Database Open Access

ScientISST MOVE: Annotated Wearable Multimodal Biosignals recorded during Everyday Life Activities in Naturalistic Environments

João Areias Saraiva Mariana Abreu Ana Sofia Carmo Hugo Plácido da Silva Ana Fred

Published: Nov. 13, 2023. Version: 1.0.0 <View latest version>


When using this resource, please cite: (show more options)
Areias Saraiva, J., Abreu, M., Carmo, A. S., Plácido da Silva, H., & Fred, A. (2023). ScientISST MOVE: Annotated Wearable Multimodal Biosignals recorded during Everyday Life Activities in Naturalistic Environments (version 1.0.0). PhysioNet. https://doi.org/10.13026/sg89-qq52.

Additionally, please cite the original publication:

Areias Saraiva, J.; Abreu, M.; Carmo, S.; Plácido da Silva, H.; Fred A., "Annotated Wearable Multimodal Biosignals recorded during Everyday Life Activities", Scientific Data (2023)

Please include the standard citation for PhysioNet: (show more options)
Goldberger, A., Amaral, L., Glass, L., Hausdorff, J., Ivanov, P. C., Mark, R., ... & Stanley, H. E. (2000). PhysioBank, PhysioToolkit, and PhysioNet: Components of a new research resource for complex physiologic signals. Circulation [Online]. 101 (23), pp. e215–e220.

Abstract

Physiological processes are affected and responsive to daily stimuli, and their daily monitoring can be key to understanding well-being and detecting diseases. Since most physiological data is acquired at rest, or in controlled scenarios, many decision algorithms based on physiological data may not perform as well during free-living conditions. ScientISST MOVE is a 3-wearable biosignal dataset recorded during natural everyday activities, including lifting a chair, greeting, gesticulating, walking and running. The wearable devices: a chestband, an armband, and the Empatica E4 wristband, recorded a double view of electrocardiography (ECG), electrodermal activity (EDA), and photoplethysmography (PPG). Additionally, they also recorded bicep electromyography (EMG), wrist temperature (TEMP) and chest and wrist actigraphy (ACC). 17 healthy volunteers participated in the data collection, resulting in a total of 10.5 sensor-simultaneous useful hours. ScientISST MOVE enables the study of the full body state in free-living environments, and the effect of daily activities on physiology. Moreover, the double view of physiological data herein provided could be used in signal quality comparison studies.


Background

The human physiology responds dynamically to factors such as age, sex, and health conditions [1-3]. It is also influenced by daily activities, such as sitting or walking [4], performing physical efforts [5] or highly cognitive tasks [6]. The human body behaves differently when at rest or at stress [7, 8], when at different temperatures [3, 9], and it is also influenced by circadian rhythms [10, 11].

Recorded biosignals provide insights into these physiological changes [12, 13]. Key biosignal modalities include ECG, PPG, SpO2, EMG, EDA, and motion sensors like ACC and GYR. Wearable devices have gained popularity for biosignal acquisition [14-16], since they offer portability, ease of use, and continuous monitoring, making it feasible to capture daily-life physiological phenomena [17].

Wearable biosignal datasets are crucial for developing and testing denoising, feature extraction, and event detection algorithms that can perform well in real-world conditions [18]. Our contribution, the ScientISST MOVE dataset, includes 7 biosignal modalities, 3 sensor-equipped devices, and over 6 everyday activities, allowing different projects in signal denoising and wearable comparison, studying the physiological differences between daily activities, or that aim at getting a full picture of the body state on daily environments.


Methods

We collected the electrocardiogram (ECG), the electrodermal activity (EDA), the photoplethysmogram (PPG), the electromyogram (EMG), the skin temperature (TEMP), the chest acceleration (C-ACC), and the wrist acceleration (W-ACC), of healthy volunteers. The following subsection describes the hardware setup.

