Multivariate signature patterns

June 2, 2026 · View on GitHub

Pre-trained multivariate fMRI brain patterns ("signatures", "neuromarkers") that can be applied to new individual-subject contrast maps to generate predicted scores (e.g., pain intensity, emotion rating, craving). Each sub-folder contains one published signature with its weights, supporting files, and a per-folder contents_description.md.

See the docs README for the documentation conventions used throughout the repo, and the CanlabCore Tools for the object-oriented MATLAB toolbox these signatures are designed to be used with.

Applying signatures to new data

Two convenience wrappers live at the root of this folder:

apply_all_signatures.m — applies a configurable set of signatures (NPS, SIIPS, PINES, VPS, Rejection, GSR, HR, …) to a cell array of fmri_data objects and returns a results table. Supports similarity_metric (dot-product / cosine / correlation), image_scaling (none / center / z-score / l2-norm), and a named image_set to choose which signature bundle to apply.
```
[SIG, sigtable] = apply_all_signatures(DATA_OBJ, ...
    'similarity_metric', 'cosine_similarity', ...
    'image_scaling', 'l2norm_images', ...
    'image_set', 'npsplus');
```
apply_siips.m — SIIPS1-specific wrapper that accepts wildcards, filename lists, or fmri_data objects and returns whole-signature responses plus local responses for each of the 44 FDR-thresholded SIIPS subregions. See the SIIPS folder (2017_Woo_SIIPS1/contents_description.md) for examples.

Both helpers internally call CanlabCore's apply_mask / canlab_pattern_similarity but are easier to drop into a batch pipeline than wiring those up by hand.

For a single signature, the most direct call is:

[obj, networknames, imagenames] = load_image_set('siips');     % keyword from load_image_set.m
new_data = fmri_data('my_contrast.nii');
sig_response = apply_mask(new_data, obj, ...
                         'pattern_expression', 'ignore_missing');

Sub-folders

Every sub-folder below contains contents_description.md (per-study overview, references, file inventory, and loading instructions), visualize_contents.m (regenerates png_images/), and the signature files themselves.

Year	Study	Topic	CanlabCore keyword
2011	Wager — Placebo prediction (J Neurosci)	Placebo response	(no keyword — load `.img` directly)
2015	Chang — PINES (PLoS Biol)	Picture-induced negative affect	`pines`
2015	Kragel — Emotion BPLS (SCAN)	7 emotion categories	`kragelemotion`
2015	Woo — Romantic Rejection (Nat Comms)	Social rejection	`rejection`
2016	Eisenbarth — Autonomic GSR/HR (J Neurosci)	Skin conductance, heart rate	`gsr`, `hr`
2016	Krishnan — VPS (eLife)	Vicarious pain	`vps`
2017	Ashar — Care / Distress (Neuron)	Empathic care vs distress	(load `.nii` directly)
2017	López-Solà — Fibromyalgia signatures (PAIN)	FM-pain & FM-multisensory patterns; NPS region masks	(load `.nii` / `.img` directly)
2017	Rosenberg — saCPM (Nat Neurosci)	Sustained attention (connectome)	(load `.mat` directly)
2017	Woo — SIIPS1 (Nat Comms)	Cerebral pain (beyond nociception)	`siips`
2018	Kragel — MFC generalizability (Nat Neurosci)	Pain × emotion × cognitive control in MFC	`kragel18`, `pain_cog_emo`
2018	Reddan — Threat ImEx (Neuron)	CS+ vs CS− threat conditioning	`csplus`
2019	Kragel — Emotion Schemas (Sci Adv)	20 emotion categories	`kragelschemas`
2019	Lee — Back pain (PAIN)	Chronic-back-pain markers (S1, PCASL, HFHRV)	(load directly)
2019	Matthewson/Woo — SCR pain (PAIN)	Skin-conductance + pain	(see `Readme.rtf`)
2019	Yu/Koban — Guilt (Cereb Cortex)	Interpersonal guilt	`guilt`
2020	Geuter — Pain PDM mediation (Cereb Cortex)	Pain mediation directions	`pdm`, `pain_pdm`
2020	Silvestrini/Rainville — aMCC pain × cognitive control (NeuroImage)	dACC pain / Stroop patterns	`stroop`
2020	Van Oudenhove/Kragel — Somatovisceral (Nat Comms)	Visceral vs somatic pain	(load `.mat` directly)
2020	Zhou — General vicarious pain (eLife)	NS / FE / general vicarious pain	(helper `load_zhouvps`)
2021	Čeko — MPA2 multiaversive (Nat Neurosci 2022)	5 aversive modalities	`mpa2`, `multiaversive`
2021	van 't Hof — BASIC (Cereb Cortex)	Sexual-image classifier	(load `.nii` directly)
2021	Zhou — Subjective fear VIFS (Nat Comms)	Fear from video	(helper `load_vifs`)
2022	Coll — Pain × money decision value	Pain / money / shock-intensity value	(load directly)
2022	Koban — NCS Craving (Nat Neurosci 2023)	Drug & food craving	`ncs`
2023	Speer — Brain Reward Signature (BRS)	Reward signature	(load `.nii` directly)
2024	FEPS — Facial Expressions of Pain (eLife)	Facial-expression-based pain	(load `.nii` directly)
2026	Açıl — Mentalizing: Self / Other (Nat Comms)	Mentalizing about self vs. other (MS, MS-Self, MS-Other, MS-SvO)	`selfother`
2026	Murillo — PiFoneM (J Pain)	Pain-induced fear of neck movement	`pifonem`, `fearofneckpain`

Conventions

Each sub-folder has the same shape: contents_description.md, visualize_contents.m, NIfTI / .mat data, a PDF of the primary reference (where redistributable), and a png_images/ directory with regenerated surface / montage / isosurface figures.
The shared rendering helpers live in ../docs/canlab_render_patterns.m (and canlab_render_atlas.m).
To regenerate every signature's PNGs in one MATLAB session, run ../docs/run_all_multivariate.m.