Impaired auditory feedback prediction in Alzheimer’s disease

doi:10.21203/rs.3.rs-2248797/v1

Download PDF

Research Article

Impaired auditory feedback prediction in Alzheimer’s disease

https://doi.org/10.21203/rs.3.rs-2248797/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Background: Alzheimer’s disease (AD) is a neurodegenerative disease involving cognitive impairment and abnormalities in speech and language. Here, we examine how AD affects the fidelity of auditory feedback predictions during speaking. We focus on the phenomenon of speaking-induced suppression (SIS), the auditory cortical responses’ suppression during auditory feedback processing. SIS is determined by subtracting the magnitude of auditory cortical responses during speaking from listening to playback of the same speech. Our state feedback control model of speech motor control explains SIS as arising from the onset of auditory feedback matching a prediction of that feedback onset during speaking – a prediction that is absent during passive listening to playback of the auditory feedback. Our model hypothesizes that the auditory cortical response to auditory feedback reflects the mismatch with the prediction: small during speaking, large during listening, with the difference being SIS. Normally, during speaking, auditory feedback matches its predictions, then SIS will be large. Any reductions in SIS will indicate inaccuracy in auditory feedback prediction not matching the actual feedback.

Methods: We investigated SIS in AD patients (n = 20; mean (SD) age, 60.77 (10.04); female (%), 55.00) and healthy controls (n = 12; mean (SD) age, 63.68 (6.07); female (%), 83.33) through magnetoencephalography-based functional imaging.

Results: We found a significant reduction in SIS at approximately 100 ms in AD patients compared to healthy controls (linear mixed effects model, F(1, 57.5) = 6.849, P= 0.011).

Conclusions: The results suggest that AD patients generate inaccurate auditory feedback predictions, contributing to abnormalities in AD speech.

Speaking-induced suppression

Alzheimer’s disease

State feedback control model

Efference copy

Magnetoencephalography

Abnormalities in speech production in Alzheimer’s disease (AD) have received scant attention in the literature. Yet AD patients exhibit anatomical abnormalities in the complex brain network associated with speech motor control, comprising the superior temporal, posterior parietal, premotor, and prefrontal regions. For instance, AD patients show degeneration of a posterior parietal network [1], volume decrease in the dorsolateral prefrontal cortex [2–9], and distinct temporal lobe atrophy patterns [10], as well as speech and language impairments [11, 12]. Some researchers have attempted to link changes in language abilities to cognitive decline in AD [13]. Several linguistic variables have been used to predict the onset of AD [14]. While these independently conducted assessments with speech and language components can be beneficial for identifying the early stage of AD, neurophysiological evidence of neural dysfunction during speaking [15] may provide a more sensitive prognostic measure of disease progression in AD.

The brain network associated with speech motor control is complex because the act of speaking is a dynamic process consisting of feedforward and feedback control. It entails the preparation and execution of speech motor programs (feedforward control), as well as the monitoring, and compensatory responses to sensory feedback fluctuations during sustained speech production (feedback control) [16–20]. Our previous behavioral study on AD indicated that abnormalities in speech motor control exist, revealing how AD patients respond to pitch perturbations in auditory feedback while hearing as they speak [21]. When pitch feedback is perturbed, AD patients initiate substantially larger compensatory responses than healthy individuals, perhaps due to a shift towards greater reliance on feedback control [22]. We also see such shifts towards feedback control in other neurological disorders. In patients with cerebellar degeneration, for example, we also observe abnormally large responses to pitch feedback perturbations, suggesting greater reliance on feedback control. This greater reliance on feedback control would be consistent with impairment of the cerebellum, which is thought to play a key role (feedback prediction) in the feedforward control of movement [22, 23]. It is plausible, therefore, that in AD patients, increased compensatory responses during unpredictable altered feedback may not only reflect a greater reliance on external auditory feedback in the control of speech, but this greater feedback reliance may be due to impaired feedforward control, possibly due to unreliable internal predictions of feedback.

