Tristan Glatard

Tristan Glatard
McGill University
Montreal, Canada

Session: B. Informatics II: Computing systems

Will talk about: Web platforms for high-throughput neuroimaging analyses: state of the art and future directions

Bio sketch:

Tristan Glatard is research scientist at the French National Centre for Scientific Research (CNRS, CREATIS unit), currently a Visiting Scholar with Alan Evans' laboratory at McGill University. His research interests include neuroinformatics, grid and cloud computing, science gateways, and performance optimization of medical imaging applications. He has published 70+ peer-reviewed papers in international journals and conference proceedings, and has been involved in various large-scale initiatives such as the on-going France Life-Imaging infrastructure ( He contributed to several software projects, in particular the Virtual Imaging Platform (VIP - and CBRAIN (, two widely used web computing platforms. Tristan Glatard also coordinated the Life-Science Grid Community (, an organization representing life-science users of the European Grid Infrastructure, and the Biomed Virtual Organization, the corresponding technical entity.

Talk abstract:

Web platforms have been playing an important role in the transparent exploitation of distributed systems for high-throughput processing of large imaging datasets. Based on examples extracted from our experience with CBRAIN and the Virtual Imaging Platform (VIP), we will present their architecture and detail strategies to improve their performance. VIP is a web platform for medical data processing and simulation on the European Grid Infrastructure. It uses an agent-based, workflow-based task scheduling model and includes several optimizations for efficient execution of applications on heterogeneous distributed systems, in particular: optimal load-balancing of Monte-Carlo simulations and autonomic fault-tolerance and fairness control. CBRAIN is a collaborative web platform for high-throughput neuroimaging, developed at the McGill Center for Integrative Neuroscience, that was recently extended to support executions on clouds based on a cost-performance optimization algorithm for the deployment of virtual machines on multiple infrastructures. On-going and future developments related to the support of open and reproducible biomedical science will also be presented, in particular our current efforts towards interoperable processing platforms and reproducible neuroimaging analyses.