Neurophysiological Dataset of Stress Resilience During Human-Computer Interaction

@article{PhysioNet-neuro-stress-resilience-hci-1.0.0,
  author = {Roy, Shotabdi and Nuamah, Joseph},
  title = {{Neurophysiological Dataset of Stress Resilience During Human-Computer Interaction}},
  journal = {{PhysioNet}},
  year = {2026},
  month = feb,
  note = {Version 1.0.0},
  doi = {10.13026/x3vc-p627},
  url = {https://doi.org/10.13026/x3vc-p627}
}

Abstract

This dataset provides multimodal neurophysiological and physiological recordings collected to investigate stress resilience. It comprises electroencephalography, functional near-infrared spectroscopy, electrodermal activity, blood volume pulse, inter-beat intervals, heart rate, and accelerometer readings from 35 participants engaged in a computer-based task under stress. The experimental design consisted of six sequential phases: Resting Baseline, Working Baseline, Stress 1, Recovery 1, Stress 2, and Recovery 2. Event markers were synchronized across all modalities to track transitions between task conditions. Additionally, subjective resilience was assessed using the 10-item Connor-Davidson Resilience Scale. The dataset also includes demographic variables such as age, gender, and ethnic background. This dataset is valuable for research on stress resilience, neuroergonomics, and machine learning applications in human factors. Researchers can utilize the data to analyze neural and physiological responses to stress, develop predictive models, and evaluate adaptive training strategies. By making this dataset publicly available, we aim to support reproducibility, interdisciplinary collaboration, and advancements in neuroergonomics.

Background

Stress resilience, the ability to maintain or recover performance under stress, is critical in dynamic operational environments such as workplaces, healthcare, and aviation [1, 2]. Identifying individuals who are susceptible to stress-related performance decline is essential for implementing targeted interventions that enhance well-being and productivity [3–6]. Despite the importance of stress resilience, current assessment methods primarily rely on self-report questionnaires, which are susceptible to bias and fail to capture the temporal dynamics of resilience as it unfolds [7–9].

Physiological and neurophysiological measures, such as heart rate variability, pupil dilation, and neural activity, offer a more objective means of studying resilience [4, 10]. However, existing studies have primarily focused on static assessments, limiting the understanding of how resilience fluctuates over time during stress exposure. The human brain, as the central organ of stress adaptation, undergoes dynamic reconfigurations during stress, yet the neural mechanisms underlying stress resilience remain poorly understood [11–14].

Electroencephalography (EEG) and functional near-infrared spectroscopy (fNIRS) are two well-established neuroimaging techniques that enable real-time monitoring of brain activity. EEG provides high temporal resolution, capturing rapid neural oscillations [15], while fNIRS offers better spatial resolution and is robust to motion artifacts [16]. Recent research suggests that combining these modalities can yield more comprehensive insights into cognitive and emotional processes during stress [17, 18].

This dataset was collected to address key gaps in resilience research by integrating multimodal neurophysiological measurements, including EEG, fNIRS, inter-beat interval (IBI), and eye tracking. This dataset provides a valuable resource for researchers studying stress resilience. Furthermore, machine learning approaches can be leveraged to extract biomarkers of resilience, potentially enabling predictive modeling of task performance under stress. The dataset is expected to facilitate novel investigations into the neurophysiological underpinnings of resilience, support the development of adaptive training interventions, and contribute to advancements in human factors research. By making this dataset publicly available, we aim to foster reproducibility, encourage interdisciplinary collaboration, and accelerate progress in stress resilience assessment.

