QA Pictures and Metrics From the QArunall SPM Module


This documentation contains information on the various quality assurance (QA)-related output created by the “QArunall” module for SPM. The QArunall module was written locally by John Graner and provides quickly-viewable pictures, movies, and plots linked by a single html file. It does not provide a QA “result” for each data set, relying on the user to interpret the information presented.

Below is a list of each element created by the module and linked to in the HTML file. Descriptions and examples are provided when relevant as well as things to look out for. Note: the word “picture” is used to refer to the visual presentation of data in the HTML file (these are just .png files); the word “image” is used to refer to the actual MRI data. An example QA HTML file is located in the documentation zip archive on Keoki in …/Graner/Data/SPM_Preprocessing_Things. After unzipping the archive, the summary file is .../QA_Summary/QA_summary_display.html.
.

Original Anatomical

Description -Three pictures of the original T1 anatomical image associated with the fMRI data. Each picture is the center slice of one of the three image dimensions (axial, sagittal, coronal). Note that the center image slice will most likely not correspond to the center of the brain. The brain may appear off-center and at a slight angle in each picture. However, the entire head should be contained within each of the three pictures.

QA Considerations – Make sure the brain image looks like a brain image. If any structural abnormalities are suggested by the pictures, open the original T1 image in a full image viewer (e.g. SPM’s “Display” module) to verify.

Processed Anatomical

Description - Three pictures of the T1 anatomical image after it has been through preprocessing. These pictures should contain only brain tissue; the skull and other surrounding tissue should have been removed. The brain should also appear to be centered and not rotated in the three pictures, as it should now be registered to a standard image template (MNI-152).

QA Considerations – Look for any areas around the edge of the brain where the skull-stripping failed and there is still non-brain tissue/bone in the image. The transformation into standard space involves a non-linear warp of the anatomical image. When this fails or goes awry portions of the image may get improperly warped and the brain will correspondingly look deformed. Look for any areas where the brain anatomy looks displaced, oddly shaped, or oddly sized.

Original fMRI

Description – Three pictures of the first volume of the original fMRI data. As with the original anatomical images, each picture should contain the brain (unless images were only acquired over a specific portion of the brain), but the brain might not be centered or aligned.

QA Considerations – Look for any spatial artifacts in the pictures, such as large regions of signal drop-out, large “protrusions” from the brain, etc. Note that these pictures are of the first volume of the original fMRI data before ANY preprocessing has been done, so they may show a pre-steady-state volume. In this case a slice-wise banding artifact (each slice has a different average signal value and is therefore distinctly visible on the two pictures not in the slice acquisition dimension) may be present but not indicative of bad data.

Preprocessed Functional

Description – Three pictures of the first volume of the preprocessed fMRI data. This version of the fMRI should be centered in the pictures and not rotated.

QA Considerations – Look for any spatial artifacts in the data (a remaining banding artifact, deformations of the brain, or places with “protrusions” from the brain). Double-check that the brain looks to be straight in each picture and that the whole brain is visible in each picture (and not cut off at an edge).

Anatomical Contours over Final fMRI

Description – Three pictures of the preprocessed fMRI with signal intensity contours of the preprocessed anatomical image overlayed in red. The red contours are created from the processed anatomical image and will usually (although not specifically) follow contours between CSF, grey matter, and white matter.

QA Considerations –The overall purpose of these pictures is to make sure the processed anatomical image and processed fMRI image are spatially aligned. Look to make sure the outline of the brain in the contours matches the outline of the brain in the fMRI image. Note that there may be some extension (maybe one or two voxels) of what appears to be the edge of the brain in the fMRI image beyond the superior edge of the contour. The corpus callosum should also be distinguishable as a contour in the sagittal picture. This contour should encapsulate a region of relatively decreased signal intensity in the underlaying fMRI data as well. The left and right ventricles should be similarly matched in both the contours and the fMRI data. If no B0 inhomogeneity correction was done there will most likely be some “leak” of fMRI signal beyond the anterior portion of the outer contour. Finally, the general outline of the gray matter should be visible in the contours and match with slightly higher signal intensities in the fMRI data. It is important for both the internal structures and the outer boarder of the brain to be aligned in these pictures.

fMRI Motion Plots: Rotation Parameters

Description – This plot shows the three rotation parameters used by SPM when motion-correcting each volume of the fMRI data. The ordinate has units of degrees (usually small) and the abscissa has units of volume number.

QA Considerations – This plot is rather straight-forward. Sharp, sudden increases or decreases in the plot represent TRs where the subject rotated his/her head suddenly. Gradual changes represent slower rotations over time. In many cases it is the sudden, sharp movements that may lead to issues. More information on the presence of these can be obtained from the other motion-related plots and movies, described below.

fMRI Motion Plots: Translation Parameters

Description – Very similar to the rotation parameter plot, this plot shows the three translation parameters used by SPM when motion-correcting each fMRI volume. The ordinate has units of millimeters.

QA Considerations – Interpretation of this plot is very similar to that of the rotation parameter plot. Make note of any sharp changes in parameters between TRs.

fMRI Motion Plots: Time Points to Censor

Description – This is a single number as well as a plot showing the TRs suggested for censoring when carrying out subsequent analysis of the fMRI data. The plot has values of either 1 or 0 for each image volume. A 0 indicates the presence of significant motion around that volume, suggesting it should be excluded from analyses. The number displayed after the text “Time Points to Censor” is simply the number of volumes labeled with “0” in the plot. A volume is suggested for exclusion if the total motion between it and the previous volume is above a certain threshold.

QA Considerations – This plot, the translation parameter plot, and the rotation parameter plot are displayed at the same size and x-scale in the HTML file to allow easy comparison between the timing of the volumes suggested for censoring and the timing of changes in the parameter plots. Make sure the volumes suggested for censoring correspond to sharp changes or peaks in the parameter plots; the three plots should sort of “tell a consistent story.”

fMRI Center Slice Movie

Description – This is a gif showing the center sagittal slice of the processed fMRI data at every TR. The white bar at the bottom of the movie represents how far through the scan time-course the current picture is.

QA Considerations – There are two main motivations for this movie. The first is to look for residual motion in the image that was not fully corrected in the motion correction step. This is often the case where there are many censored volumes or large jumps in the motion correction parameters; although some of the motion effects are removed the correction is usually not completely successful. Even if residual motion is visible in the movie (the brain appears to jump or jitter), however, it will not negatively impact analyses of the data so long as it occurs during time-points marked for censoring and censoring is in fact carried out during analysis. The second motivation behind this movie is to look for any temporal artifacts in the processed fMRI data. These can be slightly more insidious and less predictable. Possible things to watch out for include the sudden appearance of bands across the image in certain volumes, an intensity pattern that moves through the entire image as time progresses (e.g. “waves” of signal intensity moving through the brain), or the appearance of static noise inside or outside the brain in certain volumes. In general, look for any systematic or sudden change of signal that appears to be unrelated to the brain itself.

fMRI Variance Movie

Description – This is a gif showing the variance of each voxel time-course. The movie cycles through sagittal slices of the brain in the left-right direction (this could be left-to-right or right-to-left, depending on the data).

QA Considerations – The goal of this movie is to look for any systematic spatial patterns in the voxel variances across the image: straight lines of increased variance, a grid pattern throughout the movie, etc. The variances of voxels on the edge of the brain should be greater than those of voxels inside the brain (the variances of voxels on the edge of the brain will be larger due to a larger degree of signal change brought about by any motion which causes a voxel to shift from being in the brain to being outside the brain). This should manifest as a visible outline of the brain in the movie.