3. in silico Medicine
“…is the direct use of computer simulation
in the diagnosis, treatment, or prevention of
a disease.”
Predict disease
#OIA4
Paris,
France
04-‐Dec-‐13
Personalise treatment
3
4. Select
Workflow
Retrieve
ExisGng
Data
Transform
or
Infer
Data
Run
Workflow
Return
Results
Patient Data
Workflow Inputs
Workflow Outputs
VPH-Share
euHeart
@neurIST
Infostructure
Co-ordinator:
University of
Sheffield, UK
Personalised
Model
Patient Avatar
Application
Project No: 269978
Partners:
Patient Centred Computational Workflows
Knowledge
Management
Knowledge Discovery
Data Inference
Data Services:
Patient/Population
Compute
Services
HPC Infrastructure
(Public / Private)
Semantic Services
Storage
Services
Cloud Platform
(DEISA / PRACE)
VPH OP
ViroLab
CYFRONET, PL
Sheffield Teaching
Hospitals, UK
ATOS Origin, ES
Kings College
London, UK
Empirica, DE
CiNECA, IT
Nine Health
CIC, UK
INRIA, FR
IOR, IT
Open Univ., UK
Philips Elec., NL
TU Eindhoven, NL
Univ. Auckland, NZ
Uv Amsterdam, NL
UCL, UK
Univ. Vienna, AT
AATRM, ES
FCRB, ES
#SummerSchool
20-‐Jun-‐13
4
5. VPH-Share A0 Master Plan
VPH
Shar
e
Clien
t
ATM
Master Interface
Atomic Service
Manager
Atmosphere Cloud Platform
Atomic Service
Generic
Invoker
Data Browser
Workflow Composer
Semantic Services
Visualisation Tools
Service Registry
…
Silk,
LinQuer
Service
…
LOD Databases
Multi-Ontology/
Archetype
Search Services
Data Buckets
(C-DISC, CSV,
…)
Workflow
Registry
S
P
A
R
Q
L
D
is
c
o
v
e
B
r
r
y
o
w
s
e
S
r
e
LD Databases
RDB2RDF
Service
Generic
Workflow
Documen
t
Authentication
Services
libclou
d
provid
er
libclou
d
provid
er
Schema
Crawler
Workflow
Execution Service
Cloud
Clients
Atomic
Service
Registry
Atomic
Service
Description
Virtual Machine
Template
Registry
Cloud Facade
Proxy Controller
a
r
c
h
Taverna
Server
MAFEventBus
Monitoring
Controller
Atomic Service
Deployment Wizard
Database
Services
Integration Points
AHE Services API
AHE Runtime
…
Database 1 Query
Services
(SPARQL & SQL)
Database n Query
Services
(SPARQL & SQL)
Databases
(SQLServer, …)
…
External
Structured
Data Providers
REST API &
HTML Service
(Ruby) Sinatra &
Passenger
AMS
Manag
er
(Java)
OSGi
bundle
s
Apach
e
Karaf
…
Database 2 Query
Services
(SPARQL & SQL)
…
Domain Model
(Ruby)
App
State
Objec
t
Scheduler /
Optimizer
Individual Relational Databases
Mo
ng
oD
Atmosphere Internal
B
Registry
Allocation Management Service
Data
Publishing
Suite
(GUI)
VoID
VoID Document
Document
Database
VoID
Services
AHE Engine
App
Regis
try
JBP
M
Workf
low
&
Main
Logic
AHE Database
Hibernate ORM
Security Module
Storage Module
Data Reliability & Integrity Services
External
Data
Storage
Data Infrastructure Services
Connector Module
External
HPC
Platform
Load
Balancer
Extension Points
H
S
S
A
P
t
High PerformanceExecution Engine (AHE)
R
R
e
C
U
e
C
ri
E
n
g
ASIProxy
Dashboard
NOVA
API
Access & Control
Frontend
Network
Worker
Web Service Security Agent
Compute
Worker
Web Service Wrapper
Soaplab2, CXF, soap4r
Remote
Access
Service
Manager
Driver
VPH-Share Tool / App
Hypervisor
LOBCDER Federated Storage
Access
Q
ue
ue
S
ch
ed
ul
er
Request
Manager
Im
ag
es
Virtual Resource System
Cloud
Storage
Driver
Data
Volume
Raw Operating System (Linux)
Root
Volume
Atomic Service Instance Contents
PSLoade
r
External
Cloud
Data
Storage
Connection
Module
Monitoring
System
Monitoring Agent (Munin)
Data
Resource
Catalog
Object
Storage
(Swift)
P
r
A
o
c
C
c
o
x
n
y
o
u
t
O
a
n
b
t
i
j
n
e
e
c
r
t
LOBCDER Data Federation Middleware
Private Compute & Storage Cloud (OpenStack Example)
Atmosphere Cloud Platform
#OIA4
Paris,
France
04-‐Dec-‐13
5
16. PublicaGon
Data
Models,
Techniques,
Algorithms
Context
InvesGgaGon
Study
Assay
Provenance
AgribuGon
Credit
• Gathered: scattered across different
repositories/catalogues
• All necessary elements available and
accessible maybe open
• Documented sufficiently well
– Explicitly Transparent: How, Why, What,
Where, Who, When
– Comprehensive: Just Enough
– Comprehensible: Independent
understanding
• Skills and resources to repeat
– Crowd sourced? Supercomputer?
