Deficits in Prosodic Speech-in-Noise Recognition in Schizophrenia Patients and Its Association with Psychiatric Symptoms

doi:10.21203/rs.3.rs-4051474/v1

Download PDF

Research Article

Deficits in Prosodic Speech-in-Noise Recognition in Schizophrenia Patients and Its Association with Psychiatric Symptoms

https://doi.org/10.21203/rs.3.rs-4051474/v1

This work is licensed under a CC BY 4.0 License

You are reading this latest preprint version

Background

Uncertainty in speech perception and emotional disturbances are intertwined with psychiatric symptoms. How prosody embedded in target speech affects speech-in-noise recognition (SR) and is related to psychiatric symptoms in patients with schizophrenia remains unclear. This study aimed to examine the neural substrates of prosodic SR deficits and their associations with psychiatric symptom dimensions in patients with schizophrenia.

Methods

Fifty-four schizophrenia patients (SCHs) and 59 healthy control participants (HPs) completed the SR task (the target pseudosentences were uttered in neutral, happy, sad, angry, fear, and disgust prosody), positive and negative syndrome scale (PANSS) assessment, and magnetic resonance imaging scanning. We examined the deficits of the six prosodic SRs in schizophrenia patients and examined their associations with brain gray matter volume (GMV) reduction and psychiatric symptoms.

Results

Negative prosody worsened SR and reduced SR change rates across groups. SCHs had lower rates of change in prosodic SR and SR than HPs. Prosodic SR was associated with acoustic features. The GMV PLS component (covering 47 brain regions with group differences) was related to group differences in the six prosodic SRs. A happy SR was associated with the PANSS total, negative, and general scores after adjusting for covariates.

Conclusions

A better prosodic SR was related to better emotional salience, shorter duration, and lower shimmer (local) of the target sentences. The prosodic SR abnormalities in SCHs were associated with brain GMV reductions in the regions involved in sensorimotor, speech, and emotion processing. These findings suggest the possibility of improving negative symptoms by improving a happy SR in schizophrenia patients based on neuroplasticity.

gray matter volume

psychiatric symptoms

speech recognition

schizophrenia

speech prosody

Schizophrenia is a severe mental disorder with a global prevalence of approximately 1% [1, 2]. Patients with schizophrenia (SCHs) suffer from psychiatric symptoms, including positive, negative, and general syndromes. The negative syndrome is related to poor prognosis and a substantial burden of schizophrenia [3, 4]. Considering the limitations of antipsychotics in the treatment of negative symptoms [3], exploring complementary therapy is becoming urgent. Therefore, studies focusing on targeting behavioral training to improve the cognitive symptoms and personal functioning of patients with schizophrenia are emerging, but the effects of training on psychiatric dimensions are inconsistent [5–8]. Seeking intervenable behavioral targets that are directly associated with psychiatric symptoms might provide some hints for improving positive and negative symptoms. Among the potential intervention targets of schizophrenia, decreased speech recognition under auditory masking has been shown to be associated with psychiatric symptoms in schizophrenia patients [9, 10].

Previous studies have verified some theoretical assumptions about the abnormal “speech filtering” of schizophrenia patients: the perceptual integration process of suppressing auditory interference signals and capturing target speech features is disrupted in patients with schizophrenia [9, 11–16], which is associated with symptoms of thought disorders and a lack of spontaneity and fluency in speech [9, 11]. Successful speech recognition under auditory-masking (cocktail-party) environments involves multiple perceptual/cognitive processes, including target detection, selective attention, sensory/working memory, and speech production [17, 18]. This speech-in-masker recognition deficit in schizophrenia patients reflects the perceptual integration disorder they experience when processing complex auditory information, including perceptual abnormalities in speech content and form [9, 11, 19]. On the operational level, speech-in-noise recognition (SR) tasks have been applied to evaluate the effectiveness of auditory training on improving auditory cognition in patients with schizophrenia [15, 20].

Some perceptual cues (such as auditory/visual priming and perceptive spatial separation) can affect speech-on-speech recognition in schizophrenia patients [19, 21–24]. Among the cues, negative emotion embedded in target speech seems to reduce sentence-in-noise recognition in schizophrenia patients and healthy participants [22]. However, the neural mechanism underlying this process remains unknown. Previous studies have shown that the degree of SR impairment in schizophrenia patients is related to the activation intensity of the superior temporal cortex (STC) and the decrease in connection intensity between these areas and the frontal cortex [11, 19]. The AER deficits in schizophrenia patients (i.e., lower performance in recognizing and distinguishing emotional categories and components based on prosody) are related to dysfunction of the insula and its connections with the auditory cortex [25, 26]. Due to the complexity of prosodic SR tasks and the interaction between speech and emotion, the brain regions involved in prosodic SR may far exceed the neural structures mentioned above. Cognitive decompensation, emotional disturbances, and verbal thinking disorders are intertwined with psychiatric symptoms [27, 28], particularly in the case of high information uncertainty [29]. Neutral SR against speech masking is related to positive symptoms (especially thought disorders) [10, 11] and negative symptoms (especially social withdrawal) [10]. Auditory emotion recognition (AER) deficits in schizophrenia patients are associated with negative symptoms [26]. Thus, due to the synergy between abnormal speech and emotion processing in patients with schizophrenia, compared to the association between neutral SR and negative symptoms, prosodic SR may have a broader association with negative symptoms.

