Daniel Richard Leff

Correction: Physiological synchrony in heart rate and electrodermal activity predicts decreased vigilant attention induced by sleep deprivation

Ivo V. Stuldreher, Emma Maasland, Charelle Bottenheft, Jan B. F. van Erp, Anne-Marie Brouwer

Differentiating linked cognitive states under task demand and sleep restriction: toward an experimental framework for operational monitoring

Maria Rivas-Vidal, Alberto Calvo Cordoba, Fernando Daniel Farfán, Cecilia E. García-Cena

Real-world settings involve multiple interacting cognitive states such as mental workload, drowsiness, fatigue, and mind wandering, yet these are typically studied in isolation. A critical question is whether these states can be empirically differentiated given that they share overlapping behavioral and physiological signatures. We investigated whether task difficulty and sleep restriction produce dissociable effects across these states, and whether their response patterns reflect distinct underlying mechanisms or a common cognitive resource using a unified framework. Participants from operational populations (Study 1: n = 40; Study 2: n = 27) performed three complementary tasks (MATB-II, N-Back, and SART) under two manipulations: task difficulty and partial sleep restriction, targeting cognitive load and alertness. Increasing task difficulty degraded performance and increased perceived workload and fatigue across tasks. In contrast, sleep restriction primarily elevated subjective drowsiness and mind wandering, with smaller, task-dependent effects on performance. These findings provide evidence that cognitive and physiological states influencing situational awareness can be differentiated using complementary manipulations and characterized through distinct behavioral and subjective measures. This distinction provides a foundation for future predictive and monitoring approaches in applied settings.

Physiological activation regimes reveal distinct autonomic mechanisms underlying operational performance under cognitive workload and stress

Ludovica Di Pompeo, Michele Tritto, Francesco Romano, David Perpetuni, Daniela Cardone, Arcangelo Merla

Human performance in safety-critical environments depends on the interaction between mental workload, stress, autonomic regulation, and behavior. This study investigated whether associations between heart rate variability (HRV) and performance varied across physiological activation ranges and operational contexts during a multi-task flight simulation implemented through the AeroStim platform. Forty-two healthy participants (25 ± 5 years) performed either a Workload-Only condition (WO; n = 20), comprising Tracking and System Monitoring tasks, or a Stress-Oriented condition (SO; n = 22), in which additional stress-inducing tasks were concurrently administered. ECG was continuously recorded using an EqVital wearable T-shirt. Linear and non-linear HRV indices were extracted and normalized within participants using Min-Max scaling. Gaussian Mixture Modeling of normalized SDNN identified three relative, subject-specific activation ranges (Easy, Medium, and Hard), with classification showing high stability across alternative normalization procedures. HRV-performance associations were assessed using Spearman correlations with permutation-based inference and Benjamini–Hochberg correction, while performance differences were evaluated using non-parametric tests. HRV-performance correlations varied across activation ranges and contexts; none passed FDR correction, so these should be considered exploratory. In WO, associations involved LF in the Easy range, SDNN, SD2, and LF in the Medium range, and SDNN, LF/HF, and Sample Entropy in the Hard range. In SO, associations in the Medium range extended to vagal, global variability, and non-linear complexity indices, while Hard-range associations mainly involved vagal and LF-related measures. Omission rate was lower at Medium than Easy activation in WO, providing partial support for an intermediate activation advantage, whereas no consistent performance optimum emerged across contexts. Between-condition comparisons revealed a robust speed-accuracy trade-off, with WO producing faster but less accurate responses and SO producing slower but more accurate responses. Sensitivity analysis further indicated that differences in exercise windows affected HRV metrics in a metric-dependent manner, with greater deviations generally observed for frequency-domain measures. The findings suggest that HRV-performance relationships are context- and activation-dependent rather than captured by a single universal marker. The results support the potential of regime-based, multimodal HRV monitoring for adaptive human-machine systems, while highlighting the need for standardized recording durations, independent autonomic markers, and validation across individuals and real operational environments.

Visual cues in MI-BCI induce biases in EEG classification models with deep learning but not with standard machine learning

David Trocellier, Simon Kojima, Bernard N'kaoua, Fabien Lotte

Deep learning classification methods are increasingly popular in the brain-computer interface (BCI) field due to their ability to learn complex, non-linear representations from EEG data. Unlike traditional machine learning approaches, which typically rely on hand-crafted features, deep learning models autonomously extract spatio-temporal features directly from the raw or minimally preprocessed signal. However, this flexibility may come at the cost of reduced interpretability and increased susceptibility to confounding factors. This study aims to investigate whether the inclusion of task-related visual cues in the input EEG window biases deep learning models by leading them to exploit irrelevant features, thereby compromising their neurophysiological validity. This study investigates how including EEG signals measured during visual cue presentation on classification accuracy and feature extraction in motor imagery (MI)-based BCIs. Three classification pipelines were evaluated: CSP+LDA, a traditional machine learning approach; DeepConvNet, a convolutional neural network for EEG classification; and REVE, a Transformer-based EEG foundation model. All methods are trained and tested on two large scale open-access EEG datasets using two epoching strategies, with or without including the visual cue onset, to assess their impact on model performance and neurophysiological relevance. Attribution maps (e.g., Saliency maps) and CSP patterns are used to interpret the features learned by each model. Including the visual cue onset significantly improved the classification accuracy of DeepConvNet and REVE. However, attribution analysis showed that the model mainly relied on early parietal activations, most likely cue-related visual evoked potentials, rather than MI-related sensorimotor rhythms (ERD/ERS), which were more prominent when the cue onset was excluded. In contrast, CSP+LDA showed minimal accuracy changes across conditions, with consistent motor cortex activation patterns. These findings demonstrate that higher classification performance does not necessarily imply neurophysiological validity. In particular, deep learning models may exploit confounding signals, such as visual cue processing, which undermines their interpretability and applicability in asynchronous or real-world BCIs, in which there is no such cue. We advocate for systematically applying explainability tools in BCI research to ensure classifiers rely on task-relevant brain activity. Additionally, we emphasize the importance of excluding visual cue periods from EEG segments containing reposenses evoked by the visual cue onset from the input to deep learning models.

The challenges of magnetic resonance imaging and the promise of portable neurotechnology for advancing scientific discovery about neurodevelopmental disorders

Nancy Raitano Lee