220 likes | 244 Views
The CardioVascular Research Grid: A Resource Supporting National Collaborations in Heart Research NHLBI R24 HL085343 Raimond L. Winslow The Johns Hopkins University Institute for Computational Medicine. Needs of the CV Community. Need to describe both CV data and data collection methods
E N D
The CardioVascular Research Grid:A Resource Supporting National Collaborations in Heart ResearchNHLBI R24 HL085343Raimond L. WinslowThe Johns Hopkins UniversityInstitute for Computational Medicine
Needs of the CV Community • Need to describe both CV data and data collection methods • CV-specific ontologies • Semantic description facilitates re-use of data • Need standards-based representation and management of multi-scale CV data • There are unique CV data types for which no standards exist • Time-evolving EP, ECG, and multi-modal image data • Need to query across multiple types of primary and derived data located and maintained at different sites • data federation, federated querying
The CardioVascular Research Grid (CVRG) Project is Addressing These Needs • The CVRG is a community resource for the distributed management, semantic description, federation, and analysis of bothprimary and derived CV data • CVRG is • developing new CV-specific data management resources • delivering CV data analysis tools and customized workflows • creating easy to use, customized web interfaces for accessing our resources • providing these resources in a secure manner
The CVRG Approach to Technology Development • Don’t re-invent the wheel • When possible, re-use or extend software components from other BioGrid projects (e.g., caBIG and BIRN) • caBIG = cancer Biomedical Informatics Grid • BIRN = Biomedical Informatics Research Network • Address the special needs of the CV community • Time-evolving imagery • ECG and EP data • CV-specific vocabularies • Integration of multi-scale data sets • Specific data analysis workflows • Assure that the CV community drives technology development
CVRG Projects Project 1: The Cardiovascular Research Grid Core Infrastructure (PI: Joel Saltz, Emory) Project 2: Electrophysiological Data Management and Dissemination (PI: Raimond Winslow, JHU) Project 3: Mathematical Characterization of Cardiac Ventricular Anatomic Shape and Motion (PI: Laurent Younes, JHU) Project 4: Grid-Tools for Cardiac Computational Anatomy (PI: Michael Miller, JHU) Project 5: Statistical Learning With Multi-Scale Cardiovascular Data (PI: Donald Geman, JHU) Project 6: Resource Management (PI: Raimond Winslow, JHU)
The CVRG Development Team Johns Hopkins University Resource PI: Raimond Winslow (rwinslow@jhu.edu) Project PI’s and Investigators: Donald Geman (Machine Learning) Michael Miller (Cardiac CA) Daniel Naiman (Machine Learning) Tilak Ratnanather (Cardiac CA) Raimond Winslow (ECG & EP Services) Laurent Younes (Cardiac CA) Project Manager: Stephen Granite (sgranite@jhu.edu) Developers: Timothy Brown (Cardiac CA Portal) Bill Gerten (ECG Portal) David Hopkins (Website & Portal) Anthony Kolasny (Cardiac CA Portal) Luan Le (Imaging Portal) Kyle Reynolds (Systems Support) Michael Shipway (Machine Learning Portal) Mathew Toerper (Imaging) Emory University Project PI’s and Investigators: Joel Saltz (caGrid) Tahsin Kurc (caGrid) Developers: Mark Grand (Lead Developer) Michel Monsour (Study data) Tony Pan (Imaging) Ashish Sharma (Imaging) caGrid Knowledge Center Justin Permar (caGrid) Steve Langella (Security) UCSD (BIRN; Yrs 1-2) Mark Ellisman Jeff Grethe Ramil Manansala
What Data Services Does CVRG Provide? PACS (DCM4CHE), DICOM Data Service Proteomics Data Service ECG Data Service Virtual PACS CVRG Portal Portlet Access Federated Querying mRNA Expression Data Service Study Data Service (OpenClinica) Developed by CVRG Developed by NCI-caBIG Other Open Source SNP/GWAS Data Service Other CVRG Nodes
Grid-enabling Existing Systems Federated Queries CVRG Node 3 (JHU) CVRG Node 2 (Emory) CVRG Node 1 (UCSD) CVRG Portal Portlet Access Federated Querying ECG Data Service Data Service (“Lighthouse”) caGrid Introduce Virtual PACS Abstract Model (Terminology, Semantics, Objects,…) NCI EVS, NCBO BioPortal PACS 1 PACS 2 Managed Locally Real Database (Oracle, DB2,…)
What Analysis Services Does CVRG Provide? Heart Shape & Motion Analyses (LDDMM, PTA) CVRG Portal Workflow Access Machine Learning; Rweb, geWorkbench ECG Analysis Services Other CVRG Nodes
ECG Data Management and Analysis:An Example of Novel Tool Development • Problem • ECGs are the most common type of biomedical time-series data • Typically, only data derived from ECG analysis is stored • Primary data is not always saved • When saved, often difficult to collect, may be in image or paper format • Data is difficult for clinical researchers to analyze • Challenges: • There is no comprehensive data model or database implementation for managing digital ECG data • There is no ECG ontology available for use in annotation • Few web-based ECG analysis tools • Solution: • Develop these resources
