One-line: Single-subject block-design task-fMRI analysis of the classic SPM Auditory dataset, implemented independently in SPM25 and FSL FEAT for a reproducibility-oriented cross-software comparison.
- SPM25 first-level analysis: complete
- FSL FEAT first-level replication: complete
- Contrast:
Listening > Rest - Cross-software comparison: complete
- Dataset: SPM Auditory / MoAEpilot tutorial dataset
- Subject: sub-01
- Functional input: 84 analysis volumes
- TR: 7 s
- Design: block-design auditory stimulation versus rest
- Raw neuroimaging data are not committed to this repository
See DATA_SOURCES.md.
- Realignment and reslicing
- Slice-timing correction: 64 slices, descending
64:-1:1, reference slice 32 - T1-to-mean-EPI coregistration
- T1 segmentation and deformation estimation
- Functional normalization to MNI space at 3 mm isotropic resolution
- Spatial smoothing: 6 mm FWHM
- First-level GLM with the listening condition and 6 motion nuisance regressors
- Contrast: Listening > Rest
- Whole-brain voxel-level FWE inference
- Threshold: voxel-level FWE p < 0.05
- T threshold: 5.296299
- Extent threshold: 0 voxels
Largest SPM clusters:
| Cluster size | Peak T | MNI x | MNI y | MNI z |
|---|---|---|---|---|
| 403 | 11.8681 | -63 | -28 | 14 |
| 213 | 11.8743 | 57 | -22 | 11 |
- MCFLIRT motion correction
- Regular-descending slice-timing correction
- BET functional and structural brain extraction
- SUSAN smoothing: 6 mm FWHM
- High-pass filter: 128 s
- FILM first-level model with prewhitening
- Double-gamma HRF and the same block timings as the SPM model
- Six MCFLIRT motion nuisance EVs
- Functional-to-T1 FLIRT registration, 6 DOF
- T1-to-MNI152 FLIRT registration, 12 DOF
- Contrast: Listening > Rest
- Cluster inference: Z > 3.1, cluster significance p < 0.05
The two dominant FEAT clusters reproduced the expected bilateral auditory pattern:
| Cluster size | Z max | MNI x | MNI y | MNI z | Cluster p |
|---|---|---|---|---|---|
| 863 | 10.5 | -61.1 | -27.0 | 8.85 | 1.77e-35 |
| 718 | 10.3 | 58.7 | -22.8 | 5.68 | 2.85e-31 |
The full FSL cluster table is stored in results/tables/table2_fsl_clusters.txt.
Peak-coordinate correspondence is approximate because SPM MNI and FSL MNI152 normalized coordinates are not guaranteed to represent identical template geometry; no voxelwise overlap analysis is claimed.
This summary deliberately keeps the two inferential frameworks separate. SPM T statistics and voxel-level FWE results are not numerically equivalent to FEAT Z statistics and cluster-based inference.
Both software packages identified strong bilateral activation around superior temporal/auditory regions for Listening > Rest. The peak coordinates and dominant hemispheric pattern were similar, but the outputs are not numerically interchangeable because SPM and FEAT used different HRF implementations, temporal models, registration strategies, statistical images, and thresholding frameworks.
This project therefore demonstrates conceptual replication rather than exact voxelwise equivalence.
- SPM peaks:
results/tables/table1_spm_peaks.csv - SPM settings:
results/tables/spm_threshold_info.txt - FSL clusters:
results/tables/table2_fsl_clusters.txt - FSL settings:
results/tables/fsl_threshold_info.txt
- Environment:
env/TOOL_VERSIONS.md - Data provenance:
DATA_SOURCES.md - Mini-report:
reports/report.md - SPM settings:
results/tables/spm_threshold_info.txt - FSL settings:
results/tables/fsl_threshold_info.txt - SPM peak table:
results/tables/table1_spm_peaks.csv - FSL cluster table:
results/tables/table2_fsl_clusters.txt
The cross-audit verified the published numerical summaries against the lightweight result tables preserved in this repository. The original executable SPM and FEAT analysis scripts and complete derivative workspaces are not present in the audited archive. The repository therefore preserves completed analysis outputs and parameter provenance, but it is not presented as a fully executable end-to-end pipeline.
- Single-subject tutorial dataset
- No group-level inference
- SPM used voxel-level FWE inference, whereas FEAT used cluster-based inference; cluster sizes and significance values should not be compared directly
- FSL standard-space registration used affine FLIRT rather than nonlinear registration
- No atlas-based anatomical labeling was applied automatically
- Concept DOI: 10.5281/zenodo.17715105
- See
CITATION.cff