This need for feedback predictions in the control of speech is a key part of our state feedback control (SFC) model of speech motor control, which we use to interpret the results of the auditory feedback perturbation experiments with AD patients [24, 25]. This model assumes that, while speaking, incoming auditory feedback is compared with auditory predictions that are derived from efference copy of the motor commands driving production of speech output. Any mismatch between feedback and prediction results in compensatory motor responses that correct for the feedback prediction errors. Thus, under normal conditions, the onset of speech feedback in auditory cortex is predicted from motor efference copy, creating a minimal mismatch with the feedback prediction, resulting in a minimal auditory cortical response. In contrast, during passive listening to speech, the unavailability of precise predictions results in a more pronounced mismatch and auditory cortical response. Thus, a better suppression during speaking signifies good predictions, while a smaller suppression implies inaccurate predictions. By comparing auditory cortical responses to self-produced speech with those obtained during its playback, a measure of speaking-induced suppression (SIS) may be obtained. SIS indexes how accurately feedback predictions match incoming auditory feedback [16]. A reduced SIS in AD patients compared to healthy controls would support the hypothesis that internal prediction mechanisms underlying speech motor control are faulty in AD. In this study, we test this hypothesis by examining the SIS phenomenon in AD patients. Specifically, using magnetoencephalography (MEG), we compare the magnitudes of auditory cortical response around 50 ms (M50), M100 ms (M100), and 200 ms (M200), following speech stimulus onset during speaking and listening to playback of the same stimulus.

Participants

All participants (20 AD patients and 12 age-matched controls) were recruited from research cohorts at the University of California San Francisco (UCSF) Memory and Aging Center. AD patients received a complete clinical evaluation and structural brain imaging. Patients with other dementia co-pathologies, systemic medical illnesses, or those on medications impacting central nervous system function were excluded. The eligibility criteria for age-matched controls included normal performance on cognitive tests, normal structural brain imaging, a negative Aβ-PET, the absence of a crucial cognitive decline during the previous year, neurological or psychiatric illness, and other major medical illnesses. A structural magnetic resonance image was obtained for each participant. The participants had normal hearing except for age-related high-frequency hearing loss. Moreover, the participants or their assigned surrogate decision-makers signed informed consent. The UCSF Institutional Review Board for Human Research approved all experimental procedures.

Neuropsychological assessment

Each participant underwent a structured caregiver interview to determine Clinical Dementia Rating (CDR) and CDR Sum of Boxes [26] and was assessed by Mini-Mental State Examination (MMSE) [27]. Statistical tests comparing demographic characteristics and cognitive abilities for AD patients and healthy controls were conducted using SAS 9.4 (SAS Institute Inc). Patients included in this study were in the early stages of their disease depending on CDR, CDR-Sum of Boxes, and MMSE scores (see Table 1).

Table 1

Participant demographics
	AD (n = 20)	Control (n = 12)	P-value*
Age (years)	60.77 ± 10.04	63.68 ± 6.07	0.156
Female sex, n (%)	11 (55.00)	10 (83.33)	0.139
White race, n (%)†	20 (100.00)	12 (100.00)	1.000
Education (years)	16.05 ± 2.33	17.83 ± 1.40	0.029
Right handedness, n (%)	17 (85.00)	12 (100.00)	0.274
MMSE‡	23.00 ± 4.34	29.67 ± 0.65	< 0.00001
CDR	0.83 ± 0.41	0.13 ± 0.31	0.00016
CDR-SOB	4.23 ± 2.16	0.46 ± 1.30	< 0.00001
Values for age, education, MMSE, CDR, and CDR-SOB are expressed as mean ± SD. The ages were between 49.0 and 84.0 for AD patients and 56.0 and 75.6 for healthy controls.
*Statistical testing was conducted using the Mann-Whitney U test for age, education, MMSE, CDR, and CDR-SOB; Fisher’s exact test for sex, race, and handedness.
†Race was self-reported.
‡The MMSE scores denote better cognitive function with higher scores in the range of 0 − 30.

Experimental design and procedure

The MEG experiment comprised four blocks of 74 trials each, with ~ 2.5 s per trial. In the Speak condition (blocks 1 and 3), participants were instructed to phonate the “ah” sound when a dot appeared on the projection screen and terminate phonation on arrival of a visual cue to stop. After completing a Speak condition block, a Listen condition (blocks 2 and 4) followed. During the Listen condition, participants heard a playback of the auditory feedback they heard during the preceding Speak condition block, allowing isolation of speaking-specific activity. Breaks were provided after every 15 trials and the duration of each break was the participant’s choice.

A 275-channel whole-head MEG system (Omega 2000, CTF, Coquitlam, BC, Canada; sampling rate, 1200 Hz; filtering, 0.001 − 300 Hz) recorded neurophysiological responses from participants during the experiment. Each participant lays supine with their head supported near the center of the sensor array with three localizer coils affixed to the nasion and bilateral preauricular points to determine head positioning relative to the sensor array. Head movement was measured via difference in coil locations relative to the sensor array before and after each block of trials. If movement exceeded 7 mm, the block was re-run. Auditory stimuli were delivered to participants at comfortable levels via MEG-compatible earplugs (EAR-3A, Etymotic Research, Inc., Elk Grove Village, IL), with amplitudes comparable to side-tone levels during speaking. The amplitude of the auditory stimuli in the Listen condition was identical to that of the Speak condition. Participants produced speech responses via an MEG-compatible optical microphone (Phone-Or Ltd, Or-Yehuda, Israel). Visual cues to start and stop phonation were presented against a black background at the center of a projection screen situated approximately 24 inches away from the participant’s face. All stimulus and response events were integrated with MEG traces via analog-to-digital inputs in real-time using the imaging acquisition software.

