Bayesian Inference in MRI
Member-Initiated Symposium

ORGANIZERS: Matthew Cherukara, Paula Croal, Michael Germuska

Thursday, 16 May 2019

Session Number: MIS-19

Overview
A challenge across many fields of MRI is to extract meaningful results from uncertain data. Bayesian methods provide a robust and powerful statistical framework to perform inference and model comparison within this low signal to noise regime. The conceptually simple Bayes’ theorem allows for the principled incorporation of prior information in the generation of predictive statistics, increasing the stability of model fitting and enhancing the interpretability of results. Bayesian statistics are widespread and have found uses across the spectrum of MRI methods and applications; however, as a result, presentations of Bayesian methods are typically spread across multiple scientific sessions.

This symposium will highlight the current role of Bayesian inference in MRI, presenting an opportunity for increased engagement across disciplines and introducing emerging methods to a wider audience. It covers a range of Bayesian methods and their implementation in both brain and body imaging. Speakers will introduce both challenges and opportunities afforded by Bayesian statistics, in the context of cutting-edge research, along with scheduled discussion time. In relation to the wider meeting, this symposium aligns with the “Disruptors” plenary session, highlighting an alternative to the traditional frequentist approach to data analysis.

Target Audience
Scientists and clinicians interested in extracting meaningful information from noisy data.

Educational Objectives
As a result of attending this course, participants should be able to:
- Identify the basic principles of Bayesian inference and model selection;
- Describe current Bayesian approaches in MRI; and
- Assess the challenges and opportunities posed by Bayesian inference.