Network Localization of State and Trait of Auditory Verbal Hallucinations in Schizophrenia

Abstract

Background and Hypothesis

Neuroimaging studies investigating the neural substrates of auditory verbal hallucinations (AVH) in schizophrenia have yielded mixed results, which may be reconciled by network localization. We sought to examine whether AVH-state and AVH-trait brain alterations in schizophrenia localize to common or distinct networks.

Study Design

We initially identified AVH-state and AVH-trait brain alterations in schizophrenia reported in 48 previous studies. By integrating these affected brain locations with large-scale discovery and validation resting-state functional magnetic resonance imaging datasets, we then leveraged novel functional connectivity network mapping to construct AVH-state and AVH-trait dysfunctional networks.

Study Results

The neuroanatomically heterogeneous AVH-state and AVH-trait brain alterations in schizophrenia localized to distinct and specific networks. The AVH-state dysfunctional network comprised a broadly distributed set of brain regions mainly involving the auditory, salience, basal ganglia, language, and sensorimotor networks. Contrastingly, the AVH-trait dysfunctional network manifested as a pattern of circumscribed brain regions principally implicating the caudate and inferior frontal gyrus. Additionally, the AVH-state dysfunctional network aligned with the neuromodulation targets for effective treatment of AVH, indicating possible clinical relevance.

Conclusions

Apart from unifying the seemingly irreproducible neuroimaging results across prior AVH studies, our findings suggest different neural mechanisms underlying AVH state and trait in schizophrenia from a network perspective and more broadly may inform future neuromodulation treatment for AVH.

Introduction

Schizophrenia is a prevalent and complex psychiatric syndrome with a heterogeneous combination of symptoms.^1–4 Auditory verbal hallucinations (AVH), ie, auditory experiences (most typically, “voices”) in the absence of corresponding external stimuli,^5,6 have long been recognized as a core psychotic symptom affecting up to 60%–80% of schizophrenia patients.⁷ AVH severely impair patients’ daily functioning and life quality,^8,9 and about 25% of AVH are resistant to antipsychotic medication.¹⁰ While multiple AVH models (eg, inner speech, memory deficit, and bottom-up neurological models) have been developed,¹¹ there is still an urgent need for a more sophisticated understanding of the neural mechanisms underlying AVH in schizophrenia, which might be crucial for the development of effective treatment.

Neuroimaging, and magnetic resonance imaging (MRI) in particular, has strengthened its position as the most widely applied tool in psychiatry research.¹² Considerable neuroimaging effort has been devoted to investigate the neural substrates of AVH in schizophrenia utilizing 2 distinct yet complementary approaches, focusing on AVH-state and AVH-trait brain alterations, respectively.¹³ The former adopts functional neuroimaging techniques to examine within-subject contrasts of brain function in the presence vs absence of AVH in schizophrenia patients. The latter examines between-subject differences in brain structure or function that are linked to AVH, by performing group comparison between schizophrenia patients with and without AVH or correlation analysis with AVH severity across patients. Taking advantage of these approaches, numerous neuroimaging studies and meta-analyses have discovered a set of brain regions that are responsible for AVH in schizophrenia,^13–29 with the inferior frontal gyrus, insula, basal ganglia, superior temporal gyrus, and middle temporal gyrus being the relatively more affected brain regions. Nonetheless, the results of these investigations vary considerably; moreover, previous meta-analyses have also yielded inconsistent findings. This heterogeneity across studies may be reconciled by a generally accepted assumption that symptoms localize better to distributed brain networks than individual brain regions.³⁰

Conventional investigations into brain-behavior relationships have mapped neurological and psychiatric symptoms to specific brain regions; however, many symptoms correspond more closely to complex networks of highly interconnected regions.³¹ Functional connectivity network mapping (FCNM), a novel and well-validated approach that combines brain locations of interest with the human brain connectome derived from functional neuroimaging,^32–35 can link abnormalities in multiple different brain locations that cause the same symptom to a common symptom-specific brain network. This network localization approach has been successfully applied to a variety of neuropsychiatric symptoms,^36–46 many of which have eluded traditional regional localization, suggesting that this network-based framework can improve our ability to link symptoms to neuroanatomy. Remarkably, previous research using FCNM has shown that lesions causing auditory hallucinations localize to a common brain network defined by connectivity to the cerebellum and right superior temporal sulcus.⁴⁴ Nevertheless, this prior work has relied on observable lesions rather than subtle brain structural or functional alterations associated with AVH in the context of schizophrenia. Moreover, there is an open question of whether AVH-state and AVH-trait brain alterations in schizophrenia localize to common or distinct networks.

