EEG-based machine learning shows promise for neurodegenerative disease classification, but robustness to sample imbalance, center heterogeneity, and validation leakage remains a key concern for clinical translation. We developed a new framework to assess diagnostic performance, calibration, and cross-center generalizability of EEG multifeatured classifiers across CN (cognitively normal), MCI (mild cognitive impairment), AD (Alzheimer's disease), and FTD (frontotemporal dementia), while addressing imbalance, statistical uncertainty, and validation rigor across six centers. Supervised classifiers were evaluated at aggregated- and subject-level repeated cross-validation and leave-one-center-out (LOCO) schemes, and calibration was implemented via Platt scaling within strictly nested folds. CN vs AD classification showed robust performance and cross-center generalizability, with consistent AUC and calibration across cross-validation and leave-one-center-out analyses. In contrast, CN versus MCI showed moderate, heterogeneous performance and limited cross-center generalizability, with chance-level results in some cohorts, while MCI versus AD showed moderate discrimination in a single available center. FTD contrasts showed modest or limited performance due to sparse samples. Predicted probabilities were stable across validation regimes for AD, but less consistent for MCI and FTD, and correlated robustly with cognitive impairment severity only for AD. Feature importance analyses identified disease-specific signatures, including alpha-band degradation and slow-wave increases in AD, with weaker and more heterogeneous patterns in prodromal and differential dementia contrasts (FTD vs AD). EEG classifiers provided robust discrimination for CN vs AD but showed limited and heterogeneous performance for MCI and FTD across centers. These results emphasize the need for balanced sampling, strict validation of clinical and EEG protocols, and uncertainty quantification to support reliable clinical deployment.

Prognostication of critically ill children appearing acutely comatose remains challenging. Event-related potentials (ERPs) using electroencephalography (EEG) offer a noninvasive method to detect covert consciousness. These include the mismatched negativity (MMN) and P300b responses. We investigated ERPs using the auditory local–global oddball paradigm to assess inattentive or attentive cortical processing in critically ill, clinically unresponsive children. We prospectively enrolled pediatric intensive care unit (PICU) patients undergoing continuous EEG with Glasgow coma scale scores < 8. Each patient underwent testing with the local–global oddball paradigm. In each trial, five brief tones were delivered; the fifth tone was either identical (local–standard) or different (local–deviant). In total, 80

Consciousness and cognition arise from the ongoing interactions between brain regions. Synchronous fluctuations of fMRI signals may indicate that two brain regions perform similar cognitive functions, but neural interactions are also constrained by anatomical connectivity and regions' molecular, cytoarchitectonic, and metabolic profiles. Here we disentangle the respective contributions of ongoing cognition and multimodal neurobiological constraints in shaping functional connectivity. We jointly contextualise haemodynamic FC against eight distinct multimodal representations of the human connectome: (i) structural connectivity from diffusion tractography; (ii) spatial embedding; (iii) similarity of transcriptional profiles from gene expression; (iv) similarity of receptor profiles from Positron Emission Tomography; (v) laminar profile similarity from histology; (vi) correlated electrophysiological activity from magnetoencephalography; (vii) correlated metabolic activity from PET glucose uptake; (viii) coordinated activation across 123 cognitive operations from the NeuroSynth meta-analytic engine. We demonstrate that cognitive co-activation is the dominant predictor of inter-regional fMRI synchrony in the awake human brain, even when quantified using intracranial electrical stimulation. Crucially, this predominance of cognitive co-activation for shaping functional connectivity is systematically obliterated across five datasets of pharmacological and pathological perturbations of consciousness (chronic disorders of consciousness; anaesthesia with sevoflurane, propofol, or ketamine) when cognition is disconnected from the environment or altogether abolished. Altogether, we show that multimodal predictors of functional architecture shift away from cognitive co-activation and toward anatomical-molecular constraints during pharmacological and pathological perturbations of consciousness.

