DABI accommodates a wide variety of invasive neurophysiology data types, as well as imaging, behavioral, and clinical data, in order to create a comprehensive picture of the human brain in health and disease. The following table provides an overview of data types currently accepted, but due to its adaptable design, other formats will be integrated according to end-user needs.
|Data Categories||Data Types||File Formats|
|Temporal Brain Function Recordings||EEG, ECoG, multi- or single-unit microelectrode recording, EMG, TMS.||MAT, binary electrophysiology data, EDF, NEV, NS3, NS5, NS6, NSX, MPX, XLTEK.|
|Imaging||Structural MRI, resting state fMRI, DTI, PET, CSM, CT||DICOM, NIFTI, NII, MGZ|
|Behavioral and Clinical Data||Spoken acoustics, inertia recordings, audiovisual recordings, self-reported emotion, movement-related data (accelerometer to gauge activity), diary, number of channels, patient history, medical history, cognitive assessments, experimental event data, eye tracking data, behavioral data, algorithm definition files, Unified Parkinson’s disease rating.||Python scripts, R, C code, Excel, MP4, AVI, WAV, CSV, NPY*|
*The file formats listed provide a snapshot of DABI’s capabilities but do not represent a comprehensive list. Other formats will be integrated over time according to end-user needs.
Electrophysiology data is subject to amplifier noise, technical artifacts (e.g. poor electrode location, issues with electrode impedance), and physiological artifacts (e.g. eye movement). In order to ensure consistent data quality, DABI uses range checking, signal-to-noise ratio checks, artifact removal techniques, power spectrum analysis, and various filtering methods to inspect for noise. The LONI Quality Control System (LONI QC) is used for all modalities of imaging data and regularly reviewed by participating collaborators.
All information pertaining to acquisition, QC, pre-processing, and analyses is captured and retained, providing a comprehensive history and provenance to the data. When algorithms are executed within the LONI Pipeline, provenance is captured in the form of machine- and human-readable XML files.
Common data models are used to encompass differences among data collected from separate studies, including large-scale recording and manipulation of neural circuits as well as explorations of next-generation devices for sensing and stimulation in humans. This common model will comprise a common database schema, data dictionary, and code lists. Data transformation software and a data mapping tool will then be used to spatially normalize, standardize and transform data values.