Work Summary for An Open-Source Software Suite for EEG Artifact Annotation on Web Browsers |
Google Summer of Code 2022 |
From May 22nd to November 14th, I worked with Jorge Bosch-Bayard and Pedro Valdes-Sosa from the Global Brain Consortium (GBC) to produce novel web-based annotation applications for the Brain Imaging Data Structure (BIDS) under their Global EEG Norms project.
Our initial work focused on the development of gen: a Progressive Web Application (PWA) for the annotation of EEG data on the browser.
In preparation for the 2022 International Congress of Clinical Neurophysiology (ICCN), however, we decided to leverage AnyWave—an existing annotation tool developed by Bruno Columbet and Christian Benar at the The Institut de Neurosciences des Systèmes (INS) in Marseille, France.
From July through November 2022, then, the GBC and INS teams closely collaborated to update the AnyWave application to natively support annotation with Hierarchical Event Descriptor (HED) tags (anywave_hed) and be distributed to clinical neurophysiologists following ICCN 2022 (anywave_gen, in progress).
Near the end of the project, I’ve also been put into contact with a trainee who will take over the majority of the work following the GSOC 2022 project period.
All of the repos outside of AnyWave (gen, anywave_hed, anywave_gen) have been developed in the context of GSOC 2022 and are themselves an important outcome of the project.
Following our presentation at the ICCN, we received positive feedback from the clinical neurophysiology community. In fact, Pedro was able to negotiate the establishment of a joint global project between the International Federation of Clinical Neurophysiology (IFCN) and the Global Brain Consortium (GBC) for progressive harmonization and validation of advanced methods for EEG analysis for research and clinical applications.
On the software side, I was able to push the BIDS team to make improvements to their online validator by including ways to load HED schema libraries into the validator (hed-standard/hed-javascript#67). This allowed for us to validate custom HED tags in BIDS datasets on the browser—and will continue to allow others to do so.
On a personal note, I was also able able to secure a contract with the Kavli Foundation to produce a JavaScript API for Neurodata without Borders files—in part due to my work of developing a JavaScript API for reading and writing BIDS files as part of gen.
While we have been able to prototype a working version of anywave_gen, we have not been able to fully test it due to the lack of a stable version of the AnyWave application that support all of our required features. As such, we have not been able to fully test the application and have not been able to release it to the public.
As such, the next steps for the project will be to continue to work with the AnyWave team to finalize the application and release it to the public. Once this is done, we will be able to release anywave_gen to the public—as well as update the anywave_hed plugin to support the latest version of AnyWave.
Throughout this project, I’ve learned a lot about the challenges of working with large, distributed teams and the importance of clear communication and consensus-building before writing any code.
I put much time into the development of gen at the early stages of the project, we collectively decided that our time would be best spent focusing on the distribution of AnyWave to clinical neurophysiologists—not on development of the application itself. As such, the majority of the development was outsourced to the INS team, while myself and my mentors focused on determining the appropriate integrations that we’d need to acheive our goals of simple distribution of AnyWave for the collection of huge amount of EEG annotations.
While this looked significantly different than the original plan, I’m happy with the outcome and the work that I was able to do. In the coming months, I’m also excited to see the AnyWave application be used by clinical neurophysiologists around the world to annotate EEG data for the GBC project!
Artifacts are parts of a measured signal generated from sources other than those of interest. For electroencephalography (EEG) data, artifacts from a variety of physiological and environmental sources must be excluded or repaired to ensure that appropriate interpretations are applied during clinical evaluation and downstream analyses. We will develop a suite of web-based tools for facilitating artifact annotations for the Global EEG Norms project of the Global Brain Consortium (GBC). This project will enable clinical neurophysiologists to upload EEG data in standardized formats; annotate data with artifacts, such as electrical noise and movement artifacts, using a novel ontology that is interoperable with the Hierarchical Event Descriptors specification; and export this data to the Brain Imaging Data Structure standard. While the resulting libraries of this project will be integrated into a Progressive Web Application for clinical neurophysiologists, they will also be released standalone as tools to further facilitate the development of neuroscience tools on the Open Web. This could reduce the barriers to entry for neurophysiology research by enabling standalone data management and analysis applications to be released on modern browsers. In the long run, this may enable the rapid diagnosis of neurological disorders in lower to middle-income countries with unreliable internet connections.
To view my official project page, head over to the main GSOC website.
To view my early notes while working on this project, see the notes subdirectory
I’d like to thank my mentors, Jorge Bosch-Bayard and Pedro Valdes-Sosa, for their guidance and support throughout the project. I’d also like to thank Bruno Columbet and Christian Benar for their work on AnyWave and their willingness to collaborate with us on this project.