A longitudinal neuroimaging investigation has revealed that individuals with methamphetamine use disorder exhibit pronounced rigidity in whole-brain functional network dynamics during early abstinence, followed by partial recovery after sustained cessation. Published in Translational Psychiatry, the research evaluated resting-state functional magnetic resonance imaging data from forty-one healthy controls and forty individuals with methamphetamine use disorder. Clinical participants were assessed across two recovery stages: early abstinence of three and a half months or less, and prolonged abstinence lasting at least eight months. By applying energy landscape analysis to model transitions across large-scale cerebral networks, investigators demonstrated that methamphetamine dependency confines neural activity to inflexible configurations. Furthermore, baseline dynamic indicators successfully predicted longitudinal changes in subjective drug craving across prolonged abstinence, highlighting dynamic neuroimaging metrics as potential predictive recovery biomarkers.
Methamphetamine is an addictive psychoactive stimulant altering monoamine neurotransmitter systems, generating severe public health burdens and high relapse risks. Current pharmacotherapies show limited efficacy, making objective neurobiological markers essential for improving treatment strategies. Previous neuroimaging studies established that individuals with methamphetamine use disorder exhibit marked impairments in brain regions supporting reward processing, executive control, motivation, and decision-making. Longitudinal evidence indicates that neural function and cognitive capacity can partially recover after sustained abstinence, accompanied by reductions in subjective craving. However, most neuroimaging studies have focused on static functional connectivity or isolated regions of interest. These conventional static approaches aggregate neuroimaging signals over extended scanning windows, producing time-averaged snapshots that obscure how large-scale functional networks dynamically configure, interact, and transition across temporal states over time.
To investigate dynamic network behavior, the research team recruited forty-one healthy controls and forty individuals meeting DSM-5 criteria for severe methamphetamine use disorder. Clinical participants completed baseline assessments during early abstinence of three and a half months or less (MA-T1) and follow-up assessments after at least eight months of sustained abstinence (MA-T2). Participants were right-handed individuals between eighteen and forty-five years of age who possessed the cognitive capacity to complete questionnaires, without history of head injury or neuropsychiatric disorders. Tobacco dependence was quantified using the Fagerström Test for Nicotine Dependence, while alcohol use was evaluated with the Alcohol Use Disorders Identification Test. Trait impulsivity was measured using the Barratt Impulsiveness Scale. Subjective craving intensity was tracked with a visual analog scale ranging from zero to ten, which thirty-eight clinical participants completed across both baseline and follow-up sessions.
Neuroimaging was conducted on a three-tesla Siemens Magnetom Skyra scanner with a thirty-two-channel head coil. Structural three-dimensional T1-weighted images were acquired with sagittal BRAVO sequences at one-millimeter isotropic resolution. Resting-state functional scans acquired two hundred and twenty-five volumes with a gradient-echo echo-planar imaging sequence across thirty-six axial slices, using a repetition time of two thousand milliseconds, an echo time of thirty milliseconds, and a flip angle of eighty degrees. Preprocessing was executed using FSL and AFNI. The first ten volumes were discarded to ensure signal stability. Steps included slice-timing alignment, head motion correction excluding movements exceeding one point five millimeters or one point five degrees, skull stripping, coregistration, and standardization to Montreal Neurological Institute space. Time series underwent spatial smoothing, intensity normalization, nuisance signal regression, and band-pass filtering from zero point zero one to zero point zero eight hertz.
Using the Power atlas, researchers extracted time series from two hundred and fourteen spherical regions assigned to seven functional networks: default mode, frontoparietal, salience, attention, somatosensory and motor, visual, and auditory networks. Regional signals were averaged within each network to reduce transient local noise. Signals were binarized using each network’s mean activity as an objective threshold, with values above the mean representing an active state (+1) and values below indicating an inactive state (-1). Across seven networks, this yielded one hundred and twenty-eight theoretical whole-brain binary activity patterns. Investigators applied the pairwise maximum entropy model using the MATLAB Energy Landscape Analysis Toolkit to fit empirical network probabilities. The model achieved high accuracy exceeding zero point ninety-seven, with Pearson correlations between empirical and simulated pattern probabilities exceeding zero point ninety-eight across all experimental groups, confirming robust mathematical representation of brain activity patterns.
Energy landscape analysis constructed disconnectivity graphs depicting the energy barriers and hierarchical relationships among stable brain states, defined as local energy minima. Across healthy controls, early-abstinence individuals, and prolonged-abstinence individuals, the energy landscape revealed six consistent local minima, labeled A through F. These minima grouped into two major states and two minor states based on energy depth and stability. Local minima A and B formed major state number one, characterized by activation of the default mode network alongside deactivation of auditory, somatosensory and motor, and attention networks. Conversely, local minima E and F formed major state number two, displaying coactivation of auditory, somatosensory and motor, and attention networks coupled with default mode network deactivation. Minima C and D constituted minor states number one and two, displaying complementary antiphasic configurations. Major state basins encompassed ninety-two percent of total landscape area in early abstinence, compared to eighty-three percent in controls and eighty-seven percent in prolonged abstinence.
Empirical resting-state fMRI analyses demonstrated pronounced dynamic rigidity in individuals with methamphetamine use disorder during early abstinence. Compared to healthy controls, empirical appearance frequencies of both major states were significantly elevated in early abstinence (major state #1: F = 30.21, PFDR < 0.001; major state #2: F = 38.45, PFDR < 0.001). Conversely, the appearance frequencies of both minor states were significantly decreased (minor state #1: F = 85.27, PFDR < 0.001; minor state #2: F = 61.51, PFDR < 0.001). Empirical dwell time in major state #1 was significantly prolonged in early abstinence (F = 5.65, PFDR = 0.0275), while dwell times in minor states were significantly shorter (minor state #1: F = 11.76, PFDR = 0.0018; minor state #2: F = 6.23, PFDR = 0.0232). Longitudinal comparisons between early and prolonged abstinence revealed that major state appearances significantly decreased while minor state appearances significantly increased, demonstrating partial recovery.
