BIDS Manager-Pipeline: A framework for multi-subject analysis in electrophysiology

Abstract

Processing large cohorts of data is very important to increase statistical power in neuroscience and neuroimaging. The Brain Imaging Data Structure (BIDS) standard has been developed to facilitate this task, but researchers are still facing the challenge of reusing scripts to analyse data coming from different projects or different centres. Here, our goal was to develop a tool, BIDS Manager-Pipeline, to process automated analyses using open-source research software solutions on several subjects at the same time. BIDS Manager-Pipeline builds upon BIDS Manager, a software recently developed in our lab to collect, organise, and manage multimodal multicentre datasets. When processing data, BIDS Manager-Pipeline coordinates the use of several modules originating from different software environments and executable programs and can thus be considered as a bridge between BIDS datasets and software solutions. Through a graphical user interface, users can filter the subjects of interest in the dataset by ID or by other subject criteria and common input analysis parameters. BIDS Manager-Pipeline first controls whether the selected values can be applied to the selected subset, provided the inputs and the outputs, and then run the process on all selected data. The results of analyses are stored using specifications for BIDS derivatives. In addition, BIDS Manager-Pipeline creates a table gathering the different metrics resulting from the analyses across subjects, for later statistical analyses. In short, BIDS Manager-Pipeline allows processing data from large cohorts, either in basic neuroscience or in clinical research. This framework can take advantage of tools developed by the neuroscience community, by centralizing and facilitating their use.