Methods

Participants

We analyzed 35 participants (female = 13, male = 22). The mean age was 26.1 ± 4.9 years. Data from two participants were excluded due to missing values. Potential participants were invited via email and were required to complete an eligibility screening process before participation. The screening included a set of eligibility criteria, ensuring that participants had no prior experience with the Multi-Attribute Task Battery-II (MATB-II) software, did not have astigmatism that could affect eye-tracking measurements and were not using mydriatic medications that could interfere with pupillary responses. Those who met the eligibility criteria were invited to participate in the study. To assess subjective resilience, participants completed the 10-item Connor-Davidson Resilience Scale (CD-RISC), which measures psychological resilience on a scale of 0 to 40 (higher scores indicate greater resilience). All participants voluntarily provided informed consent, and compensation was offered in accordance with the Declaration of Helsinki. The study was approved by the Institutional Review Board (IRB) at Oklahoma State University.

This setup ensured synchronized neurophysiological, physiological, and behavioral data acquisition, enabling comprehensive stress resilience and cognitive workload analysis.

Experimental Design and Task Environment

The task environment was MATB-II, a validated computer-based task used in human factors research [19–21]. The MATB-II interface consists of four simultaneous tasks: System Monitoring, Tracking, Communications, and Resource Management (RESMAN). In this study, the RESMAN task served as the primary focus, requiring participants to maintain fuel levels in primary tanks A and B as close to 2,500 units as possible by transferring fuel from secondary tanks. Participants had to adjust for pump failures and fluctuating flow rates, simulating real-world operational stress. To increase cognitive load in stress conditions, the Communications (COMMS) task was introduced as the primary stressor. Participants listened to air traffic control requests and adjusted radio frequencies accordingly.

The experimental protocol included six conditions: Resting Baseline, Working Baseline, Stress 1, Recovery 1, Stress 2, and Recovery 2. During the Resting Baseline, participants relaxed with their eyes open, allowing physiological baselines to be recorded for 1 minute. The Working Baseline involved completing the RESMAN task alone. Stress conditions combined RESMAN and COMMS tasks, with increased pump failures (four pumps failing every 10 seconds) and frequent communication requests (three per minute). The Recovery conditions involved a return to RESMAN-only tasks to observe participants' recovery from stress. Each condition lasted six minutes, following a standardized timeline. Participants' task performance was continuously monitored through MATB-II log files.

Data Acquisition

To capture neurophysiological and physiological responses to stress, we collected multimodal data, including EEG, fNIRS, IBI, electrodermal activity (EDA), blood volume pulse (BVP), accelerometer (ACC), temperature, and eye tracking. EEG data were recorded using a g.Nautilus fNIRS-32 (g.tec GmbH, Austria) hybrid EEG-fNIRS system. Electrodes were placed according to the 10-20 international system, covering key brain regions associated with stress resilience, including the default mode network (DMN), salience network (SN), and central executive network (CEN). The 32 EEG channels used in the study included AF3, AF4, F7, F3, FZ, F4, F8, FT7, FT8, FC5, FC3, FC4, FC6, C5, C1, CZ, C2, C6, CP3, CP1, CP2, CP4, P7, P3, PZ, P4, P8, PO7, PO3, PO4, PO8, OZ. Hemodynamic responses were recorded using continuous wave fNIRS at 760 nm and 850 nm wavelengths. Eight fNIRS channels covered the prefrontal cortex, with emitters placed at Fp2, AFF6, AF8, AF4, Fp1, AFF5, AF3, AF7, and receivers at AF6 and AF5. These placements enabled the measurement of oxygenated and deoxygenated hemoglobin changes related to stress processing. HRV data were collected using Empatica E4 wristbands, which recorded photoplethysmography (PPG)-based heart rate variability measures. Task-evoked pupillary responses were recorded using Tobii Pro Glasses 2, a wearable eye-tracking system.

Data Description

The dataset consists of multimodal neurophysiological and physiological recordings collected from 35 participants. Participant IDs start with p followed by a numerical identifier (e.g., p01, p02, …, p37). Two participants (p14 and p23) were excluded due to missing data.

Demographic Data

The dataset includes a demographic data file, demographicCDrisk.csv, which contains information about each participant’s basic attributes. This file consists of the following columns:

• SubjectID: Unique identifier for each participant (e.g., p01, p02, …, p37).
• Gender: Self-reported gender of the participant.
• Age: Self-reported age in years.
• Ethnic Background: Self-identified ethnicity of the participant.