A bstract Spontaneous thoughts constitute most of everyday inner experience, yet long-standing methodological challenges obscure a thorough exploration of their content and neurophysiological underpinnings. Traditional approaches relying on thought probes impose strict constraints on phenomenological reports, whereas online verbal reports disrupt the natural flow of experience while interfering neural signals with motor artifacts. Here, we designed and tested an alternative approach to assess the neural basis of spontaneous thoughts combining delayed verbal retrospective free reports (RFR) with automated phenomenological ratings generated by large language models (LLMs). Twenty-two participants performed an eyes-closed free-thinking task, providing reports that were evaluated along ten phenomenological dimensions by four state-of-the-art LLMs and a panel of human raters. Machine-learning models (ML) were then trained to decode LLM-derived ratings from EEG spectral, complexity, and connectivity features. Our analyses showed that inter-rater agreement among LLMs exceeded that of human raters whereas ML models achieved above-chance accuracy for the prediction of emotional valence. These findings provide support for the use of LLMs for a scalable phenomenological annotation of spontaneous thoughts and suggest that their affective dimensions can be decoded from concurrent EEG activity.

BACKGROUND:In this study, we evaluated the potential of a network approach to electromyography and electroencephalography recordings to detect covert command-following in healthy participants. The motivation underlying this study was the development of a diagnostic tool that can be applied in common clinical settings to detect awareness in patients that are unable to convey explicit motor or verbal responses, such as patients that suffer from disorders of consciousness (DoC). METHODS:We examined the brain and muscle response during movement and imagined movement of simple motor tasks, as well as during resting state. Brain-muscle networks were obtained using non-negative matrix factorization (NMF) of the coherence spectra for all the channel pairs. For the 15/38 participants who showed motor imagery, as indexed by common spatial filters and linear discriminant analysis, we contrasted the configuration of the networks during imagined movement and resting state at the group level, and subject-level classifiers were implemented using as features the weights of the NMF together with trial-wise power modulations and heart response to classify resting state from motor imagery. RESULTS:Kinesthetic motor imagery produced decreases in the mu-beta band compared to resting state, and a small correlation was found between mu-beta power and the kinesthetic imagery scores of the Movement Imagery Questionnaire-Revised Second version. The full-feature classifiers successfully distinguished between motor imagery and resting state for all participants, and brain-muscle functional networks did not contribute to the overall classification. Nevertheless, heart activity and cortical power were crucial to detect when a participant was mentally rehearsing a movement. CONCLUSIONS:Our work highlights the importance of combining EEG and peripheral measurements to detect command-following, which could be important for improving the detection of covert responses consistent with volition in unresponsive patients.

BACKGROUND:This multi-centric study aimed to explore differences in brain activity patterns in patients with disorders of consciousness (DoC), including unresponsive wakefulness syndrome (UWS) and minimally conscious state (MCS). METHODS:Using high-density electroencephalographic (EEG) recordings from 368 DoC patients, 39 who emerged from MCS (eMCS), and 73 healthy controls, we examined instantaneous functional connectivity-based meta-states acting as attractors in a dynamical system, extracted by means of community detection algorithms and recurrence analysis. We analyzed data from two patient cohorts and included resting-state and auditory processing tasks in four frequency bands (delta, theta, alpha, beta) and from three perspectives, namely: (i) discrete activation of dominant states, (ii) a dynamical system composed of attractor states and (iii) the correlation and anticorrelation patterns of the active states. RESULTS:Findings revealed that while the overall structure of brain connectivity remained stable after injury, patients with DoC and those who emerged showed notable differences in the speed and consistency of how their brain states activated. Specifically, in higher frequencies, UWS patients exhibited faster, and less stable dynamics, shorter dwell times and decreased meta-state anticorrelation compared to those in MCS and eMCS. Moreover, a four-way combined learning classification analysis showed that the measures were able to distinguish the UWS and MCS subgroups. SIGNIFICANCE:These brain state dynamics could serve as valuable markers for assessing states of consciousness. Our results highlight the potential of using high-temporal resolution dynamic brain activity patterns to improve the understanding of altered consciousness and their application to clinical settings.

