Gestion des données scientifiques en imagerie in vivo – Journée scientifique organisée par PIV le 7 décembre 2017 au PARCC-HEGP Michael Kain

1. PIV 2017: Gestion des données
scientifiques en imagerie in vivo
07.12.2017
FLI-IAM, entry
and
access
point
to
resources
for
medical
imaging
research
Michael Kain
France
Life
Imaging
(FLI)
-­‐
Informa:on
Analysis
and
Management
(IAM)
Your
service
desk
for
research
in
medical
imaging
2017-­‐12-­‐07
PIV
2017:
Ges:on
des
données
scien:fiques
en
imagerie
in
vivo
1

2. • Intro
– Phase
1
+
2
– Architecture
– Products
• Interoperability
– CARMIN-­‐API
– OAK
–
Open
App
Catalog
• Example:
Challenges
MICCAI
2016
• Future
Agenda
2017-­‐12-­‐07
PIV
2017:
Ges:on
des
données
scien:fiques
en
imagerie
in
vivo
2

3. • France
Life
Imaging
(FLI)
is
a
large-­‐scale
research
infrastructure
project,
aimed
at
establishing
a
coordinated
and
harmonized
network
of
biomedical
imaging
in
France
• Its
mission
is
to:
– coordinate
na:on-­‐wide
research
ac:vi:es
concerned
with
in
vivo
imaging
and
combine
the
skills
to
push
the
current
technological
barriers
– provide
scien:sts
a
convenient
access
to
a
complete
range
of
imaging
technologies
(150
imaging
systems)
and
integrated
services;
in
addi:on,
the
infrastructure
will
be
open
to
collabora:ons
with
industrial
partners
• This
project
was
selected
in
2012
by
the
call
“Inves:ssements
d’Avenir
-­‐
Infrastructure
en
Biologie
et
Santé”
• 37
million
Euros
for
8
years
• Nodes
(see
next
slide)
and
work
packages:
– Molecular
Imaging
Agents
– Instrumenta:ons
&
Innova:ons
Technologiques
– Imagerie
Interven:onnelle
– Traitement
et
Analyse
en
Imagerie
Mul:modale
– Forma:on
• hdp://francelifeimaging.fr
France
Life
Imaging
(FLI)
-­‐
1
2017-­‐12-­‐07
PIV
2017:
Ges:on
des
données
scien:fiques
en
imagerie
in
vivo
3

4. 2017-­‐12-­‐07
PIV
2017:
Ges:on
des
données
scien:fiques
en
imagerie
in
vivo
4
France
Life
Imaging
(FLI)
-­‐
2

5. • Objec:ves
of
the
Node:
“Build
an
in-­‐vivo
image
management
infrastructure”
for
– Clinician:
to
conduct
clinical
and
preclinical
research
studies
involving
new
innova:ve
in-­‐vivo
medical
imaging
and
therapeu:c
procedures
– Pharma:
to
provide
pharma
and
CRO
companies
high
technological
computa:onal
solu:ons
in
medical
imaging
– Companies
and
startups:
– Medical
imaging
community:
to
allow
experimenta:on
and
valida:on
of
new
innova:ve
medical
imaging
solu:ons
• Strengths:
based
on
exis?ng
high
technological
exper:se
and
experience
– From
data
management
solu:ons:
ArchiMed,
CATI,
SHANOIR,
…
– From
medical
image
processing
solu:ons:
BrainVisa,
medInria,
VIP
+
Bou:ques,
…
– From
large
na:onal
clinical
Cohorts:
CATI,
OFSEP,
…
• Mul:ple
research
labs
involved,
around
10
• 3
Work
groups:
"Interoperability
and
data
management",
"Processing
and
workflow"
and
"Preclinical
imaging"
• hdps://portal.fli-­‐iam.irisa.fr/services
Informa:on
Analysis
and
Management
(IAM)
-­‐
1
2017-­‐12-­‐07
PIV
2017:
Ges:on
des
données
scien:fiques
en
imagerie
in
vivo
5