To address this question, we adopted the FCNM approach to investigate the brain network substrates contributing to AVH state and trait in schizophrenia, potentially unifying the heterogeneous findings across prior neuroimaging studies from a network perspective. To achieve this goal, we initially identified AVH-state and AVH-trait brain alterations in schizophrenia reported in previous literature. By integrating these affected brain locations with large-scale discovery and validation resting-state functional magnetic resonance imaging (fMRI) datasets, we then leveraged FCNM to construct AVH-state and AVH-trait dysfunctional networks. Schematic representation of the study design and analytical procedure is provided in figure 1. In addition, to assess the clinical relevance of our AVH localization, we tested whether neuromodulation targets for effective treatment of AVH were located within the AVH dysfunctional networks as it is evident that therapeutic stimulation to symptom-specific brain networks can alleviate symptom expression.⁴⁷

Fig. 1.

Study design and analytical procedure. We initially identified AVH-state (within-subject contracts of AVH presence vs absence) and AVH-trait (between-subject contrasts of AVH vs non-AVH patients) brain alterations in schizophrenia reported in previous literature. By integrating these affected brain locations with large-scale discovery (AMUD) and validation (SALD) resting-state fMRI datasets, we then adopted the FCNM approach to construct AVH-state and AVH-trait dysfunctional networks. Specifically, spheres centered at each coordinate of a contrast were firstly created and merged together to generate a contrast-specific combined seed mask. Second, based on the resting-state fMRI data, we computed a contrast seed-to-whole brain FC map for each subject. Third, the subject-level FC maps were entered into a voxel-wise one-sample t test to identify brain regions functionally connected to each contrast seed. Fourth, the resulting group-level t maps were thresholded and binarized at P < .05 corrected for multiple comparisons using a voxel-level FDR method. Finally, the binarized maps of AVH-state and AVH-trait contrasts were overlaid to produce 2 network probability maps, which were thresholded at 60% to yield AVH-state and AVH-trait dysfunctional networks, respectively. Note: AMUD, Anhui Medical University Dataset; AVH, auditory verbal hallucinations; FC, functional connectivity; FCNM, functional connectivity network mapping; FDR, false discovery rate; rs-fMRI, resting-state functional magnetic resonance imaging; SALD, Southwest University Adult Lifespan Dataset.

Materials and Methods

Study Search and Selection

Following the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines, a comprehensive and systematic literature search in the PubMed and Web of Science databases was conducted to identify relevant studies examining the neural correlates of AVH in schizophrenia, published prior to October 18, 2022. The following combination of search terms were used: (“auditory hallucinat*” OR “verbal hallucinat*”) AND (“MRI” OR “fMRI” OR “magnetic resonance imaging” OR “PET” OR “positron emission tomography” OR “SPECT” OR “single photon emission computed tomography” OR “ASL” OR “arterial spin labeling” OR “neuroimaging” OR “FC” OR “functional connectivity” OR “ReHo” OR “regional homogeneity” OR “ALFF” OR “fALFF” OR “amplitude of low-frequency fluctuation*” OR “CBF” OR “blood flow” OR “glucose metabolism” OR “VBM” OR “voxel-based morphometry” OR “DBM” OR “deformation-based morphometry” OR “GMV” OR “gray matter” OR “grey matter”). The reference lists of relevant reviews and meta-analyses were hand-searched to identify studies that were missed by the database search. We included neuroimaging studies investigating AVH-related brain alterations in schizophrenia patients. The exclusion criteria were as follows: (1) studies that did not perform voxel-based analyses; (2) studies that used region of interest rather than whole-brain analyses; (3) studies that did not report results in Talairach or Montreal Neurological Institute (MNI) space; (4) studies only comparing hallucinators to healthy controls; (5) studies on AVH outside of schizophrenia; and (6) studies with participants <10. A flow diagram of the detailed study selection process is shown in supplementary figure S1. This protocol was registered on PROSPERO (https://www.crd.york.ac.uk/PROSPERO/, registration number: CRD42023438802).