As a response to the environment and internal signals, brain networks reorganize on a sub-second scale. To capture this reorganization in patients with disorders of consciousness (DoC) and understand their residual brain activity, we investigated the dynamics of electroencephalography (EEG) microstates. EEG microstates are meta-stable topographies that last tens to a few hundreds of milliseconds and are hypothesized to reflect large-scale cortical networks. To obtain EEG‑microstate segmentation, EEG topographies per sample were clustered into four groups for the purpose of the present comparison with the existing four‑class literature. We then obtained a time series of maps with different frequencies of occurrence and duration. One such occurrence of a map with a given duration is called a microstate. The goal of this work was to study the static and dynamic properties of these topographical patterns in DoC patients. Using the microstate time series, we calculated static and dynamic markers. In contrast to the static, the dynamic metrics depend on the specific temporal sequences of the maps. The static measure map coverage showed differences between healthy controls and patients. In contrast, some dynamic markers captured inter-patient group differences. The dynamic markers we investigated are Mean Microstate Durations (MMD), Microstate Duration Variances (MDV), Microstate Transition Matrices (MTM), and Entropy Production (EP). The MMD and MDV decreased with the state of consciousness, whereas the MTM non-diagonal transitions and EP increased. In other words, DoC patients had slower and closer to equilibrium (time-reversible) brain dynamics. In conclusion, static and dynamic EEG microstate metrics differed across consciousness levels, with the latter having captured the subtler differences between groups of patients with DoC.

In the search for EEG markers of human consciousness, alpha power has long been considered a reliable marker which is fundamental for the assessment of unresponsive patients from all etiologies. However, recent evidence questioned the role of alpha power as a marker of consciousness and proposed the spectral exponent and spatial gradient as more robust and generalizable clinical indexes. In this study, we analyzed a large-scale dataset of 303 unresponsive patients and investigated etiology-specific differences in clinical markers of level of consciousness, responsiveness and capacity to recover. We compare a set of candidate EEG makers: i) absolute, relative and flattened alpha power; ii) spatial ratios; iii) the spectral exponent; and iv) signal complexity. Our results support the claim that alpha power has higher diagnostic value for anoxic patients. Meanwhile, the spectral slope showed diagnostic value for non-anoxic patients only. Changes in relative power and signal complexity occurred alongside changes in the spectral slope. Grouping unresponsive patients from different etiologies together can confound or obscure the diagnostic value of different EEG markers of consciousness. Our study highlights the importance of analyzing different etiologies independently and emphasizes the need to develop clinical markers which better account for inter-individual and etiology-dependent differences.

The resting primate brain is traversed by spontaneous functional connectivity patterns that show striking differences between conscious and unconscious states. Transcranial direct current stimulation (tDCS), a non-invasive neuromodulatory technique, can improve signs of consciousness in disorders of consciousness (DOCs); however, can it influence both conscious and unconscious dynamic functional connectivity? We investigated the modulatory effect of prefrontal cortex (PFC) tDCS on brain dynamics in awake and anesthetized non-human primates using functional MRI. In awake macaques receiving either anodal or cathodal tDCS, we found that cathodal stimulation robustly disrupted the repertoire of functional connectivity patterns, increased structure–function correlation (SFC), decreased Shannon entropy, and favored transitions toward anatomically based patterns. Under deep sedation, anodal tDCS significantly altered brain pattern distribution and reduced SFC. The prefrontal stimulation also modified dynamic connectivity arrangements typically associated with consciousness and unconsciousness. Our findings offer compelling evidence that PFC tDCS induces striking modifications in the fMRI-based dynamic organization of the brain across different states of consciousness. This study contributes to an enhanced understanding of tDCS neuromodulation mechanisms and has important clinical implications for DOCs.