#OIA4
Paris,
France
04-‐Dec-‐13
16
17. Accessible
Reusable
Capable
PublicaGon
Data
INSTRUMENTs
Samples,Specimens
Strains
Models,
Techniques,
Algorithms
Context
InvesGgaGon
Study
Assay
74%
/
26%
31%
/
8%
Provenance
AgribuGon
Credit
#OIA4
Paris,
France
04-‐Dec-‐13
17
20. Patient Avatar
• A large virtual catalogue of
every item of data, information
& knowledge of
•
•
Medical Devices
Medical Images
Patient History
Patient; or
Collection of patients
(Avatars)
Vital Signs
Genomics
• It also needs to be shared
securely, to be able
•
•
To be Searched, Browsed &
Analysed
For Healthcare, Research &
Education
Proteomics
Procedures
Results
#OIA4
Paris,
France
04-‐Dec-‐13
Histology
20
21. Patient Avatar – Goals
• Integrating fragmented data and knowledge into a single
cohesive unit of data
• Creating a centralised repository with reliable population and
patient data around which VPH tools and applications could
be built
• Providing diagnostic or prognostic decision and treatment
planning support using predictive models built around a
patient-specific avatar
• Providing a comprehensive overview of a patient with missing
data based on averages from population phenotype to explore
and test ideas virtually
#OIA4
Paris,
France
04-‐Dec-‐13
21
22. Current Access to Clinical Data
Owner or Facilitator
Description
Number (records)
Accessibility
Sheffield Teaching
Hospitals
ACS data
Clinical data, outcomes
~750
Philips
Others on request
Nine Health
HES extract
Data catalogues
~ 5 millon
On demand
FRCB
Clinical data
Ethics approved for
entire record
On demand
AATRM
Images
~ 6 million
On demand
@neurIST
Image data
CRIM data
Derived data
euHeart
3D Models
Virolab
Rule database
Publication corpus
~100 (600)
~1400
~300
216 cardiac
~90 other
4 Rule DBs
~1.2 GB
VPH-OP
Extensive clinical baseline data
Images
281
Mixed
#OIA4
Paris,
France
04-‐Dec-‐13
On demand
Registered users
Registered users
On demand
22
26. Clinical Information Systems
Tabular
Data
Non-Tabular
Data
Data Publishing Suite
Semantic Services
Reference
Data
Patient
Avatar
RDF
Graphs
Pseudo Identifier (UID)
Demographics
Height
Weight
Vital Signs
Heart Rate
Blood Pressure
Systolic
Diastolic
Aneurysm Related Health Event
Risk Factors
Aneurysm Imaging Study
Medications
Demographic
Vital Signs
Images
Pseudo Identifier (UID)
Demographics
Height
Weight
Vital Signs
Heart Rate
Blood Pressure
Systolic
Diastolic
Pulmonary Function
Risk Factors
Cardiac Imaging Study
Medications
Pseudo Identifier (UID)
Demographics
Height
Weight
Vital Signs
Heart Rate
Blood Pressure
Systolic
Diastolic
Orthopaedic Health Event
Gynaecological Information
Bone Phenotype Imaging Study
Medications
Demographic
Medications
Risk Factors
Vital Signs
Parameter
Estimation
Uncertainty
Propagation
Genomic Data
Lab Reports
Physiological
Envelope
2
3
4
1
#OIA4
Paris,
France
04-‐Dec-‐13
8
9
10
6
7
Computational Workflows and Services
26
27. Patient Avatar & Workflow Archetypes
(a) Common Shared Archetype
(b) @neurIST Workflow Archetype
Patient Pseudoidentifier (PID)
(a) Common Shared Archetype
Demographics
(a) Common Shared Archetype
(a) Common Shared Archetype
Personal & Social History
Patient Pseudoidentifier (PID)
Patient Pseudoidentifier (PID)
Fitness & Lifestyle