Coregistration of MEG data to individual MRI images was performed using the CTF software suite (MISL Ltd., Coquitlam, BC Canada; ctfmeg.com; version 5.2.1) by aligning the three fiducial locations of nasion, left, and right peri-auricular points on the individual’s MRI (3T, Siemens, Erlagen, Germany) with the corresponding coil positions placed during MEG collection, after which a single sphere head model was created. Then, the MRI was exported to Analyze format and warped to the standard T1 Montreal Neurological Institute (MNI) template via Statistical Parametric Mapping (SPM8, Wellcome Trust Centre for Neuroimaging, London, UK).

MEG data preprocessing

Condition-specific blocks were combined to create separate Speak and Listen MEG datasets for each participant. Twenty-nine reference sensors were used to correct distant magnetic field disturbance by calculating a synthetic 3rd order gradiometer [28, 29], and a dual signal subspace projection algorithm was applied to eliminate speech movement artifacts in biomagnetic measurements [30, 31]. The MEG data were then filtered using a 2 Hz high-pass filter to remove slow fluctuation and marked at voice onset. Trials were segmented − 100 ms to + 300 ms around phonation onset, corrected using DC-offset, and filtered from 2 − 150 Hz. Trials were rejected for artifacts if MEG sensor channels exceeded a threshold value of 1.5 pT, or speech was detected during Listen trials, with manual verification of all flagged artifacts. Data from seven AD patients (n = 7/27) and three healthy controls (n = 3/15) were omitted from further analysis, as less than 50 trials remained in a condition after artifact rejection. For artifact-free data (20 AD patients; 12 healthy controls), trials were averaged to produce a single time series per condition, and split into separate left and right hemisphere sensor arrays to capture the auditory response from each hemisphere.

Source reconstruction and auditory response

Individual trial-averaged data for left and right sensor array locations underwent Bayesian covariance beamforming [32, 33] focused on the MNI coordinates (left hemisphere: −54.3, − 26.5, 11.6; right hemisphere: 54.4, − 26.7, 11.7) linked to the primary auditory cortex in the corresponding hemisphere [34]. The resulting source time series was transformed into root mean square (RMS) activity for each time point, yielding a time series of positive-going evoked activity from within the voxel nearest each MNI coordinate for the primary auditory cortex. Maximal amplitude values and their corresponding latencies around the M50, M100, and M200 sensory peaks in each hemisphere were extracted from individual timeseries using a semi-automated process: First, each timeseries from left or right primary auditory cortex was averaged across all participants, from which a latency window around the maximum deflection of each peak was defined (± 20 ms for M50, ± 50 ms for M100, ± 50 ms for M200). Then, these windows were used to identify peak amplitudes and latencies for each individual time series (participant x condition) within each of the three sensory component windows. Next, extracted peaks and their latencies were visually confirmed and adjusted when necessary. Finally, SIS was calculated from each of these values: the ratio of the difference between peak amplitude in the Listen and Speak conditions, divided by the amplitude during the Listen condition (i.e., [Listen – Speak]/Listen). Subtracted peak latencies (Listen − Speak) examined differences in latency of the peak for each sensory component.

Statistical analysis

The distributions of peak latencies, amplitudes, and SIS values were examined for normality via Kolmogorov-Smirnov tests and then transformed using a Two-Step algorithm [35] when warranted, prior to statistical analyses. Linear mixed effects modeling (IBM SPSS Statistics, version 28) was employed to explore group-related (AD vs control) statistical differences in the amplitude, latency, and SIS response relative to the three peak components (PEAK), two hemispheres (HEMI) and, for un-subtracted measures, the speaking or listening condition (COND). The model included fixed factors of GROUP and its interactions (GROUP x PEAK, GROUP x HEMI, GROUP x PEAK x HEMI), with PARTICIPANT included as a random factor and repeated factors specified as PEAK, HEMISPHERE, and PARTICIPANT. Condition appeared within the fixed interaction terms (GROUP x COND, GROUP x HEMI x COND, GROUP x PEAK x COND, GROUP x PEAK x HEMI x COND) and as a repeated factor in models where peak amplitude (not SIS) was the dependent variable. Model intercepts were included in both fixed and random effect terms. Significance was assessed at P < 0.05.

