Beyond protocol standardization: The importance of data curation and software transparency

Abstract

Mancini et al.'s framework for gait assessment in Parkinson's disease (PD) is a valuable contribution, enabling a harmonization of study protocols in this research field and, consequently, a substantial improvement of data interpretation across different cohorts. However, we believe that recommendations concerning data curation and software use should be provided in more detail. To ensure data interoperability and facilitate robust data aggregation from such protocols, appropriate and harmonized data formatting and metadata standards are necessary. We further advocate for the open sharing of gait analysis algorithms, to enhance reproducibility and foster collaborative development.

Plain Language Summary Title

Why Data Sharing and Open Software Matter for Parkinson's Gait Studies

Plain Language Summary:

Understanding how Parkinson's disease affects gait is important for improving diagnosis and treatment. A recent study suggested that researchers should use the same protocol when studying walking in people with Parkinson's disease. This would help scientists compare and combine their results more easily, making research more reliable. While this is a great step forward, we believe it is also important to focus on how the data from these studies is handled and shared. This includes making sure the data is clearly organized, stored in standard formats, and described with helpful background information (called metadata). These steps make it easier for different researchers to use the data together. We also recommend that researchers share algorithms they use to analyze the data. When these tools are openly available, it is easier to doublecheck results and build better tools through teamwork. Overall, better data practices and open sharing of software can help speed up progress in Parkinson's research and lead to better outcomes for people living with the disease.

Letter

Mancini et al.'s publication ¹ marks a crucial stride towards standardized gait assessment in Parkinson's Disease (PD). To fully harness the potential of such standardized protocols, we suggest defining data curation and software aspects with greater specificity than it is implemented in the current version of the framework.

While the original framework appropriately emphasizes reporting key technical specifications, such as device type, sampling frequency, and software validation, it can benefit from designation of further details to boost true reproducibility and interoperability. Informing about device parameters or referencing algorithms is important, but does not provide sufficient transparency for others to replicate or extend analyses. ² Without access to structured raw data or actual code (particularly for ad-hoc algorithms), even well-documented studies remain difficult to reproduce.^3,4 Similarly, we argue that data and metadata standards should go beyond sharing summary variables in a .csv file or similar on an open access platform. Summary tables alone cannot meet FAIR (i.e., findable, accessible, interoperable and reusable) principles, ⁵ nor do they allow reprocessing or comparison across datasets. ⁶ Other initiatives, such as the Mobilise-D consortium, have addressed this in greater depth, identifying five key domains for standardization: file format and data structure, sensor locations and signal orientation, measurement units and sampling frequency, timing references. ⁷ To enable reproducible, cumulative science, we argue that raw data must be shared in a structured, standardized format, with clear documentation, licensing, and persistent identifiers. Moreover, the software used for analysis, ideally open-source, should be made available with version control and clear usage documentation.^8–10 Without these additional layers of openness, critical aspects of the research pipeline remain opaque, limiting verification, reuse, and collaborative progress.

It's widely recognized that motion data is recorded heterogeneously due to the diverse set of tools like marker-based motion capture systems, pressure mapping systems, or other wearable devices, which often produce different raw time-series data or parameters. However, the scientific community should advocate for the development of consensus-based specifications for motion data or parameters rather than relying on individual manufacturer formats. Drawing lessons from the neuroscience community's adoption of the Brain Imaging Data Structure (BIDS), we recognize the transformative impact of well-defined data standards.^9,11 For example, Motion-BIDS, an extension of BIDS for motion data, addresses the inherent heterogeneity of motion data by establishing a structured framework for organizing and documenting information from diverse acquisition devices, including optical motion capture, inertial measurement units, force plates, and others. It encompasses a common data format for raw motion data, specifies essential metadata such as device information and task descriptions, and organizes data within a hierarchical directory structure. Motion-BIDS could therefore serve as a valuable standard for sharing consistent and interoperable motion datasets. ¹² To support the BIDS structure, all relevant data (gait and e.g. dual-task performance) should be detected by the same software. Synchronizing motion and cognitive data ensures accurate temporal alignment, enabling more precise analysis of how physical activity correlates with mental processes. This integration enhances the validity of insights by capturing real-time interactions between movement and cognition which is a key question for example to get more information about mechanisms of freezing of gait.

To ensure that data are not only interoperable but also findable and accessible, it is essential to adopt open licensing frameworks (e.g., CC BY-NC), assign persistent identifiers (e.g., DOIs), and deposit data in respected repositories such as Open Science Framework, Zenodoo, FigShare or OpenNeuro. Some of these repositories even support the BIDS ecosystem.

Furthermore, including detailed expectations in the framework concerning software transparency, can substantially advance reproducibility and collaborative efforts in PD gait research and beyond. The emergence of rigorously validated free and open-source software (FOSS) for gait analysis offers advantages over proprietary commercial software, particularly when paired with well-curated, standardized datasets. This standardization eliminates the need for laborious, error-prone data reformatting, enabling FOSS tools to operate seamlessly on aggregated datasets. ¹³ For instance, a public algorithm for gait event detection could be deployed across studies following the proposed protocol without modification, bypassing the limitations of commercial software. Open algorithms, hosted on platforms such as those suggested by Mancini et al., permit full methodological scrutiny, while fostering community-driven development. ¹⁰ Critically, these tools avoid vendor lock-in, ensuring long-term accessibility of analytical pipelines even as commercial platforms evolve or discontinue features. By mandating open sharing of algorithms tailored to standardized data, future guidelines will motivate researchers to build upon existing work and make the development of algorithms a community effort.

By prioritizing robust data curation and open-source algorithms, reproducibility concerns in PD gait research can be directly addressed.^8,14 Open data, when structured following well documented standards and hosted on accessible platforms, allows for the synthesis of diverse datasets, enabling large-scale analyses that can yield more robust and generalizable findings. FOSS further strengthens reproducibility by exposing every step of the analysis pipeline. When both data and tools are openly shared, results become verifiable and workflows reproducible across institutions and studies, ¹³ ensuring that gait research in PD is not only standardized in protocol but also inherently reproducible. Researchers should prioritize hard- and software solutions that meet FAIR data and FOSS standards. At the same time, funding agencies should implement these standards as requirements in their calls.

An implementation of data sharing standards in PD gait studies brings important ethical considerations. The potential for widespread data sharing, especially sharing raw time-series data, must be clearly communicated with study participants. These aspects should be explicitly addressed in the information sheet and informed consent process, as well as the anonymization of data. In addition, it is important to engage participants in conversations about the broader implications of data sharing. This shift may reduce the need for future studies that duplicate already-answered research questions, thereby lessening participant burden and increasing scientific efficiency. Importantly, shared and well-documented datasets may yield additional insights over time as analytical methods evolve. For instance, data collected using foot-worn IMUs to assess step length, can later be reanalyzed with new algorithms to extract novel metrics such as foot lift or foot landing angle, which may prove even more sensitive to disease severity ¹⁵ or in combination with the dual-task, cognitive performance (e.g., reaction times) and biomechanical gait data can be examined in their interaction. This underscores the long-term scientific value of open data and reinforces the ethical imperative of transparency and participant autonomy in data sharing practices.

Mancini et al.'s ¹ framework is an important step to advance standardized gait assessment in Parkinson's disease. However, realizing its full potential necessitates more granular specifications for data curation and analytical software. While the current framework well defines task details, it insufficiently ensures analytical reproducibility and interoperability due to highlighting the importance of documenting structured raw data and analytical code.