SwipeQC

Swipe cards to sort images pass / fail

Setting up images

Convert your NIFTIs to videos

NIFTI isn't a format natively understood by web browsers but med2image can be used to convert NIFTI slices to JPEG, then imagick / ffmpeg can turn it into a video (and overlay a segmentation). In setup, mimosa.sh will work on MIMoSA outputs and most pipelines should only require editing the $flair / $seg filename. On the cluster,

singularity pull docker://pennsive/nifti2video
mkdir /path/to/mimosa_web # create output directory
dirs=$(find /path/to/mimosa_outputs -name "mimosa_binary_mask_0.25.nii.gz" | xargs dirname) # find sub/session dirs to process
for dir in $dirs; do
    outdir=/path/to/mimosa_web/${dir}
    mkdir -p $outdir
    qsub singularity run --cleanenv --bind $TMPDIR --bind /path/to/mimosa_web --bind /path/to/mimosa_outputs nifti2video_latest.sif $dir $outdir $PWD/db.sqlite
done

Using app

Launch server

SwipeQC can be run from the cluster with the images if you do port forwarding. First, start SwipeQC from an available port on the cluster

cd ..
singularity pull docker://pennsive/swipeqc-api
export SINGULARITYENV_IMAGE_PATH=/path/to/mimosa_web
# assuming 5001 is an open port
singularity run --cleanenv --bind /path/to/mimosa_web swipeqc-api_latest.sif --port=5001 # this command starts the server; it is meant to hang

Then in another Terminal window (on your local machine), forward the port you started the server on to a local port

ssh -qnNT -L 5001:127.0.0.1:5001 user@takim

And pull up http://localhost:5001 in your browser

Generate report

To list all the images sorted, run

echo "SELECT image, passed, at FROM ratings ORDER BY id ASC" | sqlite3 db.sqlite