Altered language network lateralization in euthymic bipolar patients: a pilot study

Bipolar disorder (BD) is a severe, chronic psychiatric illness characterized by mood episodes (depression, mania/hypomania) and intervening euthymic periods with residual cognitive and emotional dysregulation. BD shares features with major depressive disorder and schizophrenia, supporting a psychosis continuum in which BD occupies an intermediate position. Crow’s evolutionary theory posits that psychoses may arise from disruption of typical left-hemisphere dominance for language. Prior electrophysiological work has shown altered language lateralization across psychiatric populations. Resting-state fMRI enables investigation of intrinsic large-scale networks; BD research has identified altered connectivity in fronto-limbic and canonical networks, but results are heterogeneous and euthymic-phase correlates remain unclear. This pilot study tests whether euthymic BD shows altered resting-state Language Network (LN) organization and hemispheric asymmetry, expecting atypical right frontal (Broca’s homologue) involvement due to impaired left-hemisphere dominance and LN connectivity. Using ICA to extract the LN and fALFF to quantify spectral properties of LN activity, we further examined associations between LN spectral metrics and clinical symptoms.

Design: Cross-sectional resting-state fMRI study comparing euthymic bipolar disorder (BD) patients to healthy controls (HC), with network-level ICA and spectral (fALFF) analyses, plus brain–behavior correlations. Participants: 18 euthymic BD outpatients (10 females; mean age 54.50 ± 11.38 years; BD-I n=10, BD-II n=8) recruited at Padova University Hospital; inclusion: DSM diagnosis ≥1 year, non-remitting outpatients, euthymia at scan (Young Mania Rating Scale [YMRS] < 8). 16 age-matched HCs (8 females; mean age 51.16 ± 11.44 years) with no major psychiatric history, no psychotropic drugs, and no kinship with patients. All participants MRI-suitable, free of epilepsy/major neurological comorbidity. Ethics approval obtained; informed consent collected. Clinical assessment (BD only): Structured Clinical Interview for DSM-IV; YMRS (eligibility), Hamilton Depression Rating Scale (HAM-D), Altman Self-Rating Mania Scale (ASRM), State-Trait Anxiety Inventory Y1/Y2, and Positive and Negative Affect Schedule (PANAS). Recorded pharmacological treatments and clinical history (age at onset, duration, number of manic/hypomanic/depressive episodes, psychotic symptoms). MRI acquisition: Siemens MAGNETOM 1.5 T; head coil. Resting-state fMRI: 201 volumes; TR 2390 ms; TE 50 ms; flip angle 90°; matrix 64×64×36; voxel 1.8×1.8×6 mm³; 8:00 min; eyes open, relaxed, no stimuli. Structural T1-weighted: 160 sagittal slices; TR 2000 ms; TE 3.13 ms; flip angle 20°; matrix 320×320×160; voxel 0.656×0.656×1 mm³; 5:33 min. Neuroradiologist visually inspected scans. Preprocessing: Automated SPM12 pipeline: motion correction, alignment to T1, bias field correction, normalization to MNI space, spatial smoothing (6 mm FWHM). Language Network (LN) reconstruction via spatial ICA: Per subject, estimated number of independent components (ICs) via minimum description length; extracted 45–98 ICs using FastICA (deflation, tanh nonlinearity). Generated IC spatial maps (z-scored across voxels) and time series. Identified LN by template matching against an LN RSN template from prior work (MNI space); selected IC with highest spatial correlation to LN template. Group-level LN spatial maps and between-group comparison: Computed one-sample t-test on individual LN spatial maps within each group to obtain LN group correlation maps; significance at p<0.05 BH-FDR corrected for multiple comparisons. Between-group two-sample t-test (BD vs HC) on LN maps, p<0.05 BH-FDR, to locate regional LN differences. fALFF of LN time series: For each subject, computed FFT of LN time series. Derived fALFF within canonical 0.01–0.1 Hz and subdivided bands: Slow-5 (0.01–0.027 Hz), Slow-4 (0.027–0.073 Hz), Slow-3 (0.073–0.198 Hz), Slow-2 (0.198–0.25 Hz). Normalized band-specific fALFF values by the canonical fALFF to limit inter-individual confounds. Brain–behavior correlations (BD group): Pearson correlations between normalized fALFF bands and clinical/psychological scales (ASRM, HAM-D, PANAS-NA, STAI-Y1/Y2, etc.). Post hoc ROI fALFF analysis (asymmetry): Based on between-group LN spatial differences indicating abnormal right frontal operculum/insula engagement, defined spherical ROIs (radius 6 mm) at MNI: right BA44/insula 54, 21, 6 (abnormal region) and contralateral left -54, 21, 6 (typical LN site). For each ROI, extracted an ROI-representative time series via PCA on voxel time courses (first principal component). Computed ROI-specific fALFF across bands, normalized by whole-brain mean fALFF across the whole frequency range, and compared left vs right spectral power profiles within groups.

