Introduction
The impact of opioid use disorder (OUD) on brain white matter is a significant topic within the domain of medical research. White matter, comprising the neural fibers responsible for information transmission within the brain, is integral to cognitive function and behavioral control. The use of addictive substances such as heroin can disrupt the integrity of brain white matter, leading to cognitive impairments and neuropathological changes (Li et al., Reference Li, Li, Zhu, Qin, Zheng, Chang, Zhang, Wang, Wang, Wang and Wang2013). With the advancement of MRI technology, particularly the application of Diffusion Tensor Imaging (DTI), researchers are now able to more precisely detect and assess the microstructural changes in the white matter of OUD (Bora et al., Reference Bora, Yucel, Fornito, Pantelis, Harrison, Cocchi, Pell and Lubman2012). These techniques facilitate our understanding of the effects of OUD on brain structure and function, as well as the plasticity and potential for recovery of the brain following detoxification. Utilizing MRI technology to detect changes in the white matter of OUD before and after detoxification is of significant importance. First, it aids in identifying structural brain damage caused by OUD, including the impairment of white matter fibers (Wang et al., Reference Wang, Zhou, Liao, Tang, Liu and Hao2011). These injuries may be associated with cognitive impairments and emotional regulation issues in individuals with OUD (Wong et al., Reference Wong, Cheung, Chan, Zeng, Liu, So and Lee2015). Second, by comparing pre- and post-detoxification MRI outcomes, we can assess the plasticity and potential for recovery of the brain’s white matter. Some studies indicate that after long-term detoxification, there is an improvement in the white matter integrity of certain brain regions in OUD (Lu et al., Reference Lu, Yang, Zhao, Wen, Liu, Liu and Yuan2023a), suggesting that the brain possesses a certain degree of self-repair ability. Moreover, understanding the recovery of the brain post-detoxification is crucial for developing effective treatment and rehabilitation plans. For instance, identifying white matter integrity indicators related to relapse prediction can help optimize treatment methods (Feng et al., Reference Feng, Huang, Zhang, Yang, Guo, Xia, Peng and Li2024; Li et al., Reference Li, Zhu, Li, Ye, Chen, Liu, Li, Li, Yan, Wang and Wang2016).
In recent years, functional magnetic resonance imaging (fMRI) technology has made significant strides in exploring the functional activities within brain white matter (WM). WM, comprising the neural fiber bundles that connect various brain regions, plays a crucial role in the transmission of neural information. Traditional fMRI research has primarily focused on the activity of gray matter (GM); however, accumulating evidence suggests that the blood oxygen level-dependent (BOLD) signals within WM can also be reliably detected and are intimately associated with neural activity (Viessmann, Tian, Bernier, & Polimeni, Reference Viessmann, Tian, Bernier and Polimeni2022;Wang et al., Reference Wang, Wang, Wang, Zhang, Yang, Zhou, Qiu and Zhang2023). These findings are of paramount importance for comprehending the organizational structure and functional connectivity of brain networks. The investigation into WM BOLD signals has unveiled their role in the functional networks of the brain. Studies indicate that WM BOLD signals not only reflect neural activity but are also correlated with the structural connectivity of white matter (Huang et al., Reference Huang, Wei, Xu, Chen, Yang, Song, Yi, Jia, Wu, Fan, Cui and Zhao2023). These signals are detectable not only at rest but also exhibit structurally specific temporal correlations during the performance of specific tasks (Yang et al., Reference Yang, Wang, Liu, Zou, Jiang, Jiang, Cao and Zou2022). Furthermore, the anisotropy of WM BOLD signals suggests that the microstructural BOLD effects within white matter tracts can be quantified through Functional Connectivity Tensor (FCT) analysis (Xu et al., Reference Xu, Gao, Li, Lawless, Zhao, Schilling, Rogers, Anderson, Ding, Landman and Gore2024).
