Human brain effects of DMT assessed via EEG-fMRI

This single-blind, placebo-controlled, counterbalanced within-subjects study recruited 20 participants (mean age 33.5 y, SD = 7.9; 7 female). Each volunteer completed two scanning days separated by two weeks. On each day participants underwent a 28-min resting-state fMRI/EEG session with intravenous injection at the end of minute 8 of either 10 mL saline (placebo) or 20 mg DMT in fumarate form (dissolved in 10 mL saline), injected over 30 s and flushed with 10 mL saline. Eyes were closed throughout and an eye mask prevented opening. In a separate scan run (not the primary resting recording reported here) participants provided minute-by-minute verbal intensity ratings; these real-time ratings and previously measured DMT plasma concentrations were used as regressors for dynamic analyses. fMRI acquisition used a 3T scanner (TR = 2000 ms, voxel 3 × 3 × 3 mm3) and preprocessing followed an established pipeline (despiking, slice time and motion correction, brain extraction, registration to MNI space, scrubbing with FD threshold 0.4, spatial smoothing, band-pass filtering 0.01–0.08 Hz, detrending, nuisance regression including ventricles, draining veins and local white matter). Four participants were excluded from group fMRI analyses for excessive motion (>20% scrubbed volumes); additional subsample analyses used more stringent motion thresholds. EEG was recorded concurrently from 31 channels with MR-compatible equipment, sampled at 5 kHz and preprocessed to remove gradient and ballistocardiogram artefacts, downsampled to 250 Hz, band-pass filtered and epoched. Two participants were removed from EEG analyses for excessive artefacts and one was excluded from alpha analyses due to an atypical alpha peak; joint EEG–fMRI analyses included 12 participants who survived both modalities' preprocessing. Analyses: For fMRI, static resting-state functional connectivity (sRSFC) was assessed using ICA-derived canonical resting-state networks and dual regression to estimate within-network integrity, between-network segregation, and regional/global functional connectivity (GFC) across 112 parcellated regions (100 cortical from Schaefer atlas plus 12 subcortical). Dynamic resting-state FC (dRSFC) used tapered sliding windows (44 s, 2-s step) and related timecourses of GFC and pairwise FC to minute-by-minute intensity ratings using linear mixed-effects models, with FDR correction for multiple comparisons; analogous analyses were performed using average plasma DMT levels. Principal cortical gradients were obtained using diffusion map embedding on interregional FC similarity matrices, followed by Procrustes alignment. EEG analyses included spectral decomposition isolating oscillatory from 1/f components, band definitions (delta 1–4 Hz, theta 4–8 Hz, alpha 8–13 Hz, beta 13–30 Hz, gamma 30–45 Hz), cluster-based permutation testing, Lempel–Ziv complexity (LZs per channel, averaged as LZc) for signal diversity, and quantification of cortical travelling waves via 2D FFTs on midline electrodes. EEG–fMRI relationships were assessed by convolving EEG metrics with a canonical haemodynamic response function and relating them to region-wise GFC and pairwise FC timecourses using linear mixed-effects models. An independent in vivo 5-HT2A receptor density map from prior PET work was used to test spatial overlap with dynamic GFC effects, and NeuroSynth meta-analytic maps were used to characterise associated functions.

