A brain-based predictive signature of individual pain sensitivity

based on the Resting-state Pain susceptibility Network (RPN)

Welcome to website of the RPN-signature, a resting-state network-based predictive signature of individual pain sensitivity. The project is a joint effort of the Predictive Neuroimagiong Laboratory, (PNI-lab, Tamas Spisak) and the Bingel-lab (Ulrike Bingel), University Hospital Essen, Germany.

This site is under construction.

News

10.01.2020. Our paper describing the RPN-signature has been published by Nature Communications: Spisak, T. et al. Pain-free resting-state functional brain connectivity predicts individual pain sensitivity. Nat Commun 11, 187 (2020)

1. Summary
2. Inputs of the the RPN-signature
3. Usage via Docker
4. Output
5. Running the source code (advanced)
6. Authors and Citation

Summary

Individual differences in pain perception are of key interest in both basic and clinical research as altered pain sensitivity is both a characteristic and a risk factor for many pain conditions. Individual susceptibility to pain is reflected in the pain-free resting-state activity and functional connectivity of the brain. The RPN-signature is a network pattern in the pain-free resting-state functional brain connectome that is predictive of interindividual differences in pain sensitivity. The RPN-signature allows assessing the individual susceptibility to pain without applying any painful stimulation, as might be valuable in patients where reliable behavioural pain reports cannot be obtained. Additionally, as a direct, non-invasive readout of the supraspinal neural contribution to pain sensitivity, it may have broad implications for translational research and the development of analgesic treatment strategies.

The Resting-state Pain susceptibility Network signature consists of an fMRI image preprocessing pipeline and a prediction based on (a linear combination of) specific functional connectivity values. Its output is a single number: a predicted pain sensitivity score, to be interpreted on the scale of the QST-based (Quantitative Sensory Testing) pain sensitivity score. See the paper (under review) for details.

The list of predictive functional connections is to be found here. (Note that a sufficient predictive performance is expected only with our dedicated preprocessing pipeline, see below)
The nodes of the predictive network (with nodal predictive strength) can be downloaded here. Note that this map is not predictive on it’s own, just a spatial map of the RPN-nodes.

Inputs

This “research product” allows making predictions on the individual’s pain sensitivity based on their resting-state fMRI measurements. For the image preprocessing step, the T1-weighted anatomical images are additionally needed.

All data must be structured according to the Brain Imaging Data Structure BIDS. Consider validating your data with the BIDS validator before running the RPN-signature.

The predictive model should be robust for variations in imaging sequences. Nevertheless, we have the following suggestions (which shouldn’t be hard to meet):

In general:

3T field strength

Anatomical image:

high-resolution, “close-to-isovoxel” T1-weighted anatomical image, e.g. 1x1x1mm

Functional image:

8-12 min long resting-state fMRI scan
whole brain coverage (actually, a few millimeter can be missed from the ventral part of the cerebellum, see the RPN regional connectivity map for reference: [https://github.com/spisakt/PAINTeR/blob/master/res/RPN_predictive_network_nodes.nii.gz]
TR around 2.5 sec (the model might be robust to this, though)
interleaved slice order
approximately 3mm voxel
careful fixation to prevent motion artifacts

Usage with docker

The usage of the RPN-siganture with Docker is simple and platform-independent. You can run it like any other BIDS-app.

Get the Docker Engine (https://docs.docker.com/engine/installation/)
Have your data organized in BIDS (get BIDS specification, see BIDS paper).
Validate your data (http://incf.github.io/bids-validator/). You can safely use the BIDS-validator since no data is uploaded to the server, works locally in your browser.
Have a look at the help page to test if it works. It will start to download the docker image from docker hub (approximately 5.3Gb).
```
docker run -it tspisak/rpn-signature:latest -h
```

Run it by mounting and specifying your BIDS directory, output directory and level of analysis, like for any other BIDS-app. E.g.:

docker run -it --rm -v /data/nii-bids/:/data:ro -v /data/nii-bids/derivatives:/out \
tspisak/rpn-signature:latest /data /out participant

NOTE 1 Have a look at the help, there are some useful command line options:

E.g.:

docker run -it --rm -v /data/nii-bids/:/data:ro -v /data/nii-bids/derivatives:/out \
tspisak/rpn-signature:latest /data /out participant \
--participant_label  001 002 003 005 008 013 021 034 --mem_gb 10 --nthreads 7 --2mm

NOTE 2 Output directory must be specified as an absolute path.

NOTE 3 Note that the –2mm command line option performs spatial co-registration to a 2mm-resolution template (instead of 1mm), which is much faster (total running time is approximately 50 min instead of 8 hours per subject), but was not validated and gives slightly different (presumably less accurate) predictions.