The exploration of WM BOLD signals holds potential value for understanding the pathophysiological mechanisms of neuropsychiatric disorders. For instance, research on individuals with OUD has shown that alterations in white matter networks may be associated with the recovery of brain function following detoxification. Studies indicate that the white matter networks of individuals with OUD who have undergone detoxification tend to converge towards those of healthy controls, suggesting that the restoration of white matter networks may be related to the improvement of addictive behaviors (Qin et al., Reference Qin, Liang, Li, Hu, Wu, Qiao, Liu and Du2021). These findings afford a novel perspective for investigating white matter dysfunction in neuropsychiatric disorders such as addiction. Investigators have employed resting-state fMRI to examine the functional white matter networks in young smokers (Qin et al., Reference Qin, Liang, Li, Hu, Wu, Qiao, Liu and Du2021), providing a comparative methodology for studying white matter network alterations in OUD. In summary, research into WM BOLD signals affords us a fresh perspective for understanding the functional networks of the brain and the pathophysiological mechanisms of neuropsychiatric disorders. By integrating various MRI techniques, we can obtain a more comprehensive assessment of changes in white matter networks and explore their correlations with brain function and behavior.
In this investigation, we employed a data-driven methodology to delve into the organization and the intensity of functional neural activity within white matter functional networks (WMFNs) in individuals with OUD. K-means clustering was utilized to distinctly demarcate the WMFNs of both the OUD group and the healthy controls (HCs), as well as to differentiate between the baseline OUD group and the follow-up OUD group (Bu et al., Reference Bu, Gao, Liang, Chen, Guo and Huang2022). The functional neural activities of the WMFNs were compared between the three groups. Correlation analyses were also performed to investigate the relationships between the intensity of functional activities within WMFNs and the clinical indicators of heroin use. We posited two hypotheses: (1) the baseline OUD group would exhibit distinct WMFN patterns compared to the HC/follow-up OUD group, and (2) altered WMFNs might be correlated with clinical indicators of OUD. It is our aspiration that this study will offer a more profound understanding of the neural mechanisms underlying opioid/heroin use disorder from the perspective of WMFNs.
Materials and methods
Participants
One hundred and eight participants who had a history of opioid use were recruited from drug rehabilitation clinics in Xinkaipu, PingTang, and Zhuzhou (three cities in Hunan Province, China) between March 2017 and December 2018. All the participants (n = 108) tested positive for heroin on a urine test and satisfied the diagnostic criteria for opioid use disorder according to the fifth version of the Diagnostic and Statistical Manual of Mental Disorders (DSM-V). An additional inclusion criterion for OUD was a negative urine test for methamphetamine and ketamine. All the participants were admitted to the detoxification center and, without methadone maintenance treatment, received 8 months of education and physical training. All of them signed written consent forms and participated in the first MRI scan. Among these initial participants, 61 were invited to finish a follow-up interview and a second MRI scan ~8 months later. Fifteen participants were excluded because of excessive head movements or low-quality MRI data. Eventually, 46 individuals with OUD (23 males and 23 females, age: 43.10 ± 6.95 years) were included in this study. Years of opioid (heroin) use for each participant with OUD were recorded as clinical indicators. We recruited 42 healthy control (HC) participants via WeChat, flyers, and other methods. All HCs had a negative urine test for heroin, methamphetamine, and ketamine. All of the participants were right-handed and of Han Chinese ancestry. None of the participants had previous diagnoses of structural brain diseases, seizures, head trauma, mental or behavioral disorders, or MRI contraindications. Table 1 shows the demographic characteristics of the OUD and HC groups.
Participant demographic and clinical information
Table 1. Long description
The table consists of four columns: Variable, O U D 1 (N equals 46), H C (N equals 42), and O U D 2 (N equals 46).
* Age (years): O U D 1 and O U D 2 are 43.10 plus or minus 6.95; H C is 37.69 plus or minus 10.29.
* Gender (male/female): O U D 1 and O U D 2 are 23/23; H C is 16/26.
* Education (years): O U D 1 and O U D 2 are 7.71 plus or minus 3.04; H C is 10.70 plus or minus 2.15.
* Handedness: All groups are 100 percent right-handed (46 R for O U D groups, 42 R for H C).
* Duration of drug use (years): O U D 1 and O U D 2 are 16.21 plus or minus 6.33; H C is not applicable.
