Common and distinct neural representations of aversive somatic and visceral stimulation in healthy individuals

Pain is a primary driver of behavior, learning, and neuroplasticity, signaling current or potential bodily harm. Although human neuroimaging has advanced understanding of pain mechanisms, most work has focused on cutaneous somatic pain. Visceral pain is common clinically (e.g., IBS) yet understudied, and may differ from somatic pain in localization, relation to stimulus intensity, referral, and affective qualities. Historical models propose distinct lateral (somatosensory/insula) vs medial (cingulate/medial prefrontal) pain systems, but evidence has been mixed. Multivariate brain models have identified a Neurologic Pain Signature (NPS) that predicts perceived intensity of thermal somatic pain, with demonstrated specificity relative to other affective processes and limited generalization to other somatic pain types. Key open questions are whether the NPS generalizes to visceral pain (indicating a common core nociceptive system) and whether systematic differences exist between visceral and somatic pain representations beyond the NPS. This study addresses these questions by aggregating seven fMRI datasets across visceral and somatic stimulation, testing NPS sensitivity/specificity, and developing a classifier to distinguish visceral from somatic pain.

Prior small-sample neuroimaging studies comparing visceral and somatic pain reported mixed results, including overlaps and differences in somatosensory, cingulate, insular, prefrontal, hippocampal, and brainstem regions. The NPS, a multivariate pattern sensitive to thermal somatic pain, shows positive weights in thalamus, insula, SII, aMCC and negative weights in vmPFC, PCC/precuneus, and occipital cortex. It has been shown to be specific relative to neural signatures for social rejection, negative emotion (PINES), and vicarious pain, with little cross-correlation. However, NPS generalization to visceral pain has not been systematically evaluated. Historical lateral vs medial pain system distinctions and evidence for stronger affective components in visceral pain are debated, necessitating larger, multi-study analyses to clarify common and distinct neural representations.

Seven fMRI studies (total N=165) involving five stimulation types were retrospectively aggregated: gastric distension (Study 1), rectal distension (Studies 2–3), vulvar pressure (Study 4), esophageal stimulation (Study 5), and cutaneous thermal pain (Studies 6–7). Standard preprocessing and first-level mass-univariate GLMs were run per study. Second-level analyses combined study contrasts (pain/uncomfortable vs baseline or low-intensity control) controlling for study as a covariate. Analyses included: (1) NPS expression quantified via cosine similarity/dot product for each participant and condition; one-sample t-tests and effect sizes assessed pain vs baseline separability; robust regression tested prediction of VAS pain/discomfort ratings in visceral studies, controlling for study. (2) Spatial similarity (Pearson correlations) of activation maps with neural signatures for social rejection, negative emotion (PINES), and vicarious pain; repeated-measures ANOVA compared similarity magnitudes; FDR correction applied. (3) Specificity test using an independent archive of 18 studies (N=270; 6 pain, 12 cognitive control or negative emotion); NPS cosine similarity classified pain vs no-pain with accuracy, sensitivity, specificity, and AUROC. (4) Voxel-wise conjunction analyses (somatic>baseline and visceral>baseline; deactivations likewise) with FDR q<0.05 identified common activations/deactivations across modalities. (5) Development of a between-subject logistic regression classifier distinguishing rectal (visceral) vs cutaneous thermal (somatic) stimulation based on per-subject point-biserial correlations with seven resting-state network maps (Yeo et al., 7-network parcellation). Ten-fold cross-validation (iterated 1000 times) assessed discriminability (AUROC, sensitivity, specificity, balanced accuracy) and model stability; average coefficients were prospectively applied to five independent test studies (rectal, thermal, gastric, vulvar, esophageal). Binomial tests compared classification proportions relative to the training-derived threshold. All second-level analyses used CANlab tools (MATLAB). Data and code availability were provided (Neurovault collections 8707 and 3324; CANlab GitHub).

