multigait 0.2.1

MultiGait: Real-World Gait Pipeline for Wrist-Worn Devices

MultiGait: Real-World Gait Pipeline for Wrist-Worn Devices

MultiGait (Multimorbidity Gait) is a Python implementation of a gait analysis pipeline for real-world assessment, primarily developed for wrist-worn inertial measurement units (IMUs). This pipeline integrates signal-processing algorithms for gait detection, initial contact detection, stride-length estimation, cadence, and walking speed providing a streamlined workflow for mobility research in people with multiple long-term conditions (multimorbidity). While the pipeline is wrist-focused, fine-tuned versions for lower-back devices are also provided for use in custom workflows.

The individual algorithms included in this library have been developed and validated in multimorbidity cohorts [1]. However, the full integration of these algorithms into a complete pipeline has not yet been formally evaluated, so we do not currently recommend any specific pipeline configuration. In future releases, we plan to systematically assess combinations of algorithms and provide recommended pipelines tailored to different multimorbidity clusters.

Planned future releases include:

• Fully validated pipelines for different multimorbidity clusters.
• Novel algorithm implementations.
• Validated wear-time detection algorithms.

Note on biomechanical definitions : The biomechanical logic and gait event definitions implemented in MultiGait are based on the specifications defined within Mobilise-D.

Software developed by Dr Dimitrios Megaritis. Scientific authorship is listed in CITATION.cff.

---

Table of Contents

• Installation
• Usage
  • Pipeline Use
  • Specific Algorithms
• Citation
• Validation reference for the underlying algorithms
• Funding and Support
• License

---

Installation

From PyPI (recommended)

python3 -m pip install multigait

From GitHub:

pip install git+https://github.com/DMegaritis/multigait.git

Or clone the repository and install locally:

git clone https://github.com/DMegaritis/multigait.git
cd multigait
pip install .

---

Usage

The package is designed to be used in two main modes:

Pipeline Use

High-level pipelines allow loading raw IMU data and obtaining gait outcomes end-to-end. Example:

from multigait.pipeline.multigait_pipeline import MultiGaitPipelineSuggested

pipeline = MultiGaitPipelineSuggested()
pipeline.safe_run(long_trial)

Note: At this stage we offer a suggested pipeline with the best-performing algorithms from [1], but the full pipeline has not been fully validated yet. Please interpret results cautiously.

Additional note: While this library has been primarily developed as a complete gait pipeline for wrist-worn devices, it also includes support for lower-back algorithms. Fine-tuned versions for lower-back devices are provided and can be used in custom pipelines.

Specific Algorithms

You can also use individual algorithms separately to build custom workflows. Example modules include:

• Gait Sequence Detection (GSD)
• Initial Contact Detection (ICD)
• Stride Length Estimation (SL)
• Cadence (CAD; requires prior GSD and ICD)
• Walking Speed (Ws; requires prior GSD, CAD, and SL)

For usage examples and input/output formats, see the examples in this repository or in DMegaritis/multimobility_wrist.

Digital Mobility Outcomes (DMOs)

The following table summarizes all Digital Mobility Outcomes (DMOs) extracted by MultiGait, including walking bout types and the statistics calculated:

Walking Bout	DMO	Statistic/Measure
All WBs	Number of walking bouts	Sum
All WBs	Total Walking Duration (min)	Sum
All WBs	Initial Contacts (n_raw)	Sum
All WBs	WB Duration (s)	Mean
All WBs	Max WB Duration (s)	90th percentile
All WBs	WB Duration bout to bout variability (s)	CV
All WBs	Cadence (steps/min)	Mean
All WBs	Stride Duration (s)	Mean
All WBs	Cadence bout to bout variability (steps/min)	CV
All WBs	Stride Duration bout to bout variability (s)	CV
All WBs	Cadence within bout variability (CV)	Mean
All WBs	Stride Duration within bout variability (CV)	Mean
All WBs	Cadence within bout variability (RMSSD)	Mean
All WBs	Stride Duration within bout variability (RMSSD)	Mean
10–30s WBs	Number of walking bouts	Sum
10–30s WBs	Walking Speed (m/s)	Mean
10–30s WBs	Stride Length (m)	Mean
10–30s WBs	Walking Speed within bout variability (CV) (m/s)	Mean
10–30s WBs	Stride Length within bout variability (CV) (m)	Mean
10–30s WBs	Walking Speed within bout variability (RMSSD) (m/s)	Mean
10–30s WBs	Stride Length within bout variability (RMSSD) (m)	Mean
>10s WBs	Number of walking bouts	Sum
>10s WBs	Max Walking Speed (m/s)	90th percentile
>30s WBs	Number of walking bouts	Sum
>30s WBs	Walking Speed (m/s)	Mean
>30s WBs	Stride Length (m)	Mean
>30s WBs	Cadence (steps/min)	Mean
>30s WBs	Stride Duration (s)	Mean
>30s WBs	Max Walking Speed (m/s)	90th percentile
>30s WBs	Max Cadence (steps/min)	90th percentile
>30s WBs	Walking Speed bout to bout variability (m/s)	CV
>30s WBs	Stride Length bout to bout variability (m)	CV
>30s WBs	Cadence within bout variability (CV) (steps/min)	Mean
>30s WBs	Stride Duration within bout variability (CV) (s)	Mean
>30s WBs	Walking Speed within bout variability (CV) (m/s)	Mean
>30s WBs	Stride Length within bout variability (CV) (m)	Mean
>30s WBs	Cadence within bout variability (RMSSD) (steps/min)	Mean
>30s WBs	Stride Duration within bout variability (RMSSD) (s)	Mean
>30s WBs	Walking Speed within bout variability (RMSSD) (m/s)	Mean
>30s WBs	Stride Length within bout variability (RMSSD) (m)	Mean
>60s WBs	Number of walking bouts	Sum
All WBs	Alpha	-

CV: Coefficient of Variation; RMSSD: Root Mean Square of Successive Differences

---

Citation

If you use MultiGait in your research, please cite:

@software{megaritis2025wristmobility,
  author    = {Megaritis, Dimitrios and Alcock, Lisa and Scott, Kirsty and Hiden, Hugo and Cereatti, Andrea and Vogiatzis, Ioannis and Del Din, Silvia},
  title     = {MultiGait: Real-World Gait Pipeline for Wrist-Worn Devices for Multimorbid Populations},
  year      = {2025},
  publisher = {Zenodo},
  doi       = {https://doi.org/10.5281/zenodo.17903930},
  url       = {https://zenodo.org/records/17903930}
}

Validation reference for the underlying algorithms

[1] Megaritis, D., Alcock, L., Scott, K., Hiden, H., Cereatti, A., Vogiatzis, I., & Del Din, S. (2025). Real-World Wrist-Derived Digital Mobility Outcomes in People with Multiple Long-Term Conditions: A Comparison of Algorithms. Bioengineering, 12(10), 1108. https://doi.org/10.3390/bioengineering12101108

---

Funding and Support

This work was supported by the Medical Research Council (MRC) Gap Fund award (UKRI/MR/B000091/1).

---

License

The MultiGait library is licensed under the Apache License 2.0. It is free to use for any purpose, including commercial use, but all distributions must include the license text.