* Abstinent time (days): O U D 1 median is 30 (range 5 to 65); O U D 2 median is 337 (range 104 to 530).
* Dosage of drug use (grams per day): O U D 1 median is 0.5 (range 0.01 to 4); H C and O U D 2 are not applicable.
Note: Participant characteristics of the OUD and HC groups. HC, healthy control group; OUD1, baseline opioid use disorder group; OUD2, follow-up opioid use disorder group.
a Average ± SD.
b Median (range).
This study was approved by the Institutional Review Board of the Second Xiangya Hospital, Central South University (number: 8167071216). The experimental procedures complied with the ethical guidelines of the Declaration of Helsinki. Written informed consent was obtained from participants after the experimental procedure was explained and after they read the consent form.
Image acquisition
All MRI data were gathered at the Second Xiangya Hospital of Central South University in Changsha, China, by means of a 3T MRI scanner (MAGNETOM Skyra, Siemens) equipped with a 32-channel head coil. In order to generate high-quality MRI data, participants were told to lie still with their eyes open and their heads restrained by padding and a head-restraining belt during the MRI scan. T1-weighted imaging and T2-weighted imaging scans were employed to exclude obvious structural lesions in the brain. A 3D rapid gradient echo sequence was utilized to acquire high-resolution T1-weighted MRI scans with these parameters: repetition time (TR) = 1,450 ms, echo time (TE) = 2.0 ms, inversion time (TI) = 900 ms, field of view (FOV) = 256 × 256 mm2, slice thickness = 1 mm, flip angle = 12°, voxel size = 1 × 1 × 1 mm3, and number of slices = 176. The parameters of the resting-state fMRI were as follows: TR = 2,000 ms, TE = 30 ms, flip angle = 80°, FOV = 220 × 220 mm2, slice thickness = 4 mm, and number of axial slices = 36.
Image preprocessing
The Data Processing Assistant for Resting-State fMRI (http://rfmri.org/DPARSF), Statistical Parametric Mapping toolkits (SPM12, http://www.fil.ion.ac.uk/spm), and open MATLAB scripts (https://www.beyropsychiatrylab.com/codes_wm) were used to perform image preprocessing. The first 10 volumes of fMRI scans were discarded. Subsequently, slice-timing correction and realignment were carried out. Then, T1 anatomical images were co-registered to the functional images and further segmented into tissue probability maps of white matter, gray matter, and cerebrospinal fluid (CSF) by means of a diffeomorphic nonlinear registration algorithm in SPM12. Linear trends were removed to rectify signal drift. The mean CSF signals (95% threshold) and 24-parameter motion parameters (6 motion parameters, their values at the previous time point, and the 12 corresponding squared values) were regressed out from the functional data. The mean white matter and global brain signals were not regressed out in order to avoid eliminating the meaningful neural signals (Peer et al., Reference Peer, Nitzan, Bick, Levin and Arzy2017). To reduce the potential impacts of head motion, temporal scrubbing (framewise displacement > 0.2 mm) was also applied. Bandpass filtering (0.01–0.08 Hz) was carried out to minimize high-frequency nonneuronal noise. To lessen the partial volume effect, spatial smoothing (with a 4-mm full-width at half-maximum Gaussian kernel) was separately applied to the white matter and gray matter images. Finally, the functional images were spatially normalized into the Montreal Neurologic Institute space through T1 segmentation and were resampled into 3 × 3 × 3 mm3.
Creation of group-level white matter templates
We utilized the segmentation outcomes of each participant to get a mask for voxel selection in group clustering. Every voxel was recognized as white matter, gray matter, or cerebrospinal fluid (CSF) according to the segmented maximum probability graph. Subsequently, all the masks from the participants were averaged to acquire the proportion of participants classified as white or gray matter in each voxel.