To evaluate dynamic transitions between brain states, researchers executed random-walk numerical simulations using the Metropolis-Hastings Markov chain Monte Carlo algorithm. Simulating one hundred thousand steps per group and discarding the first one hundred steps, the simulated trajectories corroborated empirical findings regarding major state dominance and minor state attenuation in early abstinence. Transitions between the two major states were categorized into direct transitions, occurring without intermediate states, and indirect transitions passing through minor states. In empirical fMRI data, direct transitions occurred significantly more frequently in early abstinence compared to healthy controls (F = 34.78, PFDR < 0.001) and prolonged abstinence (F = 16.26, PFDR = 0.002). Meanwhile, random-walk simulations demonstrated a significantly higher frequency of indirect transitions in early abstinence relative to healthy controls (F = 112.56, PFDR < 0.001), reflecting altered transition mechanics across the energy landscape.
The investigators examined whether these dynamic abnormalities correlated with behavioral measures in early abstinence. Pearson correlation analyses revealed that visual analog scale craving scores in early-abstinence individuals were significantly positively correlated with the frequency of indirect transitions (r = 0.3389, P = 0.0348) and significantly negatively correlated with the dwell time of minor brain states (r = -0.3301, P = 0.0402). In addition, impulsivity scores on the Barratt Impulsiveness Scale exhibited a significant positive correlation with the appearance frequency of major states (r = 0.3312, P = 0.0395) and a significant negative correlation with minor state duration (r = -0.3279, P = 0.0416). These findings indicate that impaired neural flexibility and excessive confinement within dominant brain states are closely linked to clinical deficits in impulse control and craving regulation among individuals with methamphetamine use disorder.
To examine functional coordination supporting these dynamic properties, the authors analyzed network modularity across major states. Local minima A and F exhibited complementary activity patterns partitioned into two functional modules: an auditory, visual, somatosensory and motor, and attention module versus a salience, frontoparietal, and default mode module. Functional segregation strength was quantified as the difference between mean functional connectivity within modules and between modules. Across all participant groups, significant functional segregation was observed, but early-abstinence individuals displayed markedly higher segregation strength than healthy controls (F = 4.83, PFDR = 0.0379) and prolonged-abstinence patients (F = 7.22, PFDR = 0.0244). Furthermore, in early abstinence, functional segregation strength was significantly positively correlated with total years of methamphetamine use (r = 0.3348, P = 0.0428) and cumulative drug dosage (r = 0.3506, P = 0.0309), indicating that drug exposure reinforces modular isolation and exacerbates dynamic rigidity.
Following prolonged abstinence, individuals exhibited partial normalization of neurodynamic alterations, accompanied by decreasing major state frequencies and increasing minor state appearances. Building on these longitudinal observations, the researchers evaluated whether baseline dynamic indicators could predict longitudinal craving changes using support vector regression. A model with a radial basis function kernel was trained on twenty baseline dynamic metrics from thirty-eight clinical participants who completed both craving assessments. Evaluated through nested cross-validation, the model demonstrated a significant positive correlation between predicted and observed craving changes (r = 0.3718, P = 0.0215; permutation P < 0.05), with dwell times of minor state #1 and major state #1 contributing most substantially to prediction. The authors noted several limitations, including a modest sample size of forty-one controls and forty patients, two longitudinal assessment points, unassessed comorbidity factors, and parameter sensitivity inherent to energy landscape modeling.
Subject of Research: Longitudinal brain network dynamics and craving prediction in methamphetamine use disorder
Article Title: Longitudinal brain dynamics and prediction of craving changes in individuals with methamphetamine use disorder
Article References: Song, Q., Yang, W., Du, Z., Wang, D., Wen, X., Liu, J., & Yuan, K. (2026). Longitudinal brain dynamics and prediction of craving changes in individuals with methamphetamine use disorder. Translational Psychiatry, 16(1), Article 473. https://doi.org/10.1038/s41398-026-04401-y
Image Credits: AI Generated
DOI: 10.1038/s41398-026-04401-y
Keywords: methamphetamine use disorder, energy landscape analysis, resting-state fMRI, brain network dynamics, prolonged abstinence, drug craving prediction, support vector regression, default mode network, salience network, frontoparietal network, functional segregation, pairwise maximum entropy model, random walk simulation, visual analog scale, Barratt Impulsiveness Scale, neural flexibility
Cite Scienmag News
Cassandra Pierce. (November 16, 2026). Baseline Brain Dynamic Indices Predict Craving Changes Following Methamphetamine Abstinence. Scienmag. https://scienmag.com/baseline-brain-dynamic-indices-predict-craving-changes-following-methamphetamine-abstinence/
Cassandra Pierce. "Baseline Brain Dynamic Indices Predict Craving Changes Following Methamphetamine Abstinence." Scienmag, 16 November 2026, https://scienmag.com/baseline-brain-dynamic-indices-predict-craving-changes-following-methamphetamine-abstinence/. Accessed 16 November 2026.
Cassandra Pierce. "Baseline Brain Dynamic Indices Predict Craving Changes Following Methamphetamine Abstinence." Scienmag. November 16, 2026. https://scienmag.com/baseline-brain-dynamic-indices-predict-craving-changes-following-methamphetamine-abstinence/








