KinderSeg: FastSurfer Database for Age-Specific Brain Volumes in Healthy Children

Code for the paper:

Zughayyar, I., Bauer, M., Güttler, C., de Almeida Marcelino, A.L., Kühne, F., Buss, C., Heim, C., Aigner, A., Tietze, A. and Dell'Orco, A. (2025), A FastSurfer Database for Age-Specific Brain Volumes in Healthy Children: A Tool for Quantifying Localized and Global Brain Volume Alterations in Pediatric Patients. Brain Behav, 15: e70689. https://doi.org/10.1002/brb3.70689

Run with Docker

git clone https://github.com/ibrazug/kinderseg.git
cd kinderseg/Docker
cp </path/to/your/license.txt> .
docker build -t kinderseg .
docker run --gpus all --rm \
    -v <nifti_data>:/data \
    -v <output_dir>:/output \
    kinderseg \
    --age <age> \
    --threads <threads, optional, default=4>

IMPORTANT NOTE: --threads is used to define the number of threads in total. not per hemisphere.

nifti_data can be either a directory containing NIfTI files or a single NIfTI file. The output will be saved in the output directory in a subdirectory named after the input NIfTI file, without the extension. Example:

docker run --gpus all --rm \
    -v $PWD/sub-001/ses-01/anat/sub-001_ses-01_T1w.nii.gz:/data/sub-001_ses-01_T1w.nii.gz \
    -v ./output:/output \
    kinderseg --age 12

Run with Apptainer/Singularity

git clone https://github.com/ibrazug/kinderseg.git
cd kinderseg/Docker
cp </path/to/your/freesurfer/license.txt> .  # copy your FreeSurfer license into this folder
docker build -t kinderseg .
apptainer build kinderseg.sif docker-daemon://kinderseg:latest

# Run the container with proper mounting:
apptainer run --nv kinderseg.sif \
    -B <nifti_data>:/data \
    -B ./output:/output \
    --age 18 \
    --threads 14

Notes:

• Use relative paths (./data, ./output) instead of $PWD to avoid mount errors.
• Bind all -B options before the image name (kinderseg.sif).

Requirements

• At least 5 GB of RAM
• GPU (optional, but recommended)
• a valid FreeSurfer license: you can get one for free by clicking here.
• we also provided Docker.amd version for AMD chips.

Dataset Statistics

Dataset	N° in total	Ex (QC)	N° after QC	Ex (SC)	N° after SC
HBN	170	21	149	10	139
LOC	125	32	93	1	92
Kids2Health	211	5	206	1	205
Total	506	58	448	12	436

MRI Segmentation and Volumetric Analysis of the dataset subjects

MRI Post-processing

• DICOM data converted to NIfTI format using dcm2niix and Python version 3.9 was used for data manipulation

Segmentation Tools

1. FreeSurfer v7.4: Ran on a Slurm cluster with a 16-core Intel Xeon E5-2650 CPU, Utilized GNU Parallel for parallel processing

2. FastSurfer v2.3.3: Utilized FastSurferCNN pipeline on a workstation with an AMD 3970X CPU and NVIDIA GeForce RTX 3090 GPU

Data Analysis

• R version 4.1.2 used for analysis
• Libraries: ggplot2, tidyverse, viridis (R)

KinderSeg Module Files

The kinderseg directory contains the following key files and scripts:

1. metrics.py:

   • Implements metrics and evaluation methods for comparing segmented brain volumes.
   • Includes the DiceNiiRois class for calculating the Dice similarity coefficient between ROIs in NIfTI files.
   • Outputs a Pandas DataFrame with the Dice similarity between all ROIs.

2. sums.py:

   • Contains utility classes for summarizing region of interest (ROI) data and volume tables.
   • Classes:
     • SumNiiRois: Summarizes ROIs as defined in the sumroi_mapping.csv file and generates output NIfTI files.
     • SumVolTable: Summarizes volume tables (from FastSurfer and FreeSurfer) based on mapped ROIs, creating CSV output.
   • Uses parallel processing to speed up calculations and ROI summarization.

3. sumroi_mapping.csv:

   • A CSV file that contains mappings between raw ROIs and summarized ROIs.
   • Supports mappings from VINN and FreeSurfer 7 (FS7) formats.
   • Used by sums.py to transform and summarize volume data efficiently.

Article Citation

To cite this work, please use the following reference:

Zughayyar, I., Bauer, M., Güttler, C., de Almeida Marcelino, A.L., Kühne, F., Buss, C., Heim, C., Aigner, A., Tietze, A. and Dell'Orco, A. (2025), A FastSurfer Database for Age-Specific Brain Volumes in Healthy Children: A Tool for Quantifying Localized and Global Brain Volume Alterations in Pediatric Patients. Brain Behav, 15: e70689. https://doi.org/10.1002/brb3.70689

@article{https://doi.org/10.1002/brb3.70689,
author = {Zughayyar, Ibrahim and Bauer, Martin and Güttler, Christopher and de Almeida Marcelino, Ana Luisa and Kühne, Fabienne and Buss, Claudia and Heim, Christine and Aigner, Annette and Tietze, Anna and Dell'Orco, Andrea},
title = {A FastSurfer Database for Age-Specific Brain Volumes in Healthy Children: A Tool for Quantifying Localized and Global Brain Volume Alterations in Pediatric Patients},
journal = {Brain and Behavior},
volume = {15},
number = {7},
pages = {e70689},
keywords = {automated segmentation, brain volumetry, FastSurfer, FreeSurfer, hemispheric asymmetries, neuroimaging, pediatric MRI},
doi = {https://doi.org/10.1002/brb3.70689},
url = {https://onlinelibrary.wiley.com/doi/abs/10.1002/brb3.70689},
eprint = {https://onlinelibrary.wiley.com/doi/pdf/10.1002/brb3.70689},
year = {2025}
}

TY  - JOUR
T1  - A FastSurfer Database for Age-Specific Brain Volumes in Healthy Children: A Tool for Quantifying Localized and Global Brain Volume Alterations in Pediatric Patients
AU  - Zughayyar, Ibrahim
AU  - Bauer, Martin
AU  - Güttler, Christopher
AU  - de Almeida Marcelino, Ana Luisa
AU  - Kühne, Fabienne
AU  - Buss, Claudia
AU  - Heim, Christine
AU  - Aigner, Annette
AU  - Tietze, Anna
AU  - Dell'Orco, Andrea
Y1  - 2025/07/01
PY  - 2025
DA  - 2025/07/01
DO  - https://doi.org/10.1002/brb3.70689
T2  - Brain and Behavior
JF  - Brain and Behavior
JO  - Brain and Behavior
JA  - Brain Behav
SP  - e70689
VL  - 15
IS  - 7
KW  - automated segmentation
KW  - brain volumetry
KW  - FastSurfer
KW  - FreeSurfer
KW  - hemispheric asymmetries
KW  - neuroimaging
KW  - pediatric MRI
PB  - John Wiley & Sons, Ltd
SN  - 2162-3279
UR  - https://doi.org/10.1002/brb3.70689
Y2  - 2025/08/08