Samples and exclusions: 20 participants completed the protocol. Four subjects were excluded from the core fMRI group analyses for excessive motion, yielding analyses based on 16 participants for many fMRI contrasts (consistent with reported t tests). EEG preprocessing resulted in further exclusions, and joint EEG–fMRI analyses were performed on n = 12 participants who met both modalities' quality thresholds. Static fMRI connectivity: Averaged over the 8-min postinjection peak window, DMT decreased within-network integrity across canonical resting-state networks (P < 0.05, FDR corrected) with the exception of the salience (SAL) and limbic (LIM) networks. In contrast, DMT increased global functional connectivity (GFC) in the SAL, frontoparietal control (FP), and default-mode (DMN) networks (P < 0.05, FDR corrected). Between-network segregation was reduced, especially involving FP, SAL and DMN (P < 0.05, FDR corrected). Region-level GFC increases were observed in medial prefrontal cortex, dorsolateral PFC, insula and temporoparietal junction, and whole-brain average GFC increased under DMT compared with placebo [t(15) = 3.11, P = 0.007, 95% CI = 0.028 to 0.15]. Motion-control subsample analyses and global signal regression (GSR) checks yielded broadly consistent patterns. Dynamic fMRI connectivity: Using sliding-window dRSFC related to real-time intensity ratings, increases in GFC of DMN, FP, SAL and LIM networks correlated positively with intensity (P < 0.05, FDR corrected). Pairwise connectivity showed widespread positive associations with intensity, but negative associations were found between intensity and pairwise connectivity of visual–somatomotor/subcortical links (P < 0.05, FDR corrected). Time-resolved results indicated maximal increases in DMN and FP GFC during roughly minutes 1–6 postinjection (P < 0.05, cluster corrected); pairwise connectivity rose broadly in the first minute whereas segregation between lower-order unimodal sensory and motor areas became more apparent from minute two onward. Dynamic GFC changes also correlated with average plasma DMT concentrations. Spatial overlap analyses showed that regions with higher 5-HT2A receptor density related positively to the GFC model involving intensity ratings, and NeuroSynth decoding of GFC and 5-HT2A maps highlighted high-level cognitive and linguistic terms (e.g., language, semantic). Principal cortical gradient: Gradient-mapping revealed a compression of the principal sensorimotor-to-association axis under DMT. High-order TOP regions exhibited reduced gradient scores (movement toward zero) while low-level unimodal regions showed increased gradient scores toward zero, indicating reduced differentiation between unimodal and transmodal cortex. Network-wise, gradient scores increased within unimodal sensory networks (SM, VIS) and DAN, and decreased within FP, DMN and LIM, consistent with diminished macroscale hierarchical organisation. EEG spectral and complexity results: Time-averaged EEG over the 8-min postinjection window showed widespread decreases in alpha power (P < 0.01, cluster corrected) and increases in delta (P < 0.05) and gamma power (P < 0.01). Signal diversity (Lempel–Ziv complexity: LZs per channel and global LZc) increased significantly under DMT (P < 0.01). Dynamic correlations with real-time intensity indicated that greater intensity co-occurred with larger increases in delta power and LZc, and greater decreases in global alpha and posterior beta power (P < 0.05). Temporally resolved analyses found alpha reductions approximately 0–8 min (P = 0.001, cluster corrected), LZc increases ~2–14 min (P = 0.001), and delta increases largely confined to the onset phase ~0–4 min (P = 0.006). A modest beta reduction was observed in the first minute (P = 0.04). Forward travelling-wave power increased while backward wave power decreased. Subjective ratings of the richness of conscious experience correlated with LZc (r = 0.7, P = 0.008). EEG–fMRI relationships: Using simultaneous data (n = 12), frontal delta power increases related positively to widespread regional GFC increases across frontal, temporal, parietal, motor and visual association cortices (P < 0.05, FDR corrected), a result echoed in pairwise FC. Parietal alpha decreases were negatively associated with GFC increases in TOP networks and regions, and alpha reductions were linked to decreased visual–somatomotor coupling but increased coupling between sensory and TOP regions. Occipital gamma increases associated with GFC increases in FP regions and the LIM network (P < 0.05), and LZc increases associated with GFC increases in TOP and LIM (P < 0.05). These modality-crossing associations were detectable using within-subject timecourse models, whereas simple between-subject averaging reduced these effects. Other observations and validation: The authors report that head motion was significantly greater under DMT than placebo, but subsample analyses controlling for motion produced consistent results. GSR and other preprocessing checks are reported in supplementary material. The study found few robust correlations between imaging metrics and specific phenomenological features (e.g., entity encounters), suggesting limited mapping of particular subjective elements to distinct neural measures in this dataset.

Combining EEG and fMRI in a single simultaneous protocol, the study presents convergent evidence that DMT induces a multilevel dysregulation of high-order cortex consistent with a global brain action of classic psychedelics. The authors interpret the principal findings—network disintegration and desegregation, decreased alpha power, increased signal diversity and gamma power, altered travelling waves, increased delta power, elevated global functional connectivity, and a compression of the principal cortical gradient—as pointing to reduced segregation of the TOP from the rest of cortex and a collapse of the brain's hierarchical organisation. They argue that regions most affected spatially overlap with areas of dense 5-HT2A receptor expression and with neuroimaging maps related to high-level cognitive and linguistic functions, supporting a model in which 5-HT2A receptor agonism causally contributes to the observed functional effects. The authors relate increased signal entropy and hyperconnectivity to the entropic brain hypothesis, with LZc increases correlating with reported richness of experience. Delta power increases correlated with elevated global connectivity and are discussed as a marker of profound alteration of conscious content rather than merely reduced level of consciousness. Observations linking alpha reductions, gamma increases and entropy enhancements with limbic GFC increases are taken as consistent with limbic disinhibition hypotheses relevant to memory and emotion processing. Timmermann and colleagues note limitations that temper interpretation. The study is exclusive to DMT without an active drug control, leaving questions about specificity compared with other classic psychedelics. Head motion was larger under DMT, and although subsample analyses and preprocessing checks were used to validate results, the authors acknowledge that fMRI connectivity metrics remain sensitive to motion and call for larger datasets with equivalent motion between conditions. They also discuss potential confounds with arousal; while some imaging effects could be compatible with increased arousal, the spatial and temporal patterning (including delta increases and gradient compression) exceed what would be expected from arousal alone, and drowsiness ratings did not convincingly account for the observed effects. Finally, the authors observe relatively few robust links between imaging metrics and fine-grained subjective features and recommend future neurophenomenological work with richer within-subject sampling and larger datasets. In their concluding interpretation, the authors consolidate the view that DMT principally targets developmentally and evolutionarily recent transmodal cortex, increasing its communication with the rest of the brain and producing a hyperassociative, higher-entropy mode of processing. They suggest these global neural changes may relate to plasticity and to alterations in human-specific psychological faculties, while emphasising that such implications remain to be tested in future research.