Seamless Medical Image Processing on the Grid on the Example of Segmentation and Partition of the...

Seamless Medical Image Processing on the Grid on the Example of Segmentation andPartition of the Airspaces

Andrzej Rutkowski1, Michał Chlebiej1,

Marcelina Borcz2, Rafał Kluszczyński2,

Krzysztof Benedyczak2 Piotr Bała2

[email protected]

1 WMiI, Nicolaus Copernicus University, Toruń, PL2 ICM, University of Warsaw, Warsaw, PL

Outline

Motivation PL-Grid infrastructure

• resources• middleware

Segmentation and Partition of the AirspacesSolution Details of the solution Conclusions

eTELEMED 2015 Piotr Bała2

Motivation

Analysis of the medical images is important part of the medical doctor activities

Growing need for use of CPU intensive applications providing better analysis

Distributed resources available Requirement for seamless access Web portal as preferred access method Existing web portals are complicated, difficult to install and

maintain Requirement for strong security

• processing of medical images


PL-Grid

National Grid Initiative Partners:

• Polish supercomputer centers:• Cyfronet, ICM, PCSS, WCSS, TASK

Project aims:• Build and operate Polish National Grid • Provide training and user’s support• Provide support for application deployment on the grid

ICM role in PL-GRID• Operate UNICORE infrastructure• Domain applications: health, materials

Piotr Bała4 eTELEMED 2015

UNICORE 7 Architecture

Piotr Bała5

Computational Resources

Data Storages Databases

Client Layer

Target System Interface

Target System

Files

Execution Manager

UNICORE 6 Hosting

Environment

SE

CU

RIT

Y

WSRF Container

User DB

Policies

Gateway

HTTPS

RegistryAtomic Services

Other Services

HTTPS / HTTP

eTELEMED 2015

UNICORE Client framework

Eclipse-based rich client

Eclipse based workflow editor

Command line client

Web client

UNICORE Portal


UNICORE portal – Generic gridbean


UNITY IDP

Authentication• who is who• wide range of mechanisms & security levels

Federations• where do they come from?

Authorization levels• The relying system must be able to decide easily who has an

access granted

Privacy• from none to paranoid

Identity management as a service• Could be used for grids and clouds


UNICORE – UNITY Integration


CT image segmentation


Segmentation process

CT image set (DICOM-Dir) is used as starting point CT data is selected Affine registration Free Form Deformation parametrized by B-splines. The resulting deformation field is segmented Volumetric tuning of every segment boundary using patients

original data.• segmentation proces depends on the image library


Deformation


UNICORE Portal – job preparation


UNICORE Portal – job execution


UNICORE Portal – Job output


Conclusions

Grid technology became mature enough to offer reliable, high quality services designed to suit requirements of different scientific communities.

UNICORE Portal offers web interface and automation of the processing of selected applications

The medical doctors can focus on diagnosis instead of writing complicated scripts, transferring the files and mastering complicated IT infrastructure.

With the UNICORE Portal creation of the application and domain specific solution become simple and straightforward.

The software stack necessary to build full featured gateway is now small and easy to handle.


Acknowledgmens

This research was supported in part by the PL-Grid Infrastructure.

The authors thank prof. A. Kukwa from Faculty of Medical Sciences, University of Warmia and Mazury for providing inspiration and medical data for tests.


Seamless Medical Image Processing on the Grid on the Example of Segmentation and Partition of the...

Documents

Transcript of Seamless Medical Image Processing on the Grid on the Example of Segmentation and Partition of the...