MMN in  Psychosis

Auditory Mismatch Negativity (MMN) is an event related potential that is elicited by unexpected irregularities in a constant auditory stream, e.g. during an oddball paradigm, where repetitive standard sounds are interspersed with infrequent randomly presented deviants.

Reductions in MMN-responses are well established in chronic schizophrenia. However, evidence for MMN impairments in Clinical High Risk for Psychosis (CHR-P) and First Episode Psychosis (FEP) patients groups is less consistent. 

In this work, we examined recorded neuromagnetic responses to duration deviants to establish whether MMN responses are impaired during early stage psychosis. 

Contrary to our hypothesis, MMN-responses were intact, suggesting that MMN may not constitute a biomarker for early detection and diagnosis of psychosis. 

Here is my poster summarising this work

Relevant Publication

• Pradeep D, Tineke Grent-‘t-Jong, Ruchika Gajwani, Joachim Gross, Andrew I Gumley,  Rajeev Krishnadas, Stephen M Lawrie, Matthias Schwannauer, Frauke Schultze-Lutter, Peter J Uhlhaas, "Intact Mismatch Negativity Responses in Clinical High-Risk for Psychosis and First-Episode Psychosis: Evidence from Source-Reconstructed Event-Related Fields and Time-Frequency Data", Biological Psychiatry: Cognitive Neuroscience and Neuroimaging, vol. 9 (1), pp.121-131, 2024 (pdf)  https://www.biologicalpsychiatrycnni.org/article/S2451-9022(23)00244-6/fulltext

Executive Summary: Mismatch negativity (MMN) is widely used as a biomarker of psychosis and is traditionally considered pre-attentive. However, evidence suggests MMN depends on attentional context. This review shows that MMN deficits in clinical high-risk (CHR) and first-episode psychosis (FEP) often disappear when participants perform demanding visual distractor tasks. We argue that active distractor tasks reduce reliance on impaired top-down processes, supporting MMN via preserved bottom-up processing. These findings highlight the importance of task design for using MMN in early psychosis (ESZ) research.

Title: "Attentional engagement reduces mismatch negativity (MMN) deficits in early-stage psychosis: A literature review of active visual distractor task paradigms"

Background

The auditory mismatch negativity (MMN) event-related potential (ERP) has long been regarded as a pre-attentive index of auditory sensory prediction. Traditionally considered largely automatic, MMN has been widely used as a biomarker of vulnerability to and presence of psychosis. However, emerging evidence indicates that MMN is not strictly pre-attentive; rather, it is shaped by task demands and influenced by top-down attentional processes.

Given the mixed evidence from passive MMN paradigms in early psychosis samples, the aim of this review was to examine MMN findings from studies incorporating an active visual distractor task.

Methods

We carried out a systematic search of Embase, Medline, and PubMed databases with the terms  ‘mismatch negativity’ combined with 'clinical high risk for psychosis (CHR-P)' or 'first-episode psychosis (FEP)' and related variations. A total of 208 studies were identified, and after removal of duplicates, 149 studies were screened manually. Nine studies that included an active visual distractor task were included in this review. 

Results

In six of nine reviewed studies, MMN deficits, commonly observed in chronic schizophrenia, failed to appear in CHR-P and FEP samples when participants performed an engaging visual task. Three studies reported MMN deficits: one included a concurrent auditory task, another did not measure task engagement with the visual task, while a third showed intact MMN in the overall CHR-P sample employed a colour change detection task that may have imposed relatively lower attentional demands. 

Discussion

We argue that active visual distractor tasks stabilize arousal, reduce variability in task engagement, and diminish the need for top-down allocation of auditory attention, thereby shifting reliance toward relatively preserved bottom-up auditory processes. These conditions may mitigate the expression of early predictive processing impairments, thereby reducing the likelihood of detecting MMN deficits in early-stage psychosis groups. In contrast, passive or weakly controlled incidental tasks (e.g., watching a silent video or reading a book) show mixed results due to differences in top-down resource allocation required for forming and updating the internal model of auditory regularities. This could result in passive attentional fluctuations being misinterpreted as primary sensory deficits.

Conclusions

These findings call for a reconceptualization of MMN as a context-dependent phenomenon rather than a purely pre-attentive response. Careful consideration of attentional design, particularly the inclusion and quantification of active visual distractor task engagement, is critical when using MMN as a biomarker in early psychosis research.

