Intracranial Volume (ICV) Estimation

Overview

This optional step estimates the total intracranial volume (ICV) for each subject by computing the volume of the brain mask from skull stripping. ICV is sometimes used as a covariate in group-level statistical analyses to control for individual differences in head size.

Further reading: Understanding ICV / eTIV — FreeSurfer Wiki on the atlas-scaling method for estimating intracranial volume from transform matrices

Whether to control for ICV depends on your research question, analysis approach, and population. ICV is necessary in some contexts but not others — for example, developmental and aging studies where head size varies systematically across groups often include ICV as a covariate, while studies of healthy adults using tract-based metrics (e.g., mean FA along a tract via pyAFQ) typically do not. Consult your statistical analysis plan before deciding.

When ICV Matters

• Voxel-based analysis (VBA): comparing FA values voxel-by-voxel across groups where head size differences can confound results
• Whole-brain DTI metrics: computing global mean FA or total white matter volume
• Clinical populations: comparing groups with expected head size differences (e.g., neurodegeneration, developmental disorders)

Commands

# ──────────────────────────────────────────────
# Define paths
# ──────────────────────────────────────────────
ants_dir="$base_dir/ants/$subj"
output_dir="$base_dir/icv"

mkdir -p "$output_dir"

# ──────────────────────────────────────────────
# Quick ICV from brain mask volume
# ──────────────────────────────────────────────
icv=$(fslstats "$ants_dir/${subj}_BrainExtractionMask.nii.gz" -V | awk '{print $2}')
echo "${subj},${icv}" >> "$output_dir/icv_summary.csv"

# ──────────────────────────────────────────────
# Optional: tissue segmentation with Atropos (for per-tissue volumes)
# ──────────────────────────────────────────────
Atropos -d 3 \
    -a "$ants_dir/${subj}_BrainExtractionBrain.nii.gz" \
    -x "$ants_dir/${subj}_BrainExtractionMask.nii.gz" \
    -i "KMeans[3]" \
    -o "[${output_dir}/${subj}_seg.nii.gz,${output_dir}/${subj}_prob%02d.nii.gz]"

Expected Output

File	Description
icv_summary.csv	CSV with subject ID and ICV in mm³
${subj}_seg.nii.gz	3-class segmentation (optional, from Atropos)

Quality Check

• ICV values for healthy adults typically range from 1,200,000–1,800,000 mm³ (1200–1800 cm³)
• Flag subjects with values far outside this range — likely indicates failed skull stripping
• If using Atropos, overlay the segmentation on the T1 in FSLeyes to verify labels

Next Step

ICV is a covariate rather than a pipeline input, so nothing downstream blocks on it. Return to the Outputs for Tractography to confirm the rest of your outputs are in place, or see BIDS & pyAFQ if you plan to use pyAFQ's whole-brain bundle recognition.