neuGRID e outGRID: una neuGRID e outGRID: una infrastruttura elettronica globale infrastruttura elettronica globale

per le neuroscienze computazionaliper le neuroscienze computazionali

Giovanni B FrisoniGiovanni B Frisoni

Vice Scientific DirectorVice Scientific DirectorIRCCS Fatebenefratelli, The National Center for AlzheimerIRCCS Fatebenefratelli, The National Center for Alzheimer’’s and s and

Mental Diseases, Brescia, ItalyMental Diseases, Brescia, Italy

WHY DO NEUROSCIENTISTS NEED WHY DO NEUROSCIENTISTS NEED ee--INFRASTRUCTURES TO STUDY AND INFRASTRUCTURES TO STUDY AND

TREAT ALZHEIMERTREAT ALZHEIMER’’S DISEASE?S DISEASE?

Alzheimer’s dementia

Mild cognitive impairment

No cognitive symptoms

-50 -40 -30 -20 -10 -5 0 10 Years from diagnosis of AD30 40 50 60 70 80 90 Age

Am

yloi

d an

d ta

u de

posi

tion

With markers

AD path.

Clinical

Diagnosis

Current understanding of what Current understanding of what constitutes Alzheimerconstitutes Alzheimer’’ss

Impossible

Alzheimer’s

Mild cognitive impairment

No cognitive symptom

30 40 50 60 70 80 90 100 Age

Early diagnosis + Early diagnosis + antianti--amyloid drugsamyloid drugs

Am

yloi

d an

d ta

u de

posi

tion

Recipe to develop markers for AlzheimerRecipe to develop markers for Alzheimer’’s disease:s disease:1. Large databases (‘0,000s) of very mildly affected patients

T6T0 T18T12 …

…

US ADNI

≈10,000 imagesAustralian ADNI (AIBL)

≈

4,000 images

Japanese ADNI (≈

3,000 images)

European ADNI (IMI Pharma-cog)

(≈

1,500 images)

Recipe to develop markers for AlzheimerRecipe to develop markers for Alzheimer’’s disease:s disease:1.Large databases (‘0,000s) of very mildly affected patients

2.Sophisticated algorithms for image analysis

Cortical thickness at month 0

Cortical thickness at month 30 Difference analysis

1.50 mm 4.50 mm +.50 -.50

Recipe to develop markers for AlzheimerRecipe to develop markers for Alzheimer’’s disease:s disease:1.Large databases (‘0,000s) of very mildly affected patients

2.Sophisticated algorithms for image analysis3.Sophisticated statistical models

0 | 1 | 2 | 3 Years from dementia

-3 | -2 | -1 |

Gray matter change in

Healthy elders

Very early Alzheimer’s

Recipe to develop markers for AlzheimerRecipe to develop markers for Alzheimer’’s disease:s disease:1.Large databases (‘0,000s) of very mildly affected patients

2.Sophisticated algorithms for image analysis3.Sophisticated statistical models

4.Powerful computational resources

FR CH

IT NL SW

FP7 neuGRIDComputational grid-

and Géant-based e-

Infrastructure

• Includes world’s largest image databases on Alzheimer’s (US-ADNI, A-ADNI)

• Includes an advanced algorithm for cortical thickness extraction (McGill, Canada)

• Built general but to be expanded with more applications (i.e. stat models) when available

• Born with 450 cores, expandable “ad libitum”

Web Portal

• AJAX-based Portal• CAS SSO Framework

• Grid Proxy Applet• MyProxy Session

Web PortalWeb Portal

• AJAX-based Portal• CAS SSO Framework

• Grid Proxy Applet• MyProxy Session

LORIS Database

• Connected to SSO• Interfaces to Data Acq• Interfaces to Data QC

LORIS DatabaseLORIS Database

• Connected to SSO• Interfaces to Data Acq• Interfaces to Data QCOnline Shell Access

• GSISSH Applet• Access to Grid Infra.

• CIVET Pipeline gridified• SFTP Facility to Upload

Online Shell AccessOnline Shell Access

•• GSISSH Applet• Access to Grid Infra.

• CIVET Pipeline gridified• SFTP Facility to Upload

Data challenge on the whole grid infrastructure (9/2009)

Cortical thickness extraction (CIVET)6,500 image ADNI dataset

10 days

Recipe to develop markers for AlzheimerRecipe to develop markers for Alzheimer’’s disease:s disease:1.Large databases (‘0,000s) of very mildly affected patients

2.Sophisticated algorithms for image analysis3.Sophisticated statistical models

4.Powerful computational resources5.Strong link with related international efforts

FP7 outGRIDA Worldwide e‐Infrastructure 

for Computational  Neuroscientists

FP7 outGRIDA Worldwide e‐Infrastructure for 

Computational Neuroscientists

• Access to large image datasets• Image analysis algorithms• Grid-based high performance computing

• Image analysis algorithms• Grid-based high performance computing

• Home to world largest image dataset on Alzheimer’s• Image analysis algorithms• Friendly Pipeline Interface• Centralized high performance computing

FP7 outGRIDA Worldwide e‐Infrastructure for 

Computational Neuroscientists

OBJECTIVES• Define activities (including research and development)

and outline technical specifications for interoperability

• Foster, within the project’s lifetime, the maximum possible degree of interoperability allowed by this Support Action among the 3 infrastructures

• Lay the foundations for a larger research and development effort aimed to achieve full interoperability among the three infrastructures

FP7

FP7

THE GRAND VISIONA Worldwide e‐Infrastructure for Computational Neuroscientists

CAnet

Internet2.0 Géant

Markers for diagnosis and of disease activity

Diabetes Alzheimer’s

Diagnosis Glicaemia

• Atrophy greater than normal• Lower glucose utilization• Higher amyloid depositionWITH SPECIFIC TOPOGRAPHY

Monitor treatment effect

Glycated hemoglobin

• Progression of atrophy• Progression of glucose under-utilization (PET)WITH SPECIFIC TOPOGRAPHY

0 -.125 -.25