Typical 7T Laminar fMRI Data Processing Pipeline
Typical 7T Laminar fMRI Data Processing Pipeline
1. Overview
High-resolution 7T laminar fMRI requires a processing strategy that is different from a conventional whole-brain fMRI pipeline. The functional data often have sub-millimeter resolution, limited or zoomed field of view, strong EPI distortions, and only a few voxels across cortical thickness.
A common strategy across layer-fMRI workflows is:
Use the native functional EPI grid as the primary analysis space, reconstruct the anatomy and cortical surfaces in high-resolution anatomical space, estimate a robust functional-to-anatomical registration, and transfer anatomical priors such as cortical surfaces, segmentations, and ROIs into the functional space.
This approach is particularly attractive when the goal is to preserve the original functional sampling geometry and minimize unnecessary interpolation of the functional time series.
LayNii was explicitly designed for layer-fMRI analysis in the native voxel space of functional NIfTI data and is intended to work alongside other tools such as AFNI, ANTs, and FreeSurfer rather than replace them [1]. AFNI’s align_epi_anat.py similarly emphasizes combining transformations to minimize the amount of interpolation applied to the data [2].
The exact implementation should be adapted to the acquisition, distortion-correction strategy, anatomical contrast, and research question.
2. Core Design Principles
2.1 Native EPI space as the functional analysis space
The preferred conceptual organization is:
High-resolution T1w / MP2RAGE
|
v
FreeSurfer reconstruction
|
+---- cortical surfaces
+---- segmentation
+---- anatomical parcellation
+---- V1/V2 or other ROI
|
v
EPI <-> anatomy registration
|
v
Transfer anatomical information
into native EPI space
|
v
LayNii / AFNI layer analysisThe functional time series should not be repeatedly resampled merely to make it convenient to overlay with anatomical images.
Instead:
EPI grid
^
|
| anatomical information is transformed
|
T1w / surfaces / segmentation / ROIThis does not mean that EPI interpolation must never occur. Motion correction, susceptibility-distortion correction, and other preprocessing operations may require resampling. The practical goal is to:
- minimize the number of interpolations;
- combine compatible transforms when possible;
- avoid repeated interpolation of the functional time series;
- keep a clearly defined final functional reference grid.
AFNI explicitly supports concatenating transformations to reduce interpolation [2].
3. Typical Data Organization
A useful conceptual separation is:
raw/
├── functional/
│ ├── localizer/
│ ├── task/
│ └── fieldmaps_or_distortion_reference/
│
└── anatomical/
├── MP2RAGE/
└── other_structural_contrasts/
derivatives/
├── afni/
├── freesurfer/
├── registration/
├── laynii/
└── qc/The exact BIDS derivative naming convention can be adapted to the project.
Important reference datasets should be explicitly tracked:
- raw EPI;
- motion-corrected EPI;
- distortion-corrected EPI, if applicable;
- mean/reference EPI;
- high-resolution T1w or MP2RAGE;
- FreeSurfer subject directory;
- cortical surfaces;
- anatomical segmentation;
- ROI masks;
- registration transforms;
- EPI-space anatomical masks and ROIs;
- layer/depth maps;
- final layer-resolved functional data.
4. Stage 1 — Convert and Inspect the Raw Data
4.1 Convert DICOM to NIfTI
Convert the scanner data to NIfTI while preserving the relevant orientation and geometry information.
Immediately inspect:
- voxel dimensions;
- matrix dimensions;
- number of slices;
- slice orientation;
- obliquity;
- qform/sform;
- field of view;
- partial-FOV coverage;
- temporal dimension;
- TR and TE;
- phase-encoding direction;
- distortion-correction metadata.
For a layer-fMRI dataset, the geometry of the EPI is especially important because a small change in voxel grid or orientation can affect subsequent cortical-depth analysis.
4.2 Establish the reference EPI
A representative EPI volume or mean EPI can be used as the registration reference.
For example:
task/localizer EPI time series
|
v
motion correction
|
v
representative
mean EPIThe mean EPI should remain on the intended functional grid.
5. Stage 2 — Functional Preprocessing
A typical functional preprocessing sequence may include:
Raw EPI
|
+--> discard unstable initial volumes
|
+--> slice-timing correction (if required)
|
+--> motion / volume registration
|
+--> susceptibility-distortion correction (if available)
|
+--> intensity / artifact QC
|
+--> representative mean/reference EPIThe exact order depends on the acquisition and the chosen preprocessing framework.