Assessing someone's level of consciousness is a complex matter, and attempts have been made to aid clinicians in these assessments through metrics based on neuroimaging data. Many studies have empirically investigated measures related to the complexity elicited after the brain is stimulated to quantify the level of consciousness across different states. Here we hypothesized that the level of non-equilibrium dynamics of the unperturbed brain already contains the information needed to know how the system will react to an external stimulus. We created personalized whole-brain models fitted to resting state fMRI data recorded in participants in altered states of consciousness (e.g., deep sleep, disorders of consciousness) to infer the effective connections underlying their brain dynamics. We then measured the out-of-equilibrium nature of the unperturbed brain by evaluating the level of asymmetry of the inferred connectivity, the time irreversibility in each model and compared this with the elicited complexity generated after in silico perturbations, using a simulated fMRI-based version of the Perturbational Complexity Index, a measure that has been shown to distinguish different levels of consciousness in in vivo settings. Crucially, we found that states of consciousness involving lower arousal and/or lower awareness had a lower level of asymmetry in their effective connectivities, a lower level of irreversibility in their simulated dynamics, and a lower complexity compared to control subjects. We show that the asymmetry in the underlying connections drives the nonequilibrium state of the system and in turn the differences in complexity as a response to the external stimuli.

Predictive processing theories posit that the brain continuously generates expectations about incoming sensory inputs, updating them through prediction errors. While extensively studied within single modalities, it remains unclear how prediction errors unfold when expectations and violations occur across different senses. Using a local-global hierarchical oddball paradigm combined with high-density EEG in 47 participants, we contrasted unimodal and crossmodal prediction errors across auditory, visual, and somatosensory domains. We found that crossmodal violations elicit temporally sustained cortical responses which diverge from the transient, localised dynamics observed for unimodal prediction errors. Temporal decoding revealed that crossmodal effects maintain shared prolonged neural representations, suggestive of supramodal integration across all levels of cortical processing. Computational modelling further demonstrated that crossmodal prediction errors reorganize effective connectivity within and between sensory hierarchies, engaging distinct early cortical pathways depending on the sensory combination. Our findings refine hierarchical predictive coding for crossmodal transitions by demonstrating that, unlike unimodal prediction errors, crossmodal prediction errors recruit dedicated prolonged supramodal representations and flexibly adapted modality-specific networks. ### Competing Interest Statement The authors have declared no competing interest. Heinrich Böll Stiftung, https://ror.org/01vneh441, PhD scholarship

Diagnosing Disorders of Consciousness (DoC) remains a critical challenge in cognitive neuroscience. In this study we introduce Electroencephalography (EEG)-based brain states as a real-time, bedside tool for assessing dynamic brain connectivity in DoC patients. We analyze EEG data from 237 acute and chronic DoC patients across three centers, identifying five recurrent functional connectivity patterns. The probability of these patterns correlated strongly with consciousness levels, with high-entropy patterns exclusive to healthy controls and low-entropy patterns prevalent in severe DoC, predicting individual recovery outcomes. Real-time testing validated reliable bedside detection of these patterns. Our findings demonstrate EEG's potential for monitoring dynamic brain connectivity, offering insights into the neural basis of consciousness and advancing diagnostic strategies for DoC.

The study of disorders of consciousness (DoC) is very complex because patients are suffering from a wide variety of lesions, affected brain mechanisms, different symptom severity and are unable to communicate. Combining neuroimaging data and mathematical modeling can help us quantify and better describe some of these alterations. This study's goal is to provide a novel analysis and modeling pipeline for fMRI data leading to new diagnosis and prognosis biomarkers at the individual patient level. To do so, we project patient's fMRI data into a low dimension latent-space. We define the latent space's dimension as the smallest dimension able to maintain the complexity, non-linearities, and information carried by the data, according to different criteria that we detail in the first part. This dimensionality reduction procedure then allows us to build biologically inspired latent whole-brain models that can be calibrated at the single-patient level. In particular, we propose a new model inspired by the astrocyte regulation of neuronal activity in the brain. This modeling procedure leads to two types of model-based biomarkers (MBBs) that provide novel insight at different levels: (1) the connectivity matrices bring us information about the severity of the patient's diagnosis, and, (2) the local node parameters, correlate to the patient's etiology, age and prognosis. ### Competing Interest Statement The authors have declared no competing interest.

