Functional brain alterations in auditory hallucination subtypes in individuals with auditory hallucinations without the diagnosis of specific neurological diseases and mental disorders at the current stage

Abstract Background We explored common and distinct pathological features of different subtypes of auditory hallucinations (AHs) to elucidate the underlying pathological mechanisms. Methods We recruited 39 individuals with constant commanding and commenting auditory verbal hallucinations (CCCAVHs), 49 with own thought auditory verbal hallucinations (OTAVHs), 46 with nonverbal AHs (NVAHs), 32 with replay AVHs (RAVHs), and 50 healthy controls. Functional connectivity density mapping was used to investigate global functional connectivity density (gFCD) alterations in these AH groups relative to the control group. Results We observed common brain functional alterations among four subtypes of AHs, such as increased gFCD in the bilateral superior temporal gyrus and mesial frontal lobe, and decreased gFCD in the bilateral medial prefrontal cortex. Increased gFCD was detected in the bilateral insula in CCCAVH individuals, bilateral thalamus in OTAVH individuals, bilateral precuneus in NVAH individuals, and bilateral hippocampus in RAVH individuals. The common and distinct gFCD alterations among four AH subtypes were located in main components of the frontoparietal, default mode, salience, central executive, and memory networks. Different AH subtypes exhibited specific aberrant patterns. Conclusions Our findings suggest that aberrant functional activity and metabolism in the abovementioned networks play key roles in the occurrence of AHs. Our findings provide evidence for distinct gFCD alterations in specific AH subtypes.

Although each hypothesis is supported by neuroimaging evidence and can explain certain subtypes of AHs to a degree, none of them is sufficient to fully explain all AHs. These hypotheses indicate that AHs are highly complex psychotic symptoms; hence, common and different pathological features of different AH subtypes should be explored to provide insight into the mechanisms underlying AHs (Allen et al., 2008;Ćurčić-Blake et al., 2017;Diederen, Daalman, et al., 2012;Diederen, van Lutterveld, et al., 2012;Hugdahl, 2015;Upthegrove et al., 2016).
Numerous studies have investigated the pathological features of different subtypes of AHs, but relatively few studies have adopted a uniform index to investigate functional brain alterations among different subtypes of AHs. To address this gap in the literature, we conducted a study to investigate the common and distinct pathological brain features of four subtypes of AHs in AVs-NSNMCS individuals. In this study, we adopted global functional connectivity density (gFCD) to investigate brain alterations across all samples. gFCD represents the number of connections between one voxel and other voxels in the whole brain. It can also be used as a quantitative index of neural activity in specific brain regions (Lang et al., 2015). A recent PET study indicated that gFCD is a potential biomarker of quantitative changes in glucose metabolism (Thompson et al., 2016). This research indicated that gFCD alterations reflect information communication capacity of the whole brain and can be used as a quality index to assess metabolism in specific brain regions (Qin, Xuan, Liu, Jiang, & Yu, 2015;Yin et al., 2017;Zhuo et al., 2014Zhuo et al., , 2017. In this study, we adopted the gFCD method to investigate functional brain alterations in different subtypes of AHs in AHs-NSNMCS individuals. We hypothesized that patients with different AH subtypes would be associated with both common and distinct gFCD patterns.

| Samples
We recruited individuals with constant commanding and com- that might account for it?" or "Did you at any time hear some or single words but this or these words do not make sense when there was no one around that might account for it?" (Note: this is an operational definition referenced by Johns et al., 2002); according to DSM-IV Axis-I and Axis-II, participants did not completely satisfy any specific mental disorder diagnostic criteria, and SCI-D was conducted by two senior psychiatrists with more than 10 years of experience; (c) participants did not satisfy any specific neurological disease diagnostic criteria; (d) participants did not receive any antipsychotic treatment for 2 weeks before participating in this study; and (e) participants had IQ > 80. Exclusion criteria were as follows: (a) Patients had other psychotic or affective disorders, mental retardation, alcohol dependence, drug dependence, organic brain lesions, or physical and neurological diseases; (b) history of unconsciousness for more than 5 min caused by any reason; (c) contraindications for MRI examination; (d) claustrophobia; and (e) IQ < 80. All participants were right-handed. Healthy controls were distinguished by a professional psychiatrist using SCI-D NP, which was also used by two professional psychiatrists.

| Assessment of psychotic symptoms, cognition, and AH scores
All psychiatric assessments were performed by a trained psychiatrist for psychotic symptoms using Positive and Negative Symptom Scales (PANSS; Tibber et al., 2018). Global functioning was estimated using the GAF scale (Aas, Sonesson, & Torp, 2018).
GAF was scored as the highest level of functioning over the past year, defined by the lowest score in social, psychological, or professional functioning. Psychiatric disorders in family members of the participants were quantified using the Family Interview for Genetic Studies (Wahab et al., 2015). We adopted the Auditory

| Functional MRI examination
Functional magnetic resonance imaging (fMRI) was performed using a GE Healthcare Discovery MR750 3T MRI system (General Electric) with an eight-channel phased-array head coil. Participants were instructed to lie supine, quieten their thoughts, and minimize head motion during the examination. The imaging parameters were as follows: 2,000-ms repetition time (TR), 45-ms echo time (TE), 32 slices, 4-mm slice thickness, 0.5-mm gap, 220 × 220 field of view (FOV), 64 × 64 acquisition matrix, and 90° flip angle. All scans were acquired with parallel imaging using SENSitivity Encoding (SENSE), with a SENSE factor of 2. Structural images were obtained with a high-resolution three-dimensional turbo-fast echo T1-weighted sequence with the following parameters: 8.2/3.2-ms TR/TE, 188 slices, 1-mm thickness, no gap, 256 × 256 FOV, 256 × 256 acquisition matrix, and 12° FA.