Here is my talk summarizing this work

We investigated frequency deviant Mismatch Negativity (f-MMN) responses in the auditory cortex of Early Schizophrenia (ESZ) patients and matched healthy controls using high-field 7T fMRI. We aimed to assess differences in cortical laminar-specific microcircuits active during f-MMN. We also manipulated participant’s attention either towards (irrelevant auditory feature) or away (concurrent visual stimulation) from the auditory stimulation.

Abstract:

Introduction: 

The predictive processing account of brain function suggests that our brains build models of the world based on our prior experiences, and use these models to make predictions about what should happen next 1. These models are updated by confronting them with external inputs that are not explained away by internal predictions, resulting in prediction errors. An alteration of these predictive mechanisms could be at the origin of sensory symptoms observed in psychosis 2 . For example, there is evidence for a difference between Schizophrenia patients and controls in processing unexpected auditory deviants (oddballs), as represented by the mismatch negativity (MNN) response 3 . 

The reduction in the MMN response of patients  compared to controls is widely used as a biomarker of psychosis. However, most studies do not take into account the role of attention and its modulatory effects on top-down signals and MMN dynamics 4. Situations in which attention is directed towards or away from a stimulation (rather than mind-wandering) are omnipresent in everyday life,  therefore, we need to investigate how they affect MMN deficiencies in psychosis.

Furthermore, much of the current research was done using electrophysiological methods, leaving another knowledge gap, namely how MMN is manifested at the spatial scale afforded by high-field (7T) functional magnetic resonance imaging (fMRI). The use of 7T sub-millimeter fMRI for studying psychosis allows to test predictive processing accounts of psychosis on a layer-specific level 5 advancing our understanding of the neuronal circuit signatures of psychosis.   

Methods:

Experimental paradigm: We recorded BOLD signal responses in two groups of participants (17 control participants and 13 early-stage schizophrenia patients). The total scanning time was around 2 hours.

Trials were composed of a sequence of six 150 ms tones for a total duration of 1.5s. On standard trials, all tones were the same (440 Hz), but in the deviant condition (20% of trials) the last tone differed in pitch (620 Hz). Inter-trial intervals were of 10s on average, jittered between 8 and 12s. To control for attention, we also presented an unrelated visual stimulus on each trial (4 “pacman” shapes which formed a Kanizsa figure on 20% of trials and were misaligned on the remaining trials). We instructed participants to do an auditory (detecting a shorter tone in the auditory stimulation) or visual (detecting a subtle flicker in the visual stimulus) task. To confirm  task engagement, we used a performance threshold of at least 80% hit rate and less than 5% false alarm rate over the whole fMRI session.

Participants performed 4 experimental runs. Each run contained 2 parts, one for the visual task and one for the auditory task, in a counterbalanced order (feedback about task performance was displayed at the end of each part). Within each part there were 25 trials: 5 of each dedicated to the visual (flicker) and auditory (shorter tone) task, 5 had standard auditory stimulation with Kanizsa visual stimulation, 5 had deviant auditory stimulation with standard visual stimulation, and 5 had both auditory and visual standard stimulation. Only data corresponding to the fully standard and the auditory deviant trials (20 of each for each attentional conditions across all runs) were further analysed in the present work.

fMRI data acquisition: We scanned participants at the Queen Elizabeth University Hospital with a Siemens 7-Tesla (7T) Terra Magnetom MRI Scanner, equipped with a NOVA Medical 32-channel head coil. The protocol included four task-related runs (310 volumes EPI, 0.8mm voxel resolution, 2000ms TR) and four corresponding top-up runs (five volumes each), an anatomical scan (T1-weighted MP2rage, 0.6mm voxel resolution), and a combined retinotopy/tonotopy functional run (not used here).

Data analysis: We performed fMRI analyses using BrainVoyager, FSL, MATLAB and R. Anatomical data was pre-processed by applying inhomogeneity correction (IIHC), upscaling to a voxel resolution of 0.4mm isotropic (from 0.6mm), and manually aligned to ACPC space. We performed white and grey matter segmentation as well as manual correction using ITK-SNAP in the ROIs, Heschl’s gyrus (HG) and superior temporal gyrus (STG). In BrainVoyager, we generated six equivolume cortical layers for ROI-based depth sampling. Pre-processing steps for functional data included 3D motion correction, slice-time correction, high-pass temporal filtering (five cycles), and cubic spline interpolation. We aligned functional runs using inter-run spatial cross-correlations of signal intensities.