The quest to decode the complex supraspinal mechanisms that integrate cutaneous thermal information in the central system is still ongoing. The dorsal horn of the spinal cord is the first hub that encodes thermal input which is then transmitted to brain regions via the spinothalamic and thalamocortical pathways. So far, our knowledge about the strength of the interplay between the brain regions during thermal processing is limited. To address this question, we imaged the brains of adult awake male mice in resting state using functional ultrasound imaging during plantar exposure to constant and varying temperatures. Our study reveals for the first time the following: (1) a dichotomy in the response of the somatomotor-cingulate cortices and the hypothalamus, which was never described before, due to the lack of appropriate tools to study such regions with both good spatial and temporal resolutions. (2) We infer that cingulate areas may be involved in the affective responses to temperature changes. (3) Colder temperatures (ramped down) reinforce the disconnection between the somatomotor-cingulate and hypothalamus networks. (4) Finally, we also confirm the existence in the mouse brain of a brain mode characterized by low cognitive strength present more frequently at resting neutral temperature. The present study points toward the existence of a common hub between somatomotor and cingulate regions, whereas hypothalamus functions are related to a secondary network.

Brain connectivity, allowing information to be shared between distinct cortical areas and thus to be processed in an integrated way, has long been considered critical for consciousness. However, the relationship between functional intercortical interactions and the structural connections thought to underlie them is poorly understood. In the present work, we explore both functional (with an EEG-based metric: the median weighted symbolic mutual information in the theta band) and structural (with a brain MRI-based metric: fractional anisotropy) connectivities in a cohort of 78 patients with disorders of consciousness. Both metrics could distinguish patients in a vegetative state from patients in minimally conscious state. Crucially, we discovered a significant positive correlation between functional and structural connectivities. Furthermore, we showed that this structure-function relationship is more specifically observed when considering structural connectivity within the intra- and inter-hemispheric long-distance cortico-cortical bundles involved in the Global Neuronal Workspace (GNW) theory of consciousness, thus supporting predictions of this model. Altogether, these results support the interest of multimodal assessments of brain connectivity in refining the diagnostic evaluation of patients with disorders of consciousness.

Disorders of consciousness (DoC) represent a challenging and complex group of neurological conditions characterised by profound disturbances in consciousness. The current range of treatments for DoC is limited. This has sparked growing interest in developing new treatments, including the use of psychedelic drugs. Nevertheless, clinical investigations and the mechanisms behind them are methodologically and ethically constrained. To tackle these limitations, we combined biologically plausible whole-brain models with deep learning techniques to characterise the low-dimensional space of DoC patients. We investigated the effects of model pharmacological interventions by including the whole-brain dynamical consequences of the enhanced neuromodulatory level of different neurotransmitters, and providing geometrical interpretation in the low-dimensional space. Our findings show that serotonergic and opioid receptors effectively shifted the DoC models towards a dynamical behaviour associated with a healthier state, and that these improvements correlated with the mean density of the activated receptors throughout the brain. These findings mark an important step towards the development of treatments not only for DoC but also for a broader spectrum of brain diseases. Our method offers a promising avenue for exploring the therapeutic potential of pharmacological interventions within the ethical and methodological confines of clinical research. Exploring virtual pharmacological treatments using whole-brain models and deep learning reveals improvements in disorders of consciousness models when stimulating serotonergic and opioid receptors.

BACKGROUND:Disorders of consciousness (DoC) refers to a group of clinical conditions of altered consciousness. To improve their diagnosis and prognosis, multimodal assessment can be of great importance. Informal caregivers of people with DoC who are confronted with new technologies as such can benefit from interventions to expand their health literacy, i.e., the ability to use information to make health decisions for oneself and others. METHODS:We developed an information brochure on multimodal assessment for DoC in a participatory process, with decisions made by a steering group. The process was based on a methodological framework for the development of patient decision aids that built on the International Patient Decision Aid Standards (IPDAS). RESULTS:On the background of a broad variety of needs, the priority was to focus on the explanation of multimodal testing and provide information about its uncertainty. Its development aimed at enhancing informal caregivers' understanding of implications of results from multimodal assessment and its relevance for prognosis. It should avoid the portrayal of information that could lead to the impression of false hope or suboptimal rehabilitation care. Informal caregivers rated its usability and acceptability highly, though they preferred less technical language. CONCLUSION:The participatory process was crucial to the project. Future studies should investigate the effectiveness of the brochure in fostering informal caregivers' health literacy. PATIENT OR PUBLIC CONTRIBUTION:Informal caregivers of people with DoC were deliberately included in the steering group and they participated in a field test of the prototype brochure.

