Selective and replicable neuroimaging-based indicators of pain discriminability

Figures

• 

  Figure 1. Study design, experimental procedure, and discriminability index calculation (A and B) EEG and fMRI study design. Consisting of five datasets with large sample sizes, the EEG and fMRI studies aimed to reveal neural indicators of pain discriminability and assess the replicability and selectivity of these neural indicators. (C and D) EEG and fMRI experimental procedures. Participants were given transient stimuli belonging to four different sensory modalities (pain, touch, audition, and vision), each composed of two stimulus intensities (high and low, marked in purple and orange, respectively). After each stimulus, participants rated the perceived intensity, using a numerical rating scale ranging from 0 (“no sensation”) to 10 (“the strongest sensation imaginable [in each stimulus modality]”). (E) A schematic diagram describing the calculation of the discriminability index, i.e., area under the curve (AUC). Eleven integral ratings (i.e., 0–11) on the rating scale can be treated as 11 implicit response criteria participants hold. For a given criterion (e.g., 5 in the top left plot), the hit rate and false alarm rate can be estimated to define a point on the receiver operating…

• 

  Figure 2. Laser-evoked EEG responses reliably correlated with pain discriminability (A) Distributions of pain ratings evoked by nociceptive laser stimuli of high (purple) and low (orange) intensities and distributions of pain AUC values. For all three datasets, pain ratings evoked by high-intensity stimuli were significantly greater than those evoked by low-intensity stimuli. AUC values in all datasets were significantly greater than the chance level of 0.5. (B) The comparison of laser-evoked potentials (LEP) responses between high- and low-intensity stimuli in Dataset 1. Compared with low-intensity stimuli, high-intensity laser stimuli evoked larger N2/P2 (left) and N1 (right) waves. (C) Point-by-point correlations between differential LEP responses (high–low) and AUC values in Dataset 1. Gray-shaded areas denote time intervals at which the amplitude differences (left: N2 and P2 waves; right: N1 wave) significantly correlated with AUC values (false discovery rate [FDR] corrected). (D) Correlations between differential amplitudes of N1, N2, and P2 waves and AUC values. Differential N1, N2, and P2 amplitudes were consistently correlated with AUC values for all three datasets. ∗∗∗p < 0.001;…

• 

  Figure 3. EEG responses evoked by tactile, auditory, and visual stimuli did not reflect their respective sensory discriminability (A) Distributions of intensity ratings evoked by tactile (left), auditory (middle), and visual (right) stimuli of high (purple) and low (orange) intensities as well as distributions of sensory AUC values for each modality. For all three modalities, ratings evoked by high-intensity stimuli were significantly greater than those evoked by low-intensity stimuli, and AUC values were significantly greater than 0.5. Note that Datasets 1 and 2 were pooled together (Dataset 1&2) due to their identical experimental settings and stimulus parameters for the three modalities. (B) The comparison of event-related potentials (ERP) responses between high- and low-intensity stimuli in Dataset 1&2. (C) Point-by-point correlations between differential ERP responses and AUC values in Dataset 1&2. Only the differential amplitudes of P2 wave evoked by tactile stimuli weakly correlated with tactile AUC values (FDR corrected). (D and E) Correlations between differential amplitudes of N2 and P2 waves and AUC values for different sensory modalities (D: Dataset 1&2; E: Dataset 3). For both…

• 

  Figure 4. The selectivity of EEG indicators of pain discriminability (A and B) N2 and P2 selectively correlated with pain discriminability after pain and tactile intensity ratings were matched. For both Dataset 1&2 (A) and Dataset 3 (B), the matching procedure equalized ratings for pain and touch, although pain AUC values were significantly smaller than tactile AUC values. Notably, for both Dataset 1&2 and Dataset 3, differential N2 and P2 amplitudes evoked by noxious laser stimuli correlated with pain AUC values. However, no evidence could be obtained in favor of the correlation between differential N2 and P2 amplitudes and tactile AUC values. (C and D) N2 and P2 selectively correlated with pain discriminability after pain and auditory intensity ratings were matched. (E and F) N2 and P2 selectively correlated with pain discriminability after pain and visual intensity ratings were matched. ∗0.01 < p < 0.05; ∗∗∗p < 0.001; ns: p > 0.05; x: 0.1 < BF < 0.32; ne: 0.32 < BF < 3.2; ##: 10 < BF < 100; ###: BF > 100. See Tables S9–S12 for detailed statistical results.

• 

  Figure 5. Laser-evoked fMRI responses reliably correlated with pain discriminability (A) Distributions of pain ratings evoked by nociceptive laser stimuli of high (purple) and low (orange) intensities and distributions of pain AUC values. For both datasets, pain ratings evoked by high-intensity stimuli were significantly greater than those evoked by low-intensity stimuli. AUC values in both datasets were significantly greater than 0.5. (B) Brain regions showed stronger activation by high-intensity stimuli than low-intensity stimuli (top), and brain regions showed significant correlations between their differential blood-oxygen-level-dependent (BOLD) responses and AUC values (bottom) in Dataset 4. (C) Pain-related regions of interests (ROIs), defined according to the correlation results in Dataset 4, including the bilateral S1, S2, thalamus, insula, and ACC. Differential BOLD responses in several pain-related ROIs (i.e., left S1, left thalamus, left insula, and ACC) also strongly correlated with AUC values in Dataset 5, thus verifying the replicability of the findings obtained in Dataset 4. For whole brain analyses, the significance threshold was set at p = 0.001 at the voxel level and pFWE…

• 

  Figure 6. fMRI responses evoked by tactile, auditory, and visual stimuli did not reflect their respective sensory discriminability (A) Distributions of intensity ratings evoked by tactile (left), auditory (middle), and visual (right) stimuli of high (purple) and low (orange) intensities as well as distributions of sensory AUC values for each modality. For all three modalities, ratings evoked by high-intensity stimuli were significantly greater than those evoked by low-intensity stimuli, and AUC values were significantly greater than 0.5. Note that Datasets 4 and 5 were pooled together due to their identical experimental settings and stimulus parameters for the three modalities. (B–D) Brain regions showed stronger activation by high-intensity stimuli than low-intensity stimuli (top), and brain regions showed significant correlations between their differential BOLD responses and AUC values for each sensory modality (bottom). Although tactile (B), auditory (C), and visual (D) stimuli of high intensity evoked stronger activations in a series of brain regions (e.g., bilateral insula and ACC) than low-intensity stimuli, almost no brain regions showed significant correlations with AUC values. The…

• 

  Figure 7. The selectivity of fMRI indicators of pain discriminability (A) Distributions of intensity ratings and AUC values of pain and a non-pain sensation after their ratings were matched (left: pain vs. touch; middle: pain vs. audition; right: pain vs. vision). The matching procedure equalized ratings for each pair of sensory modalities. However, pain AUC values were significantly smaller than non-pain AUC values for all comparisons. (B–D) Brain regions showed significant correlations between differential BOLD responses and AUC values after pain ratings were matched with tactile (B), auditory (C), and visual (D) intensity ratings. For all comparisons, differential BOLD responses evoked by nociceptive laser stimuli significantly correlated with pain AUC values, while almost no significant correlations were observed for the three non-painful sensory modalities. Only a few areas in the occipital cortex significantly correlated with visual AUC values for the visual modality. ∗∗∗p < 0.001; ∗∗0.001 < p < 0.01; ns: p > 0.05.