Patient Pseudoidentifier (PID)
Demographics
Demographics
Employment Details
Demographics
Personal &
Personal & Social History
Vital Signs Social History
Personal &
Fitness & UseSocial History
Fitness & Lifestyle
SubstanceLifestyle
Fitness & Details
Employment Lifestyle
Employment Details
Employment Details
Vital Signs
Vital Signs
Vital Signs
Substance Use
Substance Use
Substance Use
Aneurysm Related Health Event
(b)@neurISTRisk Factors
(b) @neurISTWorkflow Archetype
Supporting Workflow Archetype
(b) @neurIST Workflow Archetype
Specimens (Blood, Tissue)
Aneurysm Related Health Event
Aneurysm Related Health Event
Medications Related Health Event
Aneurysm
Supporting Risk Factors
SupportingImaging Study
Aneurysm Risk Factors
Supporting Risk Factors
Specimens (Blood, Tissue)
Specimens (Blood, Tissue)
Imaging Details
Specimens
Medications (Blood, Tissue)
Medications
Medications
Aneurysm Imaging Study
Aneurysm Imaging Study
Aneurysm Imaging Study
Imaging Details
Imaging Details
Imaging Details (c) euHeart Workflow Archetype
Baseline History
(c) euHeart Workflow Archetype
(c) euHeart Workflow Archetype
Cardiac HealthWorkflow Archetype
(c) euHeart Event
Lab Tests
Baseline History
Baseline History
Medications History
Baseline
Cardiac Health Event
Cardiac Health Event
Cardiac Imaging Study
Cardiac
Lab Tests Health Event
Lab Tests
Imaging Details
Lab Tests
Medications
Medications
Medications
Cardiac Imaging Study
Cardiac Imaging Study
Cardiac Imaging Study
Imaging Details
Imaging Details
Imaging Details
#OIA4
Paris,
France
04-‐Dec-‐13
(e) VPHOP Workflow Archetype
Hand Grip Strength
(e) VPHOP Workflow Archetype
(e) VPHOP Workflow Archetype
Orthopeadic Health Event
(e) VPHOP Workflow Archetype
Medications
Hand Grip Strength
Hand Grip Strength Tissue)
Specimens (Blood,
Hand Grip Strength
Orthopeadic Health Event
Orthopeadic Health Event Study
Bone PhenotypeHealth Event
Orthopeadic Imaging
Medications
Medications
Medications
Specimens (Blood, Tissue)
Specimens (Blood, Tissue)
Specimens (Blood, Tissue)
Bone Phenotype Imaging Study
Bone Phenotype Imaging Study
Bone Phenotype Imaging Study
(d) ViroLab Workflow Archetype
Sexual Health
(d) Subtype Information
(d) ViroLab Workflow Archetype
HIVViroLab Workflow Archetype
(d) ViroLab Workflow Archetype
Specimens
SexualHealth
Sexual Health
Sexual Health
HIV Subtype Information
HIV Subtype Information
HIV Subtype Information
Specimens
Specimens
Specimens
27
30. Cloud-Enabled Computation Workflows
•
portal.vph-share.eu – Provides access to Clinical Data and Scientific applications
and workflows
•
Taverna Workbench – Provides Desktop tool for composition of Scientific Workflows
with VPH-Share applications and data
•
OnlineHPC.com – Allows the creation of Scientific Workflows composed of External
and VPH-Share Applications and VPH-Share data
•
Meta-Workflow Manager – Executes multiple Scientific Workflow Engines on the
VPH-Share Platform
•
Workflow Cloud Plugin – Allows execution of Scientific Web-Services and
Application on multiple cloud platforms
•
Command-Line Wrapper – Allows developers to wrap command-line applications
into a (REST/SOAP) webservice via wsme and GIMIAS
•
NoMachine RDP – Provides Remote Desktop services for applications that require
