Innovative Projects Realized

Operating Room Traffic Assessment: A Video Analysis Approach

Surgical Safety Technology aims to improve operating room safety by capturing and analyzing operation videos. Usually, operating room traffic (like people displacement) has a huge impact on surgery. Unnecessary movements can cause distraction of surgeons and pollution of the sterile environment. This project applies computer vision models to detect and track people movements in the operating room and assesses the relationship between adverse events and errors. Popular machine learning models such as Deep Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) have the capability to analyze time sequential data. Trained on the well-labeled data directly from specific hospitals, these models could work out precise operating room traffic trace and its correlation with surgical events.

Early detection of Alzheimer’s disease symptoms using speech longitudinally

An early symptom of Alzheimer’s Disease is difficulty in remembering recent events. These trends are reflected in problems in language and patterns of speech. Speech patterns of an individual can hence be used to determine the trajectories of preclinical cognitive decline. The difference in the cognitive trends over subject groups, analyzed using speech data collected over a long period of time, can be used to detect Alzheimer’s even before it can be confirmed clinically.
With the help of machine learning models, this process can be automated completely by using automatic speech recognition systems to transcribe the speech followed by analysis of these transcripts. This project will explore machine learning- based strategies to automate the early AD diagnosis pipeline.

Recommendation system for retail shopping

People rely on recommendations from other people, friends’ word, news reports, and travel guide and so forth. Recommendation systems assist people to sift through available books, web pages, restaurants, and grocery products. [16]. We want to build a recommendation flow in the retail industry to serve Canadian citizens better. The system will understand the customers and help them to make better selections and improve their shopping experience. A retail recommendation is different from e-commerce as the basket is substantially larger and customer tends to buy same product over and over again. In this project to build models to understand the existing customer base and products for shopping suggestions, robust substitutions, and search ranking. The system will make recommendations base on the users that are similar. For example, the system will learn your shopping behaviours and make product recommendation based on purchased history of other users that share the similar shopping behaviours.

Intra-operative Error Detection on Surgical Video based on Computer Vision Analysis

The intra-operative errors that occurs in adverse events have been a major concern in healthcare and surgical industry. Conventionally, error-event assessment is done by peer surgeon review, which is time consuming and costly. With the advances in machine learning and computer vision techniques, it is possible to keep track of the operation surgical procedures based on recorded surgical videos to evaluate and classify the errors occurred. With the proposed computer vision-based algorithm, it is expected to predict the error event during surgery in a scalable process to ensure a better and safer patient and surgical environment.
Since the intra-operative error detection algorithm is mainly trained on recorded surgical video data, it is expected to have an impact on improving decision-making and performance in future operations for complex patients and surgical circumstances.

Segmentation of 3D microscopy images

In-vivo imaging provides a unique opportunity to examine complex cellular activity in live tissue. Images produced by these experiments are difficult to analyze manually, typically applied to mono-layer cell culture assays (i.e. cells in a dish). Recent advances in deep learning enable the opportunity to analyze these in-vivo tissue images with greater efficiency and accuracy. This project will apply deep learning based segmentation and classification technology to a dataset provided by a collaborating pharmaceutical company. Deep learning algorithms will be developed to segment different cell types and vascular structures in the dataset and quantify features (i.e. length, volume, protrusion number, marker intensity) of these objects. These features will be used to evaluate the effectiveness of therapeutic treatments.

FINITE MODELING OF BRIDGE ELASTOMERIC POT BEARINGS

Elastomeric Pot Bearings (EPBs) are High Load Multi-Rotational bearings developed in