This study used a previously published behavioral dataset [22] and add-on neuroimaging data to explore the acoustic and neural mechanisms underlying prosodic SR and the association of prosodic SR with psychiatric symptoms in schizophrenia patients. Our hypotheses were as follows: 1) various SRs across each emotional condition might be related to the acoustic features of each emotion category; 2) in patients with schizophrenia, prosodic SR deficits are related to GMV reduction and are correlated with negative symptoms. If the hypotheses are verified, the conclusions might help in the design of behavioral intervention methods to improve psychiatric symptoms by improving the SR based on the principle of neuroplasticity.

Participants

This study was conducted in Guangzhou, China, from Sep. 2021 to Feb. 2022. Patients with schizophrenia were recruited from the Guangzhou Brain Hospital and diagnosed according to the Diagnostic and Statistical Manual of Mental Disorders-Five Edition (DSM-5)[30]. All schizophrenia patients received antipsychotic medication during this study. They were aged 18 and 59, were right-handed, had no abnormal intelligence (intelligence quotient; IQ > 70), and spoke Mandarin fluently. Exclusion criteria included hearing loss (showing any pure-tone hearing impairments for each ear at frequencies of 125, 250, 500, 1000, and 2000 Hz), vision loss, nervous system disease, alcohol or drug abuse, and treatment with electroconvulsive therapy (ECT) within the past two months. The healthy control participants (HPs) were volunteers from hospital care workers or the community around the hospital. None of the healthy participants had a history of Axis I psychiatric disorder as defined by the DSM-5. Fifty-four schizophrenia patients and 59 HPs were included in this study. Table 1 shows the characteristics of the participants [22]. Schizophrenia patients had lower IQ scores than HPs. Both the participants and the guardians of the patient participants provided written informed consent for their participation in this study. The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008. All procedures involving human subjects/patients were approved by the Medical Ethics Committee of Peking University (IRB00001052-21119).

Table 1

Characteristics of patients with schizophrenia and healthy participants
Characteristics*	SCH (n = 54)	HCs (n = 59)	t/χ2	P value	Cohen’s d [95%CI]
Age, year (SD)	31.6 (9.7)	29.6 (9.9)	1.04	0.302
Male, n (%)	27 (49.1)	29 (50)	0	1
Education year, mean (SD)	12.9 (2.3)	13.2 (2.1)	-0.71	0.479
IQ	104.3 (11.5)	114.0 (10.4)	-4.73	< 0.001	0.89 [0.5, 1.28]
Married, n (%)	13 (23.6)	19 (32.8)	1.03	0.598
Drinking, n (%)	2 (3.6)	3 (5.2)	0.62	0.430
Smoking, n (%)	8 (14.6)	13 (22.4)	0	1
BMI, mean (SD)	23.3 (4.6)	22.2 (3.7)	1.39	0.166
Chronic disease, n (%)
Hypertension	3 (5.5)	3 (5.2)	0	1
Heart Diseases	0 (0)	1 (1.7)	0	1
Diabetes	5 (9.1)	2 (3.4)	0.77	0.380
Family history (yes, %)	13 (23.6)	NA
First episode (yes, %)	7 (63.6)	NA
Age of the first onset, mean (SD)	23.9 (7.4)	NA
Illness Duration, mean (SD)	7.9 (7.2)
Chlorpromazine equivalent, mean (SD)	529.5 (263.9)	NA
Hospitalization (times)	2.1 (1.6)	NA
PANSS total, mean (SD)	71.5 (20.5)	NA
PANSS positive	17.0 (6.7)	NA
PANSS negative	17.6 (7.7)	NA
PANSS general	36.8 (9.9)	NA

* Values are mean (SD) or n (%) as appropriate. All p values are two-tailed. HCs, healthy controls; SCH, schizophrenia.

The content of this table was from part of the Table 1 of Zheng et al., (2023).

Emotional Speech Recognition under Noise Masking

Stimuli

The target speech stimuli were selected from the Mandarin Chinese auditory emotions stimulus database [31]. We selected 288 sounds (auditory sentences) from 5 speakers (Z.Y.L., Z.Q.J., C.L., T.S.S., and Z.Y.F.; three women and two men; 72 audios of each speaker) with high recognition accuracy scores. There were 48 sounds for each emotional category/block (neutrality, happiness, and four negative emotions: sadness, anger, fear, and disgust). The noise masker was steady-state speech-spectrum (Chinese speech babble) noise produced by mixing the voices of 25 speakers reading pseudosentences in neutral emotion [9, 32]. Each emotion block had 48 sentences randomly assigned to 4 signal-to-noise ratios (SNRs): -8 db, -4 db, 0 dB, and 4 dB [9, 22].

