Communications of the ACM, September 2019, Vol. 62 No. 9, pages 66-76, available here
    [3] D. Akhawe, A. Barth, P. E. Lam, J. Mitchell and D. Song, "Towards a Formal Foundation of Web Security," 2010 23rd IEEE Computer Security Foundations Symposium, Edinburgh, UK, 2010, pages 290-304, available here
    [4] H. Shimamoto, N. Yanai, S. Okamura, J. P. Cruz, S. Ou and T. Okubo, "Towards Further Formal Foundation of Web Security: Expression of Temporal Logic in Alloy and Its Application to a Security Model With Cache," in IEEE Access, vol. 7, pp. 74941-74960, 2019, available here
  • Risk Stratification for Managing Post-Partum Complications
    Supervisor: Amber Simpson
    Description:
    All pregnant and post-partum patients are at increased risk of developing a blood clot, which can lead to severe consequences including death. We have national guidelines that instruct us on how to manage pregnant patients (ie. who is at high risk and needs a blood thinner vs. who is low risk and does not), but there are numerous risk factors at play, and it can be quite convoluted to perform individualized risk stratification and get it right for every patient. The proposed work would build an app geared toward doctors and midwives and would have a series of simple yes or no questions for the user to answer. A computational model would be created based on risk factors to determine low and high risk. The app would then provide a model-based recommendation on whether or not the patient requires a blood thinner, the recommended dose, and the duration of therapy. The app could save a lot of health care providers' time and help us get this question right for every patient.
  • Transforming Bladder Cancer Diagnosis through Digital Pathology and Computing
    Supervisor: Kathrin Tyryshkin
    Description:
    Project Overview: In this independent study, you will delve into the world of cancer grading, a critical aspect of diagnosing and treating bladder cancer. Pathologists play a crucial role in this process, examining tissue samples under a microscope to determine if the cancer is low-grade or high-grade. This classification is pivotal, as it guides the intensity of monitoring and treatment for each patient.

    The Significance of Cancer Grade: Low-grade cancer cells closely resemble normal bladder cells, and pose little risk to patients. In contrast, high-grade cancers exhibit more abnormal cell shapes and arrangements and are more likely to recur after removal and to invade the bladder wall. This crucial distinction dictates the course of action for urologists in managing the cancer.

    The Challenge: Despite its importance, grading remains a subjective judgment call. Pathologists rely on extensive training and reference materials, yet there exists a considerable variation in practice. Even expert pathologists may disagree on the grade in up to 40% of cases. This inconsistency can lead to vastly different treatment approaches for patients with similar conditions.

    The Proposed Solution: This project leverages the power of digital pathology and computer-aided technologies to revolutionize cancer grading. By employing advanced algorithms, we can objectively analyze every cell in a pathology sample, providing precise measurements of cancer grade. Currently, we've successfully developed reference ranges for low-grade and high-grade cancers using small snapshots from microscope slides. This opportunity will take this initiative a step further, extending the analysis to the entire microscope slide.
    Impact and Benefits: By establishing objective measurements, we aim to transform cancer grading from a subjective opinion into an objective fact. This standardization will have a far-reaching impact, ensuring consistent care pathways across healthcare centers, and ultimately improving access to high-quality diagnostics and management. Additionally, this technology has the potential to identify a larger cohort of patients who can maintain their health with fewer invasive procedures and toxic treatment, improving both outcomes and quality of life.

    How You'll Contribute: As an undergraduate thesis student, you will work with pathologists, industry professionals, and computer science faculty and staff to train AI-driven whole slide image analysis algorithms. The team will ensure that these algorithms recognize and measure cancer cell features accurately. You will then use these feature measurements to build and validate time-dependent survival models. Your contributions will be instrumental in refining and expanding our methodology, with the ultimate goal of revolutionizing cancer diagnosis and treatment.