ECG Analysis Workflows Reynolds Workflow ECG Analysis Parameters Norav ECG Data (RDT) Reynolds ECG Analysis Algorithms Results RDT -> HL7aECG ECG Data Data Storage Services Analytical Services Analysis Results Services Physionet Workflow WFDB ECG Data Physionet ECG Data (WFDB – header, data) Physionet ECG Analysis Algorithms Results Metadata XML
CV DICOM Image Data Management:An Example of Re-Use & Extension of caBIG Tools • Problem: • Imaging (MR, CT, echo, and ultrasound) is becoming increasingly important in the assessment of heart disease. • Data often collected at multiple sites, must be shared • Challenges: • Images are usually stored as files (not searchable) • Local image data are sometimes managed using a PACs • It is difficult to manage local PACS client connections to remote PACS • Investigators want full control of data access • Solution • Use/extend existing image management tools developed in caBIG
CV DICOM Image Data Management:Re-use of caBIG & Open Source Tools LDDMM/PTA Analysis Workflow DICOM Image Mgmt Workflow 3DSlicer (Visualization) PACS (DCM4CHE) Seg3D (Segmentation) XNAT (Analysis Data) Virtual PACS Anonymization Indexing Upload Access Privileges Format Conversion PACS Client (KPACS) LDDMM/ PTA Analysis
Shape and Motion Analysis of Heart Imagery:Re-Use of Data Analysis Tools Developed in theBIRN Brain Morphometry Testbed Beg et al (2004). Computational cardiac anatomy using MRI. Magn. Reson. Med. 52(5): 1167
Statistical Inference on Disease-Induced Heart Shape Changes New Patient Heart H1 Global Average Heart Hav T(H1) Hav (preserves dot products and norms) Class 1 = ischemic cardiomyopathy Class 2 = non-ischemic cardiomyopathy Is the transformation T( ) from class 1? If yes, then ischemic cardiomyopathy Else non-ischemic cardiomyopathy
Statistical Inference on Heart Shape (cont.) septum Ant. Inferior Statistically reliable (p=.008) shape difference in ICM vs NICM Supports classification on novel test data Ardekani et al (2009). Computational method for identifying and quantifying shape features of human left ventricular remodeling. Ann Biomed Eng. 2009; in press. P value
Statistical Inference on Heart Deformation:Novel Algorithm Development Examine deformation from ED to ES Question – Are their deformation differences that discriminate ICM from NICM? ES (NICM) ES (ICM) Parallel Transport Algorithm PTA) Younes et al (2008). Transport of Relational Structures in Groups of Diffeomorphisms. J. Math. Imaging Vis. 32:41-56 Statistically Significant Difference in Wall Thickening From ED to ES PTA Preserves Strain Tensor at Each Image Voxel
Ontologies: • Emerging Strategy • When ontologies/data services are available from caBIG, we will use them. These ontologies & services are also available to BIRN • Develop new ontologies that are of special importance in CV research • Cardiovascular Electrophysiology Ontology • Version 1 complete • ECG Ontology • Version 1 nearing completion • Cardiovascular Imaging Ontologies • Beginning stages, with the National Alliance for Medical Image Computing (NAMIC) • Deposit new CV ontologies at the National Center for Biomedical Ontologies (NCBO), and also work to harmonize terms with the NCI Thesaurus/Meta-Thesaurus • These ontologies will then be available for use by BIRN and caBIG
Collaboration With the CVRG Project • CVRG Collaboration Proposals (PAR-07-426, as amended in NOT-08-103) • Development of a Hypertrophic Cardiomyopathy Consortium (Abraham) • Federation and Analysis of MESA Genetic and Imaging Data Using the CVRG (Lima) • Data Ontologies for Integrating Cardiovascular Epidemiological Studies (Rice) • Final LOI August 21 2009, final proposal submission September 21 2009 • One BRP collaboration proposal submitted • CardioVascular Imaging Informatics (CVRG & NAMIC Collaboration, PI = Kikinis) • NCBO Driving Biological Project • Ontology-Based Annotation of Biomedical Time-Series Data (Winslow) • Physionet (analytic services) • Site beta testers (request account on CVRG portal) • Contact us regarding new DBPs (rwinslow@jhu.edu) • Working Groups, Conference on CardioVascular Informatics
CVRG Web Resources • CVRG Website – Information about the CVRG • http://www.cvrgrid.org/ • CVRG Portal – Web environment for accessing all tools • Link on the CVRG Website • CVRG Wiki – Technical information on all services • Link on the CVRG Website • Software dependencies and revisions • Build information • Installation instructions