The selected studies were classified into 2 categories, focusing on AVH-state and AVH-trait brain alterations, respectively. The former studies reported within-subject contracts of brain function in the presence vs absence of AVH in schizophrenia patients. The latter studies reported between-subject contrasts of brain function or structure between schizophrenia patients with and without AVH. Since a single study may contain multiple contrasts, we focused our analysis on contrasts rather than studies. Coordinates of peak voxels of significant clusters reported in each contrast were extracted, and coordinates in Talairach space were converted to MNI space.

Discovery and Validation Datasets

Our study used Anhui Medical University Dataset (AMUD) as a discovery dataset⁴⁸ and Southwest University Adult Lifespan Dataset (SALD)⁴⁹ as a cross-scanner validation dataset. AMUD included 656 healthy adults of Chinese Han and right handedness (396 female, mean 26.57 ± 8.57 years), who were enrolled from local universities and communities through poster advertisements. Participants with neuropsychiatric or severe somatic disorders, a history of head injury with consciousness loss, MRI contraindications, or a family history of psychiatric diseases among first-degree relatives were excluded. This study was approved by the ethics committee of The First Affiliated Hospital of Anhui Medical University, and all participants provided written informed consent after being given a complete description of the study. SALD included 329 healthy adults (207 female, mean 37.81 ± 13.79 years) and full details regarding the sample (eg, informed consent, inclusion and exclusion criteria, among others) have been described in the data descriptor publication.⁴⁹ It is noteworthy that all included participants were restricted to an age range of 18–60 years to exclude the potential impact of neurodevelopment and neurodegeneration. Demographic information of the discovery and validation datasets is provided in supplementary table S1.

fMRI Data Acquisition and Preprocessing

Resting-state fMRI data of AMUD were collected on a 3.0-Tesla General Electric Discovery MR750w scanner, and those of SALD on a 3.0-Tesla Siemens Trio scanner. The fMRI parameters of the 2 datasets are provided in supplementary table S2. Participants with poor image quality (eg, visible artifacts) and incomplete brain coverage were excluded.

Resting-state fMRI data were preprocessed using Statistical Parametric Mapping software (SPM12, http://www.fil.ion.ucl.ac.uk/spm) and Data Processing & Analysis for Brain Imaging (DPABI, http://rfmri.org/dpabi).⁵⁰ The first 10 volumes for each participant were discarded to allow the signal to reach equilibrium and the participants to adapt to the scanning noise. The remaining volumes were corrected for the acquisition time delay between slices. Then, realignment was performed to correct the motion between time points. Head motion parameters were computed by estimating the translation in each direction and the angular rotation on each axis for each volume. All participants’ data were within the defined motion thresholds (ie, translational or rotational motion parameters less than 2 mm or 2°). We also calculated frame-wise displacement (FD), which indexes the volume-to-volume changes in head position. Several nuisance covariates (the linear drift, the estimated motion parameters based on the Friston-24 model, the spike volumes with FD >0.5 mm, the global signal, the white matter signal, and the cerebrospinal fluid signal) were regressed out from the data. Since global signal regression can enhance the detection of system-specific correlations and improve the correspondence to anatomical connectivity,⁵¹ we included this step in the preprocessing of resting-state fMRI data. Then, the datasets were band-pass filtered using a frequency range of 0.01–0.1 Hz. In the normalization step, individual structural images were firstly co-registered with the mean functional images; the transformed structural images were then segmented and normalized to MNI space using a high-level nonlinear warping algorithm, ie, the diffeomorphic anatomical registration through exponentiated Lie algebra technique.⁵² Next, each filtered functional volume was spatially normalized to MNI space using the deformation parameters estimated during the above step and resampled into 3-mm isotropic voxel. Finally, all data were spatially smoothed with a Gaussian kernel of 6 × 6 × 6 mm³ full-width at half maximum.