    If interested, please send me a quick note on your qualifications (e.g. courses taken) and your transcript
  • Evaluating Open-Source Large Language Model for Medical Summarization Tasks
    Supervisor: Farhana Zulkernine
    Description:
    Evaluating Open-Source Large Language Model for Medical Summarization Tasks

    Context: Given the success of ChatGPT, an increasing number of similar models have been proposed. Generation-based Large Language Model’s (LLM) successfully demonstrated its ability to deal with different general NLP tasks and engage in human-like conversation. However, for domain specific tasks, for example, healthcare, the efficacy of these models require further validation. The primary objective of this project is to conduct a comprehensive evaluation of state-of-the-art open-source large language models including Llama2 and Falcon, on medical summarization tasks.

    Objectives: In this project, students need to implement a pipeline to evaluate different open-source ChatGPT alternatives for a variety of medical summarization tasks such as medical question summarization and medical conversation summarization. Moreover, students are encouraged to investigate and apply advanced techniques in prompt engineering and leverage the principles of few-shot learning to enhance the efficacy of the chosen models. There’re also data analysis opportunities for students with medical/biology background to evaluate the model’s output and participate in RLHF projects.

    Dataset: Evaluation focuses primarily on the medical summarization in three categories: a) Medical question summarization, which involves summarizing long form medical questions provided by patients.; b) Medical conversation summarization, where the summarization is applied to the conversation between patients and physicians, and c) Question-driven summarization which involves summarizing answers based on questions and referenced documents.

    Tools: Python, pytorch library, Hugging Face transformers library.

    Deliverable: A 20–30 page double-spaced report containing the problem description, data description, related literature, flowcharts, pseudo codes, and results as applicable. Students will get an opportunity to present at the BAM Lab group meetings and collaborate with IBM US.

    Expertise Needed: CISC/CMPE 452/COGS 400 or Deep learning courses with text analytics.
    If interested, please contact Yuhao.chen@queensu.ca and Farhana.zulkernine@queensu.ca with the transcript to allow background check to ensure that you have necessary background to work on this project.
  • City of Kingston Housing Match
    Supervisor: Wendy Powley
    Description:
    The City of Kingston is looking at potentially building a tool that keeps track of housing opportunities and those looking for housing to help facilitate matches. The specific focus of the project will be matching students with available housing. In this project you would do some background research into the tools that are currently available for this task and build an implementation of a web version of such a tool complete with a database and a web-based interface.
  • Predicting Hemophagocytic lymphohistiocytosis diagnosis in paediatric patients and building an individualized treatment algorithm
    Supervisor: Hazem Abbas
    Description:
    Background
    Hemophagocytic lymphohistiocytosis (HLH) is a syndrome of uncontrolled immune activation leading to hyperinflammation and cytokine release which results in tissue damage and multi-organ failure. (1)
    HLH is classified into primary and secondary to other conditions. Without treatment, primary HLH is almost universally fatal. Primary HLH includes familial HLH, HLH associated with immunodeficiency syndromes, and other primary immune defects e.g. XLP1 and XLP2. (2) In the last two decades, two multicentre, international studies (promoted by the Histiocyte Society, namely HLH-94 and HLH-2004) conducted using well-defined treatment protocols, based on multi-agent treatment strategy and allogeneic hematopoietic stem cell transplantation (HSCT), have led to improvement of the outcome of patients affected by primary HLH.(3,4)
    Secondary HLH may develop due to cancer, infectious or immune-mediated diseases. HLH may present as an isolated central nervous system disease without systemic symptoms. (1,2)
    Diagnostic criteria have been suggested for HLH by international bodies. (4)
    Workup includes blood counts, liver function, coagulation screening, fibrinogen, triglycerides, ferritin, soluble CD25, bone marrow aspirate. CNS evaluation includes brain magnetic resonance imaging and lumbar puncture. Adjunctive studies may include computed tomography (CT) scan and positron emission tomography (PET) scan, infectious markers, bacterial, fungal and viral studies. (1-4)
    Flowcytometry screening studies (Perforin, SAP, XIAP, CD107a) are rapid screening tests to exclude major causes of HLH. Genetic testing proves definitive causes of primary HLH. (1,2)
    Various treatment modalities of HLH have been described from the gold standard HLH 94 and HLH 2004 protocols (including dexamethasone, etoposide, and ciclosporin), (3,4) to the recent French protocol including (alemtuzumab, methylprednisolone, and ciclosporin). (5) Other treatment options include single-agent alemtuzumab (6), anti-interferon gamma monoclonal antibody (Emapalumab,7), anti-interleukin 1 monoclonal antibody (Anakinra,8), Ruxolitinib (9,10) and anti-thymocyte globulin (11).
    HLH-94/HLH-2004 studies reported 5-year overall survival of ∼60%. (3,4) Poor prognostic factors include CNS involvement, malignancy-associated HLH, severe neutropenia, hypoproteinaemia, and hepatic impairment. (12-15)
    Objectives:
    Primary objective: Establishing a highly sensitive and specific diagnostic algorithm for both primary and secondary HLH using artificial intelligence tools.
    Secondary objective: exploring the feasibility of proposing individualized treatment plans for patients with HLH according to patient and disease-related variables.
    Methods:
    Systematic review of medical literatures related to HLH in children aged 0-18 years to identify relevant data that may be utilized to build a diagnostic model and exploring feasibility of building an individualized treatment model. Another source of data would be a current multi-center retrospective study exploring time to HSCT in pediatric patients with HLH. Approaching the Histiocyte society would be another option for gathering data for the project.