Europe in the early 19605 to support a bridge superstructure while transmitting large

force demands to the supporting piers and abutments, and accommodating rotation

about any horizontal axis as a function of the applied loads. EPBs have usually been

designed according to a mix of empirical and theoretical procedures. Very often, the

rationale behind some of these design rules is unclear and should be evidenced to

assure that the design still meets current engineering practice. The main objective of this

research project is to investigate and understand the structural behaviour of EPBs using

advanced numerical modeling techniques. A numerical procedure will be developed for

the design of optimized dimensions of EPBs, while satisfying the requirements of current

state of practice in Canada and Quebec. The results obtained will be thoroughly

examined to identify the influence of important design parameters and to

Automated Model Tuning for Retail

Artificial intelligence, especially Machine learning algorithms, plays important roles in building recommendation systems and promotional forecasting systems for retailers. However, training a machine learning model requires the choice of a number of significant features and requires tuning a large set of configurations. Therefore, it takes a long time for humans to find the optimal configuration for one or more predictors. However, the predictive performance of existing automated tuning models is not as good as manually tuning. Besides, the approach cannot be applied to more than one model. This project, will propose a system that can automatically come up with a set of models with corresponding features and configurations for a specific problem (e.g., promotional forecasting) that provides good or acceptable performance for the prediction.

Hand Pose Reconstruction Based on Fast Multi-Touch Sensors

Serving as the most widely-used body part for communication, hand is a very important tool for human to interact with the world. Especially with the continuing development of virtual reality and augmented reality, hand pose information has gradually become an indispensable component for improving users’ experience in interacting with computing devices. Therefore, this project aims at achieving hand pose reconstruction based on capacitive sensing technology using machine learning algorithm. The capacitive sensor that will be utilized in this project is supported by the project partner, Tactual Labs who, by the end of this project, will benefit by having its current innovative capacitive controller more intelligent.

An Artificial Agent for Light Switch

Smart home devices with artificial intelligence (machine learning and deep learning) will change our lifestyles in the near future. The objective of this project is to develop an artificial agent, which will power the smart light switches produced by ecobee. The artificial agent, a machine learning program, will use the data collected by the sensors in the smart light switches and help the users operate the light switches without the users’ manual control. The goal of this project is to develop an underlying smart program to learn the behaviors and of users with the light switches. The progress of this project will help ecobee provide better smart light switches for its clients and potentially incorporate this smart program to other similar ecobee smart home devices. Therefore, the success of this project can eventually contribute the smart home industry.

Identifying vehicle accidents and high risk drivers using Machine Learning

The primary objective of the project is to approach the problem of understanding true causality of vehicle accidents and scientifically determining which vehicles and drivers are at highest risk of an accident from a machine learning perspective. Geotab has a number of identified collisions in X, Y and Z planes, and much more. The research would be aimed at using both Geotab’s data in addition to external data such as weather and topography to develop a predictive model that can identify those drivers at highest risk of an accident. This may be based solely on current driving behavior and/or the driving history.
The results of this project are important in helping our over 20,000 commercial fleet customers understand the true safety risks that exist in their fleet leveraging a novel machine learning approach that goes beyond a generic score. TO BE CONT’D

Discovery of Endocannabinoid modulating compounds for Alzheimer’s disease therapeutics development

Alzheimer’s is the most common form of dementia which worsens over time. Current therapeutic against Alzheimer’s disease provides only symptomatic treatment. This limited effectiveness provides us with an opportunity to direct our research efforts towards developing new agents to prevent or retard the disease. Studies have shown that very small amount of tetrahydrocannabinol (THC), a chemical found in marijuana, can slow the production of Amyloid beta (A?) protein. This protein is found to be the hallmarks of Alzheimer’s disease and a key contributor in its progression. Our study aims at removing the psychoactive component from marijuana, retaining it therapeutic part and screening those compounds to find a potent drug against Alzheimer’s disease.

Outils d’aide à la décision pour la planification, l’ordonnancement et la gestion de la production de solutions textiles

Pour rivaliser avec l’offre de pays à faible coût de main d’oeuvre, l’industrie canadienne du textile a dû se doter d’une technologie à l’avant-garde et offrir à sa clientèle des produits parfaitement adaptés à leurs besoins. Planifier efficacement la production d’une gamme exhaustive de produits demeure toutefois un réel défi, c’est pourquoi cette recherche vise à proposer un ensemble d’outils permettant de soutenir les activités du fabricant nord-américain de tissus Duvaltex. En mettant sur pied différents scénarios d’aménagement de la production, un tableau de bord de gestion stratégique et un modèle avancé pour la planification et l’ordonnancement des activités, la recherche fournira ainsi à Duvaltex les outils nécessaires au maintien de sa position stratégique sur l’échiquier mondial.