Providing the cardiovascular community with rich web interfaces

to integrate cardiovascular tools, data and analysis

Stephen J. Granite, MS, MBACVRG Program Manager

NHLBI R24-HL085343http://www.cvrgrid.org

Data Management Needs of the Cardiovascular (CV) Community

• CV-specific ontologies for describing data and data collection protocols– Must address unique CV data types

• Time-evolving electrophysiology (EP), electrocardiogram (ECG), and multi-modal image data

• Software tools for managing diverse types of CV data

– Must be well documented, easy to install/use, secure, and work out of the box

• Ability to query multiple data sources located and maintained at different sites– Project design often requires that specific types of data be collected and

managed at different sites (e.g., MR Reading Centers, ECG Reading Centers, etc)

The CVRG Project

Mission:Serve the needs of basic and clinical cardiovascular researchers by providing seamless, customized, secure access to study datasets and analysis tools

Team:Institute for Computational Medicine, Johns Hopkins University (Winslow, Miller)Image Lab, Wake Forest University (Carr, Ge)Center for Comprehensive Informatics, Emory University (Saltz, Post)Computation Institute, University of Chicago (Foster, Madduri)

The CVRG Approach to Technology Development

• Don’t re-invent the wheel– When possible, re-use or extend software components from National Center

for Biomedical Computing (NCBC) & other open-source projects (e.g., i2b2, Globus Online, etc.)

– Form collaborations to develop new software components

• 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

• Let the CV community drive technology development

NHLBI-FUNDED Studies

• Multi-Ethnic Study of Atherosclerosis• Coronary Artery Disease Risk in Young Adults• Minority Health Grid• Jackson Heart• Pediatric Heart Network

ECG Tool Initial Design/Workflow

Physionet Workflow

Reynolds Workflow

Data Storage Services

Norav ECG Data (RDT)

Physionet ECG Data(WFDB – header, data)

ECG Data Analytical Services

ECG Analysis Parameters

Analysis ResultsServices

RDT -> HL7aECG

WFDB ECG Data

Reynolds ECG Analysis Algorithms

Physionet ECG Analysis Algorithms

Results

ResultsMetadata XML

CVRG Web Interfaces: Initial Approach

• Alpha CVRG Portal leveraged caGrid Portal and Web Single Sign-On (WebSSO)– caGrid Portal built upon Liferay 4.1 open-source Portal Server– caGrid WebSSO used Java Architectures Special Interest Group

Central Authentication Service (JA-SIG CAS)

• Portlets implemented:– Beta Signup– ECG Storage/Analysis Workflow (end-to-end for both workflows)– Imaging Storage & Analysis– Heart shape and motion analysis– ProteinDB utilization

Beta Signup/ECG Workflow Demo

Initial Approach Lessons Learned

• Portal (e.g., Liferay 4.1)– All the work is done on the server– Changing the tabs requires a web browser refresh

• Resets the portlet displayed so data entry starts all over again

• New requirements appeared during development– Need to dynamically display and annotate an ECG– Dynamic refresh of portlets themselves difficult

• Ability to display an embedded scrolling graph next to impossible

– Call to tools outside of a portlet required extensions to Liferay itself

• Portal environment customization should be minimized

Investigation of Web 2.0

• Web 2.0 interfaces/tools– Work can be divided between the server & web browser– Components can be placed in web pages or executed standalone, making

changes dynamic

• Yahoo User Interface (YUI)– Open-source JavaScript library– Used by i2b2 in Web Client

• Google Web Toolkit & API Library (GWT)– Open-source Java project with plugin for Eclipse

• Allows Java developers to produce rich JavaScript tools without knowledge of syntax

– Used by LabKey & WebProtege

WFDB, HL7aECGRDT, GE MUSE

ECG Tool Revised Design/Workflow

CVRG Web Interfaces: Revised Approach

• Prototype ECG Gadget (renamed ECGrid Toolkit)– Built using GWT & Google Visualization API– Can exist in a portal, a web page, a dashboard, a flash drive

and/or a handheld device (e.g., iPhone)• Expanded avenues of tool access, “putting it in the users’ hands”• Functions on multiple platforms and multiple browsers

• Functionality implemented:– Integrated the ECG workflows into one unified workflow

• Accepts multiple ECG formats stores in Physionet’s WaveForm DataBase (WFDB) format

– Extended the capability of scrolling ECG visualization• Allows for annotation within the waveform, using National Center for

Biomedical Ontology (NCBO) REST Services real-time

Prototype ECGrid Toolkit Demo

Revised Approach Lessons Learned

• Web 2.0 applications are more dynamic– Flexibility in deployment require development of

authorization/authentication components as well

• Portal/Web 2.0 combination maximizes strengths/minimizes weaknesses– Portal serves as an interface access point

• Allows addition/removal of interfaces without a portal reboot

– Interfaces can be run on servers separate from the portal• Appear centralized when they actually are distributed

Current CVRG Web Interfaces

XNAT-CVI ECGrid Toolkit

Galaxy

Web Data Services (JHU Cloud)

MR/CT/USAnalyses

Web Analysis Services(Amazon EC2 & JHU Cloud)

ECGAnalyses

Ontology(RESTful Web Service)

BioPortalalCVRG Portal Web-BrowserSoftware as a Service

Live Demonstration

• CVRG Demonstration Portal– Galaxy– eXtensible NeuroImaging Toolkit – CV

Imaging (XNAT-CVI)– ECGrid Toolkit

• CVRG website -> Portal– http://www.cvrgrid.org

Future Development

• In Process– Integration of NSF Cybersecurity Infrastructure Logon (CILogon) into

CVRG Portal• Alternative Shibboleth method developed using Johns Hopkins Enterprise

Authentication (for testing purposes)

– Deployment of Survey portlet for user feedback on CVRG tools

• Planned– Healthy Heart Imaging cohort to be made available to the community

for clinical trial controls via XNAT-CVI instance at Wake Forest• ECGs corresponding to Image datasets to be made available via ECGrid Toolkit

at Johns Hopkins

– ECG analysis services to be made available via Galaxy, to demonstrate workflow sharing and re-use

Web References

• CVRG - http://www.cvrgrid.org/• JHU Institute for Computational Medicine – http://www.icm.jhu.edu/• Wake Forest University Image Lab – http://www.wakehealth.edu/imagelab/• Emory University Center for Comprehensive Informatics –

http://cci.emory.edu/cms/index.html• University of Chicago Computation Institute - http://www.ci.anl.gov/• Liferay – http://www.liferay.com/community/welcome/dashboard• JA-SIG CAS - http://www.jasig.org/cas/• YUI - http://developer.yahoo.com/yui/• i2b2 Web Client - https://www.i2b2.org/• GWT - http://code.google.com/webtoolkit/• LabKey - https://www.labkey.org/• WebProtege - http://protegewiki.stanford.edu/index.php/WebProtege• NCBO Bioportal – http://bioportal.bioontology.org/• Physionet – http://www.physionet.org/• CILogon - http://www.cilogon.org/

AcknowledgementsJohns Hopkins University

Michael MillerRaimond WinslowLaurent Younes

Siamak ArdekaniStephen GraniteDavid HopkinsKyle ReynoldsMichael ShipwayChristian Jurado

Emory UniversityJoel SaltzAndrew PostTahsin KurcRichard WillardHimanshu Rathod

Wake Forest UniversityJ. Jeffrey CarrYaorong GeRicardo Peral

NHLBI R24 HL085343

NHLBIJennifer Larkin University of Chicago

Ian FosterRavi MadduriLiu Bo