    References:
    1- Jordan MB, Allen CE, Weitzman S, Filipovich AH, et al. How I treat hemophagocytic lymphohistiocytosis. Blood. 2011 Oct 13;118(15):4041-52. doi: 10.1182/blood-2011-03-278127.
    2- Marsh RA, Haddad E. How i treat primary haemophagocytic lymphohistiocytosis. Br J Haematol. 2018 Jul;182(2):185-199. doi: 10.1111/bjh.15274.
    3- Henter JI, Aricò M, Egeler RM, Elinder G, et al. HLH-94: a treatment protocol for hemophagocytic lymphohistiocytosis. HLH study Group of the Histiocyte Society. Med Pediatr Oncol. 1997;28(5):342-7. doi: 10.1002/(sici)1096-911x(199705)28:53.0.co;2-h.
    4- Henter JI, Horne A, Aricó M, Egeler RM, et al. HLH-2004: Diagnostic and therapeutic guidelines for hemophagocytic lymphohistiocytosis. Pediatr Blood Cancer. 2007;48(2):124-31. doi: 10.1002/pbc.21039.
    5- Despina Moshous, Coralie Briand, Martin Castelle, Laurent Dupic, et al; Alemtuzumab as First Line Treatment in Children with Familial Lymphohistiocytosis. Blood 2019; 134 (Supplement_1): 80. doi: https://doi.org/10.1182/blood-2019-124477
    6- Marsh RA, Allen CE, McClain KL, Weinstein JL et al. Salvage therapy of refractory hemophagocytic lymphohistiocytosis with alemtuzumab. Pediatr Blood Cancer. 2013 Jan;60(1):101-9. doi: 10.1002/pbc.24188.
    7- Lounder DT, Bin Q, de Min C, Jordan MB. Treatment of refractory hemophagocytic lymphohistiocytosis with emapalumab despite severe concurrent infections. Blood Adv. 2019 Jan 8;3(1):47-50. doi: 10.1182/bloodadvances.2018025858.
    8- Baverez C, Grall M, Gerfaud-Valentin M, De Gail S, et al. Anakinra for the Treatment of Hemophagocytic Lymphohistiocytosis: 21 Cases. J Clin Med. 2022 Sep 30;11(19):5799. doi: 10.3390/jcm11195799.
    9- Broglie L, Pommert L, Rao S, Thakar M, et al. Ruxolitinib for treatment of refractory hemophagocytic lymphohistiocytosis. Blood Adv. 2017 Aug 17;1(19):1533-1536. doi: 10.1182/bloodadvances.2017007526.
    10- Sin JH, Zangardi ML. Ruxolitinib for secondary hemophagocytic lymphohistiocytosis: First case report. Hematol Oncol Stem Cell Ther. 2019 Sep;12(3):166-170. doi: 10.1016/j.hemonc.2017.07.002.
    11- Mahlaoui N, Ouachée-Chardin M, de Saint Basile G, Neven B, Picard C, et al. Immunotherapy of familial hemophagocytic lymphohistiocytosis with antithymocyte globulins: a single-center retrospective report of 38 patients. Pediatrics. 2007 Sep;120(3):e622-8. doi: 10.1542/peds.2006-3164.
    12- Trottestam H, Horne A, Aricò M, Egeler RM, et al. Chemoimmunotherapy for hemophagocytic lymphohistiocytosis: long-term results of the HLH-94 treatment protocol. Blood. 2011 Oct 27;118(17):4577-84. doi: 10.1182/blood-2011-06-356261.
    13- Bergsten E, Horne A, Aricó M, Astigarraga et al. Confirmed efficacy of etoposide and dexamethasone in HLH treatment: long-term results of the cooperative HLH-2004 study. Blood. 2017 Dec 21;130(25):2728-2738. doi: 10.1182/blood-2017-06-788349.
    14- Blincoe A, Heeg M, Campbell PK, Hines M et al. Neuroinflammatory Disease as an Isolated Manifestation of Hemophagocytic Lymphohistiocytosis. J Clin Immunol. 2020 Aug;40(6):901-916. doi: 10.1007/s10875-020-00814-6.
    15- Zhang Lijun, Dai Lei, Li Deyuan. Risk factors of early death in pediatric hemophagocytic lymphohistocytosis: Retrospective cohort study.Frontiers in Pediatrics. 2022(10). DOI=10.3389/fped.2022.1031432.
  • Predicting Neutropenia Diagnosis In Pediatric Patients Utilizing Artificial Intelligence Tools
    Supervisor: Hazem Abbas
    Description:
    Predicting Neutropenia Diagnosis In Pediatric Patients Utilizing Artificial Intelligence Tools