Procedure

The experiment took place in a shielding room at Guangzhou Brain Hospital. The participants sat in front of a computer and wore headphones (Sennheiser HD 350BT). The experiment began after a participant pressed the space bar on the keyboard. The intensity of the target sound was 58 dB SPL, and the maximum noise intensity was 66 dB SPL, which is within the comfortable and safe range of human hearing. After listening to each sound, the participants needed to repeat as many characters as possible. The researchers sat behind the participant and recorded the keywords repeated by the participant. A Chinese character of each of the three keywords that was correctly repeated was recorded as one point, and a Chinese character that was incorrectly repeated or not repeated was recorded as 0 points. Each participant needed to complete the SR task with 6 emotional blocks. Each block contains 48 sounds under four different SNRs (12 audios under each SNR). The playing order of the six blocks was random across participants. Intermittently between each block, participants could rest for two minutes, and it took approximately 50 minutes to complete the entire experimental task [22].

Evaluation Index of the SR task

Each participant’s data were fitted using a logistic psychometric function, y = 1/[1 + e − s (x − µ)], with the Levenberg–Marquardt method [33], where x is the SNR corresponding to y, which is the correct recognition probability for the target keywords. The threshold µ and the slope s determine the SNR corresponding to 50% correct SR and the changing rate of SR on the psychometric function, respectively. Here, the SR performance for a participant was the µ value under each of the six emotional (neutral, happy, sad, angry, fearful, and disgust) conditions. The higher the µ is, the poorer the SR is.

Positive and Negative Syndrome Scale

Psychiatric symptoms of SCHs within one week were evaluated by senior psychiatrists using the Positive and Negative Syndrome Scale (PANSS) [34]. The scale includes three subscales with 30 items: the positive symptom scale, the negative symptom scale, and the general psychopathology scale. Each item was scored from grade 1 (absent) to grade 7 (extreme). The total score is the sum of the scores of the three subscales, ranging from 30 to 120. The higher the scores of each subscale and the total score are, the more serious the mental symptoms are.

Structural MRI Data Acquisition and Preprocessing

A 3.0-Tesla MRI system (Siemens Prisma Syngo MR E11) was used to acquire the high-resolution T1-weighted structural volumetric sequence [a 256 × 256 × 208 matrix with a spatial resolution of 0.9 × 0.898 × 0.898 mm3, repetition time (TR): 8.2 ms; echo time: 2.3 ms] covering the whole brain. The scanning duration was 8 minutes. Voxel-based morphometry (VBM) was conducted using SPM12 (Statistical Parametric Mapping, Wellcome Department of Imaging Neuroscience, London, UK; https://www.fil.ion.ucl.ac.uk/spm/software/spm12/). Specifically, we first used unified segmentation in SPM12 to segment T1-weighted anatomical images into gray matter (GM), white matter (WM), and cerebral spinal fluid (CSF) [35]. Second, the Diffeomorphic Anatomical Registration through the Exponential Lie Algebra (DARTEL) registration method was applied to construct a study-specific GM template and normalize individual GM images into the template [36]. Third, to create a Jacobian-scaled warped tissue image, the average DARTEL template was transformed to the MNI spatial template of the Montreal Institute of Neurology for standardization, and an 8 mm smooth-check image was used for spatial smoothing. Finally, based on 132 whole-brain regions of interest (ROIs), including 91 cortical areas and 15 subcortical areas from the FSL Harvard-Oxford Atlas (https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/Atlases) and 26 cerebellar areas from the Anatomical Automatic Labeling (AAL) atlas (http://www.gin.cnrs.fr/en/tools/aal-aal2/), FMRIB software (https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/) was used to extract gray matter volume (GMV) in each of the 132 brain regions of each participant.

Partial least squares regression analyses

PLS analyses were performed using the R Package PLS (https://cran.r- project.org/web/packages/pls/index.html) and the R Package Morpho (https://cran.r-project. org/web/packages/Morpho/index.html). PLS incorporates a principal component analysis (PCA) method to reduce data dimensions with linear regression and obtains the components from predictive variables (e.g., the GMV in 132 brain regions: matrix X) that have maximum covariance with the dependent variables (e.g., the six prosodic SR scores: matrix Y). The PLS components are ranked by the covariance between the independent and response variables. The first few PLS components (PLS1, PLS2, PLS3, etc.) contribute most to the optimal dimensionality-reduction representation of the covariance of the original high-dimensional data matrices [37–39]. The pairs of latent variables (LVs) are linear combinations of the raw variables created by projecting the original X and Y matrices onto their respective weights (saliences). The significance of a component was tested by comparing its covariance with the distribution of the covariance via 5000 random permutation tests. Leave-one-out cross-validation was performed to evaluate estimator performance. A 5000-time bootstrap procedure was used to test the reliability of each variable in the significant PLS component [37, 40].

Prosodic SR

Figure 1A shows group-mean percent-correct recognition of the three keywords in target sentences as a function of the SNR along with the group-mean best-fitting psychometric functions in the patient and HP groups under neutral, happy, sad, angry, fearful, and disgusted emotional conditions. A 2 * 6 group (patient, control) by emotion-condition (neutrality, happiness, sadness, anger, fear, and disgust) ANOVA on SR revealed significant group (F_1,666 = 63.2, p < 0.001; partial η2 = 0.453, power = 0.828) and emotion (F_5,666 = 45.8, p < 0.001) effects; the interaction between group and emotion-condition was insignificant (F_5,666 = 0.6, p = 0.706) (Fig. 1B). Because patients had lower IQs than HCs, this variable was entered into the model as a covariate; the ANOVA results remained unchanged (left and lower panel of Fig. 1B). Compared to the HPs, patients with schizophrenia showed a greater threshold for 50% correct recognition of target speech (the higher the threshold, the lower the SR performance) under each emotion condition except for disgust. For both the patients and controls, target speech was best recognized under neutral and happy prosody conditions, followed by anger and disgust and then sadness and fear.