Voxels recognized as white matter in over 60% of the participants were employed to generate the template. Then, we compared the resultant masks with the functional data and eliminated voxels that were identified as white matter in <80% of participants and had functional data (such as parts of the medulla and spinal cord). We used the Harvard-Oxford Atlas to label the gray matter voxels within the thalamus, caudate nucleus, putamen, globus pallidus, and nucleus accumbens (excluded from the white matter mask) as gray matter voxels. (Lorio et al., Reference Lorio, Fresard, Adaszewski, Kherif, Chowdhury, Frackowiak, Ashburner, Helms, Weiskopf, Lutti and Draganski2016; Wang et al., Reference Wang, Xue, Song, Dong, Cheng, Wang, Ma, Zou, Ding, Tao, Xin, Yu and Yuan2024).
Neural activity analysis of WMFNs
In this study, a data-driven clustering method was used to separately create WMFNs for the baseline OUD and HC groups. According to previous studies (Bu et al., Reference Bu, Liang, Lin, Gao, Qian, Chen, Chen, Wang, Yang and Huang2020; Peer et al., Reference Peer, Nitzan, Bick, Levin and Arzy2017), first, two sets of white matter masks were created to select the voxels for clustering in individuals with OUD and healthy control individuals, respectively. More precisely, we only retained voxels that were determined to be white matter and covered more than 80% of the functional data of participants. Next, a voxel-wise Pearson correlation matrix of the white matter was calculated for each participant. These individual-level correlation matrices were then averaged for each group, yielding two group-level correlation matrices. Finally, the K-means clustering method was used to cluster white matter voxels with similar functional connections for the two group-level correlation matrices (Blumensath et al., Reference Blumensath, Jbabdi, Glasser, Van, Ugurbil, Behrens and Smith2013). The number of clusters was set between 2 and 22, and the Dice coefficient was used to evaluate cluster stability, with a value > 0.9 as the evaluation criterion (Peer et al., Reference Peer, Nitzan, Bick, Levin and Arzy2017). A power spectrum analysis has been widely used to detect significant differences in signal amplitude across different neurological regions and pathological states of gray matter (Yang et al., Reference Yang, Long, Yang, Yan, Zhu, Zhou, Zang and Gong2007; Zuo et al., Reference Zuo, Di Martino, Kelly, Shehzad, Gee, Klein, Castellanos, Biswal and Milham2010). We used this method to analyze the neural activity of white matter networks in the OUD group. Fourier transform was performed for each white matter network of each participant to extract each frequency amplitude. Frequency-power plots for each network were generated by averaging the amplitudes of all participants within each group. For the baseline and follow-up OUD group, the method is the same as above.
Statistical analysis
For each white matter network, the average amplitude of all participants was taken to represent the level of network activity. A two-sample t-test was carried out to detect the differences in neural activity between the baseline OUD group and the HC group, and a paired t-test was conducted to detect the differences in neural activity between the baseline and follow-up OUD groups. The false discovery rate (FDR) correction was employed to deal with multiple comparisons among multiple white matter networks. P < 0.05 was utilized to determine significance.
Pearson correlation analysis was applied to clarify the relationships between the abnormal neural activity of white matter fiber networks (WMFNs) and the clinical indicators of years of drug use. We used the false discovery rate (FDR) correction to handle multiple correlation analyses. P < 0.05 was used to determine significance.
Results
Clustering results of WMFNs
Using the K-means clustering method with the white matter correlation matrix, eight WMFNs were discovered in the baseline OUD group; in the HC group, seven WMFNs were found; and in the follow-up OUD group, six WMFNs were identified (Figure 1a,b). Subsequently, we qualitatively defined the WMFNs based on their corresponding relationships with the known resting-state gray matter networks (Yeo et al., Reference Yeo, Krienen, Sepulcre, Sabuncu, Lashkari, Hollinshead, Roffman, Smoller, Zöllei, Polimeni, Fischl, Liu and Buckner2011). Eventually, the frontotemporal parietal, deep frontal, inferior corticospinal-cerebellar, visual, and somatomotor networks were observed in all three groups. It is worth noting that, compared to the HC and follow-up OUD groups, the baseline OUD group lacked a deep frontoparietal network and appeared to be disrupted into the superior longitudinal fasciculus network and anterior corona radiata network. The WMFN scores for each group are shown in Figure 2.