Hardware

In total, 10 sensors were used, distributed over three devices, each identified as Sx, where x is a number from 1 to 10. The following Table summarises the modalities and placement of each sensor:

Sensor Modality Unit Body Location Type of Electrode Device
S1 ECG mV Chest Gel ScientISST Chest
S2 ECG mV Chest Dry conductive textile ScientISST Chest
S3 EDA uS Left wrist Gel ScientISST Forearm
S4 EDA uS Left wrist Dry Empatica E4
S5 PPG - Left index finger - ScientISST Forearm
S6 PPG - Left wrist - Empatica E4
S7 ACC g Chest - ScientISST Chest
S8 ACC g Left wrist - Empatica E4
S9 EMG mV Left bicep Gel ScientISST Forearm
S10 TEMP ºC Left wrist - Empatica E4

Two of the devices were crafted out of ScientISST Core boards [19], one to be worn on the chest and another to be worn on the left forearm, hence herein they will be referred to as ScientISST-Chest and ScientISST-Forearm, respectively. In these two, all sensors were synchronously sampling at 500 Hz. S1 is connected to three gel electrodes, two of them were placed contralaterally on the chest - equivalent to Lead I - and the ground on the left iliac crest, all of them secured in place with surgical tape. S2 is connected to the dry contacts of a Polar chest band with no ground reference. S1 electrodes were placed right below the dry contacts, so that they capture a double-view of the same biosignal - potentially, one channel more resistant to noise (S1) and the other more prone to noise (S2). The same arrangement was prepared for the pairs S3-S4 and S5-S6. S5 was placed on the left index finger, using a gold-standard optical interface, also secured in place with surgical tape, whereas S6 recorded with the Empatica E4 wrist optical interface. The third device - the Empatica E4 wristband [20] - acquired EDA at 4 Hz, PPG at 64 Hz, temperature at 4 Hz and ACC at 32 Hz sampling frequency.

Acquisition Protocol

Each session with a volunteer was divided in five stages:

  1. Briefing and Questionnaire: Demographic and clinical information was collected from the volunteers.
  2. Wearing and Adjusting the Hardware Setup: All the devices and electrodes were placed on the volunteers' body. The researcher operating the experiment would check if every sensor was properly acquiring and sending data to the dedicated devices. Two mobile phones were used for mobility convenience. On these, both the ScientISST Sense Web App and the Empatica E4 Connect App were installed, to deal with the synchronous acquisition and data storage. Before proceeding, the initial timepoint of the session was marked on all three devices, in order to later synchronise the biosignals.
  3. Acquisition of Biosignals: The protocol was thought out to include a wide range of activities humans execute on a daily basis. During each session, upon request of the accompanying researcher, subjects would press the Empatica E4 event button to mark the onset and offset of each activity. The volunteers were asked to execute the following main activities:
    1. Lift: To repeatedly lift a chair;
    2. Greetings: To repeatedly handshake and to wave with the left hand;
    3. Gesticulate: To gesticulate with both hands while talking;
    4. Jumps: To repeatedly jump;
    5. Walk-Before: To walk outside before running;
    6. Run: To run outside;
    7. Walk-After: To walk outside after running.
  4. Uploading and De-Identifying the Data: All files from the ScientISST and Empatica E4 devices were retrieved to a computer and opened on Python with LTBio [21]. According to the initial timepoints of each device, LTBio synchronises all biosignals. Also with LTBio, regarding the subject, the following associations were made to the biosignals - age, gender, any reported medical conditions, surgical procedures, and current medication - information which was gathered in Stage 1. Moreover, a subject code was attributed to each volunteer. Subject codes were sequences of 4 alphanumeric figures randomly generated. Also, the start date and time of the biosignals was shifted to midnight of the first of January of 2000. Finally, the biosignals were saved in the .biosignal format, the serialisation format offered by LTBio, and the original files were deleted. No name, birthdate, session date, or any other unique identifier, according to the HIPAA Safe Harbor De-Identification guidelines, was present in the biosignal files.
  5. Annotating the Biosignals: Later on the same day of each session, after the volunteer had left, the research operating the experiment annotated the biosignals with LTBio's Events. Every event was reviewed and annotated with a standard set of labels (e.g. "run", "lift", etc.), which is described in Data Description. Additionally, more associations were made to the biosignal, namely, location of the electrodes and name of the channels, which was an automated equal procedure for all subjects.