| fMRI data preprocessing
Statistical Parametric Mapping 8 (SPM8; http://www.fil.ion.ucl. ac.uk/spm) was used to process resting-state fMRI scans. To allow stabilization of the scanner and acclimation of patients to the environment, the first 10 scan volumes were discarded. The remaining volumes were corrected for slice timing and motion artefacts. fMRI data were within the allowable motion thresholds (translational and rotational motion <2 mm and 2°, respectively).
Six of the motion parameters and average blood oxygen level-dependent (BOLD) signals of the ventricles and white matter were removed. Next, framewise displacement (FD) was calculated, and data with specific-volume FD > 0.5 were excluded from the study.
The datasets were filtered with band-pass frequencies ranging from 0.01 to 0.08 Hz. Individual structural images were coregistered to the mean functional image. The transformed structural images were coregistered to the Montreal Neurological Institute (MNI) space using linear registration. The motion-corrected functional volumes were spatially normalized to the MNI space using parameters estimated during linear coregistration. Finally, the functional images were resampled into 3-mm cubic voxels for further analysis.

| gFCD calculation
The gFCD was calculated for each voxel using an in-house Linux script (Waters, 2012). Functional connectivity between voxels was evaluated using Pearson's linear correlation with a correlation coefficient threshold of R > .6. gFCD calculations were limited to voxels within the cerebral gray matter mask. The gFCD for any given voxel (x0) was calculated as the total number of functional connections [k(x0)] between x0 and all other voxels using a growth algorithm. This procedure was repeated for all x0 voxels. The rsgFCD maps were spatially smoothed using a 6 × 6 × 6-mm full-width at half maximum Gaussian kernel. Each gFCD value was divided by the mean value from all included voxels to increase the normality of the distribution.

| Statistical analysis
One-way ANOVA was used to analyze sociodemographic information, severity of psychotic symptoms, or persistent time of AHs between groups of AHs-NSNMCS individuals. Differences in gFCD among groups were tested using voxel-wise one-way analysis of covariance (ANCOVA), with age, sex, education level, GAF scores, and PANSS scores as covariates, followed by post hoc intergroup comparisons. Intergroup comparisons were conducted within a mask showing gFCD differences from the ANCOVA analysis. Multiple comparisons were corrected using the family-wise error (FWE) method with a significance threshold of p < .05. Differences in gFCD between different AH subtypes were compared with t test. To investigate the relationship between gFCD and total AHRS score, a stepwise multiple regression analysis was conducted; regions showing significant gFCD differences in AHs-NSNMCS individuals were compared with the same regions in other groups. Given the importance of AH severity for neural correlations, we examined the correlation between gFCD and AHRS score in all samples with AHs, followed by FWE correction for multiple comparisons.

| Sample demographic and clinical characteristics
Sociodemographic information of participants is depicted in Table 1. Gender, age, education level, cognitive scores, AH scores, and GAF scores were significantly different. PANSS scores remained significantly different among groups.  Table 2).  Table 2).

| Association of gFCD with AH severity
No significant correlation between gFCD and AH severity (ARHS total score and ARHS frequency) was observed in AVs-NSNMCS individuals.

| D ISCUSS I ON
In this study, we firstly proposed an operational definition of AVs-NSNMCS individuals that replaces previous definitions such as "healthy individuals with auditory verbal hallucinations," "non-psychotic individuals with voice hearing," "non-psychotic adults with frequent auditory verbal hallucinations," and "healthy individuals prone to auditory hallucinations." In previous studies, operational definitions of different samples differed, and enrolment criteria of samples emphasized items such as "hearing voices in the absence of objective stimuli" and "did not satisfy any specific mental disorders and neurological diseases." However, these samples should not be considered healthy, as many studies suggested that these samples have global functional deficits and impaired cognition. Moreover, previous studies confirmed that these samples are at high risk of developing psychosis or may need medical intervention. Hence, in this study, we standardized these samples as AVs-NSNMCS individuals.
We emphasized the features of "hearing voices in the absence of objective stimuli" and "did not currently satisfy any specific mental disorders or neurological diseases." We propose that this definition will assist investigation and facilitate therapeutic interventions to reduce the rate of conversion to psychosis or deterioration of global function. Our clinical assessment demonstrated that GAF scores and MCCB performance of these samples were impaired relative to that of healthy controls. These findings support our "AVs-NSNMCS-  The second important finding of our study was that we observed four subtypes of AVs-NSNMCS individuals with common and distinct functional brain activity (Figure 3, Peak value in Table 2  are consistent with the hypothesis that "memory and thought in-

| CON CLUS IONS
To our knowledge, this report is the first to describe gFCD alterations in AVs-NSNMCS individuals. The important findings of this study were that different subtypes of AHs were underpinned by distinct functional brain activity patterns. The common and distinct gFCD alterations among these four AHs subtypes were observed mainly in components of the frontoparietal network, default mode network, salience network, central executive network, and memory network. These findings indicated that aberrant functional activity and metabolism in the abovementioned networks may be involved in the occurrence of AHs.
Our findings provide evidence to support different hypotheses of AHs based on distinct gFCD alterations in specific subtypes of AHs.

CO N FLI C T O F I NTE R E S T
The authors declared no conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The datasets generated and analyzed during the present study are available from the corresponding author on reasonable request.