- NPS generalizability to visceral pain: Significant NPS responses across all visceral studies with large effect sizes and high within-subject classification accuracy: Study 1 gastric 0.043±0.009, t(14)=4.66, P=0.0004, da=1.20, accuracy 86.7%; Study 2 rectal 0.045±0.010, t(14)=4.69, P=0.0003, da=1.21, accuracy 93.3%; Study 3 rectal 0.053±0.007, t(28)=8.08, P<0.0001, da=1.50, accuracy 96.6%; Study 5 esophageal 0.122±0.006, t(29)=20.49, P<0.0001, da=3.74, accuracy 100%. - NPS predicted subjective visceral pain/discomfort: Robust regression β=12.44±4.45, P=0.0064; sample-wise correlations r≈0.20–0.38 despite small Ns. - NPS responses in somatic studies were also strong: Study 4 vulvar 0.093±0.011, t(14)=8.61, P<0.0001, dz=2.22, accuracy 100%; Study 6 thermal 0.132±0.009, t(27)=15.22, P<0.0001, dz=2.88, accuracy 100%; Study 7 thermal 0.089±0.007, t(32)=12.11, P<0.0001, dz=2.11, accuracy 100%. - Sensitivity to pain vs other affective signatures: Across all pain data, spatial similarity was highest for NPS (mean within-person r=0.074±0.003, t(164)=24.41, dz=1.92, qFDR<0.05) and significantly lower for social rejection (r=0.017±0.0019, t=8.77), negative emotion PINES (r=0.006±0.0019, t=3.45), and negative for vicarious pain (r=−0.012±0.0021, t=−5.78), with ANOVA F(3,492)=286.23, P<0.0001; all three vs NPS P<0.0001 (Bonferroni). - Specificity of NPS to pain across independent studies: In 18-study archive (N=270), NPS discriminated pain vs control (cognitive control/negative emotion) with AUROC=0.93, sensitivity 73%, specificity 92%, accuracy 86±2.1%, dz=2.13. - Common activations/deactivations (conjunction, qFDR<0.05): Overlapping activations for somatic and visceral stimulation in midbrain, cerebellum, putamen/pallidum, hypothalamus, ventrolateral/ventral posterior lateral thalamus, parahippocampal/entorhinal cortex, posterior/mid/anterior insula, SI/SII and adjacent opercula, inferior parietal lobule, superior temporal gyrus, vlPFC, MI/premotor, aMCC/pMCC, medial frontal gyrus, and dlPFC. Overlapping deactivations in pulvinar, hippocampus/parahippocampal/perirhinal cortex, temporal pole, middle/inferior temporal gyri, occipital cortex, pACC/sACC, vmPFC/dmPFC, dlPFC, PCC/precuneus, superior parietal lobule, and dorsal pre/postcentral gyri. - Visceral vs somatic classifier (resting-state network similarity features): Cross-validated discriminability (training on rectal vs thermal): AUROC=0.92, sensitivity 94%, specificity 80%, balanced accuracy 87±4.4%, d′=2.22. Somatic indicated by stronger positive correlations with somatomotor (β=19.89±5.94, P=0.0008), dorsal attention (β=14.21±5.72, P=0.013), and ventral attention (β=12.54±5.50, P=0.023) networks; visceral indicated by more positive correlations with frontoparietal (β=−12.35±4.09, P=0.0025) and default (β=−11.89±4.02, P=0.0031) networks; limbic trend (β=−8.66, P=0.063). - Prospective generalization to independent cohorts: AUROC=0.84, sensitivity 82%, specificity 69%, balanced accuracy 76±5.7%. Esophageal classified as somatic (P_somatic=96.67±3.28%, P<0.0001), vulvar showed a somatic trend (86.67±8.78%, P=0.083), gastric intermediate (66.67±12.17%, P=0.97).

Findings indicate a brain-wide commonality in responses to visceral and somatic pain, with the NPS capturing a core nociceptive system that generalizes across modalities and accurately discriminates pain from conceptually related non-pain affective and cognitive processes. However, distinct network-level patterns differentiate visceral from somatic pain: somatic stimulation more strongly engages the somatomotor and attention networks, while visceral (rectal) stimulation more strongly engages frontoparietal control and shows reduced deactivation of default-mode regions, suggesting differences in attentional orientation, interoception, and executive evaluation. The resulting two-stage approach—first applying the NPS to classify pain vs no pain, then using a somatovisceral classifier—successfully distinguishes pain modality across independent participants and studies. These results refine traditional lateral vs medial pain system distinctions, showing stronger somatic responses in SI/SII, posterior insula, aMCC, and somatomotor network, with visceral stimulation engaging default-mode and frontoparietal regions more strongly. Variability with respect to some earlier small-sample univariate studies underscores the need for larger samples; despite inter-study differences in design and stimulation, robust, generalizable patterns and modality-specific features emerged.

The study demonstrates that the Neurologic Pain Signature provides a generalizable, sensitive, and specific marker of pain across visceral and somatic modalities, supporting a shared core pain-related network. Beyond this common core, a network-based classifier reliably distinguishes visceral from somatic pain and generalizes to independent cohorts, enabling a proposed two-stage pain classification framework (NPS pain detection followed by somatovisceral classification). Visceral pain did not appreciably correlate with neural signatures of non-pain affect, challenging the assumption of a stronger affective nature of visceral pain at the neural level. Differences among visceral types (e.g., esophageal resembling somatic, gastric/vulvar intermediate) suggest mixtures of pain-related representations that vary by tissue and individual. Future work should develop visceral-specific NPS-like models, harmonize stimulation parameters across modalities, include within-subject assessments of affective processes, and test sex differences and other individual factors.

Inter-study differences included stimulus duration (from 1 s events to 30 s blocks), intensity calibration (individually titrated vs fixed), and use of severe discomfort thresholds for rectal distension (due to urgency) rather than pain thresholds. These factors could influence neural responses and modality comparisons. Stimuli were not matched across modalities for timing or perceived intensity. Within-subject comparisons of pain and non-pain affective manipulations were not performed, limiting direct assessment of affect signatures within the same participants. Sex differences were not analyzed and require larger, dedicated samples. Despite these limitations, the multi-study design and larger sample enhanced generalizability.