Data Description

Participants

Seventeen volunteers were selected to participate in the acquisition trial. The cohort comprehends 10 (59%) male and 7 (41%) female Caucasian subjects, with a median age of 24 years old. In the file subjects_info.csv ​​, find each subject details in a tabular form. For each subject (row), you can access in each column their age at the time of acquisition, their gender, and their relevant clinical history.

All subjects completed their session without quitting and, to the present day, none has requested the deletion of any of their biosignals or associated data. However, for different reasons, not all activities were performed in all sessions. The following Table shows which activities were performed in each session and the useful recorded duration of each. By "useful" it should be understood "after discarding the periods in which no activity was being executed".

Subject Code Baseline Lift Greetings Gesticulate Jumps Walk-before Run Walk-After Total
3B8D 65 56 44 102 - 304 1649 329 2552
03FH 9 - - - - - 1300 346 1655
3RFH 184 121 43 239 64 116 1784 209 2764
4JF9 101 65 67 108 - 118 1753 - 2215
93DK 150 47 52 - 43 65 1540 - 1900
93JD 93 145 47 85 - 271 1285 239 2166
AP3H 258 62 53 81 - 94 1824 88 2463
F408 37 65 40 100 - 131 1480 7 1863
H39D 48 59 - - - 58 4333 90 4590
JD3K 43 95 31 68 - 267 1831 130 2468
KF93 177 81 31 127 - 229 1927 86 2661
KS03 171 147 64 279 - 318 - - 983
LAS2 - 91 38 75 25 164 659 - 1055
LDM5 57 60 39 102 - 234 2229 37 2726
ME93 - - - - - - 899 - 899
LK27 159 102 40 103 - 321 1896 247 2869
K2Q2 91 66 89 - - 280 564 889 1983
Total 1643 1262 767 1469 132 2970 26953 2697 37821

Additionally, as part of technical difficulties, natural to experimental data gathering studies, the ScientISST-Forearm acquisition of session ME93 and a portion of session H39D were considered invalid and, therefore, excluded from the dataset.

EDF Files Structure

The biosignals are provided in EDF+C files, which can be opened in Python or MATLAB, for instance. These files are grouped by subject, each subject having its own directory. The following tree structure is found subject/x.edf, where subject is the  subject's code, and x is any of the following {scientisst_chest, scientisst_forearm, empatica }.

Each file:

  • Includes the signals from the device after which the file is named. For instance, scientisst_chest.edf files contain S1, S2 and S7. The channels can be identified by the names { ecg-gel, ecg-dry, eda-gel, eda-dry, ppg-index, ppg-wrist, acc-chest-AXIS, acc-wrist-AXIS, emg, temp }, respectfully from S1 to S10, where AXIS can take the values { x, y, z }.
  • Includes the activities onsets and durations, in the EDF annotations channel, with the labels { baseline, lift, greetings, jumps, walk_before, run, walk_after }.
  • Includes, in the header, the subject's code and gender.

In some sessions, sub-activities were annotated, such as if the subject was going downstairs (walk_before_downstairs), or when running if the subject sprinted (sprint), or when lifting the chair if the activity was repeated (lift-1, lift-2).


Usage Notes

We provide a Jupyter notebook to get started with the dataset in our GitHub repository [22], which were developed and tested on Python 3.10. There, this notebook (edf.ipynb) let's you get a grasp on how to open the files and index activity periods of interest. At the end, it also presents some examples on how to preprocess (e.g. filter, normalise, etc.) the signals. If using EDF files, the MNE package [23] is a recommended choice for Python users, as it also offers post-processing methods.