Functional Connectivity Network Mapping

We adopted the FCNM approach to construct AVH-state and AVH-trait dysfunctional networks based on the extracted coordinates of AVH-state and AVH-trait brain alterations, respectively (figure 1). First, 4-mm radius spheres centered at each coordinate of a contrast were created and merged together to generate a contrast-specific combined seed mask (henceforth referred to as the contrast seed). Second, based on the preprocessed resting-state fMRI data of AMUD, we computed a contrast seed-to-whole brain functional connectivity (FC) map for each subject, by calculating Pearson’s correlation coefficients between time courses of the contrast seed and each voxel within the whole brain, followed by Fisher’s Z-transformation to improve normality. Third, the 656 subject-level FC maps were entered into a voxel-wise one-sample t test to identify brain regions functionally connected to each contrast seed. Note that we only considered positive FC as the biological meaning of negative FC is still a matter of debate.^51,53 Fourth, the resulting group-level t maps were thresholded and binarized at P < .05 corrected for multiple comparisons using a voxel-level false discovery rate method. Finally, the binarized maps of AVH-state and AVH-trait contrasts were overlaid to produce 2 network probability maps, which were thresholded at 60% to yield AVH-state and AVH-trait dysfunctional networks, respectively.

Relation to Canonical Brain Networks

For ease of interpretability, we examined the spatial relationships between the AVH dysfunctional networks (state and trait) and 14 well-established canonical brain networks.⁵⁴ The proportion of overlapping voxels between each AVH dysfunctional network and a canonical network to all voxels within the corresponding canonical network was calculated to quantify their spatial relationship.

Relation to Neuromodulation Targets for Effective Treatment of AVH

To assess the clinical relevance of the AVH dysfunctional networks, we tested whether neuromodulation targets for effective treatment of AVH were located within these networks. By reviewing previous studies utilizing transcranial magnetic stimulation (TMS) and transcranial direct current stimulation (tDCS) to treat AVH, we identified the neuromodulation targets for effective treatment and overlaid them onto the AVH dysfunctional networks.

Validation Analyses

Several validation analyses were performed to verify the robustness of our findings. To examine the influence of dataset selection, we conducted the same analyses using an independent validation dataset (ie, cross-scanner SALD). To ensure that our results were independent of seed size, the FCNM procedure was repeated using 1- and 7-mm radius spheres. To further exclude the influence of neurodegeneration, we repeated our analyses in the young adults within an age range of 18–30 years.

Results

Included Studies

After a comprehensive literature search and selection, a total of 48 studies with 53 contrasts from 1167 AVH and 754 non-AVH patients were included in our analysis. Specifically, 16 studies with 16 within-subject contrasts from 323 AVH patients were included in the AVH-state analysis, and 32 studies with 37 between-subject contrasts from 844 AVH and 754 non-AVH patients in the AVH-trait analysis. Sample and imaging characteristics of the included studies are summarized in supplementary tables S3 and S4.

AVH-State Dysfunctional Network

The AVH-state dysfunctional network comprised a broadly distributed set of brain regions mainly including the bilateral temporal cortex (superior, middle, and inferior temporal gyrus), lateral prefrontal cortex (inferior and middle frontal gyrus), sensorimotor cortex (precentral gyrus, postcentral gyrus, and supplementary motor area), supramarginal gyrus, insula and operculum, middle and dorsal anterior cingulate cortex, and subcortical structures (putamen, thalamus, and amygdala) (figure 2A). With respect to canonical brain networks, the AVH-state dysfunctional network primarily involved the auditory (overlapping proportion: 99.5%), posterior salience (71.3%), anterior salience (63.3%), basal ganglia (49.7%), language (41.6%), and sensorimotor networks (25.2%) (figure 3).

Fig. 2.

