Current Research in the UW Structural Informa²cs Group James F. Brinkley
Total Page:16
File Type:pdf, Size:1020Kb
Current Research in the UW Structural Informacs Group January 2011 James F. Brinkley Shared Vision • Worldwide network of integrated data and knowledge • Brings relevant informaon to point of need • Fosters translaon Long‐term Goals • Contribute to shared vision • A structural informaon framework for biomedicine • Use as a basis for organizing biomedical informaon Raonale • Structure of the body is the most useful means for organizing biomedical informaon • Most manifestaons of health and disease are properes of anatomical enes Primary Informacs Research Quesons • How to create a structural informaon framework • How to use this framework as a basis for organizing informaon Approach • Classify types of informaon • Represent and implement as different modules • Integrate within a distributed framework • Develop applicaons opportuniscally • Applicaons drive methods Current Driving Biological Applicaons • Anatomy Educaon • Brain mapping • Proteomics • Clinical trials • Craniofacial Malformaons Research Themes • Structural framework • Data management • Data integraon • Data visualizaon Projects • Foundaonal Model of Anatomy • Ontology Views • Data management • Data integraon • Visualizaon Foundaonal Model of Anatomy Onard Mejino, Nolan Nichols, Todd Detwiler, Cornelius Rosse Dimensional Anatomy Taxonomy Ontology Anatomical Structure Line (1-D) Volume (3-D) Organ Part Surface (2-D) Anterior Interventricular Organ Polyhedron -is a- Sulcus -is a- region Right Cardiac Sternocostal bounded by Coronary has Chamber Surface -is a- Sulcus super- object bounded by Infundibulum Inferior margin Right of heart has boundary of subobject Ventricle Diaphragmatic Coronary Surface bounded by Sulcus Anatomical Posterior Landmark IV Sulcus -is a- Apex Crux of heart Point (1-D) Influence of the FMA Current work on the FMA • Reorganizing the neuroantomy axis of RadLex • Incorporang Lymphacs • Represenng neural connecvity Re‐organizing the neuroanatomy axis of RadLex Enabling RadLex with the Foundational Model of Anatomy Ontology to Organize and Integrate Neuro-imaging Data Jose Leonardo V. Mejino Jr. MD1, Landon T. Detwiler MS1, Jessica A. Turner PhD3, Maryann E. Martone PhD4, Daniel L. Rubin MD, MS5 and James F. Brinkley MD, PhD1,2. 1Structural Informatics Group, 2Biomedical and Health Informatics, University of Washington, 3MIND Research Network, 4Department of Neurosciences, University of California San Diego, 5Stanford Center for Bioinformatics Research, Stanford University Abstract A. FMA B. RadLex Conclusion In this study we empowered RadLex with a We have shown how the ontological robust ontological framework and framework of the FMA explicitly defined additional neuroanatomical content the entities represented by the different derived from a reference ontology, the parcellation and naming schemes and by Foundational Model of Anatomy Ontology doing so it becomes possible to ascertain (FMA)1, with the intent of providing the relationships which correlate these RadLex the facility to correlate the different terms, a prerequisite step for sharing and standards used in annotating neuro- harmonizing data. We have started using radiological image data. It is the objective the ontology to annotate fMRI datasets of this work to promote data sharing, data Figure 2. Class taxonomy of FMA applied to RadLex with strict adherence to IS_A and derive inferences about relationships harmonization and interoperability only relationships between anatomical entities. between the datasets8. between disparate neuro-radiological Materials and Methods labeling systems. Results Enhancing the neuroanatomy ontology of RadLex Enhancement of Anatomy Taxonomy of RadLex. Acknowledgment Supported by NHLBI grant HL08770 and NIBIB involved five major steps (shown in Figure 1): Adoption of the ontological framework of the Introduction contract # HHSN268200800020C. 1.! selecting a reference ontology, the FMA FMA assures a consistent Aristotelian-type Huge amounts of neuro-imaging data are being ontology, a target application ontology, inheritance taxonomy for RadLex (Figure 2). The References produced by different groups and they are RadLex, three neuro-imaging annotation derived ontology provides explicit semantics for 1.!Rosse C, Mejino JLV. 2007. The Foundational recorded based on different, often study terminologies, Talairach Daemon Atlas3, RadLex terms . Model of Anatomy Ontology, In: Burger A, specific, brain parcellation and naming 4 Davidson D, Baldock R (eds). Anatomy FreeSurfer atlas , Anatomical Automatic Enhancement of Neuroanatomy content of FMA. schemes. Using disparate naming conventions 5 Ontologies for Bioinformatics: Principles and Labeling atlas (AAL) and a neuroanatomical Classes and spatio-structural relations were added produces incompatible terms thereby making 6 Practice, London: Springer, pp 59-117. application ontology, NeuroLex , as inputs to in the FMA to accommodate and represent the the correlation of data difficult to achieve. 