6. Phases
of
FLI-­‐IAM
2017-­‐12-­‐07
PIV
2017:
Ges:on
des
données
scien:fiques
en
imagerie
in
vivo
6
Today
Goal
Publicly
funded
Self-­‐financing
Par:ally
publicly
funded
Phase
1
Phase
2
(2-­‐3
years)
Phase
3?
Beginning
of
2018
Beginning
of
2020/21
2
Mio.
€
for
research
ins:tutes
1,2
Mio.
€
for
sub-­‐contrac:ng
?
Business
model
Prototype
development

7.
FLI@Interoperability
Shanoir
Archi
Med
CATI-­‐
DB
Retrieve
Query
Registra:on
Register
Standalone
Compute/Display
CARMIN-­‐API
Server
Download
VIP +
Boutiques
OAK
Open App
Catalog
HPC
FLI-­‐IAM
Compu?ng
Components
FLI-IAM image processing
clients and viewers
Execute
Information
System and
Search Engine
Metadata
Catalogs
SingleSignOn
Auth.service
FLI-­‐IAM
Web
Portal
FLI-­‐IAM
Distributed
Archive
Black
Board
FLI-­‐IAM
Architecture
2017-­‐12-­‐07
PIV
2017:
Ges:on
des
données
scien:fiques
en
imagerie
in
vivo
7
CARMIN-­‐API
Client
FLIWS
medInria
DTM

8. • Image
Databases:
– ArchiMed
– CATIDB
– Dopamin
– Shanoir
(Shanoir
Classic,
Shanoir
Small
Animal,
Shanoir-­‐NG)
– (XNAT)
• Image
importers/
uploaders:
– ArchiMed
Desktop
Client
– ShanoirUploader
• Processing
plaMorms:
– CATI
– FLIWS
– VIP
+
Bou:ques
Products
2017-­‐12-­‐07
PIV
2017:
Ges:on
des
données
scien:fiques
en
imagerie
in
vivo
8
• Viewers/local
processings:
– BrainVisa
– medInria
• Pipelines/processings:
– Preclinical
– Bordeaux
-­‐
Func:onal
connec:vity
– Marseilles/Paris
-­‐
Sulcal
Morphometry
– 24
MICCAI
2016
Challenger
Pipelines
• Others:
– Open
App
Catalog
(OAK)
– CARMIN
API,
+
clients
below:
• carmin-­‐widget.js
(browser)
• Janicule
(desktop)
• Portal
client
(backend)
– Dicomifier
– SegPerfAnalyzer
(SPA)
– Data
Trans
Module
(DTM)
– Portal,
based
on
LifeRay
• Single-­‐Sign-­‐On
module
– Common
catalog
– Automated
test
cases

9. Installa:ons
of
ShanoirUploader
in
France
within
the
OFSEP
project
2017-­‐12-­‐07
PIV
2017:
Ges:on
des
données
scien:fiques
en
imagerie
in
vivo
9

10. • Our
main
goal:
interoperability
• We
believe
that
the
long-­‐:me
life:me
of
today
sovware
systems
are
defined
by
the
interoperability
between
the
systems
and
their
environment
• Plays
between
the
following
standards/formats
(and
more):
– DICOM
(Digital
Imaging
and
Communica:ons
in
Medicine)
and
PACS
(Picture
Archiving
and
Communica:on
System)
– DICOM
De-­‐iden:fica:on,
Pseudonymiza:on
and
Anonymiza:on
– DICOMWeb:
WADO-­‐URI,
-­‐WS
or
-­‐RS,
QIDO-­‐RS,
STOW-­‐RS
– NIfTI
(Neuroimaging
Informa:cs
Technology
Ini:a:ve)
– NRRD
(Nearly
Raw
Raster
Data)
– XDS
(Cross-­‐Enterprise
Document
Sharing)
– HL7
CDA
(Clinical
Document
Architecture)
– HL7
FHIR
(Fast
Healthcare
Interoperability
Resources)
– BIDS
(Brain
Imaging
Data
Structure)
– NIDM
(The
Neuroimaging
Data
Model)
– Medical
Health
Record
(DMP,
EHR,
PHR,
openEHR)
Interoperability
-­‐
1
2017-­‐12-­‐07
PIV
2017:
Ges:on
des
données
scien:fiques
en
imagerie
in
vivo
10

