Skip to main content

Step 3: B0 Concatenation

Overview​

This step extracts the B0 (non-diffusion-weighted) volumes from the anterior-to-posterior (AP) and posterior-to-anterior (PA) fieldmap scans and merges them into a single 4D NIfTI file. This paired B0 dataset is the required input for TOPUP, which uses the opposite-direction distortions to estimate and correct susceptibility-induced geometric warping.

Further reading: Blip-Up-Blip-Down Distortion Correction — FMRIB Primers walkthrough of merging opposite phase-encoding B0 pairs and applying TOPUP corrections

Conceptual Background​

B0 Images​

In a diffusion MRI acquisition, most volumes are acquired with diffusion-sensitizing gradients applied (producing contrast that reflects water molecule displacement along specific directions). However, one or more volumes are acquired without diffusion weighting -- these are the B0 images (b-value = 0 s/mm^2). B0 images show the anatomy without diffusion contrast, making them ideal for estimating and correcting geometric distortions.

Opposite Phase-Encoding Directions​

Echo-planar imaging (EPI) sequences, used for diffusion MRI, are highly sensitive to magnetic field inhomogeneities near air-tissue boundaries (frontal sinuses, ear canals, base of the skull). These inhomogeneities cause geometric distortions -- voxels are shifted along the phase-encoding direction, resulting in visible stretching or compression of anatomy.

By acquiring B0 images with opposite phase-encoding directions (AP and PA):

  • In the AP image, anatomy near susceptibility boundaries is distorted in one direction (e.g., stretched anteriorly)
  • In the PA image, the same anatomy is distorted in the opposite direction (e.g., compressed anteriorly)

TOPUP exploits this symmetry: the true anatomy lies somewhere between the two distorted versions, and the field map that best explains both distortions simultaneously can be estimated and used to correct all diffusion volumes.

The Concatenation​

The first B0 volume is extracted from each of the AP and PA fieldmap scans, and the two are merged into a single 4D file along the time dimension.

Prerequisites​

  • Fieldmap AP and Fieldmap PA NIfTI files from Step 1
  • FSL installed and available on your PATH

Tool & Command Reference​

fslroi -- Extract Volumes​

fslroi extracts a subset of volumes from a 4D NIfTI file.

Syntax:

fslroi <input> <output> <first_volume> <number_of_volumes>
  • first_volume: Zero-indexed position of the first volume to extract
  • number_of_volumes: How many consecutive volumes to extract

fslmerge -- Concatenate Volumes​

fslmerge combines multiple 3D or 4D images into a single 4D file.

Syntax:

fslmerge -t <output> <input1> <input2> [input3 ...]
  • -t: Merge along the time (4th) dimension

Step-by-Step Commands​

# Define paths
input_dir="/path/to/project/nifti/${subj}/dti"
output_dir="/path/to/project/derivatives/b0_concat/${subj}"
mkdir -p "$output_dir"

# 1. Extract the first B0 volume from the AP fieldmap
fslroi "$input_dir/${subj}_fmapAP.nii.gz" \
       "$output_dir/${subj}_a2p_b0" \
       0 1

# 2. Extract the first B0 volume from the PA fieldmap
fslroi "$input_dir/${subj}_fmapPA.nii.gz" \
       "$output_dir/${subj}_p2a_b0" \
       0 1

# 3. Merge both B0 volumes into a single 4D file
fslmerge -t "$output_dir/${subj}_merged_b0s" \
    "$output_dir/${subj}_a2p_b0" \
    "$output_dir/${subj}_p2a_b0"

What each command does:

CommandAction
fslroi ... 0 1Extracts volume at index 0 (the first volume), extracting 1 volume total. This is the B0 image.
fslmerge -tConcatenates the two 3D B0 volumes along the time dimension, producing a 4D file with 2 volumes.

Batch Processing Script​

#!/bin/bash
# batch_b0_concat.sh - Extract and merge B0 volumes for all subjects
#
# Usage: bash batch_b0_concat.sh

base_dir="/path/to/project"
input_dir="$base_dir/nifti"
output_dir="$base_dir/derivatives/b0_concat"

for subj_dir in "$input_dir"/sub-*; do
    subj=$(basename "$subj_dir")

    ap_file="$subj_dir/dti/${subj}_fmapAP.nii.gz"
    pa_file="$subj_dir/dti/${subj}_fmapPA.nii.gz"

    # Check inputs exist
    if [ ! -f "$ap_file" ] || [ ! -f "$pa_file" ]; then
        echo "WARNING: Missing fieldmap for $subj -- skipping"
        continue
    fi

    out="$output_dir/$subj"
    mkdir -p "$out"

    # Skip if already processed
    if [ -f "$out/${subj}_merged_b0s.nii.gz" ]; then
        echo "Skipping $subj (already processed)"
        continue
    fi

    echo "Processing: $subj"

    # Extract B0 from AP
    fslroi "$ap_file" "$out/${subj}_a2p_b0" 0 1

    # Extract B0 from PA
    fslroi "$pa_file" "$out/${subj}_p2a_b0" 0 1

    # Merge
    fslmerge -t "$out/${subj}_merged_b0s" \
        "$out/${subj}_a2p_b0" \
        "$out/${subj}_p2a_b0"

    echo "  Done: $subj"
done

echo "B0 concatenation complete."

Expected Output​

$output_dir/
  sub-001/
    sub-001_a2p_b0.nii.gz      # Single B0 volume from AP fieldmap
    sub-001_p2a_b0.nii.gz      # Single B0 volume from PA fieldmap
    sub-001_merged_b0s.nii.gz   # 4D file containing both B0 volumes (2 volumes)

Quality Check​

Verify Volume Count​

The merged file should contain exactly 2 volumes (one AP, one PA):

fslnvols "$output_dir/${subj}_merged_b0s.nii.gz"
# Expected output: 2

You can also inspect the full header for additional verification:

fslinfo "$output_dir/${subj}_merged_b0s.nii.gz"

Look for dim4 in the output -- it should equal 2.

Visual Inspection​

Open the merged B0 file in FSLeyes and scroll through the two volumes. You should see:

  • Volume 0 (AP): Distortion in the anterior-posterior direction (e.g., frontal regions may appear stretched or compressed)
  • Volume 1 (PA): The same anatomy but with distortion in the opposite direction
fsleyes "$output_dir/${subj}_merged_b0s.nii.gz" &

The two volumes should show the same overall brain anatomy with visibly opposite distortion patterns, particularly near the frontal lobes and temporal regions.

Common Issues​

IssueCauseSolution
Merged file has wrong number of volumesExtracted wrong number of volumes with fslroi, or fieldmap has unexpected structureCheck the number of volumes in the original fieldmap files with fslnvols; adjust the volume index accordingly
AP/PA labels swappedFieldmap scan names don't match actual phase-encoding directionCheck the PhaseEncodingDirection field in the JSON sidecar files from Step 1
Distortions look identical in both volumesBoth fieldmaps were acquired with the same phase-encoding directionVerify scan protocol; re-acquire if necessary
Dimension mismatch error from fslmergeAP and PA volumes have different matrix sizes or voxel dimensionsCheck headers with fslinfo; volumes must have matching dimensions to merge

References​

Next Step​

Proceed to Step 4: TOPUP to estimate and correct susceptibility-induced distortions using the merged B0 file.