5.1 Motion correction
Motion correction is usually necessary for time-series analysis.
However, motion correction itself may involve resampling.
Therefore, for layer-fMRI, a useful strategy is to:
- estimate motion relative to a carefully selected reference;
- keep transformation matrices;
- avoid repeatedly resampling the same time series;
- combine transformations where technically appropriate.
5.2 Susceptibility-distortion correction
If field maps, reversed phase-encoding data, or another distortion model are available, distortion correction may be highly valuable.
For layer-fMRI, distortion correction is particularly important because the cortical ribbon is thin and susceptibility artifacts can cause substantial geometric displacement.
However, the correction strategy must be considered together with the anatomical registration strategy. Applying multiple independent spatial transformations sequentially can increase interpolation and potentially alter the effective spatial point-spread function.
6. Stage 3 — High-Resolution Anatomical Processing
6.1 T1w / MP2RAGE preprocessing
A high-resolution MP2RAGE or T1-weighted image is typically used as the anatomical foundation.
Possible preprocessing steps include:
- MP2RAGE reconstruction;
- denoising;
- bias-field correction;
- skull stripping;
- brain masking;
- manual QC;
- optional additional contrasts for pial-boundary optimization.
6.2 FreeSurfer reconstruction
The high-resolution anatomical image is then processed with FreeSurfer.
Conceptually:
High-resolution T1w / MP2RAGE
|
v
recon-all
|
+------+------+
| |
v v
segmentation surfaces
| |
| +-----+------+
| | |
v v v
WM white pial
surface surfaceA published 7T laminar fMRI study used a high-resolution MP2RAGE as input to high-resolution FreeSurfer reconstruction, manually curated segmentation in the occipital cortex, and used a participant-specific V1/V2 ROI definition [3].
For laminar work, automated segmentation should always be treated as a starting point rather than an unquestionable ground truth.
7. Stage 4 — Anatomical QC and Manual Correction
High-resolution layer-fMRI is particularly sensitive to errors in the cortical boundaries.
Recommended QC includes:
- inspect the WM/GM boundary;
- inspect the pial surface;
- inspect deep sulci;
- inspect the occipital cortex;
- inspect V1/V2 boundaries;
- check for topology errors;
- verify that surfaces follow the actual cortical ribbon.
Useful tools include:
- FreeSurfer / Freeview;
- ITK-SNAP;
- AFNI/SUMA;
- other surface visualization tools.
Manual correction may be justified when an automatic segmentation produces a systematic cortical-boundary error.
A published 7T laminar fMRI study specifically reported manual curation of the occipital segmentation after high-resolution FreeSurfer reconstruction [3].
8. Stage 5 — Define Anatomical ROIs
The ROI can be defined using anatomical or functional information.
Examples:
- V1;
- V2;
- V3;
- retinotopic maps;
- task-defined cortical patches;
- manually delineated regions.
A typical approach is:
FreeSurfer reconstruction
|
v
cortical parcellation / atlas
|
v
participant-specific ROI
|
v
V1 / V2 / target cortical patchFor visual laminar fMRI, participant-specific retinotopic information can provide a stronger ROI definition than relying only on a generic anatomical atlas.
The important point is that the ROI is defined using anatomical or functional priors before it is transferred to the final functional grid.
9. Stage 6 — EPI-to-Anatomical Registration
This is one of the most critical stages for partial-FOV 7T laminar fMRI.
The two images often have very different properties:
T1w / MP2RAGE EPI
------------------ ----------------
whole brain partial FOV
high anatomical contrast low anatomical contrast
high resolution lower resolution
little distortion EPI distortion
different geometry oblique geometryTherefore, registration should be treated as a dedicated high-QC step rather than a routine whole-brain fMRI alignment.
10. Option A — FreeSurfer bbregister
bbregister is a boundary-based registration tool designed to register functional images to a FreeSurfer anatomical subject.
A typical conceptual command is:
bbregister \
--s <subject> \
--mov <mean_epi.nii.gz> \
--bold \
--reg <registration.dat> \
--lta <registration.lta>The important output is the transformation describing the relationship between functional EPI and FreeSurfer anatomy.
For partial-FOV EPI, FreeSurfer documentation explicitly describes using a whole-FOV image as an intermediate registration image when needed [4].
This is relevant when the partial EPI does not contain enough anatomical context for a robust direct registration.