11. • Within
the
following
legal
environments:
– France:
• CNIL,
e.g.
Agrément
d’hébergement
de
données
de
santé
(ou
«
agrément
HDS
»)
• ANSM
– Europe:
• GDPR
(General
Data
Protec:on
Regula:on),
applies
on
25.
May
2018
– USA:
• HIPAA
(Health
Insurance
Portability
and
Accountability
Act)
• Title
21
CFR
Part
11
by
FDA
• Work
started
on
that
road:
– Common
catalog,
unified
meta-­‐data
search
index
over
all
image
databases
– Unified
access
to
processing
playorms:
CARMIN-­‐API
– OAK:
open
app
catalog
for
processing
applica:ons
of
medical
images
– VIP
+
Bou:ques:
exchange/execute
pipelines
encapsulated
in
Docker
images
(local/cluster)
• S:ll
a
very
long
way
to
go:
– Traceability
of
access
to
data
and
systems
– Assure
security
on
the
most
high
level
– Open
and
unified
access
for
(commercial)
viewers,
e.g.
Osirix,
to
databases
in
FLI-­‐IAM
– Integrate
all
databases
compa:ble
with
ShanoirUploader
and
name
to
FLI-­‐IAM
Uploader
– Unified
data
exchange
between
research
databases
Interoperability
-­‐
2
2017-­‐12-­‐07
PIV
2017:
Ges:on
des
données
scien:fiques
en
imagerie
in
vivo
11

12. • CARMIN:
a
Common
web
API
for
ReMote
pipeline
Invoca:on
• Unifies
the
remote
access
to
pipeline
hos:ng
playorms
• Core
API
features:
– Authen:ca:on
(username+password
or
API-­‐Key)
– Playorm
informa:on
– List
of
available
pipelines
and
pipeline
informa:on:
e.g.
Params
– Input
data
upload
– Pipeline
execu:on
– Output
data
download
• Implemented
in:
– FLIWS
– CATI
– VIP
– Coming
soon:
CBRAIN?
Interface
to
processing
playorms
2017-­‐12-­‐07
PIV
2017:
Ges:on
des
données
scien:fiques
en
imagerie
in
vivo
12

13. • Centralized,
open
app
catalog
for
processing
applica:ons
of
medical
images
• Based
on
CARMIN,
/pipelines
(1
method)
OAK
–
the
open
app
catalog
2017-­‐12-­‐07
PIV
2017:
Ges:on
des
données
scien:fiques
en
imagerie
in
vivo
13

14. • Connect
to
the
OAK:
– Open
for
all
partners
to
work
on
the
app
catalog
– Playorm
owners:
• Implement
one
interface
in
REST:
/pipelines,
that
gives
back
a
list
of
your
public
available
pipelines
(you
keep
control
of
all
your
code)
• 1
week
of
integra:on
work
(maximum)
• Add
your
playorm
to
the
website
and
your
pipelines
come
into
the
catalog
– Pipeline
developers/developing
companies:
• Contact
us
and
we
integrate
your
pipelines
• Send
us
a
GitHub
link
to
your
private
docker
image
• We
never
access
to
any
source
code,
we
do
not
even
have
to
see
inside
the
docker
image
• Implement
your
CARMIN
interface
• What
we
provide:
hdps://github.com/fli-­‐iam/CARMIN
– Swagger
defini:on
of
the
REST
interface:
to
create
your
code
– carmin-­‐client.js
in
JavaScript
for
web
portals
– Janicule,
C++
desktop
client
to
connect
to
CARMIN
OAK
–
the
open
app
catalog
2017-­‐12-­‐07
PIV
2017:
Ges:on
des
données
scien:fiques
en
imagerie
in
vivo
14

