Innovative Projects Realized

Evaluation of a radiomic approach based on hyperspectral retinal imaging to predict the cerebral amyloid status for the diagnosis of Alzheimer’s disease

The project will help Optina validate and further develope a novel technology to predict the presence of significant amyloid (A?) deposition in the brain from a simple, non-invasive hyperspectral retina scan in combination with an artificial intelligence algorithm. Accumulation of A? plaques in the brain is a key hallmark of Alzheimer’s disease (AD), but current methods to evaluate its presence in vivo (A? positron emission tomography imaging and quantification of A? proteins in the cerebrospinal fluid obtained) are not practically implementable as screening methods due to cost, availability and/or invasiveness nature. The continued development of the device and the design of a clinical study is an important milestone towards raising the required capital (possibly 15-20M$) to reach the AD diagnostic market. The project will train individuals who are specialized in medical device development, clinical trial design and coordination, project management and regulatory affairs.

Machine Learning and Data Mining Approaches for Smart Buildings

The goal of this project is to develop machine learning and data mining algorithms relying on non-intrusive common sensor data to estimate and predict smart buildings’ occupancy and activities. Efficient feedbacks are automatically supplied to the end user to involve occupants and increase their awareness about energy systems. This consists of generating reports helping the occupant to understand his/her energy management system and thus to be involved in the decision-making process. The project directly aligns with Ericsson’s IoT mission and expands its operator potential opportunities by exploring a dimensionality in real-time automation, monitoring and tracking, and smart surveillance. It also further strengthens Ericsson’s position in the IoT market which has a projected additional revenue potential of up to 36% (USD 619 billion) by 2026. Moreover, the machine learning techniques to be developed can be easily adapted to other problems of interest to Ericsson.

Optimisation de la qualité des images et visualisation des tissus mous en Cone Beam Computed Tomography (CBCT)

L’imagerie cone beam CT (CBCT) prend une place de plus en plus importante dans l’exploration des structures osseuses du complexe dento-maxillo-facial. Cette modalité d’imagerie, moins irradiante que l’imagerie par tomodensitométrie conventionnelle (TDM), procure des images de qualité similaire, voire supérieure, à celle des images TDM. Toutefois, les rapports Signal/Bruit ne sont pas assez élevés, et les doses, bien qu’inférieures à celles en TDM, pourraient encore être diminuées. Un des objectifs de ce projet est donc d’optimiser la qualité des images tout en cherchant à réduire davantage la dose de radiation ionisante. Pour atteindre ce double objectif, nous proposons de mettre en œuvre de nouveaux algorithmes de reconstruction/rétroprojection, notamment en faisant appel à des méthodes de reconstruction itératives. Le rehaussement qualitatif obtenu, associé à une prise en charge des propriétés spectrales du processus d’acquisition, devrait nous permettre d’accéder, pour la première fois en imagerie CBCT, à une visualisation nettement améliorée des tissus mous, laissant entrevoir ainsi de très larges perspectives pour cette modalité d’imagerie très prometteuse.

Personalizing gamification strategies to improve young adult life skills training. Part 2; support worker needs analysis and design of clinical interface

Facing Dragons is a mobile health game that gives quests and challenges to young adults (17-29) to help them figure out what to do with their lives, gain an empowered sense of self and unlock their purpose in life.

One of the key research innovations in this game is the addition of a special interface for support workers and caregivers that helps them get to know the young adults they work with.
In this project we will match young adult mobile gameplay preferences with game design elements in order to teach life skills in a fun and engaging way that maximizes the motivational power of gamification. We will also be creating a useful clinical interface to assist those who help young adults face their dragons every day.
This project is an important next step in the design of a mobile life coaching game that empowers young adults through playful learning.

Assessing the effectiveness of customer management efforts on profitability in the insurance industry

To have a strategic advantage over competitors, companies have been encouraged to adopt customer-centric, value added processes and capabilities. Firms allocate resources to train their employees in the necessary skills to build and maintain healthy relationships with their customers, yet little is understood on how investments in training impacts a firm’s performance. The objective of the proposed research is to investigate (1) Which customer management training activities have a positive impact on profitability? (2) How frequently should companies offer training to their employees? and (3) Who will benefit from more (less) training activities? To answer these questions, we will develop a model to estimate growth in profitability as a function of training efforts while controlling for economic factors. This research will help understand how investments in customer management training will enhance the firm’s overall performance and competitiveness in business markets.

Modelisation et tarification des couts en assurances dommages

La modelisation des couts par assure en assurances dommages constitue la base du processus de tarification. Les couts peuvent provenir des trois couvertures d’un contrat d’assurance automobile dommages materiels au vehicule,dommages corporels subits par I’assure et dommages corporels subits par les tierces parties. Bien qu’un accidentpeut engendrer des couts pour ces trois couvertures, la modelisation actuelle de ces couts est faite separement en supposant I’independance entre les experiences des trois couvertures.

et ignore la decomposition classique des couts en assurance automobile en fonction de la frequence (nombre de sinistres) et de la seve rite (montant d’un sinistre). L’objectif principal du projet est d’ameliorer la tarification des couts pour un assure en assurance automobile en Ontario en tenant compte des recents developpements en actuarial. L’approche proposee permettra de tarifer les produits d’assurance qui refleteront davantage Ie risque inherent aux assures et seront plus equitables.