NOTE 4 Make sure to configure Docker’s resource availability to take advantage of parallel processing.

NOTE 5 Make sure to have enough free space for storing temporary files (1.5GB per subject).

NOTE 6 Consider using the option –keep_derivatives, if you need the timeseries and connectivity data for further processing.

NOTE 7 Do quality checking (see below) before using the predicted values and adjust brain extraction parameters with the options –bet_fract_int_thr and –bet_vertical_gradient if necessary.

Output

The main output of the RPN-signature is the RPN-score, a prediction of the composite pain sensitivity score of (Zunhammer et al., 2016). Click here for more information on the composite score and its feasability as prediction target.

Additionally, the RPN-signature has a rich outpout of quality-check images and, if –keep_derivatives is specified, it outputs the processed image files and can be used as an fMRI processing workflow.

output_directory/
│    RPNresults.csv                CSV-file containing the predicted pain sensitivity scores
│    subjectsID.txt                text file linking data files to QC indices
└─── QC/                           directory for quality check images
│   └─── anat2mni/                 standardized anatomical image + the standard template
│   └─── brain_extraction/         anatomical image + the result of brain extraction 
│   └─── brain_extraction_func/    functional image + the result of brain extraction 
│   └─── carpet_plots/             carpet plots of preproicessing stages
│   └─── compcor_noiseroi/         aCompCor noise ROI overlaid on the functional image
│   └─── FD/                       framewise displacement plots
│   └─── func2anat/                functional image in anatomical space + anatomical image 
│   └─── func2mni/                 functional image in stnadard space + standard template
│   └─── motion_correction/        rotational and translational motion estimates
│   └─── regional_timeseries/      carpet plot of the atlas-based regional timeseries
│   └─── timeseries/               mean global signal timeseries of preprocessing stages
│   └─── tissue_segmentation/      tissue segmentation maximum probability images
:
:    if --keep_derivatives is specified:
:   
:    atlas.nii.gz                  brain atlas (MIST122) to define ROIs
:... anat_preproc/                 anatomical derivatives
:   :... anat2mni_std/             standard-space anatomical image
:   :... anat2mni_warpfield/       warpinf-field for standardisation (contains all steps)
:   :... bet_brain/                brain extracted anatomical image
:   :... brain_mask/               anatomical brain mask
:   :... fast_csf/                 CSF probability map
:   :... fast_gm/                  grey matter probability map
:   :... fast_wm/                  white matter probability map
:... func_preproc/                 functional derivatives
:   :    popFD_max.txt             mean FD values per subject
:   :    popFD.txt                 max FD values per subject
:   :    pop_percent_scrubbed.txt  percent of volumes scrubbed per subject
:   :... bet_brain/                brain extracted functional image
:   :... brain_mask/               functional brain mask
:   :... FD_scrubbed/              FD timeseries after scrubbing
:   :... mc_fd/                    FD timeseries
:   :... mc_frist24/               Friston-24 expansion of motion parameters
:   :... mc_func/                  motion corrected funcrtional image
:   :... mc_par/                   6 motion parameters (3 rotation, 3 translation)
:   :... mc_rms/                   root mean squared motion estimates
:... regional-timeseries           regional timeseries in tab separated format

Running the source code

Running the RPN-signature from source can be tricky for non-developers, because a lot of dependencies must be resolved and installation is undocumented.
Image preprocessing and network calculation is based on our PUMI (https://github.com/spisakt/PUMI) python module (see page for installation and dependencies), which has to be installed first.
Run pipeline script for image preprocessing and network calculation: [https://github.com/spisakt/PAINTeR/blob/master/pipeline/pipeline_PAINTeR-BIDS.py]
For help, please contact Tamas Spisak (spisak.tms@uk-essen.de)

Authors

Tamas Spisak¹, Balint Kincses², Frederik Schlitt¹, Matthias Zunhammer¹, Tobias Schmidt-Wilcke², Zsigmond Tamas Kincses², Ulrike Bingel¹

Department of Neurology, University Hospital Essen, Essen, Germany
Faculty of Medicine, Ruhr-University Bochum, Bochum, Germany
Department of Neurology, University of Szeged, Szeged, Hungary

Citation:

Tamas Spisak, Balint Kincses, Frederik Schlitt, Matthias Zunhammer, Tobias Schmidt-Wilcke, Zsigmond Tamas Kincses, Ulrike Bingel, Pain-free resting-state functional brain connectivity predicts individual pain sensitivity, Nat Commun 11, 187 (2020).

CircleCI2 Docker Build Docker Pulls