Head motion
*Motion parameter traces across all runs in one scan session. Two different subjects. Left: really good statue. Right: really bad twitcher.
*FD: framewise displacement (quantifies displacement treating the brain as a sphere). Notice the spike. Keep in mind the "severity of the motion problem" is relative to the operating zone of your experiment (if your data are low-res (3mm), a fixed amount of displacement is less of a problem than if your data are high-res (e.g. 1mm)). Also, keep in mind that FD won't be exactly 0 just because of, say, thermal noise.
*Raw data, one volume from each run, over the course of scan session. Clearly there is massive head motion (subject's fault). However, the image quality is pretty good, pretty high SNR. Notice there is a lot of coil bias. If the motion was just "between-runs" head displacement, then this should be "fixable" through standard motion correction.
s
This corresponds to the data shown in the image below. Notice that there is a large drift in trans_y over the course of a run. Head motion that is slow drifts vs. head motion that is transient.
*Sub 1 was low motion, sub 3 was high motion. These are t-value statistical quantities.
[From Faruk - Part I] Within run head motion effects on 3D anatomical MRI data (0.25 mm iso., duration 10 minutes). Left: No motion. RIght: a lot of motion.
[From Faruk - Part I] Within run head motion effects on 3D anatomical MRI data (duration 10 minutes). Left: No motion. RIght: a lot of motion.
*[From Logan] This EPI data, (phase encode A/P, which is left right in this image) involved a participant moving their hand in front of their face and moving it back. This disrupted the B0 of the 3T magnet, and created nonlinear distortions along the phase encode direction which cannot be fixed by typical motion correction (logan) 😁
[From Logan] Related to above data - This is a the result of subtracting a "hand-up" field map from the "hand down" field map - the area of distortion seen above is the bright area on the lower left (logan).