Fig. 1

EEG-IP Architecture and Data Flow. Schematic of how “acquired data” flows from the acquisition site through the platform, and ultimately to “publications” output. White boxes represent data states and colored boxes represent coded procedures. Orange boxes represent procedures that may vary from project to project often handling the harmonization of unique data set features. The green box represents the maximally standardized process that is designed to run similarly across projects and is optimized for execution on HPC resources. The red box is the manual process in which each file is inspected and annotated by a reviewer. “Acquired data” enters the platform as unmodified raw data files in the “BIDS-EEG set” as “Source data”. The “source data” is then processed with the “initiation” to become “BIDS-compliant raw data” (open file format and containing the appropriate annotation files, etc.). Once the data are BIDS-compliant they are processed into the “BIDS-derivative Lossless” state beginning with a “staging” procedure designed to harmonize properties of the data (e.g., coregister channel locations to a standard head surface) then run through the “Lossless” pipeline. Each recording session file in the resulting “preprocessed data in Lossless state” is then examined by a reviewer in an optimized QC procedure that results in “reviewed data quality annotation”. From this pre-processed state the annotations are used to guide the post processing “feature extraction & hypothesis testing” procedures that result in the output “publications”