CUIDS Distinguished Speaker Series Archive 2021-22

Saving Lives Through Visual Predictive Modelling
with Jay Woo and Lena Yam
Oct. 14, 2021, 11:00 a.m. – 12:00 p.m.

The core mission of CAA is to rescue stranded motorists and save lives. Protecting over 2.6 million Members, between 3,200 and 5,500 rescues are conducted each day. In order to maximize a finite set of rescue resources, CAA uses machine-learning algorithms to predict the precise location of motorist breakdowns. By knowing when and where breakdowns will occur, CAA is able to pre-position rescue vehicles at the right location at the right time to rescue the motorist. This lecture will provide an overview of the technology that CAA developed called Geo-Temporal Gen 2. It will also include a discussion of the mathematical and probability models that power the machine-learning algorithms. Accompanied by a live view into Geo-Temporal Gen 2, you will see mathematical and probability theory put into real-life action and witness the accuracy of the predictive models that are generated by Geo-Temporal Gen 2 to save lives on our nation’s roadways.

Watch this seminar now!

Building Digital Twins: Bringing AI into Facility Management
with Dr. Jenn McArthur
Nov. 4, 2021, 11:00 a.m. – 12:00 p.m.

Overview of the Smart Campus Integration Platform (SCIP) project (a 5-year, ~$2M NSERC Alliance Project) currently in its second year, including highlights of research from prior or related projects to give insight on its further development. This talk will cover the key steps in developing the SCIP digital twin, including the development of the federated data model, data streaming and IoT sensor integration, data analytics, preliminary event detection algorithms, and the Digital Twin integration in Autodesk Forge.

Watch this seminar now!

Modeling Clinical Outcome Variability in Duchenne Muscular Dystrophy
with Dr. Utkarsh Dang
Nov. 25, 2021, 11:00 a.m. – 12:00 p.m.

Duchenne muscular dystrophy (DMD) is a progressive disease caused by loss of dystrophin in muscle. While all patients share the primary gene and biochemical defect, there is considerable patient–patient variability in clinical symptoms at an early age as well as over time. Our objective was twofold: model early age clinical severity in steroid-naive DMD boys through circulating protein biomarkers, and infer clusters of similarly progressing boys over time. Hence, we developed multivariate models (cross-sectional) of serum protein biomarkers that explained observed variation, using functional outcome measures as proxies for severity. Furthermore, we investigated clustering clinical outcome trajectories via modeling a multivariate combination of early-age clinical outcome measurements simultaneously to better explain disease progression rates. We show that performance of DMD boys was effectively modeled with serum proteins, and that multiple groups of DMD motor trajectories progressing at a unique rate can be established at an early age.

Video will be available in early 2022