brainage/Enigma/enigma_stacking_pred.def

Bootstrap: docker                                                                                                                                                                                              
FROM: python:3.11.6-slim-bookworm                                                    
Stage: base   

%setup

#    touch ${SINGULARITY_ROOTFS}/data

%files
    save_feat.py /codes/
    predict_age_stacking.py /codes/
    analysis.sh /codes/
     
%environment
#    

%post
    pip install --upgrade pip
    pip install "scikit-learn==1.3.2" \
                "nibabel==4.0.2" \
                "nilearn==0.9.0" \
                "numpy==1.25.2" \
                "pandas==1.5.3" \
                "joblib==1.4.2" \
                "julearn==0.3.1" 
    mkdir /model/
    mkdir /data/
    mkdir /extras
    # mkdir /codes
    # cp save_feat.py /codes/
    # cp predict_age_stacking.py /codes/
    # cp analysis.sh /codes/
    chmod +x /codes/* 

    apt-get -y update && apt-get install -y unzip wget
    apt-get -y --quiet --no-install-recommends install build-essential curl ca-certificates
    curl https://gin.g-node.org/antogeo/stacking_model/raw/68c8723a15badbac0c26bf3d40aa9ba31264ea4a/extras/mask_4mm_binarized.nii --output /extras/mask_4mm_binarized.nii
    curl https://gin.g-node.org/antogeo/stacking_model/raw/68c8723a15badbac0c26bf3d40aa9ba31264ea4a/extras/mask_8mm_binarized.nii --output /extras/mask_8mm_binarized.nii
    curl https://gin.g-node.org/antogeo/stacking_model/raw/68c8723a15badbac0c26bf3d40aa9ba31264ea4a/extras/brainmask_12.8.nii --output /extras/brainmask_12.8.nii

    wget https://gin.g-node.org/antogeo/stacking_model/raw/master/stacking_enigmaModel%281%29.pkl -O /model/stacking_enigmaModel.pkl
    chmod +rw /extras/*

    # Octave 
    # add-apt-repository 'http://ftp.debian.org/debian/ stable main contrib non-free'
    DEBIAN_FRONTEND=noninteractive apt-get --yes --quiet install octave liboctave-dev
    mkdir /opt/spm12                                                             
    curl -fsSL --retry 5 https://github.com/spm/spm12/archive/r7771.tar.gz | tar -xzC /opt/spm12 --strip-components 1
    curl -fsSL --retry 5 https://raw.githubusercontent.com/spm/spm-octave/main/spm12_r7771.patch | patch -p0
    make -C /opt/spm12/src PLATFORM=octave distclean                             
    make -C /opt/spm12/src PLATFORM=octave                                       
    make -C /opt/spm12/src PLATFORM=octave install                               
    ln -s /opt/spm12/bin/spm12-octave /usr/local/bin/spm12
    rm -f /opt/spm12/src/*.{mex,o,a}                                             
    apt-get --yes remove build-essential curl liboctave-dev                      
    apt-get --yes autoremove                                                     
    apt-get --yes clean

%environment                                                                     
    export SPM_HOME=/opt/spm12

%runscript
    # # cp $1 /data
    # echo "Starting age prediction for following subject/subjects"
    # echo "predict subject's : $1"
    # exec echo "$1"
    bash /codes/analysis.sh "$@"

%startscript
    

%test
    grep -q NAME=\"Ubuntu\" /etc/os-release
    if [ "$(ls -A /data/)" ]; then
        echo "No data provided"
    else
        echo "Data found in data"
    fi

%labels
    Author: Antonopoulos Georgios https://github.com/antogeo
    Version v0.0.1

%help
    This def file builds a Singularity container with a brain-age prediction model based on
    "BrainAGE: Revisited and reframed machine learning workflow", Polona Kalc et al.
    It gets GMV and WMV nifti as inputs and predicts the age of the subject.
    You need to run the following command prior to building the container:
    > chmod -R 777 /tmp                                                                                                                                                                                          255 !

    At first, it extracts the features by performing resampling (4mm and 8mm) and smoothing
    (4mm and 8mm) to the image. Four different combinations of resampling and smoothing of 
    the two tissue images result in 8 different groups of features.