Motion of “Fast TR” sequences

Problem:

Subjects consistently show higher motion (as indexed by mean framewsie displancement (FD)) when scanned using MB1400 EPI sequence (TR = 1400, voxel size = 2mm isotropic) then MB645 EPI sequences (TR = 645ms, voxel size = 3 mm isotropic). We need to understand why this happens.

Things to do:

• [x] Plot Power's MeanFD for NKI-TRT data (See Figure 1.1)
• [x] Plot Power's MeanFD for NKI-RS enhance data (See Figure 2)
• [x] Find the subject who has the largest Mean FD difference between Rest_MB1400 and Rest_MB645 and plot their FDs. (See Figure 3)
• [x] Calculate and Plot Jenkinson's FD (Jenkinson, 2012) for the data above and plot them (See Figure 4)
• [x] Downsample the MB1400 data to 3mm voxel size and calculated FD again. See updated Figure (See Figure 1.1)
• [x] Correct 4288245's corrected data and plot the motion.
• [x] correct for physiologcial data and then do motion (On NKI-RS enhanced data) to see if this helps bring down motion (See Figure 5).
• [x] Plot cardiac, respotory noisce and FD for one subject (See Figure 6).

Results

Power's MeanFD for NKI-TRT data

Figure 1.1

Note:

1. The x-axis: From Left to right, the scans are ordered according to the order in which the participents revieving them
2. To demostrate the effect of voxel size on motion, the test data scanned by MB_1400 were downsampled to 3 mm isotropic before calculating FD. The brown box plot shows the mean FD of the downsampled data.

## Loading required package: plyr

## Attaching package: 'reshape'

## The following object(s) are masked from 'package:plyr':
## 
## rename, round_any

Figure 1.2

MD by subjects for MB1400 sequence

Power's MeanFD for NKI-RS Enhanced data

Figure 2

## Warning: Removed 24 rows containing non-finite values (stat_boxplot).

## Warning: Removed 24 rows containing missing values (geom_point).

FD of 4288245

The figure below shows the framwise displacement of subjects 4288245's resting state fMRI acquired with MB_1400 and MB_465 sequences. Subject 4288245 is from the enhanced NKI-RS dataset and has the largest mean FD difference between his/her MB_1400 and MB_645 scans.

Figure 3

Figure 4

The motion data are plotted here for resting state fMRI scans. The multiband data are cut in half (MB1400_5m, MB645_5m), down sampled with and without anti-aliasing filter (labeled as DS and DSNF, respectively).

Note: the anti-aliasing filter refers to temporal filter.

It seems that the data treated with anti-aliansing filter get the mean FD value lower to the same level as the CAP data. See below.

## Error: replacement has 0 rows, data has 470

Figure 5

Figure 5 shows the motion estimators for resting state fMRI data of 20 subjects from NKI-RS enhanced dataset. The motion estimators were generated on data corrected or uncorrected for respiratory and cardiac noise to see if the high motion in data scanned with MB_1400 sequeces are caused by aliasing with physiological noise. By visual inspection, we can see that correcting for physiological noise could not account for the high motion found in MB1400 dat.

• Data Source: 20 subjects from NKI-RS enhanced.
  
• Data type: Rest state fMRI scaned by using MB_1400, MB_645 and CAP sequences.
  
• Processing: 1. data was corrected for physiological noise by RETROICOR and then realigned to get motion estimators; 2. data was realigned without correction.
  
• Motion estimators: Power's mean FD and Jenkinson's FD.
  Note: RETROICOR was carried out by AFNI's 3dretroicor.
  

Figure 6

Subject 0112828 has the largest difference between mean FD in MB1400 scan and MB645 scan. In figure 6, this subject's mean FD, cardiac and respiratory noises during the resting state fMRI scan using MB 1400 sequence were plotted.

Figure 7

FD for the 20 subjects before you RETOICOR, and mean of them and the correlation between all of them.