AVH dysfunctional networks. Both AVH-state (A) and AVH-trait (B) dysfunctional networks are presented as network probability maps thresholded at 60%, showing brain regions functionally connected to more than 60% of the contrast seeds. Note: AVH, auditory verbal hallucinations; L, left; R, right.

Fig. 3.

AVH dysfunctional networks in relation to canonical brain networks. Polar plots illustrate the proportion of overlapping voxels between each AVH dysfunctional network and a canonical network to all voxels within the corresponding canonical network. Note: AVH, auditory verbal hallucinations; DMN, default mode network; LECN, left executive control network; RECN, right executive control network.

AVH-Trait Dysfunctional Network

In contrast to the distributed nature of the AVH-state dysfunctional network, the AVH-trait dysfunctional network manifested as a pattern of circumscribed brain regions including the bilateral inferior frontal gyrus, middle temporal gyrus, dorsal anterior cingulate cortex, and caudate (figure 2B). As to canonical networks, the AVH-trait dysfunctional network principally implicated the basal ganglia (23%) and language network (4.2%) (figure 3). Note that the overlapping proportion was rather low due to the small spatial extent of the AVH-trait dysfunctional network.

Clinical Relevance

The TMS targets for effective treatment of AVH reported in previous studies were Wernicke’s area (MNI coordinates: −69, −41, 11) and its right homologous region (71, −37, 10),⁵⁵ left temporo-parietal cortex (−55, −48, 18),⁵⁶ and left temporo-parietal junction (−51, −31, 23),⁵⁷ which were all located within the AVH-state dysfunctional network (figure 4). The tDCS targets for effective treatment of AVH were an anode over the left dorsolateral prefrontal cortex (−29, 60, 16) and a cathode over the left temporo-parietal junction (−55, −49, 18),^58,59 which were also localized to the AVH-state dysfunctional network (figure 4).

Fig. 4.

TMS (A) and tDCS (B) targets for effective treatment of AVH overlaid on the AVH-state dysfunctional network. Note: AVH, auditory verbal hallucinations; tDCS, transcranial direct current stimulation; TMS, transcranial magnetic stimulation.

Validation Analyses

The AVH dysfunctional networks derived from the validation SALD dataset were similar to those from the discovery AMUD dataset (supplementary figure S2). When repeating the FCNM procedure using 1- and 7-mm radius spheres, we found that the resultant AVH dysfunctional networks were nearly identical to those using the 4-mm radius sphere (supplementary figures S3 and S4). Our analyses in the young adults within an age range of 18–30 years yielded results similar to those of the main analyses in all participants (supplementary figure S5). These results supported the robustness of our findings to different datasets and methodological variation. For completeness, we constructed AVH-state and AVH-trait dysfunctional networks based on negative FC, with the former mainly consisting of the bilateral occipital cortex, prefrontal cortex, and posterior parietal cortex, and the latter of the bilateral occipital cortex (supplementary figure S6).

Discussion

By applying the novel FCNM approach to large-scale resting-state fMRI datasets, the present work mapped AVH-state and AVH-trait brain alterations in schizophrenia, which were neuroanatomically heterogeneous, to distinct and specific networks. The AVH-state dysfunctional network comprised a broadly distributed set of brain regions mainly involving the auditory, salience, basal ganglia, language, and sensorimotor networks. In contrast, the AVH-trait dysfunctional network manifested as a pattern of circumscribed brain regions principally implicating the caudate and inferior frontal gyrus. In addition, the AVH-state dysfunctional network aligned with the neuromodulation targets for effective treatment of AVH, indicating potential clinical significance. Apart from unifying the seemingly irreproducible neuroimaging results across prior AVH studies, our findings suggest different neural mechanisms underlying AVH state and trait in schizophrenia from a network perspective and more broadly may inform future neuromodulation treatment for AVH.

