I am encountering a severe issue with SPM12 normalization and would appreciate any guidance.
Problem:
After running “Normalize: Write”, my functional images collapse into a thin diagonal stripe instead of showing whole-brain coverage. Most voxels are nearly zero, and only a small subset (~600–700 voxels) remain nonzero.
Details:
- Functional shape: 64 × 64 × 30 × 166
- TR: 2s
Observations:
- Pre-normalization images (after slice timing + realign) look normal
- Coregistration between T1 and mean functional looks correct (visually checked with Check Reg)
- Segmentation runs without errors and produces a deformation field (y_*.nii)
- Normalized output has correct MNI dimensions (e.g., 79 × 95 × 79), but:
- mean ≈ 0.6, std ≈ 21
- only ~679 voxels > 50
- appears as a thin plane/stripe in MNI space
What I have tried:
- Checked coregistration carefully
- Reoriented T1 and functional images (set origin near AC)
- Used default bounding box and voxel size
- Tested single-volume normalization
- Confirmed SPM mex files are not blocked (macOS issue resolved)
Still, the problem persists.
Question:
Has anyone encountered a case where normalization collapses into a stripe/line like this?
Could this be due to:
- origin/header mismatch between T1 and functional?
- deformation field misalignment?
- issues with the dataset’s spatial metadata?
Any suggestions or pointers would be greatly appreciated.
Thanks in advance!