11. Option B — AFNI align_epi_anat.py
AFNI provides align_epi_anat.py for EPI/anatomical alignment.
A major advantage for a native-EPI workflow is that AFNI can align the anatomical dataset to the EPI dataset and save the transformation, rather than requiring the functional dataset to be resampled into anatomical space [2].
Conceptually:
T1w
|
| 6-DOF / cross-modal registration
|
v
EPI gridAFNI also combines transformations where possible, reducing repeated interpolation [2].
For partial-FOV data, however, the registration parameters and cost function need to be chosen carefully. AFNI documentation notes that partial data can make center-of-mass initialization problematic, and the choice of cost function should reflect whether the datasets are cross-modal or same-modality [2].
12. Option C — ANTs + ITK-SNAP
For difficult partial-FOV cases, a high-quality registration workflow can combine:
ANTs
+
ITK-SNAPANTs can be used for registration and transform estimation, while ITK-SNAP provides interactive visualization and manual quality control.
This can be particularly useful when:
- the EPI has severe distortion;
- the FOV is very small;
- automated registration is unstable;
- the anatomical boundary is difficult to identify;
- manual refinement is needed.
LayNii’s documentation specifically points users toward ITK-SNAP/ANTs and AFNI for partial-coverage alignment rather than implementing the alignment itself [1].
13. Choosing the Registration Direction
For the workflow described here, the preferred conceptual direction is:
Anatomical information
|
| transform
v
Native EPI spacerather than:
EPI time series
|
| repeated resampling
v
T1 anatomical spaceThis means that the EPI grid becomes the master grid for the final functional analysis.
For example:
EPI grid
|
+-- T1 information
+-- WM mask
+-- GM ribbon
+-- V1 ROI
+-- V2 ROI
+-- cortical surfaces
+-- layer/depth information14. Stage 7 — Transfer Anatomical Information to EPI Space
Once a reliable EPI/anatomical transform has been established, apply it to anatomical derivatives.
Typical outputs include:
T1w aligned to EPI
brain mask aligned to EPI
WM mask aligned to EPI
GM ribbon aligned to EPI
V1 aligned to EPI
V2 aligned to EPI
other ROI masks aligned to EPIFor continuous intensity images:
T1w
|
+--> linear interpolationFor discrete labels:
WM mask
GM mask
V1
V2
segmentation
|
+--> nearest-neighbour interpolationThe interpolation choice is important because a categorical label should not be converted into arbitrary intermediate values.
15. Stage 8 — Surface-to-Volume and Layer Construction
Once the anatomical information is related to the EPI grid, the next goal is to define cortical depth.
The basic geometry is:
Pial surface
|
| superficial cortex
|
| middle cortex
|
| deep cortex
|
v
WM surfaceA layer coordinate can be defined from the WM boundary toward the pial boundary.
Different methods exist, including:
- equidistant depth;
- equivolume approaches;
- surface-based interpolation;
- voxel-based layer assignment.
The appropriate method depends on the research question and cortical geometry.
16. LayNii
LayNii was designed specifically for layer-fMRI and operates in the voxel space of NIfTI data [1].
Its functions include:
- cortical layer assignment;
- columnarization;
- layer-specific smoothing;
- quality assessment;
- model-based venous signal mitigation;
- VASO-related processing.
A key feature for the present workflow is that LayNii can perform layerification in the native functional voxel space [1].
This makes the following conceptual workflow possible:
Native EPI
+
WM / pial anatomical information
|
v
LayNii
|
v
layer / depth maps
|
v
layer-resolved functional analysisLayNii does not attempt to replace all registration software. Its documentation explicitly states that partial-coverage alignment is outside its core functionality and refers users to AFNI, ANTs/ITK-SNAP, and related approaches [1].
17. AFNI + SUMA / SurfLayers
AFNI and SUMA provide another route for surface-based laminar analysis.
A representative AFNI SurfLayers workflow uses:
FreeSurfer recon-all
|
v
@SUMA_Make_Spec_FS
|
v
SUMA surfaces
|
v
layered surfaces / surface visualizationAFNI’s SurfLayers demonstration includes real 7T laminar datasets with MP2RAGE anatomy and zoomed accelerated GRASE functional data [5].
This demonstrates a complementary approach:
- FreeSurfer: anatomical reconstruction;
- SUMA: surface representation and visualization;
- AFNI: functional preprocessing and analysis;
- SurfLayers: layered surface representation.