(a) Based on the Dice coefficient, eight reliable and stable clustering results were obtained for OUD1. (b) Six reliable and stable clustering results were obtained for OUD2. (c) Seven reliable and stable clustering results were obtained for HC. HC, healthy control group; OUD1, baseline opioid use disorder group; OUD2, follow-up opioid use disorder group.
Figure 1. Long description
A three-panel layout labeled A dot O U D 1, B dot O U D 2, and C dot Control. Each panel contains a line graph above three anatomical brain views.
Line Graphs:
The Y-axis represents Dice coefficient from 0.7 to 1. The X-axis represents Number of clusters (k) from 2 to 22.
* Panel A (O U D 1): The blue line fluctuates, peaking at k equals 8 with a Y-value of 0.9203, indicated by a red arrow and a data label.
* Panel B (O U D 2): The line shows a peak at k equals 6 with a Y-value of 0.9246, highlighted by a red arrow.
* Panel C (Control): The line shows a peak at k equals 7 with a Y-value of 0.9371, highlighted by a red arrow.
Brain Maps:
Below each graph are three 3D brain models labeled Left lateral, Right lateral, and Dorsal. These models show multi-colored segmented regions in pink, yellow, green, and blue, representing the clustering results corresponding to the identified k-values for each group.
Eight WMFNs were obtained by the K-means clustering method. Among these WMFNs, eight WMFNs were found in the OUD1 group: superior longitudinal fasciculus, anterior corona radiata, frontotemporal parietal, deep frontal, inferior CST-cerebellum, forceps major, visual, and the somatomotor network; six WMFNs were found in the OUD2 group: deep frontoparietal, frontotemporal parietal, deep frontal, inferior CST-cerebellum, visual, and somatomotor network; and seven WMFNs were found in HC group: deep frontoparietal, frontotemporal parietal, deep frontal, inferior CST-cerebellum, forceps major, visual, and somatomotor network. HC, healthy control group; OUD1, baseline opioid use disorder group; OUD2, follow-up opioid use disorder group; WMFNs, white matter functional networks.
Figure 2. Long description
The diagram is organized into three vertical columns.
Column 1, O U D 1, displays eight networks:
1. Superior longitudinal fasciculus network in magenta.
2. Anterior corona radiata network in magenta.
3. Frontotemporal parietal network in cyan.
4. Deep frontal network in green.
5. Inferior C S T-Cerebellum network in yellow-green.
6. Forceps major network in purple.
7. Visual network in yellow.
8. Somaticmotor network in pink.
Column 2, O U D 2, displays six networks:
1. Deep frontoparietal network in magenta.
2. Frontotemporal parietal network in cyan.
3. Deep frontal network in green.
4. Inferior C S T-Cerebellum network in yellow-green.
5. Visual network in yellow.
6. Somaticmotor network in pink.
Note: The superior longitudinal fasciculus and forceps major networks are absent in this group.
Column 3, H C, displays seven networks:
1. Deep frontoparietal network in magenta.
2. Frontotemporal parietal network in cyan.
3. Deep frontal network in green.
4. Inferior C S T-Cerebellum network in yellow-green.
5. Forceps major network in purple.
6. Visual network in yellow.
7. Somaticmotor network in pink.
Note: The anterior corona radiata network is absent in this group.
Each network entry includes three brain orientations from left to right: a left-side view, a right-side view, and a top-down axial view, highlighting the specific spatial distribution of the colored functional regions within the white matter.
Different neural activity of WMFNs between baseline and follow-up OUD and HC groups
Eight WMFNs were obtained by the K-means clustering method in OUD1 (baseline OUD group), six WMFNs were obtained by the K-means clustering method in OUD2 (follow-up OUD group), and seven WMFNs were obtained by the K-means clustering method in the HC group. Among these WMFNs, five white matter networks were found in all three groups: frontotemporal parietal, deep frontal, visual, somatomotor, and inferior corticospinal-cerebellar networks. The deep frontoparietal network appeared to be split into the superior longitudinal fasciculus network and anterior corona radiata network in OUD1, but appeared to recover in the follow-up group. Compared with the HC/OUD2 group, the higher average amplitudes of the deep frontal network were observed in the OUD1 group; the lower average amplitudes of the frontotemporal parietal network were observed in the OUD1 group (Figure 3). Activity in the frontotemporal parietal network was negatively associated with years of opioid use (r = −0.306, p = 0.039; Figure 4).