These demographic details allow researchers to analyze potential individual differences in neurophysiological and behavioral responses based on age, gender, and ethnicity.

Subjective Resilience Data

To assess psychological resilience, participants completed the 10-item Connor-Davidson Resilience Scale (CD-RISC). The CD-RISC score is a validated measure of resilience, with scores ranging from 0 to 40, where higher scores indicate greater resilience.

The demographicCDrisk.csv file includes a column for CD-RISC scores, which can be used to analyze individual differences in stress resilience and their correlation with neurophysiological and physiological data. This measure provides valuable insights into how resilience influences cognitive and physiological responses under stress conditions.

MATB-II Data

The MATB-II task data recorded fuel level deviations from the optimal 2500-unit target for Tanks A and B every 10 seconds. The dataset is stored in CSV format in the MATB-II folder, with filenames following the convention subjectIDresman.csv (e.g., p01resman.csv). Each file contains seven columns:

• ELAPSED_TIME: Represents the time in seconds since the task initiation.
• TANK_A and TANK_B: Fuel levels recorded for the primary tanks.
• TANK_C and TANK_D: Fuel levels for auxiliary tanks.
• DIFF_A and DIFF_B: Deviations from the optimal fuel level for TANK_A and TANK_B, respectively.

Each MATB-II session lasted 30 minutes, tracking participants' task performance throughout.

Empatica E4 Data (Physiological Data, EDA, Temperature, and Acceleration)

Physiological data from the Empatica E4 wristband is stored in the PPG folder, with each participant's data contained in an individual folder (e.g., p01/). The dataset includes several physiological signals stored in CSV files, formatted as follows:

• TEMP.csv: Records skin temperature in degrees Celsius (°C).
• EDA.csv: Captures electrodermal activity (EDA) in microsiemens (μS), measuring skin conductance as an indicator of stress.
• BVP.csv: Contains photoplethysmograph (BVP) readings, which track blood volume pulse.
• ACC.csv: Records three-axis accelerometer data, with acceleration measured in the range [-2g, 2g]. The X, Y, and Z axes are recorded in separate columns.
• IBI.csv: Contains IBI extracted from the BVP signal. Each row records the time of occurrence and the duration of the interval in seconds.
• HR.csv: Stores heart rate values (in BPM) extracted from the BVP signal.

Additionally, the dataset includes tags.csv, which logs event marker times. Each participant has one event marker recorded in this file, indicating the start of the Working Baseline condition. This marker ensures synchronization between physiological data and task-based events for further analysis. All Empatica E4 sessions follow the Unix timestamp format (UTC) to synchronize recordings with other physiological signals.

fNIRS, EEG, and Eye-Tracking Data

The neurophysiological data-including EEG, fNIRS, and eye-tracking measurements, were collected using g.tec’s g.Nautilus hybrid EEG-fNIRS system interfaced with Simulink (MATLAB) for real-time data acquisition. To accommodate file size constraints, the EEG, fNIRS, and eye-tracking data are organized into participant-grouped folders (EEGfNIRSeye_p1-p5, EEGfNIRSeye_p6-p10, EEGfNIRSeye_p11-p16, EEGfNIRSeye_p17-p21, EEGfNIRSeye_p22-p27, EEGfNIRSeye_p28-p32, and EEGfNIRSeye_p33-p37). Each folder contains multimodal recordings for the corresponding participants and is maintained under 10GB in size. This segmentation ensures easier downloads and management while preserving the original structure and naming conventions of the data. No content was altered in the process of this reorganization; only the grouping was adjusted based on participant IDs to meet storage requirements.

