Integration of Semantically Annotated Neuroimaging Data via Chained Queries

We present a method for integrating semantically annotated neuroimaging data via chains of Internet accessible queries using the Foundational Model of Anatomy ontology as an example. Funding: NIH (5T15LM007442,NLM), (HL08770, NHLBI), (HHSN268200800020C, NIBIB), (1U24RR021992, NCRR/FBIRN); DOE (DE-FG02-99ER62764, MCIC) 

 

Introduction 

Human neuroimaging datasets are increasingly publicly available on the Internet, but researchers have a limited number of tools to facilitate the integration of this data. We developed a web-based Query Manager that allows users to create, save, edit, and execute queries over URL accessible XML resources. Saved queries can be executed programmatically using a RESTful API that is both URL accessible and generates XML as output. This allows a saved query to be used as input to another query, thereby permitting queries to be chained together. In previous work we used web-based queries to semantically annotate functional MRI (fMRI) datasets using neuroanatomical ontologies and "intelligently" queried annotated data by inferring relationships about granularity and connectivity using the Foundational Model of Anatomy (FMA)¹. Our current work extends this approach to integrate fMRI and VBM datasets, where each modality’s results are recorded in XML and labeled using the Talairach and AAL brain atlas parcellation schemes, respectively. Our approach generates a semantically integrated dataset within a single ontological framework that reconciles parcellation schemes across both datasets based on part relationships. 

 

In Figure 1 saved queries are in green, data sources are in orange, results are in yellow, and query engines are in blue. Query 1 (Q₁) is a SPARQL query that traverses the constitutional_part and regional_part relations of the FMA (S₂) to infer a mapping (R₁) between AAL and Talairach annotations (S₁), which represent brain regions at different levels of granularity. Query 2 (Q₂) is an XQuery that calls Q₁ and then, using the annotation mapping (R₁) and annotated dataset as its data input (S₃), generates a semantically integrated dataset, where Talairach labeled fMRI results are grouped as parts of larger AAL labeled brain regions for VBM results.

 

Conclusion 

Query chaining is an effective approach to semantically annotate, integrate, and query neuroimaging data using knowledge from ontologies, one application of which is organizing multi-modal brain imaging data into a structural framework that can facilitate data mining.

 

References 

1. Turner JA, Mejino JLV, Brinkley JF, Detwiler LT, Lee HJ, Martone ME and Rubin DL (2010) Application of neuroanatomical ontologies for neuroimaging data annotation. Front. Neuroinform. 4:10.