Ethics

This trial and respective acquisition protocol were unanimously approved by the Ethics Committee of Instituto Superior Técnico (Lisbon, Portugal), under the process of internal reference number 22/2022. There were no adverse events to declare in the course of the trial. Moreover, volunteers were informed they could take breaks between the activities of the third stage, or even during the activities if they were feeling any discomfort. Volunteers also had the right to drop out of the trial at any stage of the protocol, or even after the session had ended, in which case all their biosignals and data would be immediately and permanently deleted.

Volunteers were informed that all biosignals and data acquired from them were stored unlinked from their name, address, date of session, or any other piece of information that could be linked to their identity. They were informed that, on the other hand, their age, sex, and relevant clinical history would be linked to their data, for research purposes. By signing the consent form, volunteers also gave permission to anonymously share their biosignals and data in a public dataset for research purposes.


Acknowledgements

This work was partially funded by the IST research grants BL88/2022 and BL16/2023, under the scope of project 1018P.06071.1.01.01 "CardioLeather", by the IT research grant BI16/2021, under the project PCIF/SSO/0163/2019 ”SafeFire”, and by the Fundação para a Ciência e Tecnologia (FCT) / Ministério da Ciência, Tecnologia e Ensino Superior (MCTES) research grants 2021.08297.BD and 2022.12369.BD, through national funds and when applicable co-funded by EU funds. The authors also thank to the participants that volunteered for this trial and José Gouveia for the hardware support.


Conflicts of Interest

The authors declare that they have no known competing financial interests or personal relationships which have or could be perceived to have influenced the acquisition trials. ScientISST boards and the LTBio software were provided free-of-charge by ScientISST, a non-profitable educational organisation. H.P.S. was involved in the development of the ScientISST boards and J.A.S. and M.A. in the development of LTBio. The authors have no relation with Empatica, from which the Empatica E4 device was bought.


