Standardization of structural and functional brain integration in cannabis users
Tshetiz Dahal (MBBS)
General Physician, Clinical Researcher and Writer Lugansk State Medical University, Luhansk Oblast, 93000 Luhansk, Ukraine
Tshetiz Dahal (MBBS), General Physician, Clinical Researcher and Writer Lugansk State Medical University, Luhansk Oblast, 93000 Luhansk, Ukraine.
aAge range = 22–36 years.
Neuroimaging Data: Image data were acquired from each subject on a Siemens 3T scanner with a 32-channel coil at the University of Washington, as shown in Figure 26. 3D T1- and T2-weighted MR images were acquired at an isotropic resolution of 0.7 mm (FOV = 224 mm, matrix = 320, 256 slices). Diffusion-weighted images (DWI) were acquired isotropically at a high spatial resolution of 1.25 mm (TR/TE = 5520 ms/89.5 ms), using the high angular resolution diffusion imaging (HARDI) method, with 6 Shells with b = 1000, 2000, and 3000 s/mm2 with 270 q points distributed over three runs and three different shells. The rs-fMRI data were collected in two sessions, with EPI sequences (multiband coefficient = 8, TR/TE = 720 ms/33.1 ms, flip angle = 52°, FOV = 208 mm, spatial resolution = 2 in each session). 2 x 2mm). For rs-fMRI, participants were instructed to lie down with their eyes open, relax, look at a white cross on a dark background, think nothing, and not sleep.
Data Preprocessing: T1w images were minimally pre-processed for spatial distortion and motion correction and normalization in MNI space27. Diffusion-weighted images were also pre-processed for b0 intensity normalization, EPI distortion correction, eddy current and motion correction, and gradient non-linearity correction. All rs-fMRI data were used in 'CIFTI' format. H. Combination of cortical gray matter data modeled on the surface and subcortical gray matter data modeled on volumetric packets included in the image. All functional images were subjected to gradient equalization, EPI distortion correction, motion correction, registration of T1w scans, high-pass filtering with a cutoff of 2000 s for linear detrending, ICA-based de-noising for automatic artifact removal, Minimal preprocessing was done for bad images. Normalization of very low frequency and nonlinear components to MNI space. Details are described elsewhere28. The HCP preprocessing pipeline uses Independent Component Analysis (MELODIC, FSL-FIX) to remove artifacts and 'bad' components, as well as non-neuronal spatiotemporal components from 15 min of high-pass filtered rs-fMRI data. Did. To avoid removing interesting discrepancies from the data, a conservative, non-aggressive approach was still used in which a cutoff value of 2000 seconds was found to be better than 200 seconds in ICA-FIX29 . The rs-fMRI images were also cross-registered between subjects using the 'MSMall' algorithm30. This algorithm aligns functional networks to cortical functional maps using features derived from myelin, resting-state networks, and rs-fMRI visual field maps. Pipeline 30,31.
Network Construction: Glasser Atlas30 containing 360 regions (180 regions per hemisphere) was used to create functional and structural views of the brain. Since subcortical regions are often included in addiction studies, we used a modified version of this atlas containing 379 plots containing 19 subcortical regions. The subdivision scheme was based on changes in brain cortical architecture, function, connectivity and topography in 210 young healthy adults with HCP30. A structural connectivity matrix containing N × N elements representing normalized QA across regions was constructed for each participant. The optimal threshold was set to 0.1% of each person's maximum structural connectivity (the default threshold in DSI Studio).
We then calculated a weighted group structure matrix for each group by averaging the connectivity matrix elements for connections present in at least 75% of subjects23. In addition, a functional connectivity matrix for each individual was constructed by calculating the average time-course pairwise Pearson correlation coefficients of the 379 regions. We then thresholded the functional connectivity matrix using an optimal threshold of 0.2,27 retaining 20% of the strongest connections. The optimal threshold was determined based on a trade-off between density and overall efficiency36. The binary group function matrix for both groups was also calculated by averaging the individual matrices while retaining 20% of the strongest links. The overall procedure is shown in Figure 1.
Figure 1: A processing channel for brain structural and functional network analysis. We used fiber tractography and a subdivision scheme to construct the structural connectome for each individual. A functional connectome for each individual was also constructed by calculating the average time-course pairwise Pearson correlation coefficients of the 379 regions. Graph theory analyzes were then performed to examine the topological properties and abundant club organization of structural and functional brain networks in both healthy controls and cannabis users.
Network Topological Properties: To examine the link between cannabis use and the structural and functional connectivity of the brain, the topological properties of both structural and functional networks at the individual and group level were analyzed using the Brain Connectivity Toolbox (BCT, http://www.brainconnectivitytoolbox.org/). http://www.brain-connectivity-toolbox.net/).
To characterize the network topology of the brain, metrics of network integration (characteristic path length, global efficiency and degree), separation (clustering coefficient and modularity), and small-worldness were calculated for each network. . Details of individual properties are shown on 37 and 10.
Rich-Club Organization: In addition, we examined the effects of cannabis on abundant club tissue in the brain using methods described in 23,24. For this purpose, unweighted Rich-Club coefficients were calculated for each group mean functional network. For each k in the range [1, the maximum degree in the network], the Rich-Crab coefficient ϕ(k) was defined as the ratio of the number of connections in the sub-graph defined by nodes of less than degree k. Computes the total number of possible connections in the sub-graph
where Ek is the number of connections with a degree less than k, and Nk(Nk−1) is the total number of possible connections.
Following a similar procedure, a weighted rich-club coefficient ϕwk was computed for each group structural network. After ranking all weights of the structural network (w-ranked), ϕw(k) was computed as follows:
Where wk is the sum of the weights of links in the sub-graph of nodes with rank greater than k, and w-ranked is the vector of weights of all links in the structural network, ordered from highest to lowest weight. increase.
