Post on 22-Feb-2018
©Siemens Corporation, Corporate Research, Integrated Data Systems, 2009. All rights reserved.Page 1 © Siemens Corporation, Corporate Research, 2009. All rights reserved.
SciPort
Fabian Moerchen, Philipp Hussels, Claus NeubauerIntegrated Data Systems, Siemens Corporation, Corporate ResearchPrinceton, NJ
©Siemens Corporation, Corporate Research, Integrated Data Systems, 2009. All rights reserved.Page 2
SciPort Integrating, sharing, and searching data and meta-data
Problem
Solution
Scientific data is highly complex and diverseStructured data such as spreadsheets and SQL databasesSemi-structured data such as documents (Word, PDF, …)Images stored in PACS or file systems (DICOM, …)Multimedia files (videos, slides, …)Raw data files with scientific measurements
Scientific databases have different formats and access protocolsSQL, XMLWeb/Grid services
There is a need to integrate and share scientific data across multiple distributed sites and organizations and make it searchable through a unified interface.
SciPortTM – a web-based platform for scientific data management Integration of data from databases, file system, and gridsRapid data acquisition of meta-data through custom web formsUnified search interface for meta-data and contentLight weight data publishing data across distributed sitesFine grained access control to the database Personalization through tagging of database contents
©Siemens Corporation, Corporate Research, Integrated Data Systems, 2009. All rights reserved.Page 3
SciPort Modeling of domain model
Define domain model(s)Use SciPort web interface to define document templatesThe data fields support text, numbers, dates, tables, lists, files, etc.Each field can be constrained: required, min/max, drop down, etc.
Documents can be divided into
semantic sections
Data types include checkboxes, dates,
etc.
Specify values for drop down lists
Specify type and number of files that
can be attached
©Siemens Corporation, Corporate Research, Integrated Data Systems, 2009. All rights reserved.Page 4
SciPort Example: Annotation Document
Group
Category
File
Field
Field
Multiple instances of complex records
Tables
File attachments
©Siemens Corporation, Corporate Research, Integrated Data Systems, 2009. All rights reserved.Page 5
SciPort Modeling of permissions
Use SciPort web interface to define users and groups of users.A user has a specific role in each assigned group.A role define the permissions (read/write documents, add users, etc.)
Define users and permissions
Reader in the group MR
Administrator in the group SCR
©Siemens Corporation, Corporate Research, Integrated Data Systems, 2009. All rights reserved.Page 6
SciPort Modeling of folders
Define folders for documentsUse SciPort web interface to define folders and subfolders.Each folder is associated with document templates.Each folder can be associated with a group of users
Folder tree to sort documents as on a personal computer
This folder can store documents from three
specific templates
By creating a group of users and assigning them to this folder, data access can be
restricted to a folder branch
©Siemens Corporation, Corporate Research, Integrated Data Systems, 2009. All rights reserved.Page 7
SciPort Data acquisition
Authoring of documentsUse SciPort web interface to create new documents.Upload files with scientific data and associate it with meta-data.Constraints are checked to ensure high data quality.
Add rows to tables
Nested document structure
Attach DICOM images or other files
©Siemens Corporation, Corporate Research, Integrated Data Systems, 2009. All rights reserved.Page 8
SciPort Ontology support
Ontology supportText fields can be constrained to ontologies.RadLex – Lexicon for radiologyAuto-completion with relevant terms as the user types.
Auto-completionfrom controlled
ontology vocabulary
©Siemens Corporation, Corporate Research, Integrated Data Systems, 2009. All rights reserved.Page 9
SciPort Data integration
Import of legacy dataUse SciPort Batch Loader to populate database from folders with images and spreadsheets.Plugin of custom import modules to import data from XML or SQL databases.Unified interface to search the consolidated data through SciPort web client.
File systemimport SciPort XML
SciPortDatabase
Files
Domain model
Databaseimport SciPort XMLSQL
XMLtransformation SciPort XMLXML
SciPort Batch Loader
SciPortweb client
©Siemens Corporation, Corporate Research, Integrated Data Systems, 2009. All rights reserved.Page 10
SciPort Searching
Flexible search interface to document databaseUse SciPort web interface to search for documents combining three types of criteria:
Keyword search over all data fieldsMeta-data search (authors, creation date, etc.)Domain search: conditions on specific data fields of domain model.
