Mobgap: A State-of-the-Art Python Framework for Reproducible Estimation and Algorithm Validation of Digital Mobility...

Area of Science:

• Digital health
• Wearable technology
• Biomedical engineering

Background:

• Continuous real-world mobility assessment via wearables can transform clinical research.
• Current limitations include a lack of standardized, open-source tools for reproducible algorithm validation.
• This hinders transparency and meaningful comparison across studies.

Purpose of the Study:

• To re-implement and re-validate a comprehensive analytical pipeline for digital mobility outcomes.
• To overcome limitations of previous implementations (licensing, reproducibility, accessibility).
• To introduce the open-source mobgap Python package as a standardized benchmarking platform.

Main Methods:

• Re-implementation of a validated analytical pipeline into the open-source mobgap Python package.
• Integration and benchmarking of algorithms within the mobgap ecosystem.
• Re-validation against gold standards and reference data across six clinical cohorts under real-world conditions.

Main Results:

• The mobgap Python package demonstrated comparable or superior performance to the original implementation.
• Walking speed was estimated with an absolute error of 0.10 m/s and an intraclass correlation coefficient (ICC) of 0.81 across all cohorts.
• The package ensures reproducible validation of mobility analysis algorithms.

Conclusions:

• The mobgap ecosystem provides a standardized, open-source solution for validating wearable-based mobility algorithms.
• It enhances transparency, reproducibility, and comparability in digital mobility research.
• Mobgap serves as a reference implementation and benchmarking platform for the scientific community.