A 2 * 6 group by emotion-condition ANOVA of the slope revealed a significant main effect of group and emotional condition (F_1,666 = 6.2, p < 0.013; F_5,666 = 12.7, p < 0.001); the interaction was not significant (F_5,666 = 0.2, p = 0.951) (right panel of Fig. 1B). For SCHs and HPs, the slope was lower under the anger, fear, and sadness conditions, followed by the disgust, happiness, and neutral conditions. Compared to the HPs, patients with schizophrenia had greater slopes (the lower the slope was, the lower the rate of change in the SR along with the SNRs) under the fear and disgust emotion conditions.

Associations between the SR and the Acoustic Characteristics of Target Speech

To test whether the difference in the SR across emotional categories was related to the acoustic features of the target speech, we conducted a linear regression, with the SR accuracy (across 4 SNRs) as the dependent variable and the emotion category recognition (ECR) accuracy (30) and 11 acoustic parameters (duration, F0 mean, F0 SD, F0 max, F0 min, jitter [local], shimmer [local], root mean square (RMS) amplitude, harmonics-to-noise ratio, the spectral center of gravity, and spectral spread) [31] as independent variables (intensity and F0 range were removed from the model because there was high collinearity between intensity and RMS amplitude and between F0 range and F0 standard deviation). The results showed that for both HPs and SCHs, a better SR was positively associated with ECR accuracy and negatively associated with speech duration and local shimmer (for SCHs, adjusted R² = 0.187, F = 6.5, p < 0.001; for HPs, adjusted R² = 0.193, F = 6.7, p < 0.001), while holding the values of all other independent variables constant (Fig. 2A).

The acoustic characteristics of ECR accuracy, duration, and shimmer for sentences in the six emotion categories are compared in Fig. 2B. Sentence duration was shortest for the neutral and happy prosody conditions, followed by the anger and disgust conditions, and then the sadness and fear conditions, which matched the SR performance pattern under each emotion condition (Fig. 1). The shorter the duration is, the better the SR performance against noise masking. The ECR accuracy (an indicator of the validity of emotion salience based on data from the published speech database [31]) of happy sentences was lower than that of other emotions. Shimmer (local) represents the average absolute difference between the amplitudes of two consecutive periods, taking into account the maximum peak amplitude of the signal (41). Disgusted sentences had lower local shimmer than other emotional sentences.

Neural Correlates of Reduced Prosodic SR in Schizophrenia Patients

To test the hypothesis that the reduced SR in schizophrenia patients across prosodic categories was related to their reduced brain GMV, in the pooled sample of SCHs and HPs, we conducted a PLS regression analysis between the six prosodic SR scores (a 6 emotional categories by 113 participants matrix of the SR measures) and the GMV in the 47 brain ROIs (a 47 ROIs by 113 participants matrix of the GMV measures) showing group differences (FDR-corrected p < 0.05). This process yielded six sets of latent variables (LVs) capturing the GMV-SR associations, and only the first PLS component (PLS1) was significant (p = 0.001, 98.7% variances explained). The pair of PLS1 GMV scores (a linear combination of GMV strength) and PLS1 SR scores (a linear combination of the six prosodic SR scores) had the largest covariation (r = 0.345, p < 0.001) (the middle panel of Fig. 3A). Moreover, the PLS1 GMV and PLS1 SR scores were lower in SCHs than in HPs (the left and right panels of Fig. 3B). As the GMV and SR relationship could easily be an artifact of lower GMV and lower SR (two established findings) in patients rather than a real relationship between GMV and SR, we tested the association of PLS1 SR scores between PLS1 GMV scores by adjusting for group and IQ and found that the PLS1 GMV-SR association remained significant (p = 0.011; middle panel of Fig. 3B). The bootstrapping test showed that the GMV of the 47 ROIs reliably contributed to the PLS1-GMV LVs (all FDR corrected p < 0.05). The right posterior middle temporal gyrus (pMTG) had the greatest contribution to the PLS1 GMV profile (|Z| = 11), followed by the right posterior supramarginal gyrus (SMG), bilateral lingual gyrus (LG), left frontal orbital cortex, bilateral anterior/temporooccipital MTG and inferior temporal gyrus (ITG), posterior cingulate cortex, left anterior ITG, fusiform gyrus, right superior temporal gyrus (STG), left Heschl’s gyrus (HG), left postcentral gyrus, right putamen, bilateral thalamus, bilateral occipital pole, and cerebellum (left panel of Fig. 3A). All SRs in the six emotional conditions reliably contributed to the PLS-1 SR profile (right panel of Fig. 3A; FDR corrected p < 0.05). Thus, the reduced GMV (with 47 different regions having different contributions) may be related to the reduced SR performance across emotional categories in patients with schizophrenia.