Keyword searchMeta-data search
Domain searchRestrict to folder
Paging of results
©Siemens Corporation, Corporate Research, Integrated Data Systems, 2009. All rights reserved.Page 11
SciPort Sharing
Lightweight infrastructure for document sharingUsers can publish selected documents to a central read-only server.Users can browse the central server and receive large files from original server on demand.Shared documents are updated automatically.Templates can be shared including support for backward compatible changes.
Structured DocumentsStructured Documents
Images/FilesImages/Files
Local Server
Structured Documents
andTemplates
Central Server
Lightweight publishing of domain
data with links to attachments
Retrieve attached files of particular document
on demand
Structured DocumentsStructured Documents
Images/FilesImages/Files
Local Server
Structured DocumentsStructured Documents
Images/FilesImages/Files
Local Server
Structured DocumentsStructured Documents
Images/FilesImages/Files
Local Server
User
Browsing and searching published
documents
©Siemens Corporation, Corporate Research, Integrated Data Systems, 2009. All rights reserved.Page 12
SciPort Exporting
Export modules for documents and filesUse SciPort web interface to download documents and files as ZIP files.Export nested documents as spreadsheets with ‘flat’ structure for analysis with statistical tools or databases.Publish images and annotations to the National Biomedical Imaging Archive (NBIA) of caBIG.
SciPortDatabase
SciPortweb client
NIHClinical Trial Processor
Spreadsheets
ZIP files
SQL export
R/SAS®/Matlab®
caGRID
©Siemens Corporation, Corporate Research, Integrated Data Systems, 2009. All rights reserved.Page 13
SciPort Technology
Data storage
Medical imaging supports
Hybrid storage strategy based on XML, SQL and file system.XML enables highly customizable modeling of scientific data with arbitrarily nested structures.IBM DB2 relational database with pureXML technology for XML indexing enables storage of XML documents with efficiency, scalability, and security of a SQL databaseStoring attachments in the file system keeps the database small and enables transparent access and backup of file for administrative users on the server.
Highly interactive AJAX web based user interfaceSecure access with HTTPS and encrypted passwords.Use of proven scalable web technologies (Tomcat, JSP, JBoss, web services)
Web application
Indexing of headers and thumbnail generation for DICOM images.Integration and indexing of image annotations in AIM format.HIPAA: anonymization, encrypted passwords and protocols.caGRID: publish images and annotation to the NBIA (National Biomedical Imaging Archive)
©Siemens Corporation, Corporate Research, Integrated Data Systems, 2009. All rights reserved.Page 14
SciPort Use cases
Collaboration management
Research in emerging imaging modalities
Siemens H USA has many collaborations with hospitals.Siemens provides imaging equipment and obtains acquisitions, case reports, evaluation reports, publications.SciPort is used by several imaging divisions to manage this data centrally for research, marketing, and compliance.
SciPort can be used to rapidly perform data collection for (research) clinical trials.Setup a single server accessible to all clinical partners and design questionaires in SciPort.Automated upload of large amounts of imaging data using the SciPort batch loader.
Clinical trials data collection
Researchers in emerging imaging modalities need to compare results with conventional methods. Need to pool data from multiple research sites.SciPort can manage (anonymized) patient information, DICOM images and non-standard formats in one system.
©Siemens Corporation, Corporate Research, Integrated Data Systems, 2009. All rights reserved.Page 15
SciPort Road map summary
Hot new features on the road map
Data management
SciPort flash client with even more interaction and visualization, see next slides.Wolfram-alpha like search with statistical summaries of search results.Integration with SCR’s Extensible Imaging Platform (XIP), see next slides.Integration with SCR’s ProFound for data-driven comparative effectiveness researchIntegration with Algorithm Validation Tool (AVT), see next slides
Evaluate variability of image annotationsIntegration with data analysis tools
R, SAS, MatlabSCR’s iKDD™ (Interactive Knowledge Discovery and Data Mining)
Automatic synchronization of distributed servers for faster access of remote data.Alternative P2P architecture for searching and tagging.More interoperability with standards such as PACS, HL7, caGRID, etc.Single-sign on with LDAP etc.
