NeuroBridge: a prototype platform for discovery of the long-tail neuroimaging data

Area of Science:

• Neuroscience
• Data Science
• Biomedical Informatics

Background:

• Open science initiatives facilitate data sharing, but finding specific, underutilized datasets remains challenging.
• The long tail of science contains valuable data that is often difficult to access.
• Discovering relevant neuroimaging data for conditions like schizophrenia and addiction requires efficient search tools.

Purpose of the Study:

• To demonstrate the NeuroBridge prototype for searching full-text scientific literature.
• To identify neuroimaging data relevant to schizophrenia and addiction studies.
• To improve the accessibility of underutilized scientific data.

Main Methods:

• Developed an extensible ontology for study metadata (subject population, imaging techniques, clinical data).
• Utilized a natural-language document processor with deep-learning models for article representation.
• Integrated ontology-driven similarity search across PubMed Central and NeuroQuery.

Main Results:

• The NeuroBridge prototype successfully searched a corpus of 356 papers (2018-2021) on schizophrenia and addiction neuroimaging.
• A web-based portal with a Query Builder allows users to construct ontology-based searches.
• Returned articles provide links to abstracts, full-text papers, and listed clinical assessments.

Conclusions:

• NeuroBridge combines ontology-based search and text mining to identify relevant papers for research questions.
• The prototype represents a first step in identifying potential neuroimaging data within full-text publications.
• Future work includes expanding the corpus, enhancing the ontology, and integrating with data repositories like XNAT.