• Groups comparable demographically; no significant HC vs BD differences in age, sex, height, or weight (all t<1.0).
• LN spatial maps: In HCs, LN encompassed typical left-lateralized regions (Broca’s area/frontal operculum BA44–45, insula BA13, premotor/SMA BA6, angular gyrus BA39, superior/middle temporal gyri BA21–22). In BD, LN also included an atypical right-hemisphere region: right Broca’s homologue/frontal operculum (BA44) and anterior insula (BA13).
• Between-group LN differences (p<0.05 BH-FDR): BD > HC in right insula (BA13; MNI: 31, 21, 5). BD < HC in left middle temporal gyrus (BA21; MNI: 58, 48, 7) and right primary auditory cortex (BA41; MNI: 57, -27, 8).
• LN fALFF spectra: BD showed significantly lower amplitude than HC at frequency ranges 0.007–0.0125 Hz, 0.07–0.0825 Hz, and 0.09–0.0975 Hz.
• Brain–behavior correlations (BD group, normalized fALFF): • ASRM (mania): Slow-5 r=0.63, p=0.005 (positive); Slow-4 r=0.49, p=0.037 (positive); Slow-3 r=-0.67, p=0.003 (negative); Slow-2 r=-0.54, p=0.022 (negative). • HAM-D (depression): Slow-5 r=-0.50, p=0.034 (negative). • PANAS-NA (negative affect): Slow-5 r=-0.51, p=0.032 (negative). Interpretation: Slow-5 spectral power is lower overall in BD vs HC but within BD relates positively to mania severity and negatively to depressive symptoms/negative affect; higher-frequency bands (Slow-3, Slow-2) show the opposite association with mania.
• Post hoc ROI asymmetry (fALFF): HCs showed left>right spectral power at 0.0175–0.0275 Hz and 0.04–0.045 Hz, and right>left at 0.085–0.09 Hz; BD patients showed no left–right asymmetry across bands, indicating reduced language lateralization.
• Overall pattern supports a rightward shift with atypical right frontal operculum/insula engagement and reduced left posterior temporal activation in euthymic BD.

Findings support the hypothesis that euthymic BD exhibits altered language network lateralization consistent with Crow’s theory linking psychosis to disrupted left-hemisphere language dominance. ICA-derived LN maps revealed an atypical recruitment of right frontal operculum/anterior insula and reduced engagement of left middle temporal gyrus (MTG), a hub implicated in semantic control and language comprehension. Reduced right primary auditory cortex activity and altered temporal–frontal balance may relate to disturbed speech self-monitoring and thought organization associated with manic/psychotic symptoms. Spectrally, LN fALFF alterations were frequency-specific: reduced slow-band power in BD relative to HC, with Slow-5 showing opposite associations with mania (positive) and depression/negative affect (negative), suggesting state-relevant modulation and potential as a physiological marker of vulnerability. The absence of left–right asymmetry in ROI fALFF among BD patients mirrors patterns reported in schizophrenia and major depression, aligning with a psychosis continuum framework. Integrating spatial connectivity and spectral measures provided convergent evidence of abnormal LN hierarchy and hemispheric balance that correlate with residual symptoms during euthymia.

This pilot study demonstrates altered resting-state Language Network organization and reduced hemispheric lateralization in euthymic bipolar disorder, characterized by right frontal operculum/insula hyperactivation and left MTG hypoactivation, alongside frequency-specific LN spectral abnormalities that relate to residual mania and depression. Combining ICA-based LN mapping with fALFF offers a sensitive approach to uncover subclinical neural signatures and brain–behavior relationships in BD. Future work should replicate in larger samples, control medication effects, and extend the approach across schizophrenia and major depression to delineate shared and distinct network-level markers within a psychosis continuum, with a particular focus on insula and temporal language hubs.

• Small sample size may limit power to detect subtle effects and generalizability.
• Nearly all BD participants were medicated (mood stabilizers, antipsychotics, antidepressants, anxiolytics), which may confound BOLD and resting-state metrics.
• 1.5 T scanner and relatively coarse slice thickness may limit spatial resolution.
• Cross-sectional design precludes causal inferences about state vs trait markers.