The purpose of this document is to solicit community feedback on the Hierarchical Event Descriptors (HED) that was submitted to INCF for endorsement as a standard. The document contains the INCF standards and best practices committee's review of SDS, and the criteria in which it was evaluated (open, FAIR, testing and implementation, governance, adoption and use, stability and support, extensibility and comparison to similar standards). For the next 60 days, we are seeking community feedback on SDS.
About HED:
Hierarchical Event Descriptors (aka HED) is an open standard for describing experimental events, conditions, and experiment organization in a format that is both human- and machine-readable to enable analysis, re-analysis, and meta/mega-analysis. Its goal is to annotate participant experience and behavior as well as the environmental influence during an experiment. Thus, HED is relevant to all neuroimaging modalities as well as behavioral experiments and MOBI experiments (motion capture, eye-tracking etc.).
Summary of Discussion:
Overall, the members of the INCF Standards and Best Practices Committee could see the potential of HED to meet the criteria for INCF endorsement. It clearly supports open and FAIR science. For EEG data, HED is a good way to describe common events; it provides a specification language for events and stimuli. HED is open, supports FAIR, and has appropriate testing and implementations; it also has appropriate documentation and a decent size portfolio of open tools implementing it, as well as a pipeline of open tools under development. HED benefits from a clear governance framework as well. While the committee felt that HED is broadly applicable and was clearly designed with becoming a community standard in mind, it is concerned about its adoption and use. The evidence provided by the submitters did not indicate that there are a massive number of users; but from the committee’s perspective there could be; therefore, feedback from the community on the benefits of HED are crucial.
Recommendation:
INCF Standards and Best Practices Committee voted to put HED forward for community review.
No competing interests were disclosed
Comments
Thu, 08/08/2024 - 18:31
I believe that wider adoption of HED would enable researchers to share their studies on a more global scale. It removes the ambiguity of what constitutes a target stimulus, to a distractor, to an attentional cue.
Not only does this schema explain what events are, it also allows for researchers to succinctly report on data quality properties.
I have personally adopted HED in all of my studies and will encourage all researchers I work with to do the same.
Thu, 08/08/2024 - 19:00
I am a member of the HED working group.
Thu, 08/08/2024 - 20:49
In my opinion, if we could get more and more researchers and dataset curators to apply this standard to their data, all present and future researchers would benefit immensely (especially for efficient dataset re-use, re-analysis, and for meta and mega analyses, and all other endeavors to combine data from different sources). A potential endorsement by INCF and the linked publicity and trust would be a great step into that direction.
As a BIDS maintainer, I have collaborated with core members of the HED working group, and I am co-author one of their papers (https://doi.org/10.1016/j.neuroimage.2021.118766).
Fri, 08/09/2024 - 17:10
Tue, 08/13/2024 - 22:48
Fri, 08/23/2024 - 09:41
Based on these standards, we developed a toolbox (epilepsy2bids) to convert existing epileptic EEG datasets to a common format which can be interpreted and analysed by all.
epilepsy2bids: https://github.com/esl-epfl/epilepsy2bids
SzCORE: https://arxiv.org/pdf/2402.13005
Wed, 09/18/2024 - 19:49
None.
Mon, 09/30/2024 - 21:03
I sometimes attend the publically held HED working group meetings.
Tue, 10/01/2024 - 14:59
Tue, 10/01/2024 - 17:15
As a developer of procedures for the automation of standardized signal quality control HED is invaluable, providing the framework allowing for a common shared reference structure describing aspects of complex signals.
Finally, the future potential of AI/ML in neuroscience depends on a platform for standardized tagging of data property annotations like that provided by HED.
Wed, 10/02/2024 - 17:32
Sun, 10/06/2024 - 19:57
Involved with BIDS, NWB, DANDI, OpenNeuro.
Tue, 10/08/2024 - 14:27
In my own research, I’ve successfully applied HED alongside BIDSapps like BIDSpm and FitLins to automate and streamline reproducible fMRI analysis pipelines. These tools have made it easier to handle combined datasets, enhancing consistency across different studies. I have also worked on a new library schema for HED, HED LANG, to enable the annotation linguistic events.
Moreover, I have used HED to annotate both existing and previously shared neuroimaging datasets. These annotations not only make the datasets more discoverable but also enable automated analysis across multiple datasets. HED’s ability to standardize event-related metadata is crucial for improving the reusability and interoperability of shared datasets, making it a vital tool for advancing data sharing and collaboration.
I am a member of the HED working group
Wed, 10/09/2024 - 14:45
We offer hands-on support to our users, assisting them with HED annotations to make their data more discoverable and interoperable. Additionally, we’re developing a query interface that will allow users to search for datasets based on HED tags, significantly enhancing the findability of our data.
By adopting HED as an official INCF standard, we believe the neuroimaging community will benefit from easier data sharing, enhanced dataset discovery, and more streamlined analysis across platforms. At the Austrian NeuroCloud, our commitment to best practices in data sharing is reflected in our re3data entry (https://www.re3data.org/repository/r3d100014355), and we are confident that formalizing HED as a standard will play a pivotal role in advancing open science and fostering data reuse across the field.
Thu, 10/10/2024 - 17:22
The room for improvement for annotations in experimental paradigm in some fields is huge: one only needs to browse openneuro datasets for a few minutes to find fMRI datasets where the events file just mention 'stimulus' and 'response' as only descriptors of the what happened during the experiment with no link to even a published article with a method section describing what the task was.
One can create automated pipelines to fully reproducibly preprocess and analyze such datasets, but without proper description of the experimental context like the one HED can provide, those beautiful activation blobs on brain images are almost completely uninterpretable.
Sure one may worry that HED has not been widely adopted, but this reflects to me more a lack of structural incentive rather a lack of effort and tools on the part of the HED team to make their work usable.
Referring to the 'open science' pyramid of cultural change (https://www.cos.io/blog/strategy-for-culture-change), HED sits in the "make it possible" / "make it easy" layers. So at least when the incentive changes, the tooling will be there.
BIDS maintainers (BIDS officially supports HED)
Thu, 10/10/2024 - 20:05
Thu, 10/10/2024 - 20:18
none
Sat, 10/12/2024 - 19:56
Fri, 10/25/2024 - 18:17
Expanding the tags to incorporate more behavioral concepts used in animal research is needed, and I think the HED framework will be able to support this expansion.
member of NWB TAB