The Harvard Automated Processing Pipeline for Electroencephalography (HAPPE): Standardized Processing Software for Developmental and High-Artifact Data.

Gabard-Durnam LJ, Mendez Leal AS, Wilkinson CL, Levin AR. · 2018

Read The Harvard Automated Processing Pipeline for Electroencephalography (HAPPE): Standardized Processing Software for Developmental and High-Artifact Data. by Gabard-Durnam LJ, Mendez Leal AS, Wilkinson CL, Levin AR. (2018) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/europmc-pmc5835235.txt.

Introduction

Gabard-Durnam and co-authors present HAPPE, free software that cleans brain-wave recordings on its own. Their paper appeared in Frontiers in Neuroscience. EEG from children is short and full of artifacts. Hand cleaning is slow and varies from person to person.

The pipeline filters, rejects bad channels, and strips artifacts in fixed steps. It then issues standard quality scores for every file. In tests it beat seven rival approaches. It removed more artifact while keeping as much true signal. Reference scores from 867 files show how it handles messy real-world data.

Catalog record

CREATORGabard-Durnam LJ, Mendez Leal AS, Wilkinson CL, Levin AR.
YEAR2018
COLLECTIONPAPERS
FORMATTXT
FILE SIZE87 KB
LICENSECC BY 4.0
RIGHTSCC BY
SOURCEEurope PMC