These recordings are saved in .csv format, containing 74 rows, each representing a specific data type:

• Row 1: Timestamp for synchronization across modalities.
• Rows 2–17: Optical density data from the fNIRS device.
• Rows 18–33: Concentration data derived from the fNIRS system.
• Rows 34–65: EEG data recorded from 32 channels following the 10-20 international system.
• Rows 66–73: Eye-tracking data, including pupil diameter (left and right eyes) and gaze direction coordinates.
• Row 74: Event markers indicating task condition transitions.

EEG, fNIRS and eye tracking signals were recorded at 250 Hz sampling rate for all participants (p01 to p24). For p25 to p37, eye-tracking data was collected separately using Tobii Pro Glasses 2 at 50 Hz sampling rate, stored in CSV format under the filename subjectIDeye.csv (e.g., p26eye.csv). fNIRS and EEG recordings remain in .csv format, where Row 74 contains event markers.

EEG Channel Locations

EEG data was collected using 32 electrodes, following the 10-20 international system. The channels and their corresponding electrode numbers are:

• Ground Electrode
  • AFz (GND)
• Frontal and Prefrontal Electrodes
  • AF3 (1), AF4 (2), F3 (5), F4 (7), F7 (3), F8 (4), Fz (6)
• Frontotemporal Electrodes
  • FT7 (8), FT8 (9)
• Frontocentral Electrodes
  • FC3 (11), FC4 (12), FC5 (10), FC6 (13)
• Central Electrodes
  • C1 (15), C2 (17), Cz (16), C5 (14), C6 (18)
• Centroparietal Electrodes
  • CP1 (21), CP2 (22), CP3 (19), CP4 (20)
• Parietal Electrodes
  • P3 (24), P4 (26), P7 (23), P8 (27), Pz (25)
• Parieto-Occipital Electrodes
  • PO3 (29), PO4 (30), PO7 (28), PO8 (31)
• Occipital Electrodes
  • Oz (32)

To provide precise electrode mapping and spatial coordinates, the EEG_32_Channel_mapping.xyz file is included in this dataset. This file contains 3D Cartesian coordinates (X, Y, Z) for each EEG electrode, ensuring accurate localization for source localization, connectivity analysis, and topographical mapping.

fNIRS Channel Locations

fNIRS data was recorded using a g.Nautilus hybrid EEG-fNIRS system, with optodes positioned over the prefrontal cortex to measure hemodynamic responses.

The fNIRS optode arrangement included:

• Emitters (Light Sources):
  • Fp2 (T1), AFF6 (T2), AF8 (T3), AF4 (T4), Fp1 (T5), AFF5 (T6), AF3 (T7), AF7 (T8)
• Receivers (Detectors):
  • AF6 (R1), AF5 (R2)

This setup enables the measurement of oxygenated (HbO) and deoxygenated (HbR) hemoglobin changes at 760 nm and 850 nm wavelengths, allowing for real-time monitoring of cerebral oxygenation during stress conditions.

For precise spatial localization, the dataset includes fNIRS_frontal.xyz, which provides 3D Cartesian coordinates (X, Y, Z) for each emitter and receiver. This file ensures accurate source-detector mapping, which is crucial for topographical visualization, connectivity analysis, and multimodal neuroimaging integration.

Eye-Tracking Data Details

Eye-tracking data was collected using g.tec and Tobii Pro Glasses 2, capturing pupil diameter and gaze direction measurements. For p01 to p24, eye-tracking data was recorded synchronously at 250 Hz sampling rate alongside EEG and fNIRS. For p25 to p37, eye-tracking data was collected separately using Tobii Pro Glasses 2 at 50 Hz sampling rate. The Tobii eye-tracking data is stored in CSV format, named subjectIDeye.csv (e.g., p26eye.csv).

The eye-tracking data recorded in .mat and .csv formats includes:

• Timestamp
• Pupil Diameter (Right Eye, Left Eye).
• Gaze Direction (Right Eye X, Right Eye Y, Right Eye Z, Left Eye X, Left Eye Y, Left Eye Z).