Although neuroimaging has been a core component of psychiatry research, results are often times variable in this domain,⁶⁰ which has raised concerns around the neurobiological, clinical, and translational value of neuroimaging findings.⁶¹ This lack of reproducibility may be attributable to a number of factors, such as low statistical power from small sample sizes, heterogeneity of patient populations, varying experimental designs, and data analysis flexibility.^60,61 Traditional neuroimaging meta-analyses provide a powerful tool to delineate a convergent set of anatomical regions in association with a disease, symptom, or psychological process across different studies.⁶² However, it is quite apparent that neural processes do not act in isolation, but are instead interconnected via distributed brain networks.³¹ Accordingly, localization of a disease, symptom, or psychological process has increasingly shifted from a primary focus on single brain regions to a focus on connected brain networks, largely benefiting from the advantages of the FCNM approach that combines brain locations of interest (eg, lesion, structural damage, functional abnormality, and neural activation) with brain functional connectome data.^31–35 Indeed, earlier work has made use of FCNM to localize lesions causing auditory hallucinations to a common brain network characterized by connectivity to the cerebellum and right superior temporal sulcus,⁴⁴ which lends validity to FCNM for network localization of auditory hallucinations. However, the previous effort has focused on observable lesions, leaving open the question of whether this is also the case for subtle brain structural or functional alterations related to AVH in schizophrenia, detected by statistical analysis as opposed to visual inspection. While extensive neuroimaging research and meta-analyses of AVH in schizophrenia have found heterogeneous locations of brain abnormalities,^13–29 our analyses mapped them to common AVH-specific dysfunctional networks, further supporting the potentially broad utility of FCNM in linking psychiatric symptoms to neuroanatomy.

The AVH-state dysfunctional network, which can be used to track the onset and offset of AVH expression in schizophrenia, comprised a broadly distributed set of brain regions mainly involving the auditory, salience, basal ganglia, language, and sensorimotor networks. As the most commonly affected network,^22,29,63,64 the auditory network has been critically implicated in multiple models of AVH, prominently the auditory-perceptual and expectation-perception models,⁶⁵ both highlighting that dysfunctional auditory network might be “turned on” and “tuned in” to process internal acoustic information at the cost of processing external sounds. It is well established that the salience network is engaged in prediction error coding.^66,67 Aberrant predictive coding error could be a mechanism by which the salience network dysfunction leads to AVH.¹⁶ The basal ganglia network, particularly the ventral striatum, is densely innervated by dopaminergic neurons. Dopamine is hypothesized to modulate prediction errors in predictive coding.⁶⁸ Aberrant dopaminergic transmission in schizophrenia^69,70 could potentially alter the inhibitory signals required for prediction error signaling, giving rise to the development of AVH.⁷¹ The most prominent brain regions of the language network are Broca’s and Wernicke’s areas in the inferior frontal gyrus and temporal-parietal junction of the left hemisphere, respectively. Several cognitive mechanisms that underlie AVH, including misattribution or impaired monitoring of inner speech and perturbed interactions between bottom-up and top-down processes in auditory perception, have been closely related to disrupted connectivity between these core language regions.^20,72 Our observation of the engagement of the sensorimotor network in AVH state is coherent with findings of past investigations.^23,25,29 Aside from a mechanistic account, there is an intriguing possibility that the involvement of the sensorimotor network could be attributed to the confound that participants indicated the presence of AVH by means of actions like button presses.²³