18. Example Native-EPI Workflow
A practical implementation can therefore look like:
RAW DATA
|
+--------------+--------------+
| |
v v
7T EPI MP2RAGE/T1w
| |
v v
AFNI preprocessing anatomical QC
| |
v v
mean/reference EPI FreeSurfer
| |
| +-------+--------+
| | |
| v v
| surfaces segmentation
| | |
| +-------+--------+
| |
| v
| V1 / V2 ROI
| |
+-------------+---------------+
|
v
EPI <-> anatomy
registration
bbregister / AFNI / ANTs
|
v
registration QC
|
v
+-------------+--------------+
| | |
v v v
T1 masks ROIs
| | |
+-------------+--------------+
|
v
map anatomy to EPI
native grid
|
v
LayNii / AFNI
|
v
cortical depth /
layer maps
|
v
layer-resolved fMRI
time series
|
v
statistical
analysis19. Interpolation Strategy
The key rule should be:
Do not optimize the pipeline for the fewest commands; optimize it for the fewest unnecessary resampling operations on the functional data.
A useful conceptual table is:
| Operation | Potential interpolation | Recommended strategy |
|---|---|---|
| Initial volume selection | No | Do not resample |
| Slice timing correction | Usually temporal, not spatial | Keep separate from spatial transforms |
| Motion correction | Yes | Estimate transforms and avoid repeated resampling |
| Distortion correction | Often yes | Combine with other spatial transforms when feasible |
| EPI-to-T1 registration | Not necessarily applied to EPI | Prefer mapping anatomy toward EPI when appropriate |
| T1-to-EPI mapping | Yes | One final resampling of anatomical derivatives |
| ROI-to-EPI mapping | Yes | Nearest-neighbour for discrete labels |
| Layer construction | Method-dependent | Keep layer maps tied to the native EPI grid |
| Final functional analysis | Ideally no additional spatial resampling | Analyze on the established functional grid |
This is consistent with AFNI’s stated goal of combining transformations to minimize interpolation [2].
20. Why the Mean EPI Is Useful
A mean EPI can provide a stable reference for registration:
EPI time series
|
v
motion correction / reference selection
|
v
mean EPI
|
v
EPI-T1 registrationThe mean EPI is typically easier to register than an arbitrary single volume because it has higher temporal averaging and a more stable intensity pattern.
The resulting transformation can then be applied conceptually to anatomical derivatives.
21. Partial-FOV Registration
Partial-FOV is one of the main differences between conventional fMRI registration and laminar fMRI registration.
A partial-FOV EPI may contain only:
occipital cortex
+
small amount of surrounding tissuewhile the T1 contains:
whole brainTherefore, naïve whole-brain registration can be unstable.
A useful strategy is:
Whole-brain anatomical reference
|
v
identify corresponding anatomy
|
v
partial-FOV EPI
|
v
robust rigid alignmentFreeSurfer’s bbregister documentation gives a specific example of using a whole-FOV image as an intermediate when registering a partial-FOV image [4].
AFNI also provides options that control initialization and master grids, and its documentation warns that center-of-mass initialization may be problematic for partial data [2].
22. Registration Quality Control
Registration should not be accepted solely because the software reports convergence.
At minimum, inspect:
EPI + T1
EPI + WM boundary
EPI + pial boundary
EPI + V1/V2Useful visualization tools include:
- Freeview;
- AFNI/SUMA;
- ITK-SNAP.
A particularly important QC question is:
Does the functional cortical ribbon overlap the anatomical cortical ribbon at the scale relevant to the laminar analysis?
A registration that looks acceptable at whole-brain scale may still be unacceptable for a 0.5–1.0 mm laminar experiment.
23. Layer-FMRI-Specific QC
After layer construction, inspect:
- cortical depth maps;
- WM and pial boundaries;
- number of functional voxels across cortex;
- continuity of layers through sulci;
- layer assignment in folded cortex;
- V1/V2 coverage;
- dropout and susceptibility artifacts;
- large vessels;
- signal voids;
- partial-FOV boundaries.
Do not interpret a smooth-looking layer map as evidence that the anatomical registration is correct.
24. Surface-Based vs Native-Volume Layer Analysis
There are two major conceptual approaches.
Surface-based
EPI
|
v
surface sampling
|
v
cortical surface
|
v
depth / layer profilesAdvantages:
- natural representation of cortical geometry;
- strong integration with FreeSurfer;
- useful for surface visualization and cross-subject analysis.