Associations between the Prosodic SR and Psychiatric Symptoms

In schizophrenia patients, the PANSS total score was correlated with the SR in the happy (r = 0.340, p = 0.012, FDR-corrected p = 0.036), disgust (r = 0.340, p = 0.012, FDR-corrected p = 0.036), and neutral conditions (r = 0.3, p = 0.028, FDR-corrected p = 0.055) but not in the other conditions (Fig. 4A). After adjusting for age, sex, education, IQ, illness duration, and Chlorpromazine equivalent, only a happy SR was associated with the PANSS total score (adjusted p = 0.036). Further analysis revealed that a happy SR was associated with PANSS negative and general but not positive scores (Fig. 4B).

This study examined prosodic SR performance and its association with acoustic features of target speech, explored the neural mechanisms underlying reduced prosodic SR in schizophrenia patients, and tested the association between patients’ prosodic SR deficits and psychiatric symptoms. Schizophrenia patients performed worse in prosodic SR than in HPs. A better prosodic SR was associated with a shorter duration, lower shimmerness (local), and greater ECR accuracy for target sentences. The prosodic SR reduction in schizophrenia patients was related to a profile (a linear combination) of reduced GMV in the temporal, frontal, and lingual cortex, bilateral thalamus, and cerebellum. A happy SR was associated with total, negative, and general symptoms on the PANSS. These findings may improve negative psychiatric symptoms by improving prosodic SR based on cortical plasticity.

We found that almost all negative prosody of the target speech dampened the SR (compared to the neutral SR) in both participant groups, indicating that negative emotions embedded in speech may disturb the processing of speech meaning, making the reduced vocabulary available for thinking processing under noise masking more conspicuous [10, 11]. This might be due to the combined effect of auditory bottom-up and top-down processing. Prosodic cues may help participants follow the target speech. However, the stimuli used in this study were 12-character sentences that required auditory working memory and sustained attention against noise masking. Our results revealed that better SR was related to better ECR accuracy, shorter durations, and less shimmer target sentences. The variation in the SR under different emotional conditions may be due to the combined effect of the three acoustic features of the target sentences. For example, happy sentences had the lowest ECR accuracy; due to their short duration and medium local shimmer, the final happy SR was similar to that of neutral emotions. Moreover, the rate of change in the SR along the SNR was more easily affected by sadness, anger, and fear (relative to neutrality), indicating that participants cannot significantly improve their SR under these emotional conditions as the SNRs increase and maintain low levels of SR at various SNRs.

This study extended previous findings [11, 13, 19] by showing that SCHs showed remarkable deficits in prosodic SR and the rate of change in the SR, independent of prosodic cues within the target speech. Abnormalities in prosodic SR might be due to reduced GMV, especially in the bilateral MTG/ITG, lingual gyrus, and occipital cortex; right STG, left orbital frontal cortex, and postcingulate cortex; bilateral thalamus; and cerebellum. We also found that neutral, happy, and disgusted SR against noise masking was associated with psychiatric symptoms. A happy SR was associated with negative and general symptom dimensions after controlling for age, sex, education, IQ, illness duration, and drug dose. Previous studies have shown that neutral speech-on-speech recognition performance is associated with positive and negative symptoms on the PANSS [10], and this study suggested that involving the happy component in the speech-in-noise task seems to manifest negative and general aspects of the psychiatric symptoms of SCHs.

Happiness conveyed by vocal expressions is characterized by a faster speech rate and higher fundamental frequencies [41, 42]. An ERP study showed that SCHs may have an alteration in the processing of the happy salience of the voice [43]. Moreover, the perception of natural happiness stimuli discriminates significantly between patients with schizophrenia and healthy controls [44], and the difficulty of recognizing happy emotions is associated with the PANSS score [45]. Our study revealed that embedding happy prosody in speech did not affect SR against masking and could be a stable indicator of psychiatric symptoms. Our results supported the view that SCHs have “speech gating” deficits that are associated with psychiatric symptoms [11] and may update the understanding of this theory: happy SR deficits may reflect dysfunction of brain activities underlying the severity of psychiatric symptoms, particularly the negative and general symptom dimensions, which makes a happy SR a potential behaviorally intervention target for improving psychiatric symptoms. The underlying mechanism of the association between a happy SR and psychiatric symptoms needs further investigation.

This study has several limitations. This was a cross-sectional study, and causality could not be inferred among the associations among prosodic SR, GMV, and psychiatric symptoms. Future research examining the associations among the unaffected first-degree relatives of SCH patients may deepen the understanding of this problem. Additionally, this study did not ask the participants to rate the category or intensity of the target emotion, which may help better explain the differences in the SR among emotional conditions. The strength of this study was that we used multivariate to multivariate methods to determine the associations of prosodic SR with GMV, which is in line with the characteristics of multiple-to-multiple relationships between bioindicators and mental symptoms and tolerable to relatively small sample sizes [46–48].