    Predicting Neutropenia Diagnosis In Pediatric Patients Utilizing Artificial Intelligence Tools

    Background
    Reduction in the absolute neutrophil count (ANC) below the lower limit of the normal range for the age and ethnic origin of the affected subject. Different definitions of neutropenia have been suggested. One of the commonly used states that from the age of 1 year to adulthood the cutoff level for neutropenia is 1.5 x109/L. In neonates and preterm (28–36 weeks) infants during the first week of life (72–240 hours), the lowest ANC threshold is 2.5 × 109/L. In early preterm (less than 28 weeks) infants, neutropenia is defined as neutrophil of less than 1.0 × 109/L. (1-4)
    Isolated neutropenia may be due to inherited bone marrow failure, primary or idiopathic neutropenia (includes primary autoimmune, primary alloimmune, non-antibody-mediated, idiopathic neutropenia of infancy) or secondary to various causes including hypersplenism, infections, autoimmune diseases, nutritional deficiencies, metabolic diseases, immuno-regulatory disorders, hematologic diseases, or drug Induced. (4-6)
    The age of first detection of neutropenia, results of previous cell blood counts (CBC) could clarify the duration of neutropenia and differentiate between acute and chronic neutropenia. It is also important to know whether the neutropenia was an incidental finding or part of an acute illness. Clinical history should investigate the frequency, type, severity of infections, and need for hospitalization. Inquiries about clinical events such as fever, mouth ulcers, sore throat, gingivitis, sinusitis, otitis, skin ulcers and cellulitis, deep tissue infections, episodes of pneumonia, gastrointestinal symptoms, and perianal infections are particularly important. Periodic patterns of recurrent infections could indicate cyclic neutropenia. Examination of a patient with neutropenia should focus on growth, psychomotor development, dysmorphic features, skin, mucous membranes, upper and lower respiratory tract, lymph nodes, liver, spleen and congenital anomalies. (4)
    Laboratory testing for patients with isolated neutropenia includes viral screen, lymphocytes subsets, immunoglobulin level, anti-neutrophil antibody tests, αβ double negative T cells percentage, CBC of patient’s family, Serial patient CBC. Bone marrow testing looks for cellularity, numbers and maturation of erythroid, myeloid precursors, megakaryocytes, presence of dysplastic features, cytogenetic evaluation for acquired chromosomal abnormalities typical for MDS or AML. Genetic testing of acquired somatic gene mutations, which play a role in inducing congenital neutropenia is recommended (e.g., RUNX-1, CSF3R, TP53). (4-6)
    Fig 1: Suggested diagnostic algorithm for childhood isolated neutropenia: (4)


