Welcome!#

Hands-on quality control of human and preclinical MRI data: from acquisition to post processing#

Summary#

Ensuring the quality of the interim results and outcomes at relevant check-points of the MRI research workflow, from acquisition to secondary analyses, is critical for the reliability of the findings. Traditionally, researchers have resorted to visualization by experts to flag and exclude subpar MRI images, and to trigger corrective actions (e.g., servicing the MRI scanner). However, the deluge of data that neuroimaging is witnessing hinders the traditional manual approach to quality control and assurance (QC/QA). On one hand, visualization tools that optimize the assessment task, minimize inter-rater disagreement, and allow easy and reliable bookkeeping of annotations are fundamental to allow fast assessment of large datasets. On the other hand, there is an urgent need to at least support, and potentially replace, the experts’ assessment with more objective and reliable, automated approaches. Finally, we will also show how the quality assessment is contingent on the downstream applications, which makes its definition and quantification a difficult problem. In this session, we expose MRI practitioners to a portfolio of strategies and tools that will enable them to more effectively determine when and how they can optimize their QA/QC protocols. Although the speakers’ experience is mostly centered around MRI of the human and rodent brain, the tutorial is of interest to practitioners working in other fields and species, e.g., knee or prostate MRI, or non-human primates.

Target audience#

Human and preclinical MRI practitioners of all experience levels who want to improve their current QA/QC strategy of their research workflow.

Learning outcomes

  • Introduction (Dhritiman Das): Define QA, QC, QI and IQM; Explain QC in the context of neuroimaging

  • Tutorial 1 (Joset Etzel) : Use standards operating procedures (SOPs) and implement best pipeline and reporting practices for quality assessment across different MRI datasets in the context of reproducible research.

  • Tutorial 2 (Oscar Esteban): Understand the problems of “scanner-effects”, how they affect image quality metrics automatically extracted from images, and how to validate the statistical modeling when these effects are present.

  • Tutorial 3 (Eilidh MacNicol): Adapting QA/QC procedures and protocols to preclinical imaging - MRIQC for rodents.

  • Tutorial 4 (Satrajit Ghosh, Fidel Alfaro-Almagro): Understand “less reported” quality control issues in MRI that may arise and, through a debate, be exposed to other efficient though non-conventional methods of integrating QC/QA in the research workflow.

  • Tutorial 5 (Céline Provins, Mikkel Schöttner and Michael Dayan): Analysing and interpreting the image quality metrics generated by MRIQC.