11
A Grid Environment for Medical A Grid Environment for Medical Imaging Imaging LRMN LRMN Sorina POP, Tristan GLATARD

A Grid Environment for Medical Imaging A Grid Environment for Medical Imaging LRMN Sorina POP, Tristan GLATARD

Embed Size (px)

Citation preview

A Grid Environment for Medical ImagingA Grid Environment for Medical Imaging

LRMNLRMN

Sorina POP, Tristan GLATARD

2/10

LRMNLRMNSorina POP, Tristan GLATARDSorina POP, Tristan GLATARDCIRA, 2 juin 2009CIRA, 2 juin 2009

SummarySummary

Medical Imaging

Grid environment for medical imaging

Perspectives

Acknowledgments

3/10

LRMNLRMNSorina POP, Tristan GLATARDSorina POP, Tristan GLATARDCIRA, 2 juin 2009CIRA, 2 juin 2009

Medical ImagingMedical Imaging

• CAVIARMyocardium segmentation

Motion estimation

• ThIS/GATEMonte-Carlo simulation

Cancer treatment planning

• SIMRISimulation of MRI

Parellel MPI jobs

ThIS Simulation

Simri

Caviar Caviar

4/10

LRMNLRMNSorina POP, Tristan GLATARDSorina POP, Tristan GLATARDCIRA, 2 juin 2009CIRA, 2 juin 2009

Grid environment for medical imagingGrid environment for medical imaging

• Reliable and responsive grid execution Successful execution of one application on a grid nodeDistant grid node environment and application customization

• Application parallelization MPI (Message Passing Interface) -> SimriSplit the simulation into independent jobs (Monte Carlo) -> Gate

• Execution framework: advanced tools forJob submission, monitoring and retrievalIntegration into service platforms

• High level interface

5/10

LRMNLRMNSorina POP, Tristan GLATARDSorina POP, Tristan GLATARDCIRA, 2 juin 2009CIRA, 2 juin 2009

Reliable and responsive grid executionReliable and responsive grid execution

• DIANERegisters/removes agentsSchedules tasks on agentsStdout/err transfers

• EGEE – gLite Schedules agents on grid sitesTransfer input/result files to/from worker nodes

6/10

LRMNLRMNSorina POP, Tristan GLATARDSorina POP, Tristan GLATARDCIRA, 2 juin 2009CIRA, 2 juin 2009

Execution frameworkExecution framework

• Moteur workflow engineWorkflow execution

• Each application is described as a workflow (using Taverna)

Job submission, input selection and data piping between jobs

• Moteur – DIANE interfaceProvides generic agents

Starts DIANE master and submits DIANE agents

7/10

LRMNLRMNSorina POP, Tristan GLATARDSorina POP, Tristan GLATARDCIRA, 2 juin 2009CIRA, 2 juin 2009

Graphical interface to the gridGraphical interface to the grid

• VBrowserProvides user interface to the grid

Browses input/result files

Launches Moteur

Follows experiment progress (job monitoring)

• Application: the Gate-LabGATE-Lab client (VBrowser plugin)

• Parses simulation (mac) file, zip inputs• Stores inputs on the grid, submits workflow• Keeps track of simulation history

GATE-Lab server• Launches workflow engine• Starts DIANE pilot-job master• Submits agents when necessary

VBrowser

Gate-Lab client

8/10

LRMNLRMNSorina POP, Tristan GLATARDSorina POP, Tristan GLATARDCIRA, 2 juin 2009CIRA, 2 juin 2009

Perspectives (I)Perspectives (I)

• VIP – Virtual Imaging Platform

• GoalEnable heavy simulations: multimodal, dynamicStore and retrieve data (organ models and simulated images)

• Multi-platform executionLarge-scale grids (EGEE, NorduGrid)Local clusters (Creatis, IN2P3)GPU

9/10

LRMNLRMNSorina POP, Tristan GLATARDSorina POP, Tristan GLATARDCIRA, 2 juin 2009CIRA, 2 juin 2009

Perspectives (II) Perspectives (II)

• Distributed database for medical imagingStorage, sharing, indexing and search on metadata

Technologies• iRODS

– Access and uniform management of heterogeneous data distributed among different sites

• MDM (Medical Data Manager)– Management of distributed DICOM images, high security level

Medical Images Database

10/10

LRMNLRMNSorina POP, Tristan GLATARDSorina POP, Tristan GLATARDCIRA, 2 juin 2009CIRA, 2 juin 2009

AcknowledgementsAcknowledgements

• University of Lyon, CREATIS-LRMN• Hugues Benoit-Cattin, David Sarrut, Patrick Clarysse

• I3S, CNRS• Johan Montagnat

• University of Amsterdam, Academic Medical Centre• Silvia D. Olabarriaga

• CERN• Jakub T. Mosciki

Thank you for your attention!Thank you for your attention!

Questions?