Comparison Mode¶

Comparison mode (-C [name]) writes all outputs into a named subdirectory within each subject/session's derivatives folder. This lets you run the pipeline with different preprocessing options and compare outputs side-by-side without overwriting each other.

How It Works¶

When -C [name] is passed:

Preprocessing: All output files are written to derivatives/dceprep-<name>/sub-##/ses-##/ instead of the standard derivatives/dceprep/ directory.
Analysis: If a comparison run's preprocessed data does not exist, essential files (T1 maps, brain masks, registration transforms) are copied from the standard run before analysis proceeds. This avoids redundant computation for steps that are unchanged between comparisons.

What gets copied vs. regenerated¶

Copied from standard run	Regenerated in comparison run
Registration transforms (`.mat` files)	DCE preprocessing (bias correction, z-norm)
Brain masks	Ktrans maps
T1 maps (if unchanged)	QC reports
Segmentation masks	Population report

Example: Comparing Motion Correction¶

Run the pipeline twice — once with motion correction, once without:

# Run 1: with motion correction
./preprocess_all.sh -d /data/rawdata -b -Z -m -C withMC
./DCE_all.sh -d /data/rawdata -C withMC

# Run 2: without motion correction
./preprocess_all.sh -d /data/rawdata -b -Z -C noMC
./DCE_all.sh -d /data/rawdata -C noMC

Outputs land in:

derivatives/
├── dceprep-withMC/
│   └── sub-01/ses-01/
│       ├── anat/
│       └── dce/
└── dceprep-noMC/
    └── sub-01/ses-01/
        ├── anat/
        └── dce/

Compare the Ktrans maps and population reports between the two runs to assess the impact of motion correction on your dataset.

Example: Comparing Z-Normalization¶

Test whether z-axis normalization improves inter-scanner consistency:

# Without z-normalization
./preprocess_all.sh -d /data/rawdata -b -C noZnorm
./DCE_all.sh -d /data/rawdata -C noZnorm

# With z-normalization
./preprocess_all.sh -d /data/rawdata -b -Z -C withZnorm
./DCE_all.sh -d /data/rawdata -C withZnorm

Compare the population reports: the z-normalized run should show reduced variance in Ktrans across subjects scanned on different machines.

Targeting Specific Subjects¶

Use -T alongside -C to run comparison mode on a subset of subjects:

./preprocess_all.sh -d /data/rawdata -b -Z -C withZ -T sub-01/ses-01

Notes¶

Comparison mode does not require a clean run (-c); both named runs can coexist.
The population report will aggregate whichever run is active — pass -C [name] to DCE_all.sh to report on a specific comparison.