The AVH-trait dysfunctional network, which can be used to differentiate schizophrenia patients with and without AVH, manifested as a pattern of circumscribed brain regions principally implicating the caudate and inferior frontal gyrus. Alterations in these regions might constitute a stable vulnerability factor for AVH independent of the presence, absence, and severity of this symptom. The caudate has been thought to play a pivotal role in linguistic processes⁷³ and regulating language switch.^74,75 Interestingly, a considerable proportion of proficiently bilingual patients with AVH report alternating hallucinatory voices in more than one language.⁷⁶ Given that functional and structural abnormalities in the caudate in schizophrenia patients with AVH are evident,^76,77 it is reasonable to assume that caudate dysfunction contributes strongly to the development and maintenance of AVH. The inferior frontal gyrus is multifaceted construct ie broadly involved in speech comprehension, production of inner speech, and imagination of the speech of others.^78,79 Converging evidence points toward the central role of the inferior frontal gyrus in the pathophysiology of AVH trait. For example, Raij et al demonstrated that the subjective reality of AVH correlated strongly and specifically with the hallucination-related activation strength of the inferior frontal gyrus.⁸⁰ A more recent study by Fovet et al showed that decoding activity in the inferior frontal gyrus could predict the occurrence of AVH across schizophrenia patients.⁸¹ It is noteworthy that the selected AVH-trait studies used heterogeneous functional neuroimaging measures (eg, ReHo, ALFF/fALFF, static FC, dynamic FC, and CBF), which may characterize distinct aspects of brain function. For instance, dynamic FC is thought to reflect dynamic changes in brain functional organization and nonstationary switching of discrete brain states,⁸² and dynamic FC alterations have been evident in schizophrenia.^83,84 Thus, this heterogeneity might introduce potential biases given the possibility that changes in different measures may map to distinct brain networks.

Noninvasive neuromodulation techniques, such as TMS and tDCS, have shown promise in treating AVH.^58,85,86 Although heterogeneously distributed, the TMS^55–57 and tDCS^58,59 targets for effective treatment of AVH were found to fall within the AVH-state dysfunctional network, suggesting that these neuromodulation approaches may exert their treatment effects on AVH via modulating this network. This finding could have implications for future clinical research on important topics, eg, optimizing neuromodulation targets for AVH treatment from the network perspective and identifying network-based biomarkers for selecting individuals with AVH who will maximally benefit from neuromodulation treatment.

There are several limitations to the present study. First, we used resting-state fMRI data from large samples of healthy adults to determine network localization of AVH. It appears more reasonable to use fMRI data better matched to the demographic and clinical characteristics of the patients in the selected studies. However, prior work suggests that use of well-matched data makes little difference with respect to network localization.^36,87,88 Second, this study was retrospective rather than prospective. We used coordinate data from published studies to minimize selection bias, so future studies are required to prospectively test our findings. Third, we used a less conservative overlapping threshold (60%) to find brain regions that were connected to 60% of the contrast seeds. Individual studies selected for our analyses differed in their power and statistical approach, which may bias us against finding a common brain network. Future work accounting for these sources of variance may further improve the current methodology. Fourth, we did not consider each study’s sample size and effect size in our FCNM analysis, since there has been no consensus yet on how to account for these factors. Further investigation of their influences, in concert with analytical advances in the future, will help address this issue. Finally, the present results should not be used to minimize concerns regarding small sample size, clinical heterogeneity, experimental variability, and analytic flexibility that limit reproducibility of neuroimaging findings.^60,61 Continued effort to address these concerns will be critical to advance the field.

In conclusion, the current study combined the novel FCNM approach and large-scale resting-state fMRI datasets to map the neuroanatomically heterogeneous brain alterations associated with AVH in schizophrenia to a broadly distributed AVH-state dysfunctional network and a circumscribed AVH-trait dysfunctional network. By reconciling the inconsistent results across previous AVH studies, network localization may offer a generic unifying framework that could potentially resolve concerns around the reproducibility of neuroimaging findings pertinent to psychiatric symptoms. Moreover, our findings suggest different neural mechanisms underlying AVH state and trait in schizophrenia from a network perspective. In addition, the AVH-state dysfunctional network aligned with the neuromodulation targets for effective treatment of AVH, which may be informative about future neuromodulation treatment for AVH more broadly.

Supplementary Material

Supplementary material is available at https://academic.oup.com/schizophreniabulletin/.

Funding

This work was supported by the Anhui Provincial Natural Science Foundation (2308085MH277), the Outstanding Youth Support Project of Anhui Province Universities (gxyqZD2022026), the Scientific Research Key Project of Anhui Province Universities (2022AH051135 and 2023AH053298), the Scientific Research Foundation of Anhui Medical University (2022xkj143), the Health Scientific Research Project of Anhui Province (AHWJ2023A20032), and the Postgraduate Innovation Research and Practice Program of Anhui Medical University (YJS20230012).