2.!Langlotz CP. RadLex 2006. a new method for the system; entities referenced by the different annotation Current terminologies for neuro-imaging lack indexing online educational materials. 2.! application of the high level class taxonomy of terminologies (Figures 3 and 4). Explicit Radiographics 26:1595–1597. the semantic framework to explicitly declare the the FMA to re-organize the anatomy hierarchy ontological representation therefore allowed for 3.!Talairach, J., Tournoux, P., 1988. Co-planar precise meanings of the terms and therefore 7 of RadLex ; the correlation of the different terms by using stereotaxic atlas of the human brain. Thieme neuro-imaging data and information represented '''''!"#$%&'BC'D8%%&:.4"8='87'=&$%8.=.489"A.:'&=4"4"&>'%&7&%&=A&/';<'/"77&%&=4'4&%9"=8:8#"&>C 3.! enhancement of the neuroanatomy content of FMA properties such as IS_A and PART_OF Medical Publishers, New York. by the terms cannot be readily associated and the FMA to capture the representations that (Figure 5). 4.!Lancaster JL, Woldorff MG, Parsons LM, Liotti applied across different studies. RadLex2 different terminologies intend to use for M, Freitas CS, Rainey L, Kochunov PV, (Radiology Lexicon from RSNA) is a annotating neuroimaging data as well as other Extraction of neuroanatomical “view”, Nickerson D, Mikiten SA, Fox PT. Automated controlled terminology for radiology and seeks forms of neuroscientific data; NeuroFMA, for incoporation into RadLex. Talairach Atlas Labels For Functional Brain to provide the needed semantics for correlating 4.! extraction of the enhanced neuroanatomy View extraction is performed via a procedural Figure 3 Mapping. Human Brain Mapping 10:120-131(2000). the diverse terminologies used for annotating component of the FMA, the NeuroFMA, as an program that is written in JAVA, utilizing the neuro-imaging data. In this work we leveraged Protégé ontology API. Rather than creating a view 5.!N. Tzourio-Mazoyer, B. Landeau, D. ontology “view” for incorporation into Papathanassiou, F. Crivello, O. Etard, N. Delcroix, the Foundational Model of Anatomy Ontology by starting from an empty ontology and then RadLex; Bernard Mazoyer and M. Joliot (January 2002). adding classes, the process starts with a complete (FMA) to re-structure and reinforce the 5.! merging of the extracted NeuroFMA with the "Automated Anatomical Labeling of activations in anatomical domain of RadLex so it can ontologically re-organized RadLex . copy of the FMA and then eliminates everything SPM using a Macroscopic Anatomical incorporate, accommodate and correlate the not required in the NeuroFMA (Figure 6). Parcellation of the MNI MRI single-subject different annotation terminologies. Merging of NeuroFMA into RadLex. brain". NeuroImage 15 (1): 273–289. Technical details for this step are beyond the scope 6.!http://neurolex.org/wiki/Main_Page The approach we describe here serves two 7.!Mejino JLV, Rubin DL, Brinkley JF. FMA- of this presentation. However we found that we practical purposes: RadLex: An application Ontology of Radiological 1)! reference ontologies provide a principled could coalesce classes from the two ontologies in Anatomy derived from the Foundational Model of and robust framework on which to build RadLex. The merging produced “FMA-like” Anatomy Reference Ontology. Proc. AMIA Symp applications specific to a particular field structure to RadLex. A total of 12, 579 classes and 2008:465-469. 33,361 property values were imported into !"#$%&'()'!*+,-./0&1'2%"#345'/&%"6&/'7%89'43&'!*+'2:&745';<'/&:&4"8='2>4%"?&8$4>5'.=/'.//"4"8='87' 8.!Turner JA, Mejino JLV, Brinkley JF, Detwiler 2)! underlying ontological framework :"=?>'.>'>38@=';<'43&'!"#$':"=?'87'Anatomical structure,'@3"A3'@.>'/&:&4&/'7%89'Material facilitates and promotes integration, RadLex from the NeuroFMA. It would have been anatomical entity';$4'.//&/'48'Anatomical entity. LT, Lee HJ, Martone ME and Rubin DL (2010) interoperability and reuse of knowledge very difficult and time-consuming to implement Application of neuroanatomical ontologies for among application ontologies. these changes manually. neuroimaging data annotation. Frontiers in Neuroinformatics 4(10):pp. 1-12. Incorporang Lymphacs FMA class taxonomy Nakagawa et. al. CANCER 1994, Vol. 73, No. 4, 1155‐1162. Represenng Neural Connecvity Representing Neural Connectivity in the Foundational Model of Anatomy Ontology BIC Nolan Nichols1, Aaron Perlmutter1, Jose L. V. Mejino Jr. 2, Daniel L. Rubin3, and James F. Brinkley1,2,4 Departments of Medical Education and Biomedical Informatics1, Biological Structure2, Computer