Hello. I have a question for the resampling.
As background, we have a dot probe task with 3 groups (SR/DP/HC) and 3 (DL: death/life/neutral) X 3 (CON: congruent/incongruent/neutral) conditions. By using the SPM, we specified nifti files in each condition (like 5 nifti files in each individual because the neutral conditions are overlapped; deathXcongruent, deathXincongruent, lifeXcongruent, lifeXincongruent, and neutral).
After the preprocessing using SPM, I resampled the individual imaging files using the structure atlas (T1 in MNI_N27). In addition, for the 3dLMEr analysis, I changed format of the brainmask (brainmask.nii.gz in MNI_N27), from nifti to afni, and its atlas. Then, I ran the 3dLMEr model estimation, including age as covariate.
The code is shown below:
3dresample -input ffespmeeg_111_sdotprobe-f_1_t0_1000_f30_58_1.nii -master T1.nii.gz -prefix ffespmeeg_111_sdotprobe-f_1_t0_1000_f30_58_1_resampled.nii
3dcopy brainmask.nii.gz MNI_N27_mask
3drefit -view tlrc MNI_N27_mask_orig
3dLMEr -prefix XXX \
-resid XXX \
-model 'Group*Cond1*Cond2+Age+(1|Subj)+(1|Subj:Group)' \
-qVars 'Age' \
-SS_type 3 \
-bounds -2 2\
-gltCode Cong_Con_S 'Cond2 : 1*con Group : 1*SR Age :' \
-gltCode Cong_Con_D 'Cond2 : 1*con Group : 1*DP Age :' \
-gltCode Cong_Con_HC 'Cond2 : 1*con Group : 1*HC Age :' \
-gltCode Cong_INCon_S 'Cond2 : 1*incon Group : 1*SR Age :' \
-gltCode Cong_INCon_D 'Cond2 : 1*incon Group : 1*DP Age :' \
-gltCode Cong_INCon_HC 'Cond2 : 1*incon Group : 1*HC Age :' \
-gltCode Cong_Neu_S 'Cond2 : 1*neut Group : 1*SR Age :' \
-gltCode Cong_Neu_D 'Cond2 : 1*neut Group : 1*DP Age :' \
-gltCode Cong_Neu_HC 'Cond2 : 1*neut Group : 1*HC Age :' \
-gltCode DL_Death_S 'Cond1 : 1*death Group : 1*SR Age :' \
-gltCode DL_Death_D 'Cond1 : 1*death Group : 1*DP Age :' \
-gltCode DL_Death_HC 'Cond1 : 1*death Group : 1*HC Age :' \
-gltCode DL_Life_S 'Cond1 : 1*life Group : 1*SR Age :' \
-gltCode DL_Life_D 'Cond1 : 1*life Group : 1*DP Age :' \
-gltCode DL_Life_HC 'Cond1 : 1*life Group : 1*HC Age :' \
-mask MNI_N27_mask+tlrc \
-jobs 16 \
-dataTable \
Subj Group Cond1 Cond2 Age InputFile \
111 SR death con 29.2 ffespmeeg_111_sdotprobe-f_1_t0_1000_f30_58_1_resampled.nii \
111 SR life con 29.2 ffespmeeg_111_sdotprobe-f_1_t0_1000_f30_58_3_resampled.nii \
111 SR death incon 29.2 ffespmeeg_111_sdotprobe-f_1_t0_1000_f30_58_2_resampled.nii \
111 SR life incon 29.2 ffespmeeg_111_sdotprobe-f_1_t0_1000_f30_58_4_resampled.nii \
111 SR neut neut 29.2 ffespmeeg_111_sdotprobe-f_1_t0_1000_f30_58_5_resampled.nii \
112 SR death con 21.7 ffespmeeg_112_sdotprobe-f_1_t0_1000_f30_58_1_resampled.nii \
112 SR life con 21.7 ffespmeeg_112_sdotprobe-f_1_t0_1000_f30_58_3_resampled.nii \
112 SR death incon 21.7 ffespmeeg_112_sdotprobe-f_1_t0_1000_f30_58_2_resampled.nii \
112 SR life incon 21.7 ffespmeeg_112_sdotprobe-f_1_t0_1000_f30_58_4_resampled.nii \
112 SR neut neut 21.7 ffespmeeg_112_sdotprobe-f_1_t0_1000_f30_58_5_resampled.nii \
113 SR death con 23.6 ffespmeeg_113_sdotprobe-f_1_t0_1000_f30_58_1_resampled.nii \
113 SR life con 23.6 ffespmeeg_113_sdotprobe-f_1_t0_1000_f30_58_3_resampled.nii \
...
After the resampling, there was a huge computational demand, which is probably because the increased voxel frequencies in the resampled individual data.
My questions are:
- Do you think my code seems adequate?
- How can I deal with a huge computational overloading after the resampling?
- I am wondering what if I resample the reference file (e.g., T1.nii.gz) using the individual file (opposite approach) rather than resampling the individual files using the reference file. Would it be different?
I appreciate your consideration in advance.