user interaction
#OIA4
Paris,
France
04-‐Dec-‐13
30
32. CLIENT-‐SIDE
SERVER-‐SIDE
External
ApplicaGon
Taverna
Workbench
Clinical
Researcher
Taverna
Server
Workflow
Manager
API
VPH-‐Share
plugin
VPH-‐Share
plugin
Web
services
…
Cloud
Façade
GIMIAS
CLPs
VPH-‐Share
Workflow
Taverna
On-‐line
VPH-‐Share
plugin
…
Web
services
Web-‐based
Remote
Desktop
#OIA4
Paris,
France
04-‐Dec-‐13
AS
AS
AS
AS
AS
AS
AS
with
interacGon
AS
without
interacGon
STORAGE
32
34. Scientific Cloud Platform
VPH
Shar
e
Clien
t
ATM
Master Interface
Atomic Service
Manager
Atmosphere Cloud Platform
Atomic Service
Generic
Invoker
Data Browser
Workflow Composer
Semantic Services
Visualisation Tools
Service Registry
…
Silk,
LinQuer
Service
…
LOD Databases
Multi-Ontology/
Archetype
Search Services
Data Buckets
(C-DISC, CSV,
…)
Workflow
Registry
S
P
A
R
Q
L
D
is
c
o
v
e
B
r
r
y
o
w
s
e
S
r
e
LD Databases
RDB2RDF
Service
Generic
Workflow
Documen
t
Authentication
Services
libclou
d
provid
er
libclou
d
provid
er
Schema
Crawler
Workflow
Execution Service
Cloud
Clients
Atomic
Service
Registry
Atomic
Service
Description
Virtual Machine
Template
Registry
Cloud Facade
Proxy Controller
a
r
c
h
Taverna
Server
MAFEventBus
Monitoring
Controller
Atomic Service
Deployment Wizard
Database
Services
Integration Points
AHE Services API
AHE Runtime
…
Database 1 Query
Services
(SPARQL & SQL)
Database n Query
Services
(SPARQL & SQL)
Databases
(SQLServer, …)
…
External
Structured
Data Providers
REST API &
HTML Service
(Ruby) Sinatra &
Passenger
AMS
Manag
er
(Java)
OSGi
bundle
s
Apach
e
Karaf
…
Database 2 Query
Services
(SPARQL & SQL)
…
Domain Model
(Ruby)
App
State
Objec
t
Scheduler /
Optimizer
Individual Relational Databases
Mo
ng
oD
Atmosphere Internal
B
Registry
Allocation Management Service
Data
Publishing
Suite
(GUI)
VoID
VoID Document
Document
Database
VoID
Services
AHE Engine
App
Regis
try
JBP
M
Workf
low
&
Main
Logic
AHE Database
Hibernate ORM
Security Module
Storage Module
Data Reliability & Integrity Services
External
Data
Storage
Data Infrastructure Services
Connector Module
External
HPC
Platform
Load
Balancer
Extension Points
H
S
S
A
P
t
High PerformanceExecution Engine (AHE)
R
R
e
C
U
e
C
ri
E
n
g
ASIProxy
Dashboard
NOVA
API
Access & Control
Frontend
Network
Worker
Web Service Security Agent
Compute
Worker
Web Service Wrapper
Soaplab2, CXF, soap4r
Remote
Access
Service
Manager
Driver
VPH-Share Tool / App
Hypervisor
LOBCDER Federated Storage
Access
Q
ue
ue
S
ch
ed
ul
er
Request
Manager
Im
ag
es
Virtual Resource System
Cloud
Storage
Driver
Data
Volume
Raw Operating System (Linux)
Root
Volume
Atomic Service Instance Contents
PSLoade
r
External
Cloud
Data
Storage
Connection
Module
Monitoring
System
Monitoring Agent (Munin)
Data
Resource
Catalog
Object
Storage
(Swift)
P
r
A
o
c
C
c
o
x
n
y
o
u
t
O
a
n
b
t
i
j
n
e
e
c
r
t
LOBCDER Data Federation Middleware
Private Compute & Storage Cloud (OpenStack Example)
Atmosphere Cloud Platform
#OIA4
Paris,
France
04-‐Dec-‐13
34
35. Platform Architecture
Admin
Modules
available
in
first
prototype
Developer
Data
and
Compute
Cloud
Planorm
ScienGst
Deployed
by
AMS
on
available
resources
as
required
by
WF
mgmt
or
generic
AS
invoker
VPH-‐Share
Master
UI
AS
mgmt.
