Researchers in cognitive neuroscience routinely collect large datasets with tens to hundreds of participants. Yet the processes of data integration, curation, version control, and publishing are rarely automated. Hence, we present such an automated workflow, implemented and distributed through our open-source Python package, LSLAutoBIDS. In our workflow, we rely on LabStreamingLayer (LSL) to integrate different streamable modalities (e.g., EEG, VR, experimental events) with <1ms precision and use more custom data collection scripts for other secondary data (e.g., experimental logs, proprietary eye-tracking). Next, we curate the recorded data into the Brain-Imaging-Data-Structure (BIDS) format, adding extracted and user-specified metadata, which forms the entire dataset. To version this dataset and keep track of modifications, we use DataLad, based on git and git-annex. Finally, we directly archive (and potentially publish) the dataset on Dataverse or other open-data repositories. All these steps are automated and should be done immediately after data collection to ensure systematic data integration, curation, version control, and archiving - and ultimately making data sharing easy. The core idea of such a workflow is to change the infrastructure of large data collections in science in a way that they are open from the point of creation. Consequently, this frontloads the conversion of the large-scale datasets to a citable data publication and thus ensures Open-Science by design.

These are the slides from the Presentation (15+5 min presentation).More information about the project can be found in https://www.s-ccs.de/LSLAutoBIDS/ (including the preprint of the work)