Caution
mobgap is currently under active development and not ready for production use. Do not use any of the algorithm results for actual research purposes. Most of them are not in their final state and are not properly validated yet.
Learn more about this in our blog post about the alpha release.
Caution
v0.7.0 has a critical performance update for the default Impaired Walking pipeline!
We highly recommend updating! See more in the Changelog.
A Python implementation of the Mobilise-D algorithm pipeline for gait analysis using IMU worn at the lower back (Learn more about the Mobilise-D project). This package is meant as reference implementation for research and production use.
We are open to contributions and feedback, and are actively interested in expanding the library beyond its current scope and include algorithms and tools, that would allow mobgap to grow into a general purpose library for gait and mobility analysis.
First install a supported Python version (3.9 or higher) and then install the package using pip.
pip install mobgapFor all optional features to work, you might need to additionally install the following packages:
pip install openpyxl pingouinThis can also be done in one step by installing the package with the all extra:
pip install "mobgap[all]"If you need the latest unreleased version of mobgap, install the package using pip (or poetry) with the git repository URL
pip install "git+https://github.com/mobilise-d/mobgap.git" --upgradeIf you run into problems, clone the repository and install the package locally.
git clone https://github.com/mobilise-d/mobgap.git
cd mobgap
pip install .Or the equivalent commands of the python package manager you are using to install local dependencies.
The package is designed to be used in two modes:
-
Usage as a full end-to-end pipeline:
We provide high level pipelines that take in raw sensor data and output final gait parameters on a walking bout level, and on various aggregation levels (e.g. per day or per week). These pipelines were validated as part of the Technical Validation Study of Mobilise-D and are the recommended way to obtain gait parameters according to the Mobilise-D algorithms. Depending on the clinical cohort and the amount of gait impairment, we recommend different pipelines. When using the pipelines in the recommended way, you can expect error ranges as reported in [1]. Outside, this recommended use case, we cannot provide any supported evidence about the correctness of the results.
If you are using the pipelines in this way, we recommend citing [1] and [2] as follows:
Gait parameters were obtained using the Mobilise-D algorithm pipeline [1, 2] in its official implementation provided with the mobgap Python library version {insert version you used}.
When appropriate, include the link to the mobgap library as a footnote or as an "online resource" in the reference list.
In general, we would like to ask you to be precise about the version of the mobgap library you used and only use the term "Mobilise-D algorithm pipeline" if you used the pipelines as described in the technical validation study and not when you just use individual algorithms (see point 2) or use the pipelines with modified parameters.
In the latter case, we recommend the following citation:
Gait parameters were obtained using an approach inspired by Mobilise-D algorithm pipeline [1, 2]. The algorithm pipeline was implemented based on {name of Pipeline class} available as part of the mobgap Python library version {insert version you used} with the following modifications: {insert modifications you made}.
[1] Kirk, C., Küderle, A., Micó-Amigo, M.E. et al. Mobilise-D insights to estimate real-world walking speed in multiple conditions with a wearable device. Sci Rep 14, 1754 (2024). https://doi.org/10.1038/s41598-024-51766-5 [2] Micó-Amigo, M., Bonci, T., Paraschiv-Ionescu, A. et al. Assessing real-world gait with digital technology? Validation, insights and recommendations from the Mobilise-D consortium. J NeuroEngineering Rehabil 20, 78 (2023). https://doi.org/10.1186/s12984-023-01198-5 -
Usage of individual algorithms:
Besides the pipelines, we also provide individual algorithms to be used independently or in custom pipelines. This can be helpful to build highly customized pipelines in a research context. But be aware that for most algorithms, we did not perform a specific validation outside the context of the official pipelines. Hence, we urge you to perform thorough validation of the algorithms in your specific use case.
If you are using individual algorithms in this way, we recommend citing the original papers the algorithms were proposed in and mobgap as a software library. You can find the best references for each algorithm in the documentation of the respective algorithm.
Gait parameters were obtained using the {name of algorithm} algorithm [algo-citation] as implemented in the mobgap Python library version {insert version you used}.
When appropriate, include the link to the mobgap library as a footnote or as an "online resource" in the reference list.
While mobgap is a relatively young project, it is used in multiple projects and by multiple companies. Below a list (in no particular order) of projects and companies that use mobgap:
- Mobilise-D: (obviously)
- ActiGraph: Learn more
- Empatica: Learn more
If you are using mobgap in your project or company and would like to be listed here, please let us know via Github Issues or Email.
The library was developed as part of the Mobilise-D project under the lead of the Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU). The original copyright lies with the Machine Learning and Data Analytics Lab (MAD Lab) at the FAU (See NOTICE). For any legal inquiries regarding copyright, contact Björn Eskofier. Copyright of any community contributions remains with the respective code authors.
The mobgap library is licensed under an Apache 2.0 license. This means it is free to use for any purpose (including commercial use), but you have to include the license text in any distribution of the code. See the LICENSE file for the full license text.
Please note that this software comes with no warranty, all code is provided as is. In particular, we do not guarantee any correctness of the results, algorithmic performance or any other properties of the software. This software is not a medical product nor licensed for medical use.
Neither the name "Mobilise-D" nor "mobgap" are registered trademarks. However, we ask you to use the names appropriately when working with this software. Ideally, we recommend using the names as described in the usage recommendation above and not use the name "Mobilise-D algorithm pipeline" for any custom pipelines or pipelines with modified parameters. If in doubt, feel free ask using the Github issue tracker or the Github discussions.
If you are planning to make any changes to the code, follow this guide
To run typical development tasks, you can use the provided poethepoet commands:
>>> poetry run poe
...
CONFIGURED TASKS
format
format_unsafe
lint Lint all files with ruff.
ci_check Check all potential format and linting issues.
test Run Pytest with coverage.
test_ci Run Pytest with coverage and fail on missing snapshots.
docs Build the html docs using Sphinx.
docs_clean Remove all old build files and build a clean version of the docs.
docs_linkcheck Check all links in the built html docs.
docs_preview Preview the built html docs.
version Bump the version number in all relevant files.
conf_jupyter Add a new jupyter kernel for the project.
remove_jupyter Remove the project specific jupyter kernel.
update_example_data Update the example data registry.
Before you push, you should run the format, lint and test task to make sure your code is in a good state.
Some of the tests can only be executed when certain data is available. To make sure that the tests concerning the TVS dataset are run, you need to export an environment variable with the path to the TVS dataset.
MOBGAP_TVS_DATASET_PATH="/path/to/tvs/dataset" poe testThis work was supported by the Mobilise-D project that has received funding from the Innovative Medicines Initiative 2 Joint Undertaking (JU) under grant agreement No. 820820. This JU receives support from the European Union‘s Horizon 2020 research and innovation program and the European Federation of Pharmaceutical Industries and Associations (EFPIA). Content in this publication reflects the authors‘ view and neither IMI nor the European Union, EFPIA, or any Associated Partners are responsible for any use that may be made of the information contained herein.
And of course, this development was only made possible by the joint work of all Mobilise-D partners.
