Johanna Schwarz

Advisors

Ann-Kristin Seifer (M. Sc.), Dr.-Ing. Felix Kluge, Prof. Dr. Björn Eskofier

Duration

02 / 2023 – 08 / 2023

Abstract

Digital mobility outcomes (DMOs), such as walking speed and step length, show promise as clinical measures in many medical conditions. They can indicate functional movement disability, health decline, and predict hospitalization and mortality. Typically, gait characteristics are assessed in laboratory or clinical environments. However, those assessments suffer from artificial measurement conditions, white coat effects, and are performed only infrequently. Thus, recent research such as in the Mobilise-D project [1] focuses on transferring gait assessment into real-world environments to assess a patient’s everyday walking performance, investigate treatment and medication effects, and monitor fluctuating disease symptoms over long and continuous periods of time.

While this new measurement paradigm promises new insights into everyday movement limitations, it poses new challenges to the acquisition and analysis of the underlying data. Typically, waist or lower limb worn inertial measurement units are used to acquire movement data, but those devices can be inconvenient to the patients in terms of wearability. Wrist worn devices might be more acceptable to patients and, thus, better usable in large-scale studies, but there is only limited validation for real-world DMO analysis across a diverse set of pathological conditions. Also, upper limbs are complex locations to assess DMOs due to the high movement variability [2].

Most approaches assessing step length and walking speed focus on the application of machine learning models [3]–[10]. While the respective studies provide valuable insights into the assessment of DMOs from wrist-worn devices, they are partly based on small datasets, exhibit limited heterogeneity regarding patient populations, and use spatially confined reference systems for validation (e.g., gait carpets or treadmills limiting assessment of free walking). Some approaches use additional information from GPS systems which are fused to enhance and personalize models, but those are typically not available in combination with wrist-only systems. Overall, most current solutions are not generalizable to diverse patient populations.

The aim of this project is the implementation of a robust algorithm for the extraction of walking speed and step length from wrist-worn inertial sensor data in a data set with diverse patient populations and reference data in free-living conditions. The Mobilise-D technical validation study dataset consisting of six different populations (total n=120), including free-living walking, will be used for testing the algorithms. An independent data set should be identified for training the algorithms. Overall, the results of this study will advance the field of sensor-based real-world gait assessment.

References

[1] L. Rochester et al., “A Roadmap to Inform Development, Validation and Approval of Digital Mobility Outcomes: The Mobilise-D Approach,” Digit. Biomarkers, vol. 4, no. 1, pp. 13–27, Nov. 2020, doi: 10.1159/000512513.
[2] L. E. Diez, A. Bahillo, J. Otegui, and T. Otim, “Step Length Estimation Methods Based on Inertial Sensors: A Review,” IEEE Sens. J., vol. 18, no. 17, pp. 6908–6926, 2018, doi: 10.1109/JSEN.2018.2857502.
[3] A. Soltani, H. Dejnabadi, M. Savary, and K. Aminian, “Real-World Gait Speed Estimation Using Wrist Sensor: A Personalized Approach,” IEEE J. Biomed. Heal. Informatics, vol. 24, no. 3, pp. 658–668, 2020, doi: 10.1109/JBHI.2019.2914940.
[4] J. Liu, T. Lockhart, and S. Kim, “Prediction of the spatio-temporal gait parameters using inertial sensor,” J. Mech. Med. Biol., vol. 18, no. 7, pp. 1–8, 2018, doi: 10.1142/S021951941840002X.
[5] L. R. Peraza et al., “An Automatic Gait Analysis Pipeline for Wearable Sensors: A Pilot Study in Parkinson’s Disease.,” Sensors (Basel)., vol. 21, no. 24, Dec. 2021, doi: 10.3390/s21248286.
[6] S. Zihajehzadeh, O. Aziz, C. Tae, and E. J. Park, “Combined Regression and Classification Models for Accurate Estimation of Walking Speed Using a Wrist-worn IMU,” in 2018 40th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), Jul. 2018, vol. 3, pp. 3272–3275, doi: 10.1109/EMBC.2018.8513013.
[7] B. Fasel et al., “A wrist sensor and algorithm to determine instantaneous walking cadence and speed in daily life walking,” Med. Biol. Eng. Comput., vol. 55, no. 10, pp. 1773–1785, 2017, doi: 10.1007/s11517-017-1621-2.
[8] V. D. R. Seethi and P. Bharti, “CNN-based Speed Detection Algorithm for Walking and Running using Wrist-worn Wearable Sensors,” Proc. – 2020 IEEE Int. Conf. Smart Comput. SMARTCOMP 2020, pp. 278–283, 2020, doi: 10.1109/SMARTCOMP50058.2020.00064.
[9] J. Park, A. Patel, D. Curtis, S. Teller, and J. Ledlie, “Online pose classification and walking speed estimation using handheld devices,” in Proceedings of the 2012 ACM Conference on Ubiquitous Computing – UbiComp ’12, 2012, p. 113, doi: 10.1145/2370216.2370235.
[10] M. Altini, R. Vullers, C. Van Hoof, M. Van Dort, and O. Amft, “Self-calibration of walking speed estimations using smartphone sensors,” 2014 IEEE Int. Conf. Pervasive Comput. Commun. Work. PERCOM Work. 2014, pp. 10–18, 2014, doi: 10.1109/PerComW.2014.6815158.
[11] E. Warmerdam et al., “Proposed mobility assessments with simultaneous full-body inertial measurement units and optical motion capture in healthy adults and neurological patients for future validation studies: Study protocol,” Sensors, vol. 21, no. 17, pp. 1–13, 2021, doi: 10.3390/s21175833.
[12] C. Mazzà et al., “Technical validation of real-world monitoring of gait: a multicentric observational study,” BMJ Open, vol. 11, no. 12, p. e050785, Dec. 2021, doi: 10.1136/bmjopen-2021-050785.