The normalized Rich-Club coefficients ϕnorm(k) of the structural and functional networks in each group were then calculated with respect to ϕrandom(k). It was computed as the average rich club coefficient over 1000 random networks of the same size and similar connection distribution. We use 23 μm to test whether the rich clubs of the real network significantly exceed those of the null model p < 0.05. For structural and functional networks of cannabis users and healthy controls, ϕnorm(k) is greater than 1 and within k with p < 0.05 indicated the presence of abundant club nodes. In this study, we chose k levels such that 30% of the network nodes are identified as rich club nodes.
Statistical Analysis: Differences in global and local plot metrics between cannabis users and healthy controls were assessed using t-tests. In addition, we used node-level linear regression analysis to examine the relationship between cannabis users' structural/functional network measures (grade and clustering coefficients) and time of cannabis use (TUC). Results were presented using a range of statistical significance thresholds (p < 0.05, p < 0.02, p < 0.01, and p < 0.005). Mainly due to multiple corrections, the false discovery rate (FDR) was used, uncorrected and corrected for multiple comparisons. Comparisons can be overly conservative when dealing with large numbers of nodes.
Informed Consent: Informed consent was obtained from all subjects involved in the study.
Figure 2 and Tables S1–S4 show significant differences (p < 0.05, p < 0.02, p < 0.01, and p < 0.005, uncorrected) in node degrees and clustering coefficients of structural and functional networks between groups. increase. As shown, the structural networks of cannabis users were less central (p < 0.01 , unmodified). Several nodes in the left parieto-occipital region, including V3CD, showed increased structural grade in cannabis users compared with controls. In functional networks, the left anterior cranial cranium showed a significant reduction in grade (p>0.005, uncorrected) in cannabis users.
Cannabis users also showed higher regional segregation (clustering coefficient, p<0.01 uncorrected) within frontoparietal regions, including the premotor cortex, the anterior cranial cortex, and the inferior frontal cortex of the structural network. Several areas in the posterior visual cortex, including the ventral visual cortex and V3CD, showed lower clustering coefficients in cannabis users.
Functional networks in cannabis users were also characterized by increased clustering coefficients in the left inferior frontal cortex, ventral visual cortex, FST and TG dorsal regions. Compared with the control group, the cannabis group showed less regional functional segregation within the right hemisphere in the dorsolateral prefrontal cortex, parahippocampal cortex 2, and the ventral region of the diencephalon.
Overall, none of the above significant differences between cannabis users and healthy controls survived after FDR correction.
Figure 2 : Regions showing differences in degree and clustering coefficient between cannabis users and healthy controls in (a) structural networks and (b) functional networks. The color of nodes indicates significant increases (red) or decreases (blue) in degree and clustering coefficient for cannabis users (CB) compared to healthy controls (HC). The size of nodes represents between group differences with p < 0.05, p < 0.02, p < 0.01 and p < 0.005 (uncorrected) with larger nodes showing smaller p values.
Rich-Club organization of structural and functional networks: Figure 3 and Tables S9, S10 show the spatial distribution of structurally and functionally abundant club nodes in both groups. As shown, the structure-rich clavate in both groups was mainly distributed in left bilateral frontal, temporal and occipital lobe regions and deep brain structures. Compared with controls, the structural networks of cannabis users showed higher and lower numbers of abundant club nodes within the superior and inferior temporal gyri, respectively.
Feature-rich club nodes were mainly located in the parietal and posterior regions of both groups, with minor differences. Cannabis users showed slightly fewer and more abundant club nodes within centrotemporal and parietal regions, respectively.
Figure 3: Rich club organization of (a) structural networks and (b) functional networks for cannabis users and healthy controls. The common rich club nodes in two groups are shown in blue. Few rich club nodes were only found for healthy controls (in red) or cannabis users (in green).
Post hoc Analysis: Figure 4 and Tables S5–S8 show regression results showing regions where plotted measures were significantly (p < 0.05, uncorrected) associated with TUC for structural (SN) and functional (FN) networks. In this figure, the nodes exhibit a rate of change in node degree (β coefficient) and a clustering coefficient higher than mean + 2SD with increasing TUC. In several regions of the posterior region, structural networks (within the bilateral frontal cortex, left parieto-parieto-occipital junction, right V3CD) and functional networks (within the left parahippocampal region, left ventral-medial). Visual field, left superior parietal cortex, left inferior parietal cortex, right hippocampus, right medial temporal cortex). Grades in several regions of the SN (in the left dorsolateral prefrontal cortex) and FN (in the right inferior frontal cortex, right premotor cortex) were positively correlated with TUC. Clustering coefficients of frontal and occipital multiple nodes were also positively correlated with TUC in functional and structural networks, respectively (p < < 0.01, uncorrected). The left interparietal sulcus region in the SN and the left anterior abutment, anterior cingulate gyrus and medial temporal cortex in the FN were found to be negatively associated with TUC (p<0.01, uncorrected). The left inferior frontal cortex and right intra-parietal area in the SN and the right orbital and pole-frontal cortex, right anterior cranial cortex and left tail in the FN showed opposite trends (Table S5). The above important associations between network measurements and TUC did not survive the FDR amendment. After FDR correction, only a significant association between grade and TUC within the presubiculum region persisted.
Figure 4 : Regions showing significant association with times used cannabis in (a) structural and (b) functional networks. Nodes in red and blue show a negative (NEG) and positive (POS) association with times used cannabis, respectively. The node size represents the significant level (p < 0.05, p < 0.02, p < 0.01 and p < 0.005, uncorrected) with larger nodes showing smaller p values. After FDR correction, only the PreS region was found to be statistically significant.