Conceptual disorganization and redistribution of resting-state cortical hubs in untreated first-episode psychosis: A 7T study

Conceptual disorganization (CD)—a positive aspect of formal thought disorder characterized by illogical, tangential, and loosely associated speech—is a core syndrome of schizophrenia and closely linked to real-world functional outcomes. While prior work has emphasized abnormalities in language-related regions (notably the superior temporal gyrus, STG), findings have been inconsistent and often limited to medicated, chronic samples or task-based paradigms. Few studies have examined whole-brain, network-level resting-state connectivity correlates of disorganization in first-episode psychosis (FEP). The present study asks whether acute CD in minimally treated FEP is associated with alterations in whole-brain hub organization, and whether these alterations are specific to CD versus other positive or negative symptoms. Using 7T resting-state fMRI and voxel-wise degree centrality, the authors aimed to identify brain regions with altered centrality in FEP versus healthy controls and assess how the severity of CD relates to these alterations.

Systematic reviews of neuroimaging in formal thought disorder (FTD) highlight structural and functional aberrations in language-processing regions, especially the STG, but also report null results and findings outside the language network. Limited resting-state fMRI studies in chronic, medicated schizophrenia have linked disorganization to reduced thalamo-somatosensory connectivity, disrupted frontoparietal-cerebellar coupling, and altered DLPFC connections with insula, Wernicke’s area, and premotor regions. However, several studies find no association between disorganization and resting-state connectivity, and nearly all prior work focuses on established, medicated schizophrenia. Consequently, the network-level, whole-brain functional correlates of acute disorganization in untreated FEP remain unclear. This gap motivates a discovery-based, voxel-wise approach without a priori regional assumptions.

Design and participants: Consecutive referrals to the Prevention and Early Intervention for Psychosis Program (PEPP) in London, Ontario (April 2017–July 2018) were screened. Inclusion: first-episode psychosis, lifetime antipsychotic exposure <14 days, age 16–39. Exclusion: recent substance use disorder, major head injury, significant uncontrolled medical illness, MRI contraindications. Healthy controls (HCs; n=31) were recruited via advertisement, matched on age, gender, and parental SES; no personal psychiatric history or family history of psychosis. Of 51 FEP recruits, 13 were excluded (motion/quality/anxiety/non-rest), yielding 38 FEP and 31 HCs for analysis. Patients were acutely psychotic and minimally treated at scan (mean exposure <4 days). Clinical/cognitive assessment: PANSS-8 assessed positive (P1, P3, G9) and negative (N1, N4, N6) symptoms, with P2 indexing CD. Additional measures included BNSS, CGI-S, SOFAS (functioning), and a modified digit symbol substitution test (DSST) for processing speed. High P2 (scores 4–7) denoted clinically significant CD; low P2 (1–3) denoted minimal CD, based on remission criteria conventions. Antipsychotic exposure was harmonized using WHO defined daily dose (DDD). MRI acquisition: 7T Siemens Magnetom with 32-channel head coil at CFMM, Western University. Resting-state EPI: 360 volumes, TR=1000 ms, TE=20 ms, flip angle=30°, FOV=208 mm, 63 slices, 2 mm thickness, 6 min scan, eyes open. High-res T1: TR=6000 ms, TE=2.83 ms, flip angles=4°/5°, slice thickness 0.75 mm. Preprocessing: DPARSF/DPABI pipelines with SPM12. Steps included slice-timing correction, realignment, T1 co-registration and tissue segmentation, DARTEL for normalization. Nuisance regression removed 24 motion parameters, WM and CSF signals, and linear/quadratic trends; global signal was not regressed. Temporal band-pass filter 0.01–0.1 Hz; detrending and standardization applied. Motion scrubbing used framewise displacement (FD) threshold 0.5 mm with nearest-neighbor interpolation (1 frame before, 2 after). Subjects exceeding 3 mm translation or 3° rotation were excluded; mean FD was computed for group comparisons. Degree centrality (DC): Within a gray matter mask (SPM12 tissue prior resliced to 104×104×63 and warped to native space), voxel-wise time series correlations generated a whole-brain matrix. Threshold r>0.25 yielded an undirected adjacency matrix. Binarized DC (bDC, number of suprathreshold edges) and weighted DC (wDC, sum of suprathreshold weights) were computed and z-transformed per subject (subtract whole-brain mean DC, divide by SD). Maps were registered to MNI space (3 mm isotropic) via DARTEL and smoothed (4 mm FWHM). Because bDC and wDC results were similar, bDC was emphasized. Statistical analysis: Two main contrasts: (1) FEP vs HC to identify psychosis-related centrality changes; (2) High P2 vs Low P2 to identify CD-related changes. Cluster-forming threshold p<0.001, cluster-level family-wise error (FWE) corrected p<0.05. Gender adjustment applied in FEP vs HC contrasts. Seed-based functional connectivity (FC) used 6 mm spheres centered at cluster centers-of-mass within the FEP group to map network affiliations. Specificity analyses: Univariate GLM regressed individual bDC eigenvariate from the CD-related cluster on symptom scores (P1, P2, P3, G5, G9, N1, N4, N6), SOFAS, and DSST. Spearman’s correlation tested continuous P2-bDC association. Potential confounds (mean FD, age, CGI-S) were examined for correlations with bDC.

