7  Visualization and reporting

ROI analysis

See also: Planning

If you were planning a ROI analysis, you will typically have one value per ROI, subject and condition. You can report descriptive statistics graphically using your favourite approach (violin plots, bar plots, boxplots, etc.).

Usually we also visualize ROIs at their locations.

Example

(Faber et al. 2022)

(Faber et al. 2022)

Run the segmentation

Deep-MI/FastSurfer: PyTorch implementation of FastSurferCNN

(Henschel et al. 2020)
# create the directory
mkdir /home/jovyan/2026_ANI/lobule9_bids_correct/derivatives/fastsurfer

# run the fastsurfer as a singularity container
singularity exec --nv \
                 --no-mount home,cwd -e \
                 -B $HOME/2026_ANI/lobule9_bids_correct:$HOME/my_mri_data \
                 -B $HOME/2026_ANI/lobule9_bids_correct/derivatives/fastsurfer:$HOME/my_fastsurfer_analysis \
                 -B $HOME/license.txt:$HOME/my_fs_license.txt \
                 /shared/singularity/fastsurfer-gpu.sif \
                   /fastsurfer/run_fastsurfer.sh \
                     --fs_license $HOME/my_fs_license.txt \
                     --t1 $HOME/my_mri_data/sub-001/ses-1/anat/sub-001_ses-1_acq-mprage_T1w.nii.gz \
                     --sid sub-001 --sd $HOME/my_fastsurfer_analysis \
                     --3T \
                     --threads 4 \
                     --seg_only

View output

right-click and open the file with freeview

set the color map to “Lookup Table”

Multiple comparison correction

Reminder: group analysis

Family-wise error rate

Image credit: unknown

If the data is not smoothed, equivalent to Bonferroni correction

Other correction types

- Cluster threshold

- False discovery rate

- Cluster-level inference

Uncorrected

# interactive plot (you can browse the activations)
from nilearn import plotting

# Use subject's anatomy as background
bg_img = '/home/jovyan/gambling/bids/derivatives/fmriprep/sub-001/ses-1/anat/sub-001_ses-1_acq-mprage_desc-preproc_T1w.nii.gz'
plotting.view_img(zmap, threshold=1.96, vmax=10, 
    bg_img=bg_img,
    cut_coords=[0, 0, 0],
    width_view=600,
    title=contrast_string)

Cluster-thresholded

from nilearn.glm import threshold_stats_img

# note: fpr with alpha=0.05 is the same as uncorrected z>=1.96 (p<0.05)
thresholded_map1, threshold1 = threshold_stats_img(
    zmap,
    alpha=0.05,
    height_control="fpr",
    cluster_threshold = 100,
    two_sided=True,
)

# Use subject's anatomy as background
bg_img = '/home/jovyan/gambling/bids/derivatives/fmriprep/sub-001/ses-1/anat/sub-001_ses-1_acq-mprage_desc-preproc_T1w.nii.gz'
plotting.view_img(thresholded_map1, threshold=threshold1, vmax=10, 
    bg_img=bg_img,
    cut_coords=[0, 0, 0],
    width_view=600,
    title=contrast_string)

FDR-thresholded

from nilearn.glm import threshold_stats_img

thresholded_map1, threshold1 = threshold_stats_img(
    zmap,
    alpha=0.05,
    height_control="fdr",
    two_sided=True,
)

# Use subject's anatomy as background
bg_img = '/home/jovyan/gambling/bids/derivatives/fmriprep/sub-001/ses-1/anat/sub-001_ses-1_acq-mprage_desc-preproc_T1w.nii.gz'
plotting.view_img(thresholded_map1, threshold=threshold1, vmax=10, 
    bg_img=bg_img,
    cut_coords=[0, 0, 0],
    width_view=600,
    title=contrast_string)

FWE-thresholded

from nilearn.glm import threshold_stats_img

thresholded_map1, threshold1 = threshold_stats_img(
    zmap,
    alpha=0.05,
    height_control="bonferroni",
    two_sided=True,
)

# Use subject's anatomy as background
bg_img = '/home/jovyan/gambling/bids/derivatives/fmriprep/sub-001/ses-1/anat/sub-001_ses-1_acq-mprage_desc-preproc_T1w.nii.gz'
plotting.view_img(thresholded_map1, threshold=threshold1, vmax=10, 
    bg_img=bg_img,
    cut_coords=[0, 0, 0],
    width_view=600,
    title=contrast_string)

Non-parametric inference

Permutation-based inference

(Nichols and Holmes 2001)

An example difference between conditions A and B. The labels of each trial are permuted and each time a test statistic (in this example mean difference) is computed. The empirical likelihood to observe a specific value is derived.

For the multiple comparison case, we use the same logic, but instead of single voxel we consider the maximum of each statistical map.

Result reporting

Whole-brain analysis

  • P-value threshold, MC correction method

  • All clusters with

    • size

    • peak statistic value

    • MNI coordinates

    • anatomical area (see atlasreader and the associated notebook)

Region-of-interest analysis

  • MCC for the number of areas examined

References

Faber, Jennifer, David Kügler, Emad Bahrami, et al. 2022. “CerebNet: A Fast and Reliable Deep-Learning Pipeline for Detailed Cerebellum Sub-Segmentation.” NeuroImage 264 (December): 119703. https://doi.org/10.1016/j.neuroimage.2022.119703.
Henschel, Leonie, Sailesh Conjeti, Santiago Estrada, Kersten Diers, Bruce Fischl, and Martin Reuter. 2020. “FastSurfer - a Fast and Accurate Deep Learning Based Neuroimaging Pipeline.” NeuroImage 219 (October): 117012. https://doi.org/10.1016/j.neuroimage.2020.117012.
Nichols, Thomas E., and Andrew P. Holmes. 2001. “Nonparametric Permutation Tests for Functional Neuroimaging: A Primer with Examples.” Human Brain Mapping 15 (1): 1–25. https://doi.org/10.1002/hbm.1058.