    Objectives:
    Primary objective: Establishing a reliable diagnostic algorithm for childhood neutropenia utilizing artificial intelligence tools.
    Methods:
    A systematic review of medical literature related to isolated neutropenia in children aged 0-18 years to identify relevant data that may be utilized to build a diagnostic model. Another source of data would be a current single-center retrospective study in pediatric patients with isolated neutropenia. Approaching the severe congenital neutropenia registry is another route to gather relevant data as needed.


    References:
    1- Fioredda F, Onofrillo D, Farruggia P, et al. Diagnosis and management of neutropenia in children: The approach of the Study Group on Neutropenia and Marrow Failure Syndromes of the Pediatric Italian Hemato-Oncology Association (Associazione Italiana Emato-Oncologia Pediatrica- AIEOP). Pediatr Blood Cancer. 2022;69:e29599. https://doi.org/10.1002/pbc.29599
    2- Dinauer MC, Newburger PE, BorregaardN. The phagocyte system and disorders of granulopoiesis and granulocyte function, Orkin SH, Fisher DE, Look AT, Lux SE, Nathan DG, Eds. Nathan and OsKi’s Hematology of Infancy and Childhood. 8th edition. Philadelphia: Elsevier Saunders Company; 2015:773-850.
    3- Christensen RD,Henry E, Jopling J, Wiedmeier SE. The CBC: reference ranges for neonates. Semin Perinatol. 2009;33:3-11.
    4- Severe Congenital Neutropenia International Registry. www.severechronic-neutropenia.org
    5- Farruggia P, Puccio G, Fioredda F, et al. Autoimmune neutropenia of childhood secondary to other autoimmune disorders: Data from the Italian neutropenia registry. Am J Hematol. 2017 Sep;92(9):E546-E549. doi: 10.1002/ajh.24803.
    6- Donadieu J, Bellanné-Chantelot C. Genetics of severe congenital neutropenia as a gateway to personalized therapy. Hematology Am Soc Hematol Educ Program. 2022 Dec 9;2022(1):658-665. doi: 10.1182/hematology.2022000392. PMID: 36485107; PMCID: PMC9821599.

  • Deep Learning-based Simulation of Texas Hold ‘em.
    Supervisor: Yuan Tian
    Description:
    Apply several deep learning models and data analytics techniques with the goal of predicting a players Hand Rank based on how they played any given hand. Hand Rank is a numerical metric of how strong their hand is - it represents the percentage of hands that the player beats.
  • Vulnerability Detection in Firmware Images
    Supervisor: Steven Ding
    Description:
    This 12-week project aims to enhance firmware security by detecting vulnerabilities within firmware images. The focus will be on collecting CVE data, compiling projects, analyzing compiled binaries, and evaluating non-AI CWE detection tools.

    Project Details
  • Firmware Catalogue Development
    Supervisor: Steven Ding
    Description:
    This project spans 12 weeks and focuses on creating crawlers to collect commercial firmware images for the purpose of building a comprehensive firmware catalogue.

    Project Details
  • Adversarial Attacks Against Generative AIs such as ChatGPT
    Supervisor: Steven Ding
    Description:
    This project spans 12 weeks and focuses on exploring and understanding adversarial attacks specifically targeted at generative AIs such as ChatGPT. The aim is to study vulnerabilities, develop attack strategies, and propose defense mechanisms to enhance the robustness of these models against adversarial manipulation.

    Project Details
  • Intrusion Detection System (IDS) for CAN Bus in Automobiles
    Supervisor: Steven Ding
    Description:
    This 12-week project aims to develop an Intrusion Detection System (IDS) specifically tailored for the Controller Area Network (CAN) bus in automobiles. The project involves designing algorithms, implementing monitoring mechanisms, and testing for anomalous behavior detection within the CAN bus system.