Potential concern:
- mapping functional voxels to surfaces may involve interpolation;
- the relationship between surface coordinates and native EPI voxels must be carefully controlled.
Native-volume / voxel-based
EPI native grid
|
v
WM/pial boundaries
|
v
layerification
|
v
layer labelsAdvantages:
- directly preserves the functional voxel grid;
- naturally fits LayNii’s design;
- convenient for restricted-FOV layer-fMRI;
- avoids forcing the functional data onto a conventional whole-brain grid.
For the specific goal of preserving the native EPI sampling geometry, the second approach is particularly attractive.
25. Recommended Division of Labor Between Tools
FreeSurfer
Best suited for:
- anatomical reconstruction;
- cortical segmentation;
- WM and pial surfaces;
- cortical parcellation;
- surface topology;
- anatomical ROI definition.
AFNI
Best suited for:
- functional preprocessing;
- motion correction;
- EPI reference generation;
- EPI/anatomical alignment;
- transformation management;
- volume-based analysis;
- SUMA integration;
- visualization and QC.
bbregister
Best suited for:
- boundary-based EPI-to-FreeSurfer anatomical registration;
- estimating a robust rigid transformation between functional and anatomical spaces.
ITK-SNAP
Best suited for:
- visual registration QC;
- manual segmentation;
- manual correction;
- difficult partial-FOV registration cases;
- ROI inspection.
LayNii
Best suited for:
- native-voxel layerification;
- cortical depth assignment;
- columnarization;
- layer-specific processing;
- layer-fMRI-specific QC;
- layer-dependent analysis.
26. Recommended Conceptual Pipeline
For a high-resolution 7T visual laminar-fMRI experiment, a robust conceptual workflow is:
7T RAW DATA
|
+--------------+--------------+
| |
v v
EPI data MP2RAGE/T1w
| |
v v
AFNI preprocessing structural QC
| |
v v
motion/distortion QC FreeSurfer
| |
v surfaces + labels
native EPI reference |
| |
| V1 / V2 ROI
| |
+-------------+---------------+
|
v
EPI-T1 registration
bbregister / AFNI / ANTs
|
v
registration QC
|
v
anatomical information -> EPI grid
|
+-------------+-------------+
| | |
v v v
WM/GM ROI surfaces
| | |
+-------------+-------------+
|
v
LayNii / AFNI
|
v
cortical depth map
|
v
layer assignment
|
v
layer-resolved BOLD
|
v
statistical analysis27. The Most Important Practical Rule
For this type of experiment, keep the following distinction explicit:
ANATOMICAL SPACE
|
| high-resolution anatomical reconstruction
| surfaces
| segmentation
| ROIs
|
| registration transform
v
FUNCTIONAL EPI SPACE
|
| native voxel grid
| original functional sampling
| layer maps
| functional time series
|
v
LAYER ANALYSISThe anatomical data are allowed to be resampled into the EPI grid.
The functional time series should be protected from unnecessary resampling.
28. How This Relates to a Typical FreeSurfer + AFNI + LayNii Pipeline
A practical implementation can therefore be divided into separate scripts:
01_convert_and_validate.sh
02_prepare_anatomical.sh
03_run_freesurfer.sh
04_prepare_epi_afni.sh
05_create_epi_reference.sh
06_register_epi_to_fs.sh
07_map_anatomy_to_epi.sh
08_create_layer_maps.sh
09_extract_layer_timeseries.sh
10_qc_all.shThis separation has an important advantage:
registration transforms are treated as explicit analysis products rather than hidden side effects of preprocessing.
For reproducibility, save:
- registration matrices;
- LTA files;
- AFNI
.aff12.1Dtransforms; - reference EPI;
- anatomical reference;
- ROI masks;
- layer maps;
- QC snapshots.
29. Relation to the Example bbregister Script
A representative implementation of this philosophy is a workflow that:
- uses a mean localizer EPI as the functional reference;
- runs
bbregister; - saves both
.datand.ltatransformations; - uses the LTA to map T1, masks, and V1/V2 into the EPI grid;
- uses trilinear interpolation for continuous T1 intensity;
- uses nearest-neighbour interpolation for masks and ROIs.
This is consistent with the native-EPI-space strategy described above.