15. • 3 usage scenarios identified for FLI-IAM
– Software as a Service (SaaS)
• 1.a: turn-key service
• 1.b: resource service
– Platform as a Service (PaaS)
• Show case: organization of two scientific challenges
– Storage + Computing as a Service (SCaaS)
• MICCAI 2016
– Medical Image Computing and Computer Assisted Intervention
– SATELLITE EVENTS: Challenges MSSEG + PETSEG
– 15
challengers
/
23
pipelines
integrated
– 5
engineers
hired
for
the
challenges
– More
than
25
task
force
mee:ngs
organized
since
October
2015
– More
than
600
emails
exchanged
between
challengers
and
pipeline
integrators
(1-­‐2
weeks
average
:me
of
pipeline
integra:on)
– Around
85
man-­‐months
of
work
Example:
Challenges
MICCAI
2016
15
PIV
2017:
Ges:on
des
données
scien:fiques
en
imagerie
in
vivo
2017-­‐12-­‐07

16. People network for MSSEG
16
PIV
2017:
Ges:on
des
données
scien:fiques
en
imagerie
in
vivo
2017-­‐12-­‐07

17. Key numbers - 2
17
Defini:on/
Planning/Orga,
11
Recruitment,
1
Data
Prepara:on,
22
Portal,
6
Shanoir/DTM,
15
VIP/Bou:ques,
3
Pipeline
Integra:on,
13
Evalua:on/
SPA,
10
Tes:ng/QC,
4
Man-­‐months
PIV
2017:
Ges:on
des
données
scien:fiques
en
imagerie
in
vivo
2017-­‐12-­‐07

18. Architecture for challenges
18
Portal
FLI-­‐IAM
Shanoir
VIP/Bou?ques
Database
for
medical
images
Processing
plaMorm
DTM
SFTP
PIV
2017:
Ges:on
des
données
scien:fiques
en
imagerie
in
vivo
2017-­‐12-­‐07

19. VIP/Boutiques: tool integration
Tool description
(JSON)
Tool implementation
Workflow Management System
for Science Automation
http://pegasus.isi.edu
Virtual Imaging Platform
VIP
http://vip.creatis.insa-lyon.fr
Your tool ?
Niak / PSOM
http://boutiques.github.io
http://cbrain.mcgill.ca
2017-­‐12-­‐07

20. • We
did
a
market
study
(TMTG)
at
the
beginning
of
2017
• Organiza:on
of
phase
2
– How
will
FLI-­‐IAM
be
structured?
• As
structure
itself,
and
in
rela:on
to
FLI,
phase
2
(in
valida:on)
• Partners,
research
and
industrial
(ongoing
exchanges)
• Poten:al
service
customers
(ongoing
presenta:ons)
– Order
of
services
to
propose?
• First
service
offer
to
start:
Shanoir
Small
Animal
• What
else
aver?
• Some
of
the
next
steps
to
prepare:
– More
interoperability!!!
– Agrément
d’hébergement
de
données
de
santé
(ou
«
agrément
HDS
»)
– Service
for
clinical
research
studies:
hos:ng
of
electronic
health
records
– Project
OpenAIRE:
integrate
open
science
directly:
• DOI
(digital
object
iden:fier)
managment
within
databases
and
processing
playorms
within
FLI-­‐IAM
• Any
ideas
or
proposals:
please
contact
me
J
Future
PIV
2017:
Ges:on
des
données
scien:fiques
en
imagerie
in
vivo
20
2017-­‐12-­‐07

21. Thank
you
for
your
aVen?on!
2017-­‐12-­‐07
PIV
2017:
Ges:on
des
données
scien:fiques
en
imagerie
in
vivo
21