Where is Pain in the Brain? Unraveling Pain-Specific Functional Neurosignatures via 7 Tesla fMRI
Poster No:
2349
Submission Type:
Abstract Submission
Authors:
Henning Reimann1, Jurjen Heij2, Thomas Gladytz1, Thoralf Niendorf1
Institutions:
1Berlin Ultrahigh Field Facility, Max Delbrück Center for Molecular Medicine, Berlin, Germany, 2Spinoza Centre for Neuroimaging, Amsterdam, Netherlands
First Author:
Henning Reimann
Berlin Ultrahigh Field Facility, Max Delbrück Center for Molecular Medicine
Berlin, Germany
Berlin Ultrahigh Field Facility, Max Delbrück Center for Molecular Medicine
Berlin, Germany
Co-Author(s):
Thomas Gladytz
Berlin Ultrahigh Field Facility, Max Delbrück Center for Molecular Medicine
Berlin, Germany
Berlin Ultrahigh Field Facility, Max Delbrück Center for Molecular Medicine
Berlin, Germany
Thoralf Niendorf
Berlin Ultrahigh Field Facility, Max Delbrück Center for Molecular Medicine
Berlin, Germany
Berlin Ultrahigh Field Facility, Max Delbrück Center for Molecular Medicine
Berlin, Germany
Introduction:
Functional MRI (fMRI) has failed to identify unambiguous and reliable signatures unique to pain1,2. Painful stimuli cause dynamic formation of large-scale functional networks related to stimulus prioritization (salience), attention, memory, interoception, autonomic control, and the initiation of motor patterns – all of which are highly characteristic but not specific to pain1,2. Here, we employ enhanced blood oxygen level-dependent (BOLD) contrast and spatial fidelity of 7T fMRI to disentangle pain-specific from domain-general representations via fine-grained cortical mapping. By comparing painful versus non-painful salient stimuli, we dissect the functional architecture of pain by specifying its domain-general core, probe its pain-specific features, and detail the large-scale orchestration of pain versus salience evoked networks.
Methods:
fMRI of 8 volunteers (4/4, w/m), 4 runs, each 12 min. Stimuli (4 stimulus types, 8 per type and run) were applied briefly (duration ~100 ms) and pseudorandomized (interstimulus spacing 18-24s): contact heat (51°C), cutaneous electrical stimulation in 2 stimulus intensities: mildly painful (10 mA) and aversive but non-painful (6mA), and flashes of light (Fig. 1A). T2*-weighted fMRI (GE-EPI, TR/TE/FA = 2s/33.2ms/66°, FOV/matrix = 240x240mm/160x160, number of slices = 80, spatial resolution = 1.5mm isotropic, 220 volumes) acquired on MAGNETOM 7T scanner (Siemens Healthcare, Erlangen, Germany), single-channel-transmit/32-channel receive head coil (Nova Medical, Wilmington, MA, USA). fMRI data were motion/distortion corrected, smoothed, registered to MNI152, and statistically analyzed using FSL FLAME. Partial least squares (PLS) model trained and cross-validated on individual z maps (total 128 maps) to differentiate between stimulus modalities. 16 maps per volunteer (4 per stimulus) were normalized by their singular vector length in a singular value decomposition truncated after the 24th component. Leave-one-subject-out cross validation was performed to ensure transferability.
Results:
Cortical representations across stimuli (Fig. 1B) intersected in a domain-general functional core GS(∩) (Fig. 1C,D). PLS regression revealed voxels that drive the classification of pain versus other salient stimuli (Fig. 2). The mean covariance coefficient (CC) map depicts voxels that drive classification across all stimulus modalities (Fig. 2A); reflecting the shared domain-general patterns of GS(∩). The mean CC map may be regarded as an offset from which the CC maps of each individual stimulus modality deviate (Fig. 2B). Comparing CC of painful heat with mildly painful electrical stimulation shows very similar patterns that highly classify for both conditions, but the orchestration of those patterns (i.e., how they vary from the offset) differs substantially. Particularly in insular (74-78) and opercular regions (42-48), the weighting was even inverted. Since some of these areas were positioned within GS(∩), we masked all individual z-maps with GS(∩) and calculated the area (AUC) under the receiver operator curve (ROC) for painful heat versus other salient stimuli. We found an AUC of 0.9 and higher for painful heat when only considering domain-general areas in GS(∩).
·Figure 1
·Figure 2
Conclusions:
By combination of a set theory inspired intersection analysis of mass-univariate and multivariate analyses of painful versus non-painful salient fMRI data at 7T, we dissected the functional architecture of pain. Its domain-general functional core GS(∩) displays unique connectivity patterns for specific pain modalities and salient non-painful stimuli. GS(∩) exhibits great similarity to the recently introduced allostatic-interoceptive brain system3. This work deepens the understanding of pain-specific activity and its underlying mechanisms, operating upon an evolutionary well-conserved domain-general machinery for survival.
Modeling and Analysis Methods:
fMRI Connectivity and Network Modeling
Novel Imaging Acquisition Methods:
BOLD fMRI 1
Perception, Attention and Motor Behavior:
Perception: Multisensory and Crossmodal
Perception: Pain and Visceral 2
Perception: Tactile/Somatosensory
Keywords:
HIGH FIELD MR
Multivariate
Pain
Somatosensory
1|2Indicates the priority used for review