BRAINformat: A Data Standardization Framework for Neuroscience Data

Oliver Rübel, Prabhat, P. Denes, David S. Conant, Edward F. Chang, Kristofer E. BouchardbioRxiv 2015 · DOI

TL;DR

The BRAINformat is a new data standardization framework for neuroscience, built on HDF5, that addresses the "extreme data" challenge by providing a modular, extensible, and self-describing way to manage complex, multi-modal neuroscience data. It introduces novel concepts like "Managed Objects" and "Relationship Attributes" to enable robust data organization, sharing, and analysis.

What Problem Does This Paper Address?

Neuroscience is drowning in data – the "extreme data" era. This deluge of information, characterized by high volume, velocity, variety, and veracity, poses a significant hurdle for data sharing between labs and for leveraging powerful computational resources for discovery. Existing data formats are often domain-specific, proprietary, or lack clear specifications, leading to errors, misinterpretations, and hindering collaboration. This makes it difficult to integrate data from diverse sources (like neural recordings, video, eye-tracking) and to effectively annotate and analyze it based on its semantic meaning.

Methods

BRAINformat tackles this by building upon the robust HDF5 format, which provides a solid foundation for storing large-scale scientific data. The core innovations lie in:

  • Managed Objects: These are formal, self-contained specifications for data objects (files, groups, datasets) that include a human-readable description, a unique identifier, and a formal specification. This modular approach allows for the independent specification and reuse of data format components, simplifying the creation of complex formats.
  • Format Specification Language and API: BRAINformat provides a standardized way to define data formats using Python dictionaries, which can be serialized to JSON. This ensures that specifications are computer-interpretable and human-readable, facilitating both programmatic access and understanding.
  • Relationship Attributes: This novel concept allows for the explicit modeling of semantic and structural relationships between data objects. This goes beyond simple HDF5 links by defining types of relationships (e.g., "order," "indexes," "equivalent") and enabling complex queries and data mapping.
  • Index Map Relationships: A specialized form of relationship attributes that allows for the explicit description of complex mappings between data objects, such as those arising from image registration or downsampling. This simplifies tasks like finding corresponding data points across different modalities.
  • Annotation Modules: Dedicated modules for describing data selections and managing collections of annotations. This is crucial for associating semantic metadata with specific subsets of data, enabling advanced analyses.

Datasets

The paper demonstrates BRAINformat's application to electrocorticography (ECoG) data collected from neurosurgical patients during speech production. This specific use case highlights the framework's ability to handle multi-modal data streams, complex task descriptions, and the need for rich metadata. No new benchmarks were introduced; rather, existing ECoG data was used to showcase the framework's capabilities.

Key Insights and Ideas

  • Modularity is Key: By breaking down data formats into "Managed Objects," BRAINformat allows for a more manageable and extensible approach to standardization, avoiding the need to reinvent the wheel for every new data type.
  • Semantics Matter: Simply storing data isn't enough; understanding the relationships between data objects is critical for advanced neuroscience analysis. Relationship Attributes provide a formal way to capture this semantic richness.
  • Bridging the Gap: BRAINformat aims to bridge the gap between general-purpose storage formats (like HDF5) and the specific, often complex, needs of neuroscience data analysis, particularly in areas like data annotation and inter-dataset relationships.
  • Open Source and Accessible: The availability of the BRAINformat library with detailed documentation and tutorials is crucial for its adoption and for fostering collaboration within the neuroscience community.

Key Results

The paper demonstrates the implementation of BRAINformat by designing a data standard for ECoG data. This standard organizes raw and processed ECoG recordings, along with associated metadata, in a structured and extensible manner. The framework's ability to handle complex data selections and annotations is highlighted, enabling efficient querying and analysis of events within the ECoG data. The paper also showcases the practical application of index map relationships for tasks like mapping between processed and original image data.

Limitations and Future Work

The authors acknowledge that neuroscience data presents an "incredible big data challenge" and that while BRAINformat provides a strong foundation, continuous development is needed. Future work includes:

  • Advanced Metadata Ontology and Data Type Specification: Enhancing the framework's ability to define and manage complex metadata ontologies and data types.
  • Efficient Metadata Search: Developing more sophisticated search capabilities for metadata.
  • Additional Data Selection Schema and Representations: Expanding the support for various data selection methods.
  • Linking and Interaction with External Data: Enabling seamless integration with data stored in third-party formats (e.g., movies, images).
  • Broader Use Case Coverage: Developing additional data modules to support a wider range of neuroscience research applications beyond ECoG, including model species data.

Summary updated

Abstract

ABSTRACT Neuroscience is entering the era of ‘extreme data’ with little experience and few plans for the associated volume, velocity, variety, and veracity challenges. This is a serious impediment for both the sharing of data across labs, as well as the utilization of modern and high-performance computing capabilities to enable data driven discovery. Here, we introduce BRAINformat, a novel file format and model for management and storage of neuroscience data. The BRAINformat library defines application-independent design concepts and modules that together create a general framework for standardization of scientific data. We describe the formal specification of scientific data standards, which facilitates sharing and verification of data and formats. We introduce the concept of Managed Objects , enabling semantic components of data formats to be specified as self-contained units, supporting modular and reusable design of data format components and file storage. The BRAINformat is built off of HDF5, enabling portable, scalable, and self-describing data storage. We introduce the novel concept of Relationship Attributes for modeling and use of semantic relationships between data objects, and discuss the annotation of data using dedicated data annotation modules provided by the BRAINformat library. Based on these concepts we implement dedicated, application-oriented modules and design a data standard for neuroscience data. The BRAINformat software library is open source, easy-to-use, and provides detailed user and developer documentation and is freely available at: https://bitbucket.org/oruebel/brainformat .
BibTeX
@misc{rubel2015brainformat,
  author = {Oliver Rübel and Prabhat and P. Denes and David S. Conant and Edward F. Chang and Kristofer E. Bouchard},
  title = {{BRAINformat: A Data Standardization Framework for Neuroscience Data}},
  year = {2015},
  doi = {10.1101/024521},
  howpublished = {bioRxiv},
}
Figure