Averaged amplitude of baseline OUD, follow-up OUD, and HCs. The left panel shows the frequency activity within each white matter functional network. The right panel shows the differences in mean amplitude for each WMFN between the three groups. An asterisk indicates a significant difference (*p < 0.05, FDR-corrected). HC, healthy control group; OUD1, baseline opioid use disorder group; OUD2, follow-up opioid use disorder group.
Figure 3. Long description
The figure consists of eight panels. Seven panels are line graphs showing amplitude of f as a function of frequency f in Hertz from 0.00 to 0.15. Each graph represents a specific white matter functional network: Deep frontoparietal, Frontotemporal parietal, Deep frontal, Inferior C S T-Cerebellum, Forceps major, Visual, and Somaticmotor. In these graphs, data is plotted for con (blue), Heroin 1 (pink), and Heroin 2 (red). Most networks show a peak in amplitude between 0.01 and 0.05 Hertz, followed by a sharp decline.
The eighth panel is a bar chart in the bottom right summarizing the mean amplitude for each network across groups: con (blue), O U D 1 (light pink), O U D 1-acr (medium pink), O U D 1-slp (purple), and O U D 2 (red). The x-axis lists the seven networks. Significant differences marked with an asterisk and brackets are shown for:
* Deep frontal: O U D 1 is significantly higher than con.
* Frontotemporal parietal: con is significantly higher than O U D 2.
* Deep frontoparietal: O U D 1-acr is significantly higher than con.
Error bars represent standard deviation. The O U D 1 group generally shows higher amplitudes in the Deep frontal and Deep frontoparietal networks compared to controls, while the con group shows higher amplitude in the Frontotemporal parietal network compared to the O U D 2 group.
The frontotemporal parietal network activity shows a significant negative correlation with the years of opioid use (p < 0.05, FDR-corrected).
Figure 4. Long description
At the top left is a 3 D translucent model of a human brain with a teal-colored region highlighted in the frontotemporal parietal area. In the upper right corner, statistical values are listed as r equals minus 0.306 and p equals 0.039. Below these elements is a scatter plot. The y-axis is labeled Frontotemporal parietal network with numerical increments from 0.00 to 0.08. The x-axis is labeled suck time (year) with numerical increments from 5 to 30. The data is represented by orange circular points scattered across the graph. A solid purple regression line starts at approximately 0.038 on the y-axis at year 0 and shows a steady linear decrease as it moves toward the right, ending near 0.025 at year 27. The highest concentration of data points is clustered between 5 and 12 years and again around 20 to 25 years.
Discussion
In the current study, we explored the functional organization of white matter functional networks (WMFNs) in OUD1, OUD2, and healthy controls (HCs); we also examined the alterations in neural activity and the interrelations of the WMFNs. A total of eight WMFNs were obtained in OUD1 (baseline OUD group), six WMFNs were obtained by the K-means clustering method in OUD2 (follow-up OUD group), and seven WMFNs were obtained by the K-means clustering method in the HC group. The deep frontoparietal network appeared to be split into the superior longitudinal fasciculus network and anterior corona radiata network in OUD1, but appeared to recover in the follow-up group. Compared with the HC/Her2 group, the higher average amplitudes of the deep frontal network were observed in the OUD1 group; the lower average amplitudes of the frontotemporal parietal network were observed in the OUD1 group.
Abnormal WMFN organization in OUD
Previous studies have identified structural abnormalities in the white matter of individuals with OUD (Lu et al., Reference Lu, Yang, Zhang, Tang, Du, Fan, Luo, Yan, Zhang, Li, Liu, von, Yu, Liu and Yuan2022). The current study further investigated potential alterations in white matter functional activity in individuals with OUD from the perspective of white matter functional networks (WMFNs).