    Project Details
  • AI-Based Communication Protocol Reverse Engineering
    Supervisor: Steven Ding
    Description:
    This 12-week project focuses on leveraging AI to reverse engineer communication protocols from raw bit streams. The objective is to develop models capable of inferring fields within communication protocols using artificial intelligence techniques.

    Project Details
  • AI for Video-Based Counter Uncrewed Aerial Systems Concept Development
    Supervisor: Steven Ding
    Description:
    This 12-week project focuses on leveraging AI for the development of a video-based system to counter uncrewed aerial systems (UAS). The project aims to collect data from open-source repositories, utilize machine learning algorithms, and develop a concept for identifying and countering UAS through video analysis.

    Project Details
  • AI-Based Attack on Drone Swarm Algorithms
    Supervisor: Steven Ding
    Description:
    This 12-week project aims to assess the vulnerabilities of drone swarm algorithms to AI-based attacks. The focus will be on analyzing drone swarm behavior, devising AI-driven attacks, and evaluating their impact on the swarm's functionality and security.

    Project Details
  • ERS0 - Firmware Software Bill of Material (SBOM) Analysis Platform Development
    Supervisor: Steven Ding
    Description:
    ERS0 - Firmware Software Bill of Material (SBOM) Analysis Platform Development

    This project spans 12 weeks and focuses on the development of a Firmware SBOM Analysis Platform. The existing system is built using Python, Docker Compose, Socket.IO, React, Chakra UI, and PyTorch. Students will contribute to the project using an agile development process, delivering tested features every three weeks. Scrum meetings will be held weekly to decide on tasks and ensure project alignment.

    Project Details
  • Menu Item Recognition from Images
    Supervisor: Steven Ding
    Description:
    This 12-week project focuses on developing an AI-based solution to recognize menu items, ingredients, and pricing from menu pictures. The project involves data collection, proposing a recognition method, implementing the solution, and rigorous testing for accuracy.

    Project Details
  • AI for Swarm Counter-Swarms
    Supervisor: Steven Ding
    Description:
    The project aims to develop a simulated environment for testing and refining an intelligent drone swarm AI algorithm. The objective is to create a network of smaller drones that can detect, intercept, and neutralize incoming swarms of hostile UAS (Unmanned Aerial Systems). This simulation will employ swarm intelligence algorithms to effectively neutralize threats.

    Project Details
  • IMDb Knowledge Graph
    Supervisor: Yuanzhu Chen
    Description:
    In this project, you will collect data from IMDb, construct a knowledge graph using Neo4j, and analyze patterns using this graph database.
  • Course-Curriculum Map
    Supervisor: Yuanzhu Chen
    Description:
    In this project, you will create a mapping between degree plans offered by School of Computing and courses administered by the School. You will visualize the mapping and analyze it using graph theoretic terms.
  • Explanatory AI (XAI) for Improving Deep Learning Performance
    Supervisor: Hesham ElSawy
    Description:
    XAI has been identified as a key factor for adoption of AI systems in sensitive sectors applications such as military, automation healthcare, finance and banking, etc. The concept of XAI revolves around analyzing and understanding the input, the internals, and outputs of Black-box models. That is by obtaining human-interpretable models, and increase the trustworthiness in their decision-making mechanisms. This project will further utilize XAI outputs (i.e., as a feedback) to improve existing deep learning models. The developed XAI techniques can be applied to different fields, and hence, the application domain will be selected according to the mutual interest between students and instructor.
  • Generating a Multi-Sensor Dataset for Human Activity Recognition
    Supervisor: Farhana Zulkernine
    Description:
    Context: Multi-sensor data fusion has received much attention over the recent years in computer vision. The aim of multi-sensor data fusion is to overcome the limitations of individual sensors and combine the advantages of using multiple sensors to advance machine perception from a more comprehensive representation of the data of different modalities (e.g., RGB video, integer, audio, point cloud). After the collection of data, preprocessing is a vital task, especially for multi-sensor data, to extract useful data features and fuse them effectively to improve the performance of machine learning models by training them using the curated multimodal data. Typically, supervised machine learning approach is used where data need to be labeled manually or using rule-based systems.