Implications for clinical practice

Patients with schizophrenia exhibit multidimensional impairments in perception, thinking, emotion, and volitional behaviors. Antipsychotic drugs are currently the mainstream biological treatment for schizophrenia, but there are significant individual differences in the response of patients to antipsychotic drugs [1]. Antipsychotics can effectively alleviate positive psychiatric symptoms (such as hallucinations, delusions, agitation, etc.) in some patients [49, 50], but most of these patients have poor adherence to antipsychotic drugs due to tolerance issues or other reasons, leading to recurrent and persistent disease [1, 2]. Approximately 30% of patients, although able to benefit from medication treatment, have limited efficacy and continue to experience varying degrees of psychiatric symptoms; approximately 10–30% of patients are not sensitive to drugs [1, 50]. Therefore, exploring behavioral intervention methods that help improve the psychiatric symptoms and cognitive function of patients with schizophrenia is highly important for caring for patients with schizophrenia. This study revealed that deficits in prosodic speech-in-noise recognition, based on decreases in gray matter volume in the brain regions involved in sensory-motor, speech, and emotion processing, were associated with negative syndrome, poor insight, and emotional disturbances in patients with schizophrenia. The deficits in prosodic speech-in-noise recognition might be an intervenable behavioral target to improve psychiatric symptoms based on neuroplasticity.

This study revealed that patients with schizophrenia have lower prosodic SR, which is related to reduced brain GMV in the cerebellum and multiple temporal, occipital, and subcortical regions involved in sensory-motor, speech, and emotion processing. Schizophrenia patients and HPs have similar SR patterns across emotional categories. Better prosodic SR against noise masking was associated with better emotional salience, shorter sentence duration, and lower amplitude variation of the sound wave. A happy SR deficit in schizophrenia patients was related to negative and general psychiatric symptoms. This study sheds light on the neurocognitive and psychopathological mechanisms underlying the psychiatric symptoms of schizophrenia. Future interventional studies focusing on improving a happy SR may be a promising way to improve the negative and general symptoms of schizophrenia.

Conflict of interest

The authors declare that there are no conflicts of interest.

Funding/Acknowledgments

This work was supported by the National Natural Science Foundation of China General Project (grant number: 32271138), the Beijing Natural Science Foundation, China (grant number: 7202086), and the Planned Science and Technology Projects of Guangzhou, China (grant numbers: 202201011338 and 2023A03J0836). The funding agencies had no role in the design, analysis, interpretation, or writing of this study.

Author Contribution

Conceptualization: C. Wu, S. She, and Y. Zheng; Methodology, Data Curation, and Formal analysis: C. Wu, B. Gong, and Q. Li; Designing computer programs and visualization: C. Wu; Validation: C. Wu and B. Gong; Investigation: S She, B. Gong, Q. Li, X. Lu, Y. Liu, and Y. Xia; Resources: C. Wu, S. She, H. Wu and Y. Zheng; Writing - Original Draft: C. Wu and S She.

Data availability statement

The data that support the findings of this study are available upon request from the corresponding authors.

Statement of Ethics

The study was approved by the Medical Ethics Committee of Peking University (IRB00001052-21119). The investigation was carried out in accordance with the latest version of the Declaration of Helsinki. All participants provided written informed consent prior to participation in this study.