Deep frontoparietal network
Addiction is considered to be associated with functional changes in the frontoparietal network (FPN) (Taebi et al., Reference Taebi, Becker, Klugah-Brown, Roecher, Biswal, Zweerings and Mathiak2022), which is closely related to cognitive control, motivation, and goal-directed behavior. Studies have shown that individuals addicted to cocaine exhibit reduced activity in the left frontoparietal network when processing nondrug-related reward stimuli (Costumero et al., Reference Costumero, Rosell-Negre, Bustamante, Fuentes-Claramonte, Llopis, Avila and Barros-Loscertales2018). Additionally, this network shows increased activity when processing cocaine-related images, and this activity is positively correlated with the number of years the individual has been using cocaine. This suggests that abnormal activity in the frontoparietal network may be related to deficits in cognitive control in individuals with addiction.
Alcohol dependence is thought to result from an overactive neural motivational system and an insufficient cognitive control system, and rebalancing these systems may reduce excessive drinking. Differences in functional connectivity between the frontoparietal cognitive control network and the motivational network (including the striatum and orbitofrontal cortex) have been found between individuals with alcohol dependence and healthy controls (Jansen, van Wingen, van den Brink, & Goudriaan, Reference Jansen, van Wingen, van den Brink and Goudriaan2015; Wu et al., Reference Wu, Yi and Zhou2025). This suggests that addiction may be associated with abnormalities in the functional connectivity within the frontoparietal network and between it and other brain networks. Addiction and the Role of the Frontoparietal Network in Inhibitory Control: Research on inhibitory control in individuals with addiction has shown inconsistencies in the activity of the frontoparietal network (FPN) and the ventral attention network (VAN), including results of both overactivity and underactivity (Qiu & Wang, Reference Qiu and Wang2021). These inconsistent results may reflect complex changes in cognitive control and motivational processing in individuals with addiction.
In summary, addiction is associated with abnormal activity and functional connectivity within the frontoparietal network (FPN). These abnormalities may be related to changes in cognitive control, motivational processing, and reward responses in individuals with addiction. The findings provide important insights into the neural mechanisms underlying addiction and may facilitate the development of targeted neural intervention strategies for treating addictive disorders.
Frontotemporal parietal network
The finding that activity in the frontotemporal parietal network is negatively associated with years of opioid use in the current study suggests that prolonged exposure to opioids may have detrimental effects on the function of this neural network. The frontotemporal parietal network (McClintock et al., Reference McClintock, Worhunsky, Xu, Balodis, Sinha, Miller and Potenza2019) is involved in a variety of cognitive processes, including attention, working memory, and executive functions, which are crucial for decision-making and behavioral control (Huang et al., Reference Huang, Liu, Mwansisya, Pu, Zhou, Liu, Chen, Rohrbaugh, Marienfeld, Xue and Liu2017). The reduced activity in the frontotemporal parietal network may indicate a decline in cognitive function related to long-term opioid use. This could manifest as difficulties in attention, memory, and planning, which are important for everyday functioning and sobriety. Additionally, chronic opioid use can lead to neuroadaptive changes in the brain (Volkow, Michaelides, & Baler, Reference Volkow, Michaelides and Baler2019), including alterations in neurotransmitter systems and neural circuits. The decreased activity in the network could be a result of these adaptations, which may be necessary for the brain to cope with the constant presence of opioids.
Brain plasticity after OUD withdrawal
In our study, we have observed a disintegration of the deep frontotemporal parietal network in OUD, a network that typically functions as a unified entity in healthy individuals. Specifically, this network in individuals with OUD has been split into the superior longitudinal fasciculus network and the anterior corona radiata network. This disintegration may be due to the damage that OUD inflicts on the white matter fibers of the brain, leading to interruptions in neural connections and functional dysregulation (Lu et al., Reference Lu, Yang, Zhao, Wen, Liu, Liu and Yuan2023a). The superior longitudinal fasciculus (SLF) (Janelle, Iorio-Morin, D’Amour, & Fortin, Reference Janelle, Iorio-Morin, D’Amour and Fortin2022) is a major white matter pathway that connects the frontal, parietal, and temporal lobes, playing a crucial role in language, attention, and the integration of information across these regions. The anterior corona radiata (Karababa et al., Reference Karababa, Bayazit, Kilicaslan, Celik, Cece, Karakas and Selek2015) is a bundle of fibers that projects from the cortex to other parts of the brain, involved in motor and sensory functions. The observed splitting of these networks could have significant implications for the cognitive and behavioral deficits seen in OUD, potentially affecting the ability of individuals to perform tasks that require complex cognitive processing and executive functions. This finding suggests that the integrity of white matter pathways is crucial for maintaining normal cognitive function, and that disruption of these pathways may contribute to the pathophysiology of addiction. Understanding these changes could lead to the development of new interventions aimed at restoring or bypassing these disrupted connections, potentially improving treatment outcomes for heroin addiction.