Event Markers

• Event markers were sent after the start of the Working Baseline every 6 minutes to track transitions between conditions.
• The final event marker corresponds to the start of Recovery 2.
• Each participant has at least five event markers, though some have additional markers due to technical errors. Any discrepancies are documented in the notes file.

Usage Notes

This dataset offers a unique opportunity to explore the neurophysiological and physiological correlations of stress resilience during human-computer interaction. While no preprocessing scripts or visualization code are provided with this release, the structure and content of the dataset are designed to facilitate reuse in a wide variety of research applications.

Data Utility and Reuse Potential

The data were collected using a standardized experimental protocol that includes both baseline and stress-inducing task conditions. The synchronized multimodal recordings (EEG, fNIRS, eye tracking, and physiological signals such as EDA, HR, BVP, IBI, and ACC) enable researchers to:

• Investigate dynamic brain and physiological responses under stress and recovery.
• Analyze individual differences in stress resilience using both subjective and objective markers.
• Apply machine learning methods to classify resilience levels or cognitive states.
• Develop or validate algorithms for signal preprocessing, feature extraction, and classification.

The dataset is compatible with major signal processing platforms such as EEGLAB, MNE-Python, FieldTrip, and Homer3, and includes spatial coordinate mapping files for EEG and fNIRS sensor locations to support topographical and connectivity analyses.

Known Limitations

Users should be aware of the following considerations when working with the dataset:

• Eye-Tracking Synchronization: For participants p25–p37, eye-tracking data were recorded separately and sampled at 50 Hz, which may limit precise alignment with EEG and fNIRS signals collected at 250 Hz.
• Event Marker Variability: Although task condition transitions were marked consistently, some participants may have extra or missing event markers due to logging errors. These instances are documented in the accompanying notes.
• Sample Size: The dataset includes 35 participants, which supports exploratory and machine learning research but may not reflect broader population diversity.
• No Starter Code: At this time, preprocessing or analysis scripts are not included, but MATLAB scripts used for synchronized data acquisition are available upon request.

Suggested Use Cases

This dataset is well-suited for:

• Research on stress resilience and adaptive systems.
• Educational demonstrations involving multimodal signal analysis.
• Benchmarking new algorithms in multimodal data fusion or physiological computing.
• Exploratory studies on resilience prediction and performance monitoring in applied settings.

Additional Notes

For spatial analysis, the dataset includes:

• EEG_32_Channel_mapping.xyz: 3D Cartesian coordinates for each EEG electrode.
• fNIRS_frontal.xyz: Coordinate file for fNIRS emitters and detectors positioned over the prefrontal cortex.

Researchers are encouraged to cite the dataset appropriately and contact the authors with questions, collaboration ideas, or requests for acquisition code.

Release Notes

Version 1.0.0: Initial public release of the dataset.

Ethics

This study was approved by the Institutional Review Board at Oklahoma State University (IRB-23-326) and conducted in accordance with the Declaration of Helsinki ethical guidelines. All participants provided written informed consent before participation and were informed of their right to withdraw at any time without consequences. Data collection procedures ensured anonymization and de-identification to protect participant confidentiality. The dataset does not contain any Protected Health Information (PHI).

The study carries minimal risk, as it involves non-invasive physiological and neurophysiological recordings during a computer-based task. The potential benefit of this research is improving our understanding of stress resilience mechanisms, which can contribute to designing safer work environments, adaptive training protocols, and cognitive workload monitoring systems.

Acknowledgements

This research was supported in part by the U.S. National Science Foundation under Award Number 2232869.

Conflicts of Interest

The authors declare that there are no conflicts of interest related to this dataset or its publication.