References

  1. Hunter, S. K., Critchlow, A. & Enoka, R. M. Influence of aging on sex differences in muscle fatigability. J. Appl. Physiol. 97, 1723–1732, 10.1152/japplphysiol.00460.2004 (2004).
  2. Macfarlane, P. W. The influence of age and sex on the electrocardiogram. In Kerkhof, P. L. M. & Miller, V. M. (eds.) Sex-Specific Analysis of Cardiovascular Function, Advances in Experimental Medicine and Biology, 93–106, 10.1007/978-3-319-77932-4_6 (Springer International Publishing, Cham, 2018).
  3. Cramer, M. N., Gagnon, D., Laitano, O. & Crandall, C. G. Human temperature regulation under heat stress in health, disease, and injury. Physiol. Rev. 102, 1907–1989, 10.1152/physrev.00047.2021 (2022).
  4. Cook, G., Burton, L., Hoogenboom, B. J. & Voight, M. Functional movement screening: The use of fundamental movements as an assessment of function - Part 1. Int J Sports Phys Ther 9, 396–409 (2014).
  5. Seshadri, D. R. et al. Wearable sensors for monitoring the physiological and biochemical profile of the athlete. NPJ Digit. Med. 2, 1–16, 10.1038/s41746-019-0150-9 (2019).
  6. Posada-Quintero, H. F. & Bolkhovsky, J. B. Machine learning models for the identification of cognitive tasks using autonomic reactions from heart rate variability and electrodermal activity. Behav. Sci. 9, 45, 10.3390/bs9040045 (2019).
  7. McArdle, W. D., Katch, F. I. & Katch, V. L. Exercise physiology: nutrition, energy, and human performance (Wolters Kluwer Health/Lippincott Williams & Wilkins, Philadelphia, 2015).
  8. Ladakis, I. & Chouvarda, I. Overview of biosignal analysis methods for the assessment of stress. Emerg. Sci. J. 5, 233–244, 10.28991/esj-2021-01267 (2021).
  9. Majumder, R., Mohamed Nazer, A. N., Panfilov, A. V., Bodenschatz, E. & Wang, Y. Electrophysiological characterization of human atria: The understated role of temperature. Front. Physiol. 12 (2021).
  10. Black, N. et al. Circadian rhythm of cardiac electrophysiology, arrhythmogenesis, and the underlying mechanisms. Hear. Rhythm. 16, 298–307, 10.1016/j.hrthm.2018.08.026 (2019).
  11. Miklošíková, M., Tomaszek, L. & Malcˇík, M. Circadian rhythm and skin conductivity, their measurement and use of obtained data to plan daily activities. In Proc. of the 20th Int’l Conf. on Emerging eLearning Technologies and Applications (ICETA), 422–427, 10.1109/ICETA57911.2022.9974888 (2022).
  12. Penzel, T., Kesper, K. & Becker, H. F. Biosignal monitoring and recording. In Zielinski, K., Duplaga, M. & Ingram, D. (eds.) Information Technology Solutions for Healthcare, Health Informatics, 288–301, 10.1007/1-84628-141-5_13 (Springer, London, 2006).
  13. Mukhopadhyay, S. C. Wearable sensors for human activity monitoring: A review. IEEE Sensors J. 15, 1321–1330, 10.1109/JSEN.2014.2370945 (2015).
  14. Prieto-Avalos, G. et al. Wearable devices for physical monitoring of heart: A review. Biosensors 12, 292, 10.3390/ bios12050292 (2022).
  15. Ates, H. C. et al. End-to-end design of wearable sensors. Nat Rev Mater 7, 887–907, 10.1038/s41578-022-00460-x (2022).
  16. Kim, M. et al. Emerging bio-interfacing wearable devices for signal monitoring: Overview of the mechanisms and diverse sensor designs to target distinct physiological bio-parameters. Adv. Sens. Res. n/a, 2200049, 10.1002/adsr.202200049 (2023).
  17. Stuart, T., Hanna, J. & Gutruf, P. Wearable devices for continuous monitoring of biosignals: Challenges and opportunities. APL Bioeng 6, 021502, 10.1063/5.0086935 (2022).
  18. Yoon, D., Jang, J.-H., Choi, B. J., Kim, T. Y. & Han, C. H. Discovering hidden information in biosignals from patients using artificial intelligence. Korean J Anesth. 73, 275–284, 10.4097/kja.19475 (2020).
  19. ScientISST Sense: https://scientisst.com/sense [Accessed 20/09/2023]
  20. Empatica E4: https://empatica.com/research/e4 [Accessed 20/09/2023]
  21. LTBio: https://pypi.org/project/LongTermBiosignals/ [Accessed 20/09/2023]
  22. ScientISST MOVE Repository: https://github.com/jomy-kk/ScientISST-MOVE/tree/main/usage [Accessed 20/09/2023]
  23. MNE Library: https://mne.tools [Accessed 20/09/2023]

Share
Access

Access Policy:
Anyone can access the files, as long as they conform to the terms of the specified license.

License (for files):
Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International Public License

Corresponding Author
You must be logged in to view the contact information.
Versions

Files

Total uncompressed size: 446.9 MB.

Access the files

Visualize waveforms

Folder Navigation: <base>
Name Size Modified
03FH
3B8D
3RFH
4JF9
93DK
93JD
AP3H
F408
H39D
JD3K
K2Q2
KF93
KS03
LAS2
LDM5
LK27
ME93
LICENSE.txt (download) 0 B 2023-11-10
RECORDS (download) 1.2 KB 2023-10-23
SHA256SUMS.txt (download) 4.6 KB 2023-11-13
subject-info.csv (download) 1.4 KB 2023-10-23