Sample: Final analyses included 38 FEP and 31 HCs. Patients were acutely psychotic, minimally treated (<4 days mean exposure). Across the three groups (High P2, Low P2, HC), SOFAS and DSST differed significantly (SOFAS: F(2,57)=109.17, p<0.001; DSST: F(2,66)=12.259, p<0.001), while age, gender, and SES did not differ significantly. Psychosis-related centrality differences (FEP vs HC): • Reduced bDC in a right STG/insula/Heschl’s gyrus cluster in FEP vs HC (cluster size=148 voxels; peak MNI 48 −9 3; T=5.00; cluster-level FWE p=0.001). Subcomponents included Temporal_Sup_R (75 voxels), Insula_R (46), Heschl_R (22). • No regions showed increased bDC in FEP>HC. • Seed-based FC of the right STG/insula cluster in FEP (n=38) revealed connectivity with bilateral insula, middle cingulate cortex, supplementary motor area (SMA), bilateral thalamus, calcarine and lingual cortex extending into cerebellar lobules 4–6, and bilateral middle frontal gyri (largest cluster centered at MNI 51 −6 3; T=13.23; p<0.001). No group differences in seed-based FC were detected between FEP and HC. CD-related centrality differences (High P2 vs Low P2): • Increased bDC in a medial superior parietal lobule (mSPL) cluster encompassing paracentral lobule and precuneus in High P2 (n=12) vs Low P2 (n=26): cluster size=88 voxels; peak MNI −12 −30 69; T=4.93; cluster-level FWE p<0.001. • Group means (bDC eigenvariate) for this cluster: HC=0.35 (SD 0.33); Low P2=0.22 (0.23); High P2=0.68 (0.33). Comparisons: High P2 > HC (t[36]=4.94, p<0.001); Low P2 vs HC (t[55]=1.6, p=0.11). • Specificity: In a univariate GLM, only P2 significantly predicted mSPL bDC (t=2.24, p=0.035); other symptom scores, SOFAS, DSST were not significant (p=0.09–0.84). P2 correlated with mSPL bDC (Spearman’s rho=0.42, p=0.009). mSPL bDC showed no significant relations with mean FD, age, or CGI-S (p=0.17–0.58). No High vs Low P2 differences in seed-based FC were found. • Seed-based FC of the mSPL cluster in FEP showed coupling with postcentral/precentral regions, middle cingulate cortex and SMA, bilateral STG, and right middle temporal gyrus, indicating links with sensorimotor and, to a limited extent, language-associated regions. Motion and confounds: No significant differences in head motion (FD) between High and Low P2 groups (t=0.03, p=0.98). Cluster bDC measures were not systematically related to FD in patients or controls.

This study provides the first voxel-wise, whole-brain 7T resting-state characterization of acute conceptual disorganization (CD) in minimally treated first-episode psychosis (FEP). Three principal observations emerged: (1) FEP patients display reduced centrality in the right superior temporal/insula/Heschl’s gyrus cluster, a hub functionally affiliated with salience and executive systems, consistent with network-level hub disruption early in psychosis; (2) among patients, higher CD is associated with increased centrality of a medial superior parietal/paracentral-precuneus cluster, functionally connected to sensorimotor and, to a lesser extent, language regions; and (3) this redistribution of centrality is specific to CD severity and not explained by other positive/negative symptoms, functioning, processing speed, motion, age, or CGI-S. The reduction of STG/insula centrality aligns with prior findings in schizophrenia, indicating early hub attenuation without compensatory hub increases at the group level in acutely ill, minimally treated FEP. The absence of increased centrality in other regions in FEP vs HC contrasts with observations in medicated/chronic samples, suggesting that the emergence of alternative hubs (e.g., in visual or temporal regions) may relate to chronicity or treatment effects rather than acute psychosis per se. The CD-associated increase in mSPL centrality—a region not typically considered a language hub—supports a model of cortical redistribution where peripheral sensory-association nodes temporarily assume greater network roles to maintain information transfer in the face of hub disruption. Such compensatory rerouting may inadvertently shunt associative processes and contribute to disorganized speech. The functional coupling of mSPL with both sensorimotor regions and superior/middle temporal areas hints that downstream effects on the language network could mediate CD. Overall, the findings highlight symptom-specific network reconfiguration in psychosis, emphasizing that distinct pathophysiological processes may underlie different symptom dimensions.

In untreated first-episode psychosis, reduced centrality of a right temporal-salience hub (STG/insula/Heschl’s) coexists with symptom-specific increases in centrality within medial superior parietal regions among patients with pronounced conceptual disorganization. This pattern suggests a redistribution of network hubs, potentially reflecting compensatory rerouting that affects associative and language processes, thereby contributing to disorganization. The study advances understanding of the neural substrates of CD in acute psychosis using an unbiased, whole-brain approach at ultra-high field. Future research should longitudinally track symptom course to delineate the temporal dynamics of hub redistribution, assess treatment effects on network topology, employ comprehensive multidimensional assessments of FTD (linguistic and clinical), and ensure balanced sex representation to enhance generalizability.

Key limitations include: (1) lack of comprehensive, multidimensional instruments specifically quantifying all facets of formal thought disorder beyond PANSS P2 and related items; (2) potential gender imbalance with relatively fewer female participants, limiting generalizability; (3) analytic choices (e.g., correlation threshold r>0.25, cluster-extent inference) may influence spatial localization and sensitivity; (4) resting-state design cannot establish causal mechanisms; (5) the cross-sectional nature precludes inferences about the evolution of network changes with treatment or illness progression; and (6) although motion was stringently controlled and examined, residual motion-related confounds can never be entirely excluded in rs-fMRI.