Skip to content. | Skip to navigation


Personal tools

This is SunRain Plone Theme
You are here: Home / Training / 2017 / CREATE-MIA 2017 Retreat

CREATE-MIA 2017 Retreat

CREATE-MIA 2017 Retreat

  • CREATE-MIA Event
  • Student presentation
  • Seminar
  • Professional skills
When Oct 27, 2017
from 07:45 AM to 05:30 PM
Where McGill Gault Nature Reserve, Mont-Saint-Hilaire
Add event to calendar vCal

The CREATE-MIA Retreat will be held at the McGill Gault Nature Reserve in Mont-Saint-Hilaire ( A bus has been hired to transport the Montreal-based participants there and back. Please meet at 3480 University Ave. (in front of McConnell Engineering) at 7:45am for a  prompt 8:00am departure  At the end of the day, the bus will leave the Gault Estate at 4:00pm to return to McGill at approximately 5:00pm.

The day will start with research talks by CREATE-MIA trainees, be followed by an opportunity for trainees from the different labs to network informally and discuss their research, and finish with talks by two invited speakers, Prof. Hervé Lombaert and Mr. Andrew Doyle.

Trainee Research Presentations

"Connectivity of the motor strip", Jasmeen Sidhu, Nuclear Medicine and Radiobiology, U. Sherbrooke

"Bundle specific tractography: Enhanced fiber orientation from white matter fascicle priors", Francois Rheault, Computer Science, U. Sherbrooke

"MRI based 3D printed models for aligning histology to MRI data", Sethu Boopathy Jegathambal, Biomedical Engineering, McGill

"Liver tumor segmentation", Ozan Ciga, Computer Science, McGill

"Cartan frames for Heart Wall Fiber Motion", Babak Samari, Computer Science, McGill

"Improving the existing methodology of TCD", Shreyank Gupta, Electrical Engineering, ÉTS

"LATIS: Ultrasound imaging as a non invasive and versatile modality", Arnaud Brignol, Electrical Engineering, ÉTS

Invited talks

Predicting MS disease activity and drug responders based on a Bag of Lesions representation
Mr. Andrew Doyle, Centre for Integrative Neuroscience, McGill


Multiple Sclerosis is characterized by the formation of brain lesions, which present with wide variability and are markers of disease activity. This variability is poorly understood, and a neuroinformatics approach has been adopted to automatically determine lesion types, which are then used to represent patients with a Bag of Lesions histogram. Using this novel representation inspired by the Bag of Words model in text analysis, supervised classifiers can be trained to automatically predict future disease activity up to two years in the future for groups of patients in separate treatment arms of a clinical trial. By combining different models’ predictions, potential responders to two MS treatments are automatically identified with sensitivity of 92% and 94% and specificity of 82% and 84%, showing that this is a very promising approach towards personalized treatment for MS patients. 


Andrew Doyle is a hacker working as Research Software Developer at the McGill Centre for Integrative Neuroscience. After completing his Bachelor’s degree in Electrical Engineering at McGill University, he worked in image processing at the National Film Board of Canada before returning to McGill as a CREATE-MIA trainee to complete a Master’s in Professor Tal Arbel’s Probabilistic Vision Group. He is a strong believer in open source and open science, and enjoys interdisciplinary collaborations.

Spectral Matching & Learning of Surface Data - Example on Brain Surfaces
Prof. Hervé Lombaert, Department of Software and IT Engineering, ÉTS

How to analyze complex shapes, such as of the highly folded surface of the brain?  In this talk, I will show how spectral representations of shapes can benefit learning problems where data lives on surfaces.  Key operations, such as segmentation and registration, typically need a common mapping of surfaces, often obtained via slow and complex mesh deformations in a Euclidean space.  Here, we exploit spectral coordinates derived from the Laplacian eigenfunctions of shapes and also address the inherent instability of spectral shape decompositions.  Spectral coordinates have the advantage over Euclidean coordinates, to be geometry aware and to parameterize surfaces explicitly.  This change of paradigm, from Euclidean to spectral representations, enables a classifier to be applied *directly* on surface data, via spectral coordinates.  


The talk will focus, first, on spectral representations of shapes, with an example on brain surface matching, and second, on the learning of surface data, with an example on automatic brain surface parcellation.  


Hervé Lombaert is a Starting Research Scientist at Inria Sophia-Antipolis, France, and Associate Professor at ETS, Montreal - with research interests in Statistics on Shapes, Data & Medical Images.  He had the chance to work in multiple centers, including Microsoft Research (Cambridge, UK), Siemens Corporate Research (Princeton, NJ), Inria Sophia-Antipolis (France), McGill University (Canada), and Polytechnique Montreal (Canada)  -  more at []
« May 2019 »
Funded by NSERC

Funding provided by NSERC