Neuroimaging techniques such as MRI, fMRI and PET provide a wealth of quantitative information related to anatomy, function, development, and disease etiology of the nervous system. Experimental studies utilizing neuroimaging techniques have generated and reported growing amount of research results in the literature. The availability of these results allows neuroscience researchers to perform comparative and integrative studies to discover new knowledge and generate new hypotheses. However, because that much of these results are published in the full text body of the articles as tables and figures, in addition to the textural format, they are not readily available for systematic retrieval and access via most online databases of scientific literature, such as PubMed, that cover only the abstracts and citations. To address this problem, we have developed the Internet Brain Volume Database (IBVD), a web-based searchable database of neuroanatomic volumetric observations manually curated from the journal articles. However, it is becoming apparent that our current labor-intensive manual approach is inadequate to keep pace with the growing number of publications. We propose to automate the development of IBVD with information extraction (IE) techniques. The primary long-term goal of this project is to develop a software system called NeuroIE that automatically search, extract, integrate and share research results from the published neuroimaging articles.
The specific aims of the proposal are: (1) Identifying and develop volumetric neuroimaging domain lexicon and ontology. Ontological structures specific to the volumetric neuroimaging domain will provide both guidance and constraints for analyzing tables, document structure and English sentences, as well as the intermediate framework for representing the extracted knowledge. The ontology will also link to the volumetric neuroimaging lexicon that provides necessary vocabulary for language processing. (2) Developing NeuroIE's core information extraction component. NeuroIE will be able to extract the following neuroimaging study results from full text of published journal articles: demographic information of subjects, volumetric measurements, disease diagnosis, and the neuroimaging methods used in the study. (3) Integrating NeuroIE with IBVD. By focusing on the neuroanatomic volumetric observations, the NeuroIE project will develop algorithms and software for mining the neuroimaging studies from the published literature, resulting in automated process of acquiring and entering data for neuroimaging databases, in support of advancements in understanding the underlying neuroscience in health and disease.
This project proposes to develop software for mining neuroimaging studies from the literature, resulting in automated curation of the Internet Brain Volume Database, in support of advancements in understanding the underlying neuroscience in health and disease.
|Asman, Andrew J; Smith, Seth A; Reich, Daniel S et al. (2013) Robust GM/WM segmentation of the spinal cord with iterative non-local statistical fusion. Med Image Comput Comput Assist Interv 16:759-67|