Demographic characteristics of participants

AD patients exhibited mild disease with CDR of 0.83 ± 0.41 (mean ± standard deviation (SD)), CDR-SOB of 4.23 ± 2.16 (mean ± SD), and MMSE of 23.00 ± 4.34 (mean ± SD). Healthy controls had similar age and sex, race, and right-handedness percentages; however, they were more educated than AD patients (Table 1).

Time course of auditory cortical activity during speaking and listening

To examine SIS in AD we contrasted evoked auditory responses in the speaking and listening conditions between AD patients and healthy controls. Individual evoked response power timeseries were extracted from regions of interest in the primary auditory cortex [34] for each condition and hemisphere. The group average of each timeseries is presented in Fig. 1, where expected peaks representing the M50, M100, and M200 are observed.

Abnormal speaking-induced suppression in Alzheimer’s disease

We first analyzed group differences in peak amplitude for the speaking and listening conditions (Fig. 2C − F). While no overall main effect of GROUP was observed in the peak amplitude, significant interactions with GROUP were observed with peak component (GROUP x PEAK, F(4,212.8) = 16.999, P < 0.001), hemisphere (GROUP x HEMI, F(2,295.8) = 7.137, P < 0.001), and speaking/listening condition (GROUP x COND, F(2,295.8) = 4.247, P = 0.015), as well as a 3-way interaction with condition and peak component (GROUP x PEAK x COND, F(4,212.8) = 3.610, P = 0.007).

Further analyses within each of the three component peaks revealed that amplitude was, across condition, not statistically different between the two participant groups (GROUP; for M50, F(1,29.6) = 0.437, P = 0.514; for M100, F(1,33.4) = 0.662, P = 0.422; for M200, F(1,31.0) = 0.288, P = 0.595). However, when separating speaking (GROUP; for M50, F(1,32.8) = 0.739, P = 0.396; for M100, F(1,19.1) = 6.617, P = 0.019; for M200, F(1,29.0) = 0.100, P = 0.754) and listening (GROUP; for M50, F(1,28.0) = 0.030, P = 0.864; for M100, F(1,27.5) = 0.263, P = 0.612; for M200, F(1,28.8) = 0.378, P = 0.544) conditions, we observed significant group effects for M100 amplitude during speaking for the patient group as compared to healthy participants. No substantial M100 amplitude reduction during speaking is apparent in AD patients (Fig. 1A and B; Fig. 2C and D). We directly compare this effect at the M100 peak component using the SIS measure below.

To determine whether SIS amplitudes in AD differ from SIS amplitudes in healthy controls, the amplitude difference in auditory cortical responses was analyzed using SIS. Consistent with the findings above, a group difference in SIS was found across the three peaks of M50, M100, and M200 (GROUP x PEAK, F(4,111.8) = 4.245, P = 0.003), again with no hemispheric difference between AD patients and healthy controls (GROUP x HEMI, F(2,139.3) = 1.189, P = 0.308). Post hoc tests confirmed that reduced SIS for AD patients relative to healthy age-matched participants occurred specifically at the M100 (GROUP; for M50, F(1,30.0) = 0.326, P = 0.573; for M100, F(1,57.5) = 6.849, P = 0.011; for M200, F(1,30.0) = 0.259, P = 0.615). In healthy controls, SIS was extant at M100 in both hemispheres (Fig. 2A and B; Supplementary Table 1), confirming previous findings that SIS originated primarily from M100 responses [16, 34, 36, 37]. In the left hemisphere, SIS values at M100 were − 0.06 ± 0.10 (mean ± standard error (SE)) for AD patients and 0.41 ± 0.13 (mean ± SE) for healthy controls (Fig. 2A; Supplementary Table 1). Likewise, in the right hemisphere, healthy controls had a substantially higher SIS than AD patients (AD patients, − 0.04 ± 0.12 (mean ± SE); healthy controls, 0.16 ± 0.16 (mean ± SE); Fig. 2B; Supplementary Table 1). A diminished SIS in AD patients appears to be due to the substantial contribution of unsuppressed peak amplitudes at M100 during speaking, rather than the impact of decreased peak amplitudes at M100 during listening (see above).

Peak latency of auditory cortical activity

Group differences in peak latency between AD patients and healthy controls (Fig. 3C − F) varied across the three peaks (GROUP x PEAK, F(4,180.6) = 320.142, P < 0.001). However, this effect was largely due to differences in the within-group latency pattern across peak rather than a between-group difference at particular peaks (GROUP, M50, F(1,106.5) = 1.095, P = 0.298; M100, F(1,30.1) = 1.474, P = 0.234; M200, F(1,38.2) = 0.023, P = 0.879).