Marder SR, Cannon TD: Schizophrenia. N Engl J Med 2019, 381(18):1753-1761.
McCutcheon RA, Reis Marques T, Howes OD: Schizophrenia-An Overview. JAMA Psychiatry 2020, 77(2):201-210.
Aleman A, Lincoln TM, Bruggeman R, Melle I, Arends J, Arango C, Knegtering H: Treatment of negative symptoms: Where do we stand, and where do we go? Schizophr Res 2017, 186:55-62.
van der Meer L, Kaiser S, Castelein S: Negative symptoms in schizophrenia: reconsidering evidence and focus in clinical trials. Br J Psychiatry 2021, 219(1):359-360.
Molina JL, Thomas ML, Joshi YB, Hochberger WC, Koshiyama D, Nungaray JA, Cardoso L, Sprock J, Braff DL, Swerdlow NR et al: Gamma oscillations predict pro-cognitive and clinical response to auditory-based cognitive training in schizophrenia. Transl Psychiatry 2020, 10(1):405.
Mothersill D, Donohoe G: Neural Effects of Cognitive Training in Schizophrenia: A Systematic Review and Activation Likelihood Estimation Meta-analysis. Biol Psychiatry Cogn Neurosci Neuroimaging 2019, 4(8):688-696.
Lejeune JA, Northrop A, Kurtz MM: A Meta-analysis of Cognitive Remediation for Schizophrenia: Efficacy and the Role of Participant and Treatment Factors. Schizophr Bull 2021, 47(4):997-1006.
Yeo H, Yoon S, Lee J, Kurtz MM, Choi K: A meta-analysis of the effects of social-cognitive training in schizophrenia: The role of treatment characteristics and study quality. Br J Clin Psychol 2022, 61(1):37-57.
Wu C, Cao S, Zhou F, Wang C, Wu X, Li L: Masking of speech in people with first-episode schizophrenia and people with chronic schizophrenia. Schizophr Res 2012, 134(1):33-41.
Wu C, Wang C, Li L: Speech-on-speech masking and psychotic symptoms in schizophrenia. Schizophr Res Cogn 2018, 12:37-39.
Wu C, Zheng Y, Li J, She S, Peng H, Li L: Cortical Gray Matter Loss, Augmented Vulnerability to Speech-on-Speech Masking, and Delusion in People With Schizophrenia. Front Psychiatry 2018, 9:287.
Abdul Wahab NA, Zakaria MN, Abdul Rahman AH, Sidek D, Wahab S: Listening to Sentences in Noise: Revealing Binaural Hearing Challenges in Patients with Schizophrenia. Psychiatry Investig 2017, 14(6):786-794.
Ross LA, Saint-Amour D, Leavitt VM, Molholm S, Javitt DC, Foxe JJ: Impaired multisensory processing in schizophrenia: deficits in the visual enhancement of speech comprehension under noisy environmental conditions. Schizophr Res 2007, 97(1-3):173-183.
Swerdlow NR, Bhakta SG, Talledo J, Kotz J, Roberts BZ, Clifford RE, Thomas ML, Joshi YB, Molina JL, Light GA: Memantine effects on auditory discrimination and training in schizophrenia patients. Neuropsychopharmacology 2020, 45(13):2180-2188.
Swerdlow NR, Bhakta SG, Talledo J, Benster L, Kotz J, Vinogradov S, Molina JL, Light GA: Auditory discrimination and frequency modulation learning in schizophrenia patients: amphetamine within-subject dose response and time course. Psychol Med 2021:1-9.
Ramage EM, Klimas N, Vogel SJ, Vertinski M, Yerkes BD, Flores A, Sutton GP, Ringdahl EN, Allen DN, Snyder JS: Concurrent sound segregation impairments in schizophrenia: The contribution of auditory-specific and general cognitive factors. Schizophr Res 2016, 170(1):95-101.
Formisano E, Hausfeld L: The Dialog of Primary and Nonprimary Auditory Cortex at the 'Cocktail Party'. Neuron 2019, 104(6):1029-1031.
Peelle JE: Speech Comprehension: Stimulating Discussions at a Cocktail Party. Curr Biol 2018, 28(2):R68-r70.
Wu C, Zheng Y, Li J, Wu H, She S, Liu S, Ning Y, Li L: Brain substrates underlying auditory speech priming in healthy listeners and listeners with schizophrenia. Psychol Med 2017, 47(5):837-852.
Molina JL, Joshi YB, Nungaray JA, Thomas ML, Sprock J, Clayson PE, Sanchez VA, Attarha M, Biagianti B, Swerdlow NR et al: Central auditory processing deficits in schizophrenia: Effects of auditory-based cognitive training. Schizophr Res 2021, 236:135-141.
Zheng Y, Wu C, Li J, Wu H, She S, Liu S, Mao L, Ning Y, Li L: Brain substrates of perceived spatial separation between speech sources under simulated reverberant listening conditions in schizophrenia. Psychol Med 2016, 46(3):477-491.
Zheng Y, Li Q, Gong B, Xia Y, Lu X, Liu Y, Wu H, She S, Wu C: Negative-emotion-induced reduction in speech-in-noise recognition is associated with source-monitoring deficits and psychiatric symptoms in mandarin-speaking patients with schizophrenia. Compr Psychiatry 2023, 124:152395.
Li J, Han ZR, Gao MM, Sun X, Ahemaitijiang N: Psychometric properties of the Chinese version of the Difficulties in Emotion Regulation Scale (DERS): Factor structure, reliability, and validity. Psychol Assess 2018, 30(5):e1-e9.
Wu C, Zheng Y, Li J, Zhang B, Li R, Wu H, She S, Liu S, Peng H, Ning Y et al: Activation and Functional Connectivity of the Left Inferior Temporal Gyrus during Visual Speech Priming in Healthy Listeners and Listeners with Schizophrenia. Front Neurosci 2017, 11:107.
Kantrowitz JT, Hoptman MJ, Leitman DI, Moreno-Ortega M, Lehrfeld JM, Dias E, Sehatpour P, Laukka P, Silipo G, Javitt DC: Neural Substrates of Auditory Emotion Recognition Deficits in Schizophrenia. J Neurosci 2015, 35(44):14909-14921.
Gong B, Li Q, Zhao Y, Wu C: Auditory emotion recognition deficits in schizophrenia: A systematic review and meta-analysis. Asian J Psychiatr 2021, 65:102820.
Suhr JA: Executive functioning deficits in hypothetically psychosis-prone college students. Schizophr Res 1997, 27(1):29-35.
Liu J, Subramaniam M, Chong SA, Mahendran R: Maladaptive cognitive emotion regulation strategies and positive symptoms in schizophrenia spectrum disorders: The mediating role of global emotion dysregulation. Clin Psychol Psychother 2020, 27(6):826-836.
Limongi R, Silva AM, Mackinley M, Ford SD, Palaniyappan L: Active Inference, Epistemic Value, and Uncertainty in Conceptual Disorganization in First-Episode Schizophrenia. Schizophr Bull 2023, 49(Supplement_2):S115-s124.
Association AP: Diagnostic and statistical manual of mental disorders, fifth edition (DSM-5). Arlington, VA: American Psychiatric Association. 2013.
Gong B, Li N, Li Q, Yan X, Chen J, Li L, Wu X, Wu C: The Mandarin Chinese auditory emotions stimulus database: A validated set of Chinese pseudosentences. Behav Res Methods 2023, 55(3):1441-1459.
Yang Z, Jing C, Huang Q, Wu X, Wu Y, Schneider BA, Liang L: The effect of voice cuing on releasing Chinese speech from informational masking. Speech Communication 2007, 49(12):892-904.
Wolfram S: Mathematica : [electronic resource] a system for doing mathematics by computer. The American Mathematical Monthly 1989, 96(9):855.
Wu BJ, Lan TH, Hu TM, Lee SM, Liou JY: Validation of a five-factor model of a Chinese Mandarin version of the Positive and Negative Syndrome Scale (CMV-PANSS) in a sample of 813 schizophrenia patients. Schizophr Res 2015, 169(1-3):489-490.
Ashburner J, Friston KJ: Unified segmentation. Neuroimage 2005, 26(3):839-851.
Ashburner J: A fast diffeomorphic image registration algorithm. Neuroimage 2007, 38(1):95-113.
Krishnan A, Williams LJ, McIntosh AR, Abdi H: Partial Least Squares (PLS) methods for neuroimaging: a tutorial and review. Neuroimage 2011, 56(2):455-475.
Vértes PE, Rittman T, Whitaker KJ, Romero-Garcia R, Váša F, Kitzbichler MG, Wagstyl K, Fonagy P, Dolan RJ, Jones PB et al: Gene transcription profiles associated with intermodular hubs and connection distance in human functional magnetic resonance imaging networks. Philos Trans R Soc Lond B Biol Sci 2016, 371(1705).
Wu H, Wu C, Wu F, Zhan Q, Peng H, Wang J, Zhao J, Ning Y, Zheng Y, She S: Covariation between Childhood-Trauma Related Resting-State Functional Connectivity and Affective Temperaments is Impaired in Individuals with Major Depressive Disorder. Neuroscience 2021, 453:102-112.
McIntosh AR, Lobaugh NJ: Partial least squares analysis of neuroimaging data: applications and advances. Neuroimage 2004, 23 Suppl 1:S250-263.
Teixeira JP, Oliveira C, Lopes C: Vocal Acoustic Analysis – Jitter, Shimmer and HNR Parameters. Procedia Technology 2013, 9:1112-1122.
Juslin PN, Laukka P: Communication of emotions in vocal expression and music performance: different channels, same code? Psychol Bull 2003, 129(5):770-814.
Pinheiro AP, Niznikiewicz M: Altered attentional processing of happy prosody in schizophrenia. Schizophr Res 2019, 206:217-224.
Chaturvedi R, Kraus M, Keefe RSE: A new measure of authentic auditory emotion recognition: Application to patients with schizophrenia. Schizophr Res 2020, 222:450-454.
Tseng HH, Chen SH, Liu CM, Howes O, Huang YL, Hsieh MH, Liu CC, Shan JC, Lin YT, Hwu HG: Facial and prosodic emotion recognition deficits associate with specific clusters of psychotic symptoms in schizophrenia. PLoS One 2013, 8(6):e66571.
Sui J, Adali T, Pearlson G, Yang H, Sponheim SR, White T, Calhoun VD: A CCA+ICA based model for multitask brain imaging data fusion and its application to schizophrenia. Neuroimage 2010, 51(1):123-134.
Beaton D, Dunlop J, Abdi H: Partial least squares correspondence analysis: A framework to simultaneously analyze behavioral and genetic data. Psychol Methods 2016, 21(4):621-651.
Syeda WT, Wannan CMJ, Merritt AH, Raghava JM, Jayaram M, Velakoulis D, Kristensen TD, Soldatos RF, Tonissen S, Thomas N et al: Cortico-cognition coupling in treatment resistant schizophrenia. Neuroimage Clin 2022, 35:103064.
Jauhar S, Johnstone M, McKenna PJ: Schizophrenia. Lancet 2022, 399(10323):473-486.
Kane JM, Agid O, Baldwin ML, Howes O, Lindenmayer JP, Marder S, Olfson M, Potkin SG, Correll CU: Clinical Guidance on the Identification and Management of Treatment-Resistant Schizophrenia. J Clin Psychiatry 2019, 80(2).

