Creatis (insa lyon) J. Montagnat

ACI GRID 2002 MEDIGRID: high performance medical image processing on a computational grid (accepted 3 years project)

• CREATIS (INSA Lyon)

• J. Montagnat

• LISI (INSA Lyon)

• L. Brunie

• ERIC (University Lyon 2)

• S. Miguet

Purposes

• Medical image storage and processing on the grid

• Structured data
• Strong semantic (distributed medical files)
• Sensitive (security issues)

• Process complex algorithms with large computing power and memory requirements

• parallel processing

• Handle very large data sets

• distributed storage
• massive distributed processing

• Context: datagrid testbed.

Scientific challenges

• Shared and distributed data management

• data hierarchy, dynamic indices, optimization, caching

• Hybrid content-based and metadata queries

• content-based queries, pre-processing, distributed queries

• Complex modeling of anatomical structures

• anatomical and functional models, parallelisation

• Simulation of MRIs

• MRI modeling, artifacts modeling, parallel simulation

Data and users

• Medical Data

• Images and metadata
• Nominative (critical) and non-nominative (private) data
• DICOM3 for medical images

• Users

• Patient: has free access to its medical data.
• Physician: has complete read access to its patients data. Few persons have read/write access.
• Researchers: may obtain read access to anonymous medical data for research purposes. Nominative data should be blanked before transmission to these users.

Data management requirements

• Medical data requirements

• Very large databases (Tb/year/hospital) and very long term storage (20 years for all images, up to 70 years in some cases)
• Large amount of metadata
• Access right on a medical department basis
• Log data processing

• Security related requirements

• Data access security (Read-only access for most users, access right on a medical department basis, no read access to private data by any third party including system administrators)
• Explicit control of sites where private data may be replicated
• Nominative metadata should be stored in secure databases
• Images should not be replicated outside hospital without blanking headers
• Images content should be encrypted

Data infrastructure testbed

• 4 image sources:

• Acquired from Lyon cardiological hospital
• Bone structure database from ESRF Grenoble
• Mammographies from the DDSM
• Simulated images

Foreseen medical data infrastructure

• Split nominative and anonymous data to allow data replication on unsecured sites.

Applications

• Image database importation

• Pre-processing of images for database indexing
• Data distribution

• Hybrid content-based and metadata queries

• Triggering parallel processing, distributed queries, DICOM server interface
• pipeline processing

• MRI simulation

• parallel simulation kernel

• Heart segmentation from MRI sequences

• heart modeling, parallel job execution

Application example: Metadata query and image processing

Schedule

• year 1

• setting up databases
• statistical parameters computation
• image indexing
• spitfire-based security
• MRI simulator
• grid interface to DICOM servers

• year 2

• interface with datagrid / remote processing
• distributed data management
• spitfire extensions
• content-based and metadata queries
• pipeline processing

Schedule

• year 3

• distributed data management
• biomedical modeling
• mammography studies
• cardiac studies
• first medical assessment