Quality Control in preclinical MRI: Where do artifacts come from and how to fix them

Synopsis

This talk will introduce additional considerations for quality assurance in small animal MRI. For instance, we have anecdotally found that equipment for stabilising physiology under anaesthetic protocols can impact image quality, particularly air heaters. There are further considerations to make after the completion of each image processing step, as many software tools are designed for humans and have implicit voxel size/shape/contrast assumptions that are not valid for rodents.

The tutorial will highlight where in the processing pipeline these are most likely to arise, examples of these degraded results, and suggestions to mitigate them. Finally, we will frame these problems in the context of our experience extending MRIQC and fMRIPrep for rodents, to conclude with an interactive session where the participants will be challenged with identifying quality issues on a set of images that conform a “gallery of horrors,” which we will build by gathering examples showing artifacts and quality issues from in-house databases, open datasets, and examples provided by the community. The tutorial will demonstrate the utility of the visual reports MRIQC generates to identify quality issues on preclinical MRI data.

What is preclinical imaging?

Preclinical imaging (i.e., imaging of experimental animal models) bridges the gap between basic science and medical science by applying techniques from both fields in the same individual. For example, MRI can be combined with invasive procedures, such as optogenetic techniques, in a single rodent subject but not in human participants.

Why is preclinical imaging useful for conversations about quality?

In priniciple, preclinical MRI is equivalent to human MRI, which makes it very attractive as a translational technique.

However, in practice, preclinical MRI has historically trailed its human equivalent in some respects. This likely reflects its novelty relative to human imaging, but there are a number of factors including acquistion idiosyncrasies for a given study’s design, and the availability of software resources.

We will discuss how both of these factors can contribute to data quality, regardless of the target species. Following this, we will work through the stages of acquisitional and processing standard operating procedures to visualise what happens when things go wrong and how automated tools can objectively identify problematic images.

Learning outcomes

• Understand the similarities and differences between preclinical and human MRI

• Become familiar with the common sources of error in image acquisition and processing

• Understand the importance of visualisation for data quality

• Use different visualisation methods to identify problems in image quality