    Objectives: In this project, students need to collect data using multiple sensors such as camera, radar, and audio (RGB, Skeleton, Point Cloud) for Human Activity Recognition (HAR). Depth cameras can be used with predefined algorithms or our own algorithms to generate skeleton data of humans for pose detection from still images or activity recognition from a sequence of video frames. Audio recorded during the activity can be used to generate labels. In this project, students will be required to 1) set up the sensors and recruit participants to generate Activity of Daily Living (ADL) and other activities for detecting health problems such stomach pain, headache, unstable walking, and falling. 2) Then necessary software must be developed to ingest the data simultaneously from multiple sensors and store that with labels which can be generated using microphones. 3) Finally, the data must be preprocessed to feed to existing machine learning models for activity recognition. Enthusiastic students can extend their work after the term to implement their own models by extending existing models and publish their work in conferences or journals.

    Tools: Python, Pytorch library, different types of sensors.

    Deliverable: A 20–30 page double-spaced written report containing the problem description, data description, related literature, flowcharts, pseudo codes, and results as applicable. Students will get an opportunity to present at the BAM Lab group meeting and public their own created dataset.

    Expertise Needed: CISC/CMPE 452/COGS 400 or Deep learning courses about computer vision.
    If interested, please contact 17zz55@queensu.ca, aman.anand@queensu.ca or farhana.zulkernine@queensu.ca with your transcript to allow background check to ensure that you have necessary knowledge to work on this project.
  • Race Car Telemetry Visualization
    Supervisor: James Stewart
    Description:
    Race car drivers analyse telemetry from many sensors to improve their lap times. This telemetry is usually presented as 2D graphs that are functions of time or distance and include position, orientation, acceleration in various directions and around various axes, brake pressures, steering angle, and so on.

    The project will develop an application to display a 3D model of a real car travelling around a real racetrack: Canadian Tire Motorsport Park. 3D models for the car and track can be found online. Telemetry from the car will be provided from a CSV file and will be displayed on the moving car with 3D graphical widgets. Implmentation is to be done in C++ with OpenGL.
  • Video Enhancement for Lightboard Lectures
    Supervisor: James Stewart
    Description:
    In a lightboard lecture video, the instructor writes with a dry-erase pen on a glass surface. But the ink deposition is uneven, making it difficult to read.

    This project will enhance the faded writing by detecting and tracking the pen strokes in the video. The pen strokes will be kept in a separate buffer and processed to make them clearer. Video cuts must be detected. Implementation is to be done in C++ with OpenGL.
  • Elevator Arrival Prediction
    Supervisor: James Stewart
    Description:
    NOT AVAILABLE YET AS WE HAVE NOT INSTALLED THE ELEVATOR HARDWARE.

    We have hardware to track the movement of the elevator in Goodwin and to display its current position and direction.

    This project will take telemetry from the elevator and will predict its arrival time at a particular floor. The arrival time and other information will be displayed in a graphical interface on the wall at the 5th floor elevator. Various prediction models will be developed and tested.
  • Animations of Algorithms and Data Structures
    Supervisor: James Stewart
    Description:
    For CISC365 we would like to visualize data structures and algorithms such as: median finding, convex hull (divide & conquer and randomized), minimum area contour, edit distance, randomized binary space partition.

    The project will involve the development of interactive animations of some of these topics. Implementation is to be done in C++ with OpenGL.

    See also the related project by Burton Ma.
  • Fourier Transform Exploration Tool
    Supervisor: James Stewart
    Description:
    This project will build a tool to demonstrate the Fourier Transform, for use in CISC457. A user will use the tool to see the individual waves that constitute an image, to see the effect of zeroing low-magnitude waves or waves corresponding to peaks in the spectral domain, to see convolutions and de-convolutions, and to see the effect of various band-pass filters. Implementation is to be done in C++ with OpenGL.