To examine the temporal relationship between the placement of peaks during speaking and listening conditions, we subtracted peak latency of the speaking condition from that of the listening condition for each peak component (Supplementary Table 1; Fig. 3A and B). The negative latency differences — occuring across all groups, peak components, and hemispheres — support an overall delay in peak activity during speaking relative to listening condition. Group differences in this temporal delay occurred across peaks (GROUP x PEAK, F(4,96.6) = 6.149, P < 0.001) and, as with un-subtracted latencies, reflected a different within-group pattern across peaks rather than differences between groups at the individual M50, M100, or M200 components.

This is the first study to demonstrate that SIS of the M100 responses from auditory cortex is absent in AD patients, while SIS is evident in matched older adult controls. The reduced SIS in AD patients compared to healthy controls supports the hypothesis that internal prediction mechanisms underlying speech motor control are faulty in AD. Here we discuss why impaired auditory feedback prediction processes would lead to reduction in SIS.

Speakers appear to monitor their sensory feedback during speaking, comparing incoming feedback with feedback predictions – a process that is predominantly automatic, unconscious and prospective [19, 24, 34, 36, 38, 39]. Speakers experience self-agency only when auditory feedback minimally deviates from predicting what they expect to hear [40, 41]. When speakers hear minimal perturbations of their auditory feedback while speaking, they typically make compensatory corrective responses that oppose the perturbation direction, showing that they judge the perturbations to be errors in their speech output [19, 21, 24, 34, 36].

These compensatory responses to feedback perturbations are accounted for in our SFC model of speech motor control. In the SFC model, the state of the vocal tract articulators is continually being estimated during speaking. The estimated state is compared with the desired state of the articulators appropriate for the current speech sound being produced, and controls are issued to the vocal tract to make the estimated state track the desired state. The current articulatory state is estimated via a prediction/correction process, where the next state of the articulators is first predicted from the previous estimate and efference copy of the controls currently being issued to the vocal tract. This state prediction is then used to predict the current sensory feedback expected from the vocal tract. Incoming sensory feedback is compared with these predictions, and any prediction errors are converted to corrections to the predicted state, resulting in an updated estimate of the current articulatory state. If the updated state estimate differs from the current desired state, controls are issued to the vocal tract generating a compensatory response.

Thus, in auditory cortex, the SFC model supposes that, during speaking, incoming auditory feedback is compared with efference-copy derived predictions of that feedback, and measuring SIS should provide an index of the accuracy of the auditory feedback predictions. The above discussion also suggests that inaccurate predictions would lead to large prediction errors, causing large state corrections, ultimately leading to large compensatory responses. In this way, the abnormal SIS in AD patients is consistent with the abnormally large compensations in their response to auditory perturbations.

Although the hemispheric difference in SIS amplitudes between AD patients and healthy controls did not reach statistical significance (see Results), SIS at M100 was higher in the left hemisphere in controls (Fig. 2A and B; Supplementary Table 1). The left hemisphere dominance of SIS in healthy participants also aligns with our SFC model [24, 25] which proposes that the left hemisphere primarily detects auditory feedback prediction errors, whereas the right hemisphere converts these errors into state corrections [42, 43].

Previous studies established that AD patients have overactivity in the posterior temporal lobe (pTL) and underactivity in the medial prefrontal cortex (mPFC). These are associated with abnormally large responses to pitch perturbations [15, 21]. The degree of compensation and mPFC activity during compensation are also correlated with measures of cognitive abilities in AD patients, particularly those of executive function [15]. These findings are particularly important because other studies have shown evidence that activity in mPFC appears to index confidence in the accuracy of the prediction of what subjects expect to hear when they speak. Several previous studies have shown that activity within mPFC correlates with successful self-predictions [40, 44–46], indicating a neural correlate of self-agency. High confidence in prediction accuracy is reflected in high mPFC activity, while low one is reflected in lower mPFC activity. Our SFC model says that mPFC inhibits the state correction process in pTL that ultimately drives perturbation responses. Thus, in AD, underactivity in mPFC would disinhibit pTL, resulting in overactivity in pTL and large perturbation responses. If the auditory feedback predictions of AD patients were more variable and inaccurate, this would result in underactivity of mPFC. In this way, our finding that AD patients are compromised at predicting auditory feedback could also account for low activity in mPFC in AD during speech production.

To finish our discussion, we would like to acknowledge some limitations of this study. The first limitation is the smaller sample size, which could affect the results’ reliability by reducing the power of the study and increasing the margin of error. However, this study’s sample size is similar to previous studies involving AD patients [15, 21] with consistent results. Another limitation is that we used an MNI template brain for specifying the primary auditory cortex based on the anatomical atlases, which was reverse transformed into the individual’s MRI coordinates for source reconstruction. It may be better to use a cohort specific (AD or older adults) template for reduction of anatomical variability in our cohorts. Furthermore, it is possible that the functional location of auditory cortex is variable across subjects independent of the anatomy. Instead of specifying the location of auditory cortex anatomically, an alternative approach would be functional specification of auditory cortex in individual subjects, for example by localizing the auditory evoked field response to simple tones in each participant’s brain [37].

