MNI Normalization Usage

MRSIPrep registers MRSI maps to T1w space with ANTs and, optionally, normalizes them further into MNI space.

docker run --rm \
  -v /path/to/bids:/data:ro \
  -v /path/to/derivatives:/out \
  mrsiup/mrsiprep:cpu \
  /data /out participant \
  --participant-label S001 \
  --session-label V1 \
  --metabolites CrPCr,GluGln,GPCPCh,NAANAAG,Ins \
  --ref-met CrPCr \
  --mode mni-norm \
  --output-spaces MNI152NLin2009cAsym \
  --mni-resolution origres \
  --nthreads 16

--output-spaces (default MNI152NLin2009cAsym only) selects which space(s) the final MRSI maps are resampled into as permanent derivatives: MRSI, MNI152NLin2009cAsym (aliases mrsi, mni accepted). --mni-resolution selects the MNI template resolution used for both T1w→MNI registration and final resampling: origres (MRSI native resolution, default — avoids implying spatial precision the MRSI acquisition never had, and matches the resolution mrsiprep's own spatial-smoothness benchmark evaluates at), t1wres (T1w resolution), or an explicit <N>mm. In --longitudinal mode, the shared subject-template-to-MNI registration stage always uses t1wres regardless of this flag, since the template spans multiple sessions that may have different native MRSI resolutions and there is no single well-defined "native" choice there; per-session outputs are unaffected.

T1w-space resampling of every metabolite (+ CRLB/SNR/FWHM/spikemask) is opt-in via --output-mrsi-t1w, since nothing downstream (regional extraction, connectivity, metprofiles) consumes it — only the registration-overview QC report needs one reference-metabolite map in T1w space, and it generates that itself into --work-dir regardless of this flag, so the QC figure is always available even without --output-mrsi-t1w:

docker run --rm \
  -v /path/to/bids:/data:ro \
  -v /path/to/derivatives:/out \
  mrsiup/mrsiprep:cpu \
  /data /out participant \
  --participant-label S001 \
  --session-label V1 \
  --metabolites CrPCr,GluGln,GPCPCh,NAANAAG,Ins \
  --ref-met CrPCr \
  --mode mni-norm \
  --output-mrsi-t1w \
  --nthreads 16

With --output-mrsi-t1w, the full per-metabolite T1w-space maps are written to <out>/mrsiprep/sub-*/ses-*/mrsi/t1w/ as permanent derivatives.

CRLB, SNR, and FWHM(linewidth) maps for the configured --metabolites are transformed into T1w/MNI space alongside the signal maps whenever present; per-metabolite spike masks are only transformed if --transform-spikemask is passed (the combined QC mask stays in MRSI space).

ANTs-SyN normalization

--normalization controls the T1w→MNI strategy. The default simple path uses a fast affine fit; ants-syn runs a full deformable SyN registration for higher anatomical accuracy at additional compute cost:

docker run --rm \
  -v /path/to/bids:/data:ro \
  -v /path/to/derivatives:/out \
  mrsiup/mrsiprep:cpu \
  /data /out participant \
  --participant-label S001 \
  --session-label V1 \
  --metabolites CrPCr,GluGln,GPCPCh,NAANAAG,Ins \
  --ref-met CrPCr \
  --mode parc-con \
  --normalization ants-syn \
  --output-spaces MNI152NLin2009cAsym \
  --nthreads 16

existing reuses a precomputed transform instead of registering — useful when a T1w→MNI transform was already produced by a prior run or an external pipeline.

See Longitudinal (Subject-Template) Normalization for --longitudinal, which registers one unbiased ANTs template built across a subject's sessions to MNI instead of registering each session directly.

See Basic Usage for the full CLI reference, including --registration-backend, --normalization, --output-spaces, --output-mrsi-t1w, --mni-resolution, --ref-met, --registration-t1-target, --transform-spikemask, --overwrite-t1-reg, --overwrite-mni-reg, and --overwrite-transform.