References

1. Colligan TW, Higgins EM. Workplace Stress: Etiology and Consequences. J Workplace Behav Health. 2006 Jul 25;21(2):89–97.
2. Melamed S, Shirom A, Toker S, Berliner S, Shapira I. Burnout and risk of cardiovascular disease: Evidence, possible causal paths, and promising research directions. Psychol Bull. 2006;132(3):327–53.
3. Winslow BD, Carroll MB, Martin JW, Surpris G, Chadderdon GL. Identification of resilient individuals and those at risk for performance deficits under stress. Front Neurosci. 2015;9:328. doi:10.3389/fnins.2015.00328.
4. Walker FR, Pfingst K, Carnevali L, Sgoifo A, Nalivaiko E. In the search for integrative biomarker of resilience to psychological stress. Neurosci Biobehav Rev. 2017 Mar;74:310–20.
5. Nuamah J. Effect of recurrent task-induced acute stress on task performance, vagally mediated heart rate variability, and task-evoked pupil response. International Journal of Psychophysiology. 2024;112325.
6. Uba J, Nuamah J. Investigating Human Physiological Responses to Work-Related Stress. Proceedings of the Human Factors and Ergonomics Society Annual Meeting. 2023 Sep;67(1):2285–90.
7. Pascual-Leone A, Bartres-Faz D. Human Brain Resilience: A Call to Action. Ann Neurol. 2021 Sep;90(3):336–49.
8. Hill Y, Den Hartigh RJR, Meijer RR, De Jonge P, Van Yperen NW. The temporal process of resilience. Sport, Exercise, and Performance Psychology. 2018 Nov;7(4):363–70.
9. Smeets T. Autonomic and hypothalamic–pituitary–adrenal stress resilience: Impact of cardiac vagal tone. Biological Psychology. 2010 May;84(2):290–5.
10. Joshi S, Li Y, Kalwani RM, Gold JI. Relationships between Pupil Diameter and Neuronal Activity in the Locus Coeruleus, Colliculi, and Cingulate Cortex. Neuron. 2016 Jan;89(1):221–34.
11. Paban V, Modolo J, Mheich A, Hassan M. Psychological resilience correlates with EEG source-space brain network flexibility. Network Neuroscience. 2019 Jan;3(2):539–50.
12. Lü W, Wang Z, You X. Physiological responses to repeated stress in individuals with high and low trait resilience. Biological Psychology. 2016 Oct;120:46–52.
13. McEwen BS, Gianaros PJ. Stress- and Allostasis-Induced Brain Plasticity. Annu Rev Med. 2011 Feb 18;62(1):431–45.
14. McEwen BS, Gray JD, Nasca C. Recognizing resilience: Learning from the effects of stress on the brain. Neurobiology of Stress. 2015 Jan;1:1–11.
15. Nuamah J, Uba J. Graph analysis of functional brain networks under varying working memory load: An EEG-based minimum spanning tree approach. Cognitive Systems Research. 2024 Aug 1;86:101238.
16. Nuamah JK, Mantooth W, Karthikeyan R, Mehta RK, Ryu SC. Neural efficiency of human-robotic feedback modalities under stress differs with gender. Front Hum Neurosci. 2019;13:287. doi:10.3389/fnhum.2019.00287.
17. Al-Shargie F, Kiguchi M, Badruddin N, Dass SC, Hani AFM, Tang TB. Mental stress assessment using simultaneous measurement of EEG and fNIRS. Biomed Opt Express. 2016 Oct 1;7(10):3882.
18. Khan MJ, Hong KS. Hybrid EEG-fNIRS-based eight-command decoding for BCI: Application to quadcopter control. Front Neurorobot. 2017;11:6. doi:10.3389/fnbot.2017.00006.
19. Santiago-Espada Y, Myer RR, Latorella KA, Comstock JR Jr. The Multi-Attribute Task Battery II (MATB-II) software for human performance and workload research: A user's guide. Hampton (VA): NASA Langley Research Center; 2011.
20. Nixon J, Charles R. Understanding the human performance envelope using electrophysiological measures from wearable technology. Cogn Tech Work. 2017 Nov;19(4):655–66.
21. Liu S, Nam CS. Quantitative modeling of user performance in multitasking environments. Computers in Human Behavior. 2018 Jul;84:130–40.