The important conceptual distinction is:
Do not:
EPI -> T1 -> ROI -> EPI
Prefer:
T1 / surfaces / ROI
|
v
single well-defined transform
|
v
native EPI grid30. Industry/Community Common Logic
Although different laboratories use different software combinations, the common logic across representative workflows is remarkably consistent:
Common principle 1 — High-resolution anatomy comes first
FreeSurfer or an equivalent anatomical pipeline provides:
- cortical boundaries;
- surfaces;
- segmentation;
- ROI definitions.
Common principle 2 — Functional and anatomical spaces are explicitly linked
A dedicated EPI/anatomical registration is performed using tools such as:
- bbregister;
- AFNI;
- ANTs;
- ITK-SNAP-assisted workflows.
Common principle 3 — Partial-FOV requires special care
A standard whole-brain registration assumption may not work well for zoomed EPI.
Common principle 4 — Native functional geometry matters
Layer-fMRI software such as LayNii is explicitly designed to operate in native functional voxel space [1].
Common principle 5 — Interpolation should be minimized
AFNI explicitly supports combining transformations to reduce interpolation [2].
Common principle 6 — Anatomical boundaries require aggressive QC
Published 7T laminar studies report manual inspection or correction of anatomical segmentation, particularly in the cortical regions of interest [3].
Common principle 7 — Layer analysis should be performed only after registration and anatomical QC
The order is therefore approximately:
functional preprocessing
+
anatomical reconstruction
|
v
registration
|
v
registration QC
|
v
ROI / surface mapping
|
v
layer construction
|
v
layer-resolved analysis31. References and Representative Resources
[1] Huber et al. — LayNii: A software suite for layer-fMRI
Huber R, Poser BA, Bandettini PA, Arora K, et al. LayNii: A software suite for layer-fMRI. NeuroImage. 2021;237:118091.
Key relevance:
- native voxel-space processing;
- layerification;
- columnarization;
- layer-specific processing;
- partial-FOV-aware design philosophy.
https://pmc.ncbi.nlm.nih.gov/articles/PMC7615890/
[2] AFNI — align_epi_anat.py
AFNI documentation for EPI/anatomical registration.
Key relevance:
- EPI-to-anatomy or anatomy-to-EPI alignment;
- transformation concatenation;
- reduced interpolation;
- master-grid control;
- partial-data considerations.
https://afni.nimh.nih.gov/pub/dist/doc/program_help/align_epi_anat.py.html
[3] 7T laminar fMRI using arterial spin labeling
Representative 7T sub-millimeter laminar-fMRI workflow using:
- MP2RAGE;
- high-resolution FreeSurfer;
- manual cortical QC;
- V1/V2 ROIs;
- nearest-neighbour functional sampling;
- no surface/intracortical smoothing.
https://pmc.ncbi.nlm.nih.gov/articles/PMC8075193/
[4] FreeSurfer — bbregister
FreeSurfer documentation for boundary-based registration, including partial-FOV registration using a whole-FOV intermediate image.
https://freesurfer.net/fswiki/bbregister
[5] AFNI/SUMA SurfLayers
AFNI’s SurfLayers demonstration includes 7T laminar-fMRI examples using MP2RAGE, FreeSurfer, SUMA, and AFNI.
https://afni.nimh.nih.gov/pub/dist/doc/htmldoc/tutorials/surflayers/surflayers.html
[6] A protocol for ultra-high field laminar fMRI in the human brain
A practical overview of ultra-high-field laminar-fMRI preprocessing challenges, including functional/anatomical coregistration, FreeSurfer segmentation, and cortical-layer construction.
https://pmc.ncbi.nlm.nih.gov/articles/PMC8039727/
32. Summary
The most defensible general strategy for a 7T partial-FOV laminar-fMRI dataset is:
1. Preserve the original EPI geometry.
|
2. Preprocess the functional data carefully.
|
3. Reconstruct high-resolution anatomy independently.
|
4. Generate accurate WM/pial surfaces and anatomical ROIs.
|
5. Perform dedicated EPI-to-anatomy registration.
|
6. Perform rigorous registration QC.
|
7. Transfer anatomical information into the native EPI grid.
|
8. Construct cortical depth/layer maps in or tied to EPI space.
|
9. Extract layer-resolved functional signals.
|
10. Perform statistical and neuroscientific analysis.The central philosophy is:
Anatomy defines where the cortex and layers are; the native EPI defines where the functional measurements are. Registration establishes the correspondence between them, while minimizing unnecessary resampling of the functional data.
This is the key conceptual framework for designing a robust FreeSurfer + AFNI + ITK-SNAP + LayNii workflow for high-resolution 7T laminar fMRI.