Rapid assessment of decision biases using reach-decision tasks in web-based applications

Imagine being asked two questions during a job interview: 1) Are you more collaborative or more individual? 2) Would you prefer working from home or working in the office? Now imagine that you feel strongly that you are collaborative, and slightly prefer working from home. An interviewer might look at those two responses and feel they are contradictory. However, if they knew that you were more indecisive about working at home, it would make more sense. Here, we propose to use movement dynamics recoded via mobile apps to provide this more detailed decision information. For Paradigm, our partner organization that specializes in assessment, this new assessment platform will mean they can more easily access this rich decision information, resulting in better information collected from more people in less time.

Hydrogeochemical investigation of elevated geogenic uranium in a subarctic region

Northern Canada faces environmental changes from growing resource extraction and global warming, which make an understanding of baseline conditions critical. In the Dawson Range, Yukon, naturally elevated concentrations of uranium have recently been discovered in groundwater at levels that exceed federal water-quality guidelines. This region is also the focus of advanced mineral exploration and falls within traditional territories of several First Nations. Mining activities may enhance uranium mobilization through the generation of waste rock and tailings. Thawing of permafrost might cause a similar effect by altering hydrological and geochemical conditions in groundwater. This project’s goal is to understand and communicate the baseline controls on uranium mobilization in the Dawson Range through analyses of water and rock samples and involvement with local industry, First Nations, and government.

Towards an Intelligent and Secure 5G Ecosystem for the Transformation and Digitalization of Societies Through Artificial Intelligence

Artificial intelligence (AI) has transformed our way of perceiving and interacting with technology, by providing state-of-the-art solutions for challenging problems across the tech-spectrum. The main objective of this cluster of projects is to investigate, develop, adapt, integrate and evaluate state-of-the-art machine learning (ML) techniques, which are suitable for modeling and prediction using datasets collected for complex real-world telecommunications applications. Given the applications of interest for Ericsson Inc., we will focus on ML techniques:
1. to process complex operational data (time series or high dimensional) from real time large-scale wireless and IoT networks;
2. to enable intelligent decision making and data sharing and provenance, and modeling using technologies, such as blockchain, that can scale for real-time systems;
3. for lifecycle management of operating 4G and 5G wireless networks, by addressing the need for long-term deployment, self-profiling, and anomaly detection; and
4. to augment human-computer interactions for real-time decision in support of operation and management of large-scale industrial systems.
Training ML models in such cases typically leads to complex optimization problems, using massive amounts of noisy and incomplete training data.

Intégration des technologies et thérapies associés aux soins de santé personnalisés (SSP)

Le secteur des soins de santé personnalisés (SSP) soulève de nombreux défis en matière d’adhésion, d’intégration et de réglementation. Il est alors nécessaire de faire l’état des défis et des enjeux concernant les SSP dans le système de santé.
Lorsqu’on parle d’intégration ou d’adhésion des SSP, on fait référence à leur utilisation par les professionnels de la santé et plus particulièrement par les médecins. Pour qu’une nouvelle technologie ou thérapie soit intégrée au système de santé, il faut qu’elle soit bien connue et bien comprise par les médecins. Ainsi, ils seront en mesure de l’utiliser en pratique courante. Cette intégration et cette adhésion sont d’autant plus facilitées par l’obtention du remboursement par les assureurs privés et par les assureurs publics. Le secteur public se fait via le régime d’assurance médicament du Québec (RAMQ). Pour obtenir le remboursement d’une nouvelle thérapie ou technologie, des agences d’évaluations des technologies doivent évaluer la valeur clinique et économique d’un produit. Ces agences se basent sur certains critères d’évaluation. Ils émettent ensuite une recommandation au ministre de la Santé et des services sociaux (MSSS) à savoir si oui ou non l’innovation devrait être remboursée par la RAMQ.

Transparent and Trustworthy Deep Feature Learning for Cyber-Physical System Security

The latest artificial intelligence (AI) technologies have effectively leveraged the wealth of data from cyber-physical systems (CPSs) to automate intelligent decisions. However, for safety-critical CPS like smart grids and smart cities, the conversion of massive data into actionable information by the AI must be not only effective but also reliable. To this end, this project will develop innovative feature learning methods that can distill raw spatiotemporal data, integrate with establish expert knowledge and system models, and present decision-supporting information with transparency and trustworthiness. With a focus on security monitoring applications in the safety-critical CPS, new scientific tools and practice guides developed by the project will benefit the research and development of AI-based & 5G-enabled CPS products and solutions for Ericsson while enhancing the smart infrastructure security for the general public of Canada.

Data driven energy efficient base station sleep control for 5G systems

The objective of this project is to develop a software system which can optimally control the base station sleep states in 5G networks to save energy. The 5G wireless networks are required to be green and yield very low carbon dioxide emissions. Compared with that of 4G wireless networks, the power efficiency of 5G is expected to be increased to 100-fold. High energy efficiency is a critical requirement in 5G network design and operation. We propose station sleep strategies based on machine learning, stochastic programming and robust optimization models which, by leveraging demand patterns learned from historical load data, provide statistically optimal energy efficiency and delay-bounded QoS.