This competition on inference in complex physiological models is organised as part of the SofTMech Statistical Emulation and Translation Hub.
This competition aims to assess computational tools for accurate, robust and computationally efficient inference of unknown parameters in complex cardiovascular biophysical models from physiological data. We would expect most of the submitted techniques to be based on surrogate modelling and emulation. Two state-of-the-art computational fluid dynamics (CFD) models of the pulmonary and systemic circulation, respectively, will be provided. These models will serve as benchmark examples to assess your methods for fast model parameter inference and uncertainty quantification (UQ). To evaluate the performance of your method, synthetic data simulated from the aforementioned CFD models with pre-set parameter values will be also provided subject to additive observational noise. The true parameter values will be revealed after all participants have submitted their answers.
You will be invited to submit (i) estimates of the biophysical parameters, (ii) the corresponding “best” data prediction and (iii) a measure of estimation uncertainty (standard deviation). The assessment of your method will be based on accuracy and UQ in parameter and function space.
Team entries: entries from a team (without restrictions on the number of participants) are accepted; please note that there will only be one award to be shared by the team and reimbursement costs would only apply to one member of the team. Each Team should have a name and a Team Lead. When registering you will be asked for that information.
The competition consists of two stages.
Stage one: you will be given the opportunity to familiarise yourselves with the model and construct a statistical emulator for the quantity of interest well in advance of the data being released.
Stage two: after the data have been released, you will be given a limited time interval (one week) to conduct the parameter estimation and UQ analysis and submit your predictions. This limitation is with clinical translation in mind, to mimic clinical practice and decision support.
The participants with the best three entries for each model will be invited to disseminate their work in a one-day conference hosted at the University of Glasgow. The best three entries for each model will also receive a certificate and a financial award, details of which will be announced later. The final goal of the competition is a research publication in a high-impact journal with all six selected submissions, which will provide a platform to disseminate the statistical methods and results obtained. The independent assessment of your method by the competition organisers based on an objective gold standard will be an opportunity for you to promote your work and increase its visibility.
For full information please visit here