Mobilise-D Technical Validation Study (TVS) dataset

Mobilise-D Technical Validation Study (TVS) Dataset

This dataset was recorded as part of the Mobilise-D project, a comprehensive initiative aimed at developing and validating digital solutions for assessing mobility in real-world environments. The Mobilise-D project seeks to address the critical need for accurate, reliable, and scalable tools to monitor and evaluate gait and mobility patterns, particularly in populations with mobility impairments.

The dataset comprises recordings from a diverse cohort of participants, including healthy individuals and patients with various mobility-related conditions. Data collection was conducted using state-of-the-art wearable sensors and devices, capturing a wide range of gait parameters and contextual information in both controlled and free-living settings. The primary objective was to ensure the robustness and precision of digital mobility assessment tools under real-world conditions.

Key features of the dataset include:

Demographic & Clinical Data: Age, gender, height, weight, and clinical diagnoses.

Sensor Data: Raw and processed data from accelerometers, gyroscopes, and other wearable sensors.

Reference Gait Parameters: Stride length, stride frequency, gait speed, and variability measures.

The dataset has undergone rigorous validation processes to confirm its accuracy and reliability. It serves as a critical resource for researchers and developers aiming to enhance digital health technologies and improve clinical assessments of mobility. The TVS dataset paves the way for future innovations in digital biomarkers and personalized healthcare solutions.

Brief Overview

This dataset contains data from 108 participants from six cohort groups that included older healthy adults (HA) and participants with potentially altered mobility due to Parkinson's disease (PD), multiple sclerosis (MS), proximal femoral fracture (PFF), chronic obstructive pulmonary disease (COPD) or congestive heart failure (CHF). Data was recorded across five measurement sites. Data availability varies between participants, and some tests might be missing for some participants.

The recording was split into a comprehensive in-lab assessment and a 2.5 hour unsupervised free living conditions. For the in-lab measurements reference information from marker-based motion capture systems and the multi-device wearable INDIP system are provided. For the free-living recording, only the INDIP system is available as a reference.

Participants wore a McRoberts MM+ IMU at the lower back. Some participants additionally wore a custom wrist-IMU at the non-dominant hand. The IMUs were synchronized with the reference system.

The following tests were performed as part of the In-Lab data capture (Name in recording file in parentheses):

Timed-Up-and-Go (Test4)

Straight Walk Comfortable (Test5)

Straight Walk Slow (Test 6)

Straight Walk Fast (Test 7)

L-Test (Test8)

Surface Test (Test9)

Hallway Test (Test10)

Simulated daily activities (Test11)

For some tests multiple trials are available. Additional trials were performed in the case of technical or performance issues. Hence, the last trial of each test should always be preferred.Tests not listed above (e.g. Test 1-3) are non-walking tests used for calibration. When performing gait related operations, these tests should be excluded.

For the free-living recording Recording4 corresponds to the actual recording. Recording 1-3 only contains calibration recordings that are usually not required.

Learn more about the data collection protocol:

C. Mazzà, L. Alcock, K. Aminian, C. Becker, S. Bertuletti, T. Bonci, P. Brown, et al. "Technical Validation of Real-World Monitoring of Gait: A Multicentric Observational Study." BMJ Open 11, 12 (2021): e050785 (https://doi.org/10/gt55p7).

S. Kirsty, T. Bonci, F. Salis, L. Alcock, E. Buckley, E. Gazit, C. Hansen, et al. "Design and Validation of a Multi-Task, Multi-Context Protocol for Real-World Gait Simulation." Journal of NeuroEngineering and Rehabilitation* 19, 1 (2022):141 (https://doi.org/10/gt55t6).

Files

/data/: Raw data files sorted by cohort/patientId/measurement_condition./participant_information.xlsx: Basic demographic and clinical information of all participants and "data quality" overview for all recordings

For each recording the following files are provided:

infoForAlgo.mat: Reduced set of relevant demographic information that is required to process the data with the Mobilise-D algorithmic pipeline

data.mat: Core data file following the Mobilise-D file format. For each trial the data contains the raw sensor data of the lower-back IMU (SU), the raw data of all reference sensors, and calculated gold standard parameters for all relevant gait parameters based on the reference system. For some participants, data from a wrist worn sensor is included.

test_list.json: A json file containing all the available tests and trials including the data.mat file. This information is also available via the data.mat file, but the json file is faster to parse and should help with identifying the correct data files to load.

Tips and Notes

Data Quality

Depending on the use case, specific trials should not be used. participant_information.xlsx file contains a sheet named data quality, that indicates for each system used, if the data was recorded properly. "0" indicates that the data is not usable at all, "1" indicates that some issues remain. This usually indicates partial or full data loss in a single test or unreliable reference information. These recordings might be usable for certain types of analysis, but should not be used for proper algorithm validation on the dataset. Only recordings with data quality >=2 for all required systems should be used.

Walking Aid Use

The participant_information.xlsx file contains 3 columns with information about walking aid use. The two columns self_reported_indoors and self_reported_outdoors describe the use of walking aids independent of the study context as reported by the patients of the day of the recording. This information might be different from the actual walking aids used during the assessment. This information can be found use_during_lab_assessment column. This information was recorded by the study conductor. For the free-living tasks patients were allowed to use any assistance they needed. Actual use was not recorded for this assessment.