Severely brain-injured patients may enter a spectrum of conditions collectively known as disorders of consciousness (DoC). This spectrum includes clinical categories such as unresponsive wakefulness syndrome or minimally conscious state, where the behavioral assessment of consciousness can often be deceptive. To bridge this dissociation, neuroimaging techniques are employed to look for the residual brain functions. Each neuroimaging modality imperfectly captures distinct aspects of brain preservation - functional, anatomical, or both. In this study, we adopt a comprehensive approach by integrating the neurophysiology and neuroimaging modalities available from the standard and advanced clinical assessment through interpretable machine learning (ML). The electrophysiological modalities included high-density electroencephalography (EEG) (resting state and task), whereas neuroimaging modalities included anatomical and resting-state functional magnetic resonance imaging (MRI), diffusion MRI, and 18F-fluoro-deoxy-glucose positron emission tomography (FDG PET). Our investigation reveals that specific modalities, such as functional assessments provide comprehensive insights into the currently evaluated state of consciousness - the diagnosis of the patients. Conversely, structural modalities offer valuable information about the patient's evolution within the consciousness spectrum. We validate the proposed analysis with data coming from other centers with different acquisition parameters. Importantly, we show that there is an improved model performance with the increase in the number of modalities. We observe a higher inter-modality disagreement for MCS patients and those patients who improve. Lastly, we observe a difference in feature importances in diagnosis and prognosis. This integrative multimodal and ML methodology presents a promising avenue for a more nuanced understanding of DoC, contributing to enhanced diagnostic precision and prognostic capabilities in clinical practice. ### Competing Interest Statement Jacobo D. Sitt and Lionel Naccache are scientific co-founders of NeuroMeters (have scientific advisory activity but no executive or management activity). ### Funding Statement This work was supported by the Ecole Doctorale Frontieres de l'Innovation en Recherche et Education-Programme Bettencourt (to D.M.). This project is part of the multicentric application for the EU ERAPerMed Joint Translational Call for Proposals for "Personalised Medicine: Multidisciplinary research towards implementation" (ERA PerMed JTC2019). It is funded by local funding agencies of the participating countries (for France it is the Agence Nationale de Recherche ANR, funding code: ANR-19-PERM-0002, for Germany the Federal Ministry of Education and Research BMBF, funding code: 01KU2003). This project is supported by the Human Brain Project (HBP) MODELDxConsciousness Consortium (Agence Nationale de Recherche ANR, funding code: S.1600.ANR.HBPR). ### Author Declarations I confirm all relevant ethical guidelines have been followed, and any necessary IRB and/or ethics committee approvals have been obtained. Yes The details of the IRB/oversight body that provided approval or exemption for the research described are given below: Pitie-Salpetriere Hospital (Comite de Protection des Personnes, Ile de France 1, Paris, France) under the French label of routine care research (protocol number 2013-A01385-40, code Recherche en soins courants, protocol number M-Neuro-DOC, CE SRLF 20-2). Medical Faculty of Ludwig-Maximilians-Universitat Munchen (Germany) (protocol numbers 20-634 and 20-635). IRCCS Fondazione Don Carlo Gnocchi (IRCCS Regione Lombardia, Italy) (protocol number 32/2021/CE_FdG/FC/SA). I confirm that all necessary patient/participant consent has been obtained and the appropriate institutional forms have been archived, and that any patient/participant/sample identifiers included were not known to anyone (e.g., hospital staff, patients or participants themselves) outside the research group so cannot be used to identify individuals. Yes I understand that all clinical trials and any other prospective interventional studies must be registered with an ICMJE-approved registry, such as ClinicalTrials.gov. I confirm that any such study reported in the manuscript has been registered and the trial registration ID is provided (note: if posting a prospective study registered retrospectively, please provide a statement in the trial ID field explaining why the study was not registered in advance). Yes I have followed all appropriate research reporting guidelines, such as any relevant EQUATOR Network research reporting checklist(s) and other pertinent material, if applicable. Yes The data is not publicly available.