Source: https://europepmc.org/articles/PMC5835235

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

The Harvard Automated Processing Pipeline for Electroencephalography (HAPPE): Standardized Processing Software for Developmental and High-Artifact Data By Laurel J. Gabard-Durnam, Adriana S. Mendez Leal, Carol L. Wilkinson, April R. Levin Published in: Frontiers in Neuroscience DOI: 10.3389/fnins.2018.00097 ABSTRACT Electroenchephalography (EEG) recordings collected with developmental populations present particular challenges from a data processing perspective. These EEGs have a high degree of artifact contamination and often short recording lengths. As both sample sizes and EEG channel densities increase, traditional processing approaches like manual data rejection are becoming unsustainable. Moreover, such subjective approaches preclude standardized metrics of data quality, despite the heightened importance of such measures for EEGs with high rates of initial artifact contamination. There is presently a paucity of automated resources for processing these EEG data and no consistent reporting of data quality measures. To address these challenges, we propose the Harvard Automated Processing Pipeline for EEG (HAPPE) as a standardized, automated pipeline compatible with EEG recordings of variable lengths and artifact contamination levels, including high-artifact and short EEG recordings from young children or those with neurodevelopmental disorders. HAPPE processes event-related and resting-state EEG data from raw files through a series of filtering, artifact rejection, and re-referencing steps to processed EEG suitable for time-frequency-domain analyses. HAPPE also includes a post-processing report of data quality metrics to facilitate the evaluation and reporting of data quality in a standardized manner. Here, we describe each processing step in HAPPE, perform an example analysis with EEG files we have made freely available, and show that HAPPE outperforms seven alternative, widely-used processing approaches. HAPPE removes more artifact than all alternative approaches while simultaneously preserving greater or equivalent amounts of EEG signal in almost all instances. We also provide distributions of HAPPE's data quality metrics in an 867 file dataset as a reference distribution and in support of HAPPE's performance across EEG data with variable artifact contamination and recording lengths. HAPPE software is freely available under the terms of the GNU General Public License at https://github.com/lcnhappe/happe. INTRODUCTION Electroencephalography (EEG) is a sensitive means to noninvasively capture neurophysiological activity with clinical and basic science utility across a number of fields. However, during acquisition, the EEG signal is contaminated by both experimental artifacts like electrical interference and electrode displacement, and participant-induced artifacts like eye and muscle movements. These artifact signals are in many cases far more prominent than the neurophysiological signal, significantly distorting EEG as a measure of brain function if left uncorrected (Cuevas et al., 2014; Keil et al., 2014). Therefore, a series of post-acquisition processing transformations are typically applied to the EEG signal to address these artifacts and prepare the data for analyses, including filtering, artifact removal, and signal re-referencing (Keil et al., 2014). However, the pipelines and parameters for EEG processing vary across studies with little standardization. For example, one common artifact removal approach is labor and training-intensive researcher selection of uncontaminated EEG data, the criteria for which is partially subjective and therefore inconsistent across individuals. Standardized, automatable EEG processing pipelines thus offer several advantages, including uniform application of artifact removal criteria, efficient workflow with large sample sizes, and the facilitation of data comparisons across studies, labs and sites in multi-institution projects. Accordingly, software tools automating various stages of EEG processing have become a methodological focus (e.g., PREP, Bigdely-Shamlo et al., 2015, FASTER, Nolan et al., 2010, ADJUST, Mognon et al., 2011, TAPEEG, Hatz et al., 2015, ASR, Mullen et al., 2013, MARA, Winkler et al., 2014, SASICA, Chaumon et al., 2015), but they have largely been developed and tested on healthy adult EEG data with low levels of artifact contamination (Nolan et al., 2010; Mognon et al., 2011; Mullen et al., 2013; Bigdely-Shamlo et al., 2015; Chaumon et al., 2015; Hatz et al., 2015). Notably, EEG data from developmental populations like infants, young children, and people with neurodevelopmental disorders present further challenges to extracting uncontaminated signal. EEG signals from these populations have the highest levels of artifact contamination (e.g., infants cannot follow instructions to refrain from moving their mouth or eyes during data collection), and protocols typically have far shorter EEG collection times than those with healthy adults to accommodate reduced tolerance for testing (Tran et al., 2004; Cuevas et al., 2014). Moreover, additional polygraphic measurements used to identify physiological artifacts, like EOG electrodes, are not typically used during EEG acquisition with developmental populations due to both reduced tolerance of their acquisition and decreased signal quality. These factors combined have made it difficult to directly apply contemporary processing approaches from the adult EEG literature, like independent component analysis (ICA), that require longer recordings to most effectively parse artifact from signal (Makeig et al., 1996; Delorme and Makeig, 2004; Albera et al., 2012; Grandchamp et al., 2012) (although see Zima et al., 2012; Piazza et al., 2016). However, the typical manual artifact rejection approaches used for these developmental EEG data routinely remove the majority of the EEG recording, reducing experimental power and sacrificing the neurophysiologically relevant aspects of EEG also contained within the rejected segments (Tran et al., 2004; Tierney et al., 2012; Cuevas et al., 2014; Gabard-Durnam et al., 2015). Moreover, the emerging focus on collecting larger datasets through repositories and large-scale studies along with the use of higher-density EEG nets make manual data selection increasingly impractical as a processing strategy (Bigdely-Shamlo et al., 2015). Currently, there is an unmet and growing need for automated processing tools suitable for EEG recordings like those generated by these populations. The purpose of the Harvard Automated Preprocessing Pipeline for EEG (HAPPE) is to provide an automated, standardized approach for processing these classes of EEG data. Specifically, HAPPE is designed for data with high levels of artifact or very short recording length (on the scale of several minutes), although the pipeline may appropriately be used with longer or less-contaminated data. HAPPE integrates Matlab-based (The Mathworks, Inc.) code with freely available academic software, including EEGLAB functions (Delorme and Makeig, 2004), to automatedly batch process resting-state and event-related EEG data from raw format to corrected signal prepared for analyses in the frequency domain. HAPPE comprises both a fully-automated and semi-automated setting so that users may visualize processing performance on individual EEG files at multiple stages in the semi-automated setting, and adjust user inputs if desired, before running the complete dataset through the fully-automated pipeline. The following sections detail each of the proposed HAPPE pipeline's processing steps and post-processing report metrics, demonstrate HAPPE's effectiveness under variable conditions of EEG artifact and recording length by including an analysis of 10 developmental files, and compare HAPPE's performance to that of seven alternative, widely-used processing approaches. THE HARVARD AUTOMATED PREPROCESSING PIPELINE FOR EEG (HAPPE) HAPPE EEG inputs HAPPE accommodates multiple types of EEG with different acquisition parameters as inputs. HAPPE reads data in from EGI-exported (Electrical Geodesics, Inc.) Matlab files for resting-state EEG, and the data may have differing identifying variable names across files. HAPPE reads EGI-exported simple binary files for event-related EEG presently. However, users can easily modify the importing code to read any file format for any resting-state or event-related EEG that EEGLAB accepts. An individual HAPPE run should include only resting-state data or only event-related data, and users must specify one or the other as the input file type. HAPPE is currently

Read the complete text