Quality Control of Structural MRI Images Applied Using FreeSurfer-A Hands-On Workflow to Rate Motion Artifacts

Abstract

In structural magnetic resonance imaging motion artifacts are common, especially when not scanning healthy young adults. It has been shown that motion affects the analysis with automated image-processing techniques (e.g., FreeSurfer). This can bias results. Several developmental and adult studies have found reduced volume and thickness of gray matter due to motion artifacts. Thus, quality control is necessary in order to ensure an acceptable level of quality and to define exclusion criteria of images (i.e., determine participants with most severe artifacts). However, information about the quality control workflow and image exclusion procedure is largely lacking in the current literature and the existing rating systems differ. Here, we propose a stringent workflow of quality control steps during and after acquisition of T1-weighted images, which enables researchers dealing with populations that are typically affected by motion artifacts to enhance data quality and maximize sample sizes. As an underlying aim we established a thorough quality control rating system for T1-weighted images and applied it to the analysis of developmental clinical data using the automated processing pipeline FreeSurfer. This hands-on workflow and quality control rating system will aid researchers in minimizing motion artifacts in the final data set, and therefore enhance the quality of structural magnetic resonance imaging studies.

Keywords: attention-deficit/hyperactivity disorder (ADHD); head motion; quality control; rating system; structural MRI; volumetry.

Figure 1

QC workflow. Boxes in light blue represent QC steps using T1-weighted images and boxes in dark blue represent QC steps using processed images via automated processing pipeline. Exclude/Include: exclusion/inclusion of data set in further preprocessing and analysis.

Figure 2

Examples of the quality of T1-images according to rating categories. The images in category C1 (pass) are clear, no motion artifacts or ringing can be seen and subcortical structures as well as gray matter/white matter can be well differentiated. In category C2 (check) ringing can be seen but the contrast of the structures themselves is good and they can be differentiated. In category C3 (fail) ringing as well as motion artifacts (distortion) are present. The contrast is very poor and structures blend into each other.

Figure 3

Structural volume differences across QC rating categories for total gray matter (A), cortex (B), left amygdala (C), and white matter hypointensities (D). Significance values refer to post-hoc t-tests. ^**p < 0.01, ^*p < 0.05. Error bars denote SEM.