No competing interests reported.

Download PDF

Editorial decision: Revision requested
31 May, 2024
Reviews received at journal
29 May, 2024
Reviews received at journal
16 May, 2024
Reviewers agreed at journal
13 May, 2024
Reviewers agreed at journal
08 May, 2024
Reviewers agreed at journal
08 May, 2024
Reviewers agreed at journal
19 Mar, 2024
Reviewers agreed at journal
19 Mar, 2024
Reviewers invited by journal
19 Mar, 2024
Editor assigned by journal
14 Mar, 2024
Editor invited by journal
11 Mar, 2024
Submission checks completed at journal
11 Mar, 2024
First submitted to journal
08 Mar, 2024

You are reading this latest preprint version

Deficits in Prosodic Speech-in-Noise Recognition in Schizophrenia Patients and Its Association with Psychiatric Symptoms

Status:

Version 1

Abstract

Background

Methods

Results

Conclusions

Figures

Introduction

Methods

Participants

Emotional Speech Recognition under Noise Masking

Stimuli

Procedure

Evaluation Index of the SR task

Positive and Negative Syndrome Scale

Structural MRI Data Acquisition and Preprocessing

Partial least squares regression analyses

Results

Prosodic SR

Associations between the SR and the Acoustic Characteristics of Target Speech

Associations between the Prosodic SR and Psychiatric Symptoms

Discussion

Implications for clinical practice

Conclusions

Declarations

References

Additional Declarations

Status:

Version 1