We performed deconvolution general linear model analysis across all cortical layers, in all groups, to acquire peak BOLD response beta values (AR2 corrected and z-scored). We then fit a model predicting peak responses from stimulation (deviant vs non deviant tone), attentional condition (visual vs auditory), ROI (HS vs STG), group (control vs patient), cortical layer (as a linear parameter), all interaction between these factors as well as a random intercept and slope (over cortical depths) by participant. We also fit a model predicting the difference in peak response for deviant - standard trials from attentional condition, area, group, layer, all interaction between these factors and a random intercept by participant (random slopes would induce a singular fit). From these models, we could derive comparisons between parameter levels using estimated marginal means.

Results:

From the peak response model, we found that (1) the response to deviant trials is always (i.e. across all combinations of other factors) higher than to standard (non-deviant) trials (all t(2529) > 2.3, all p < .05), (2) the response in the early auditory area (Heschl’s gyrus) is always higher than in the late auditory area (superior temporal gyrus) (all t(2529) > 4, all p < 10-4), (3) control participants always have a higher response than patients (all t(2531) > 4.4, all p < 10-4), (4)  the response is higher when paying attention to the auditory compared to the visual modality (all t(2529) > 2.27, all p < .05), except in the controls’ early auditory area where there is no significant difference (control trials: t(2529) = 1.488, p > .05, deviant trials: t(2529) = 1.8, p > 0.05).

Furthermore, taking cortical layers into account, there is no laminar trend difference for deviant vs standard or visual vs auditory attention in either control or schizophrenia groups.There is a significant laminar effect, under which the peak response control participants in the early auditory area (HG) is significantly more sloped towards superficial layers compared to the later area (STG) (t(2529) = 3.89, p < 10-4).

From the mismatch response model (deviant – standard peak), we found that patients have a higher mismatch response than controls in the auditory attention condition (HG: t(1279) = -2.87, p < 10-2, STG: t(1279) = -2.06, p < .05), however switching to visual attention increases mismatch response in the controls’ late auditory area (STG) (t(1264) = -2.24, p < .05), while it decreases in the patients’ early area (HG) (t(1264) = 3.46, p < 10-3). There is no significant effect of laminar trend in this model.

Conclusions:

We examined responses to a mismatched tone in the auditory cortex of schizophrenia patients, and control participants, using high-field 7T fMRI. We directed participants' attention either toward or away from the auditory stimulation. We found results differentiating the effect of attentional modulation between patients and control participants. Specifically, it seems that there is less differentiation of early vs late auditory areas in patients. In control participants, switching from auditory to visual attention increases mismatch response in the late auditory area, while in patients mismatch response starts higher in both areas but decreases in the visual attentional condition. Overall, these results expand the idea of distinctive feedback mechanisms as neural correlates of psychosis to include contrasting effects of attentional modulation.

References:

1. Clark, A. (2013) ‘Whatever next? Predictive Brains, situated agents, and the future of Cognitive Science’, Behavioral and Brain Sciences, 36(3), pp. 181–204. doi:10.1017/s0140525x12000477.

2. Sterzer, P., Adams, R. A., Fletcher, P., Frith, C., Lawrie, S. M., Muckli, L., ... & Corlett, P. R. (2018). The predictive coding account of psychosis. Biological psychiatry, 84(9), 634-643.

3. Erickson, M. A., Ruffle, A., & Gold, J. M. (2016). A meta-analysis of mismatch negativity in schizophrenia: from clinical risk to disease specificity and progression. Biological psychiatry, 79(12), 980-987

4. Näätänen R, Paavilainen P, Tiitinen H, Jiang D, Alho K. Attention and mismatch negativity. Psychophysiology. 1993 Sep;30(5):436-50. doi: 10.1111/j.1469-8986.1993.tb02067.x. PMID: 8416070.

5. Haarsma, J., Kok, P. & Browning, M. The promise of layer-specific neuroimaging for testing predictive coding theories of psychosis. Schizophrenia Research 245, 68–76 (2022).