This study discovered abnormalities in speech motor control in AD patients, characterizing their reduced SIS. Our SFC model of speech motor control suggests that the diminished SIS is consistent with the impaired auditory feedback predictions in AD, contributing to generating overly large compensatory changes in articulatory controls. Uncovering the specific patterns of speech-motor-control network dysfunctions relating to early speech and language impairments in AD will enable us to identify some of the earliest network abnormalities in this disease.

Alzheimer’s disease

SIS

Speaking-induced suppression

SFC

State feedback control model

MEG

Magnetoencephalography

CDR

Clinical Dementia Rating

CDR-SOB

CDR Sum of Boxes

MMSE

Mini-Mental State Examination.

Acknowledgements

The authors thank all participants and their families for supporting this research.

Authors’ contributions

K.X.K. and C.L.D. analyzed the data, while J.F.H. and S.S.N. designed the study. Moreover, K.G.R., H.K., A.J.B, H.L., D.M., M.L.G.-T., and K.V. recruited participants, conducted experiments, and handled the data. K.X.K., C.L.D., S.S.N., and J.F.H. interpreted the data and wrote the manuscript. All authors read and approved the final manuscript.

Funding

This study was funded by the National Institutes of Health grants: R01DC017696 (J.F.H., S.S.N.), 3R01DC017696-01S1 (J.F.H., S.S.N.), R01DC017091 (J.F.H., S.S.N.), R01DC010145 (J.F.H.), R21NS076171 (S.S.N.), R01NS100440 (J.F.H., S.S.N., M.L.G.-T.), R01EB022717 (S.S.N.), F32AG050434-01A1 (K.G.R.), K08AG058749 (K.G.R.), and K23AG038357 (K.V.); National Science Foundation Grant BCS-1262297 (S.S.N.); a grant from John Douglas French Alzheimer’s Foundation (K.V.); a grant from Larry L. Hillblom Foundation, 2015-A-034-FEL (K.G.R); University of California San Francisco Alzheimer’s Disease Research Center pilot project grant (K.V); grants from the Alzheimer’s Association, and made possible by Part the Cloud: PCTRB-13-288476 (K.V.), and ETAC-09-133596 (J.F.H.); a gift from Ricoh Inc. (S.S.N.); and a gift from the S.D. Bechtel Jr. Foundation (K.V.).

Availability of data and materials

The data supporting the findings of this study are available upon request.

Ethics approval and consent to participate

This study was approved by the UCSF Institutional Review Board for human research. All participants or their legal representatives provided written informed consent.

Consent for publication

Not applicable.

Competing interests

The authors report no competing interests.