In general, only a small number of participants used walking aids within the study. Therefore, we do not recommend analyzing walking aid users as a different group or including walking aid use as a stratifier.

General Notes

The participant IDs are "double pseudonymized" and do not correspond to data-ids used within the Mobilise-D project or previously published example data

The first digit of the participant IDs identifies the recording center. This information might be helpful to identify systematic domain shifts in the data, as different centers used slightly different measurement setups.

In case multiple trials are available for a single test, only use the last one when performing algorithm validation to keep the data between the participants balanced.

Reference Parameters

Below some notes and general recommendation regarding the reference parameters:

Reference parameters are provided on a MicroWb and ContinousWalkingPeriod level. In most cases, you will likely want to work with the information in "ContinousWalkingPeriod" (if you are using mobgap to load the data, this information is simply called "Wb"). Learn more here

For in-lab measurements, the Stereophoto (aka. marker-based Mocap system) should be the preferred reference, as parameters are expected to be more accurate. However, due to limitations of the field-of-view of these systems, some walking trails are not completely covered by the references.

Neither reference system includes turning information, as no established reference definition could be identified, that would allow for unbiased comparison of parameters.

Learn more about the methods for extracting reference parameters:

T. Bonci, F. Salis, K. Scott, L. Alcock, C. Becker, S. Bertuletti, E. Buckley, et al. “An algorithm for accurate marker-based gait event detection in healthy and pathological populations during complex motor tasks” Frontiers in Bioengineering and Biotechnology, section Biomechanics, 10:868928, 2022 (DOI: 10.3389/fbioe.2022.868928).

F. Salis , S. Bertuletti, T. Bonci, M. Caruso, K. Scott, L. Alcock, E. Buckley, et al. “A multi-sensor wearable system for the assessment of diseased gait in real-world conditions”. Frontiers in Bioengineering and Biotechnology, 11, 2023 (https://doi.org/10.3389/fbioe.2023.1143248).

Usage Recommendation

This dataset is designed to validate algorithms and NOT to derive clinical insights from the patient cohorts.

This dataset was used to validate the algorithms of the Mobilise-D computational pipeline for lower trunk IMU data. Details on the publications are reported below.

Per-Block Validation:

M.E. Micó-Amigo, T. Bonci, A. Paraschiv-Ionescu, M. Ullrich, C. Kirk, A. Soltani, A. Küderle, et al. "Assessing Real-World Gait with Digital Technology? Validation, Insights and Recommendations from the Mobilise-D Consortium." Journal of NeuroEngineering and Rehabilitation 20, no. 1 (June 14, 2023): 78 (https://doi.org/10/gt55qb).

Full Pipeline Validation:

K., Cameron, A. Kuederle, M.E. Mico-Amigo, T. Bonci, A. Paraschiv-Ionescu, M. Ullrich, A. Soltani, et al. "Estimating Real-World Walking Speed from a Single Wearable Device: Analytical Pipeline, Results and Lessons Learnt from the Mobilise-D Technical Validation Study." Scientific Reports, 14,1, 1754 2024 (https://doi.org/10.21203/rs.3.rs-2965670/v1).

Implementation of these validation procedures are also available via the open-source library mobgap.

We recommend the use of this library in all use cases, as it provides high level tools to load and process the dataset. Documentation for this can be found in the following examples:

The TVS dataset class

Working with data in mobgap

Working with reference data in mobgap

Suggested Citation

When you are working with the data, we suggest the following citation:

Küderle, A. (2024). Mobilise-D Technical Validation Study (TVS) dataset [Data set]. Zenodo. http://doi.org/10.5281/zenodo.13899385

Please cite our paper in your publications if our repository helps your research.

C. Mazzà, L. Alcock, K. Aminian, C. Becker, S. Bertuletti, T. Bonci et al. "Technical Validation of Real-World Monitoring of Gait: A Multicentric Observational Study". BMJ Open 11, 12 2021): e050785. (https://doi.org/10/gt55p7).

License and Legal Information

Mobilise-D Technical Validation Study Dataset © 2024 by Mobilise-D Consortium is licensed under CC BY-NC-ND 4.0

Acknowledgments

We extend our gratitude to all participants who contributed to the Mobilise-D project, enabling the comprehensive collection and analysis of mobility data. This work would not have been possible without the dedication and collaboration of the Mobilise-D Consortium members, including researchers, clinicians, and technical staff.

We also acknowledge the funding and support provided by the European Union's Horizon 2020 research and innovation program under grant agreement No 820820. Special thanks to our partner institutions and organizations for their invaluable contributions and continued support.

Disclaimer

The Mobilise-D Technical Validation Study Dataset is provided for research purposes only. The Mobilise-D Consortium makes no warranties, express or implied, regarding the accuracy, completeness, or reliability of the dataset. Users of the dataset assume all responsibility for any conclusions drawn from the data.

The dataset must be used in accordance with ethical guidelines and applicable laws and regulations. Any publications or presentations based on this dataset should appropriately cite the source. The Mobilise-D Consortium is not liable for any misuse of the dataset or for any direct, indirect, incidental, or consequential damages arising out of the use of the dataset.

ChangeLog

V1.0.0 - Initial release

V1.0.1 - README file updates only