In conclusion, the abnormal organization of white matter functional networks (WMFNs) in heroin addicts reveals the impact of addiction on brain structure and function, while the recovery phenomena observed after withdrawal not only enhance our understanding of the neural mechanisms underlying heroin addiction but also offer a scientific basis for developing more effective rehabilitation strategies. The abnormalities in WMFNs highlight the profound changes that addiction can induce in the brain’s connective pathways, affecting cognitive, emotional, and behavioral functions. The observation of recovery after withdrawal underscores the brain’s capacity to reorganize and heal, suggesting that targeted interventions can harness this plasticity to facilitate recovery. Several limitations should be noted. First, the potential effectiveness of education- and exercise-based treatment in sustaining abstinence and recovery of WMFN requires further validation. Second, the sample size is small owing to the specificity of the study population, and imaging data were obtained only at baseline and the 8-month follow-up. Third, longitudinal designs with multiple time points are needed to confirm WMFN recovery. Finally, age- and gender-matched controls are required to exclude confounding by time-related effects.
Conclusion
This study reveals that opioid use disorder disrupts white matter functional networks. The deep frontoparietal network splits into two separate networks. The frontotemporal parietal network shows reduced activity linked to longer drug use. After 8 months of detoxification, partial recovery occurs. These findings demonstrate brain plasticity in white matter. By leveraging these insights, researchers and clinicians can design interventions that specifically address the neural disruptions caused by heroin addiction, potentially leading to more personalized and effective treatment approaches. These strategies may include cognitive rehabilitation, pharmacological treatments, and other therapeutic modalities aimed at restoring normal brain function and supporting long-term recovery.
Data availability statement
The raw fMRI data that support the findings of this study are not publicly available because they contain information that could compromise the privacy of participants with opioid use disorder. Processed data and analysis scripts are available from the corresponding authors upon reasonable request, subject to institutional ethics approval and a data access agreement.
Author contributions
Conceptualization: W.Y., K.Y., J.L.; Data curation: W.Y., Y.L., L.Y.; Formal analysis: W.Y., Y.L.; Funding acquisition: K.Y., J.L.; Investigation: W.Y., Y.L., X.W., J.Z.; Methodology: W.Y., Y.L., L.Y.; Project administration: K.Y., J.L.; Resources: X.W., L.Y., K.Y., J.L.; Supervision: K.Y., J.L; Software: Y.L., J.Z.; Validation: W.Y.; Visualization: W.Y.; Writing – original draft preparation: W.Y., Y.L. All authors have approved the final submitted version of the article.
Funding statement
This study was supported by grants from the National Natural Science Foundation of China (grant numbers: U22A20303, 82501862, 82571705, and 82671950), Clinical Research Center for Medical Imaging in Hunan Province (grant number: 2020SK4001), and Leading Talents of Scientific and Technological Innovation in Hunan Province in 2021 (grant number: 2021RC4016). Natural Science Basic Research Program of Shaanxi (Program No. 2026JC-JCQN-097). Medical Discipline Construction Project of Hunan Province (Internal number: 078401).
Competing interests
The authors declare no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Ethical standards
This study was conducted in accordance with the Declaration of Helsinki and approved by the Institutional Review Board of the Second Xiangya Hospital, Central South University (approval number: 8167071216). Written informed consent was obtained from all participants before their inclusion in the study. All authors have approved the manuscript in its current form.
Open access