Rabinovici GD, Seeley WW, Kim EJ, Gorno-Tempini ML, Rascovsky K, Pagliaro TA, et al. Distinct MRI atrophy patterns in autopsy-proven Alzheimer’s disease and frontotemporal lobar degeneration. Am J Alzheimers Dis Other Demen. 2007;22:474–88.
Baron JC, Chételat G, Desgranges B, Perchey G, Landeau B, de la Sayette V, et al. In vivo mapping of gray matter loss with voxel-based morphometry in mild Alzheimer’s disease. Neuroimage. 2001;14:298–309.
Frisoni GB, Testa C, Zorzan A, Sabattoli F, Beltramello A, Soininen H, et al. Detection of grey matter loss in mild Alzheimer’s disease with voxel based morphometry. J Neurol Neurosurg Psychiatry. 2002;73:657–64.
Busatto GF, Garrido GEJ, Almeida OP, Castro CC, Camargo CHP, Cid CG, et al. A voxel-based morphometry study of temporal lobe gray matter reductions in Alzheimer’s disease. Neurobiol Aging. 2003;24:221–31.
Kawachi T, Ishii K, Sakamoto S, Sasaki M, Mori T, Yamashita F, et al. Comparison of the diagnostic performance of FDG-PET and VBM-MRI in very mild Alzheimer’s disease. Eur J Nucl Med Mol Imaging. 2006;33:801–9.
Du A-T, Schuff N, Kramer JH, Rosen HJ, Gorno-Tempini ML, Rankin K, et al. Different regional patterns of cortical thinning in Alzheimer’s disease and frontotemporal dementia. Brain. 2007;130:1159–66.
Ishii K, Kawachi T, Sasaki H, Kono AK, Fukuda T, Kojima Y, et al. Voxel-based morphometric comparison between early- and late-onset mild Alzheimer’s disease and assessment of diagnostic performance of z score images. AJNR Am J Neuroradiol. 2005;26:333–40.
Grossman M, McMillan C, Moore P, Ding L, Glosser G, Work M, et al. What’s in a name: voxel-based morphometric analyses of MRI and naming difficulty in Alzheimer’s disease, frontotemporal dementia and corticobasal degeneration. Brain. 2004;127:628–49.
Singh V, Chertkow H, Lerch JP, Evans AC, Dorr AE, Kabani NJ. Spatial patterns of cortical thinning in mild cognitive impairment and Alzheimer’s disease. Brain. 2006;129:2885–93.
Chan D, Fox NC, Scahill RI, Crum WR, Whitwell JL, Leschziner G, et al. Patterns of temporal lobe atrophy in semantic dementia and Alzheimer’s disease. Ann Neurol. 2001;49:433–42.
Klimova B, Maresova P, Valis M, Hort J, Kuca K. Alzheimer’s disease and language impairments: social intervention and medical treatment. Clin Interv Aging. 2015;10:1401–8.
Szatloczki G, Hoffmann I, Vincze V, Kalman J, Pakaski M. Speaking in Alzheimer’s Disease, is That an Early Sign? Importance of Changes in Language Abilities in Alzheimer’s Disease. Frontiers in Aging Neuroscience [Internet]. 2015 [cited 2022 Jan 25];7. Available from: https://www.frontiersin.org/article/10.3389/fnagi.2015.00195
Mueller KD, Hermann B, Mecollari J, Turkstra LS. Connected speech and language in mild cognitive impairment and Alzheimer’s disease: A review of picture description tasks. J Clin Exp Neuropsychol. 2018;40:917–39.
Eyigoz E, Mathur S, Santamaria M, Cecchi G, Naylor M. Linguistic markers predict onset of Alzheimer’s disease. eClinicalMedicine [Internet]. Elsevier; 2020 [cited 2022 Mar 26];28. Available from: https://www.thelancet.com/journals/eclinm/article/PIIS2589-5370(20)30327-8/fulltext
Ranasinghe KG, Kothare H, Kort N, Hinkley LB, Beagle AJ, Mizuiri D, et al. Neural correlates of abnormal auditory feedback processing during speech production in Alzheimer’s disease. Sci Rep. 2019;9:5686.
Niziolek CA, Nagarajan SS, Houde JF. What does motor efference copy represent? Evidence from speech production. J Neurosci. 2013;33:16110–6.
Herman AB, Houde JF, Vinogradov S, Nagarajan SS. Parsing the phonological loop: activation timing in the dorsal speech stream determines accuracy in speech reproduction. J Neurosci. 2013;33:5439–53.
Kort NS, Cuesta P, Houde JF, Nagarajan SS. Bihemispheric network dynamics coordinating vocal feedback control. Hum Brain Mapp. 2016;37:1474–85.
Chang EF, Niziolek CA, Knight RT, Nagarajan SS, Houde JF. Human cortical sensorimotor network underlying feedback control of vocal pitch. Proc Natl Acad Sci U S A. 2013;110:2653–8.
Findlay AM, Ambrose JB, Cahn-Weiner DA, Houde JF, Honma S, Hinkley LBN, et al. Dynamics of hemispheric dominance for language assessed by magnetoencephalographic imaging. Ann Neurol. 2012;71:668–86.
Ranasinghe KG, Gill JS, Kothare H, Beagle AJ, Mizuiri D, Honma SM, et al. Abnormal vocal behavior predicts executive and memory deficits in Alzheimer’s disease. Neurobiology of Aging. 2017;52:71–80.
Parrell B, Agnew Z, Nagarajan S, Houde J, Ivry RB. Impaired Feedforward Control and Enhanced Feedback Control of Speech in Patients with Cerebellar Degeneration. J Neurosci. 2017;37:9249–58.
Houde JF, Gill JS, Agnew Z, Kothare H, Hickok G, Parrell B, et al. Abnormally increased vocal responses to pitch feedback perturbations in patients with cerebellar degeneration. J Acoust Soc Am. 2019;145:EL372.