Connecting, publishing and sharing
Support drafts and versioning of documents generating an audit trail for HIPAA compliance.Data quality reporting (number of missing values per template / folders)
©Siemens Corporation, Corporate Research, Integrated Data Systems, 2009. All rights reserved.Page 16
SciPort Road map: Flash client
WYSIWYG form design with drag & drop
Flash technology for web clients enables more interactivity, visualization and improved browser compatibility.
©Siemens Corporation, Corporate Research, Integrated Data Systems, 2009. All rights reserved.Page 17
SciPort Road map: XIP integration
XIP is an open source framework and platform for rapidly developing medical imaging applications that are "plug-and- play" across multiple computing environments. It offers an extensible set of modular objects for researchers to develop and evaluate software solutions for research and clinical problems. XIP also provides an integrated development environment (IDE) for imaging software development through the use of modular Visual Programming, without the need to write and compile any code.
SciPortDatabase
Results
Annotations
Images
SCR ImagingGianluca Paladini Gianluca.Paladini@siemens.com
SciPort can provide images to XIP and derived annotations and results.
©Siemens Corporation, Corporate Research, Integrated Data Systems, 2009. All rights reserved.Page 18
SciPort Road map: ProFound integration
Modular construction of disease screening workflowsUsers can select and connect disease, screening, and treatment modules or select pre-constructed workflows.
Rapid customization of model parametersUsers edit selected parameters, such as the sensitivity, specificity, and cost of a module.
1- and 2-D Sensitivity analysisEnables the user to explore cost- effectiveness over a range of parameter values
SCR IDSGayle Wittenberg Gayle.Wittenberg@siemens.com
ProFound is an IT framework for comparative effectiveness modeling, simulation, optimization and visualization of complex healthcare workflows. ProFound includes modules describing disease progression, screening, and treatment of patients and help to evaluate workflows regarding life-years saved, quality of life, and cost.
SciPortDatabase
Results
Models
EHR
SciPort can provide high quality electronic health records to ProFound to perform data-driven comparative effectiveness analysis and store ProFound workflow definitions, workflow parametrizations, trained workflow models, and effectiveness results back into SciPort.
©Siemens Corporation, Corporate Research, Integrated Data Systems, 2009. All rights reserved.Page 19
SciPort Road map: AVT integration
SCR SEScott Keneman Scott.Keneman@siemens.com
SciPortDatabase
Results
Models
Images
AVT is a project to develop caBIG community tools for analyzing the variabilility of annotations. The vision of AVT is to facilitate comparison of annotation methods from different research groups by implementing an efficient workflow and infrastructure for statistical analysis of annotations. The SciPort team is implementing the database components of AVT.
Annotations
SciPort can provide images and annotations from the electronic health records containing additional meta-data directly to AVT to utilize the statistical analysis and store results (not a current capability of AVT) for archiving and reporting.
©Siemens Corporation, Corporate Research, Integrated Data Systems, 2009. All rights reserved.Page 20
SciPort Road map: iKDD™ integration
Siemens CT SE 5Bernd Wachmann Bernd.Wachmann@siemens.com
iKDD™ is a platform for workflow based data mining and optimization. Users can compose complex workflows of data processing operators including preprocessing, feature selection, predictive modeling.
SciPortDatabase
Results
Models
EHR
SciPort can provide high quality integrated electronic health records including quantitative features derived from images with XIP or other tools to train predictive models. Results such as ROC curves can be stored back into SciPort.
©Siemens Corporation, Corporate Research, Integrated Data Systems, 2009. All rights reserved.Page 21
SciPort Contact
Fabian Moerchen, Ph.D.
Siemens CorporationCorporate ResearchIntegrated Data Systems755 College Road East08540 Princeton, NJ, USATel.: +1 (609) 734-3529Fax: +1 (609) 734-3346Mobile: +1 (609) 481-9712 fabian.moerchen@siemens.com