Houde JF, Nagarajan SS. Speech production as state feedback control. Front Hum Neurosci. 2011;5:82.
Houde JF, Chang EF. The cortical computations underlying feedback control in vocal production. Curr Opin Neurobiol. 2015;33:174–81.
Morris JC. The Clinical Dementia Rating (CDR): current version and scoring rules. Neurology. 1993;43:2412–4.
Folstein MF, Folstein SE, McHugh PR. “Mini-mental state”. A practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res. 1975;12:189–98.
Vrba J, Robinson SE. Signal Processing in Magnetoencephalography. Methods. 2001;25:249–71.
Weinberg H, Brickett PA, Vrba J, Fife AA, Burbank MB. The use of a SQUID third order spatial gradiometer to measure magnetic fields of the brain. Annals of the New York Academy of Sciences. US: New York Academy of Sciences; 1984;425:743–52.
Sekihara K, Kawabata Y, Ushio S, Sumiya S, Kawabata S, Adachi Y, et al. Dual signal subspace projection (DSSP): a novel algorithm for removing large interference in biomagnetic measurements. J Neural Eng. 2016;13:036007.
Cai C, Kang H, Kirsch HE, Mizuiri D, Chen J, Bhutada A, et al. Comparison of DSSP and tSSS algorithms for removing artifacts from vagus nerve stimulators in magnetoencephalography data. J Neural Eng. IOP Publishing; 2019;16:066045.
Vrba J, Robinson SE. SQUID sensor array configurations for magnetoencephalography applications. Supercond Sci Technol. IOP Publishing; 2002;15:R51–89.
Cai C, Chen J, Findlay AM, Mizuiri D, Sekihara K, Kirsch HE, et al. Clinical Validation of the Champagne Algorithm for Epilepsy Spike Localization. Front Hum Neurosci. 2021;15:642819.
Kort NS, Nagarajan SS, Houde JF. A bilateral cortical network responds to pitch perturbations in speech feedback. Neuroimage. 2014;86:525–35.
Templeton G. A Two-Step Approach for Transforming Continuous Variables to Normal: Implications and Recommendations for IS Research. Communications of the Association for Information Systems [Internet]. 2011;28. Available from: https://aisel.aisnet.org/cais/vol28/iss1/4
Houde JF, Nagarajan SS, Sekihara K, Merzenich MM. Modulation of the auditory cortex during speech: an MEG study. J Cogn Neurosci. 2002;14:1125–38.
Ventura MI, Nagarajan SS, Houde JF. Speech target modulates speaking induced suppression in auditory cortex. BMC Neuroscience. 2009;10:58.
Ford JM, Mathalon DH. Anticipating the future: automatic prediction failures in schizophrenia. Int J Psychophysiol. 2012;83:232–9.
Hickok G, Houde J, Rong F. Sensorimotor integration in speech processing: computational basis and neural organization. Neuron. 2011;69:407–22.
Subramaniam K, Kothare H, Mizuiri D, Nagarajan SS, Houde JF. Reality Monitoring and Feedback Control of Speech Production Are Related Through Self-Agency. Front Hum Neurosci. 2018;12:82.
Korzyukov O, Bronder A, Lee Y, Patel S, Larson CR. Bioelectrical brain effects of one’s own voice identification in pitch of voice auditory feedback. Neuropsychologia. 2017;101:106–14.
Kort N, Nagarajan SS, Houde JF. A right-lateralized cortical network drives error correction to voice pitch feedback perturbation. The Journal of the Acoustical Society of America. Acoustical Society of America; 2013;134:4234–4234.
Tourville JA, Reilly KJ, Guenther FH. Neural mechanisms underlying auditory feedback control of speech. Neuroimage. 2008;39:1429–43.
Franken MK, Eisner F, Acheson DJ, McQueen JM, Hagoort P, Schoffelen J-M. Self-monitoring in the cerebral cortex: Neural responses to small pitch shifts in auditory feedback during speech production. Neuroimage. 2018;179:326–36.
Khalighinejad N, Schurger A, Desantis A, Zmigrod L, Haggard P. Precursor processes of human self-initiated action. Neuroimage. 2018;165:35–47.
Subramaniam K, Luks TL, Fisher M, Simpson GV, Nagarajan S, Vinogradov S. Computerized cognitive training restores neural activity within the reality monitoring network in schizophrenia. Neuron. 2012;73:842–53.

No competing interests reported.

AdditionalFile1.pdf
Additional file 1: Supplementary Table 1. Speaking-induced suppression and peak latency difference between the speaking and listening conditions.

Download PDF

Version 1

posted

You are reading this latest preprint version

Impaired auditory feedback prediction in Alzheimer’s disease

Status:

Version 1

Abstract

Figures

Background

Methods

Participants

Neuropsychological assessment

Experimental design and procedure

MEG data preprocessing

Source reconstruction and auditory response

Statistical analysis

Results

Demographic characteristics of participants

Time course of auditory cortical activity during speaking and listening

Abnormal speaking-induced suppression in Alzheimer’s disease

Peak latency of auditory cortical activity

Discussion

Conclusions

Abbreviations

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1