interface
Atomic
Service
Instances
VPH-Share Tool / App.
AM
Service
Generic
AS
invoker
VM
templates
Workflow
descripGon
and
execuGon
Security
mgmt.
interface
ComputaGon
UI
extensions
DRI
Service
Data
mgmt.
interface
AS
images
101101
101101
011010
101101
011010
011010
111011
111011
111011
Available
Managed
cloud
datasets
infrastructure
Atmosphere
persistence
layer
(internal
registry)
Raw OS (Linux variant)
LOB Federated storage access
Web Service cmd. wrapper
Web
Service
security
agent
Generic VNC server
Generic
data
retrieval
Data
mgmt.
UI
extensions
Security
framework
LOB
federated
storage
access
Cloud
stack
clients
HPC
resource
client/backend
Custom
AS
client
Remote
access
to
Atomic
Svc.
UIs
#OIA4
Paris,
France
04-‐Dec-‐13
Physical
resources
35
36. Cloud Platform Architecture
•
Developer
Admin
ScienGst
The platform provides a set of APIs for the VPH-Share Master
Interface and other applications, enabling Atomic Services to be
developed.
VPH-‐Share
Core
Services
Host
Cloud
Facade
(secure
RESTful
API
)
VPH-‐Share
Master
Int.
Cloud
Manager
Atmosphere
Management
Service
(AMS)
Cloud
stack
plugins
(JClouds)
Development
Mode
Atmosphere
Internal
Registry
(AIR)
Generic
Invoker
Workflow
management
ComputaGonal
Cloud
Site
External
applicaGon
Cloud
Facade
client
Customized
applicaGons
may
directly
interface
the
Cloud
Facade
via
its
RESTful
APIs
#OIA4
Paris,
France
04-‐Dec-‐13
Head
Node
Image
store
(Glance)
Worker
Worker
Worker
Worker
Node
Node
Node
Node
Worker
Worker
Worker
Worker
Node
Node
Node
Node
36
37. Accessible HPC Execution Platform
—
Provides
virtualized
access
to
high
performance
execuGon
environments
—
Seamlessly
provides
access
to
high
performance
compuGng
to
workflows
that
require
more
computaGonal
power
than
clouds
can
provide
—
Deploys
and
extends
the
ApplicaGon
HosGng
Environment
–
provides
a
set
of
web
services
to
start
and
control
applicaGons
on
HPC
resources
Invoke
the
Web
Service
API
of
AHE
to
delegate
computaGon
to
the
grid
ApplicaGon
-‐-‐
or
-‐-‐
Present
security
token
(obtained
from
authenGcaGon
service)
ApplicaGon
HosGng
Environment
Auxiliary
component
of
the
cloud
planorm,
responsible
for
managing
access
to
tradiGonal
(grid-‐based)
high
performance
compuGng
environments.
Provides
a
Web
Service
interface
for
clients.
AHE
Web
Services
(RESTlets)
GridFTP
WebDAV
Tomcat
container
Workflow
environment
-‐-‐
or
-‐-‐
End
user
QCG
CompuGng
Job
Submission
Service
(OGSA
BES
/
Globus
GRAM)
RealityGrid
SWS
User
access
layer
Resource
client
layer
Delegate
credenGals,
instanGate
compuGng
tasks,
poll
for
execuGon
status
and
retrieve
results
on
behalf
of
the
client
Grid
resources
running
Local
Resource
Manager
(PBS,
SGE,
Loadleveler
etc.)
#OIA4
Paris,
France
04-‐Dec-‐13
37
38. Unstructured Data Storage
Ticket
validaGon
service
LOBCDER
host
(149.156.10.143)
Auth
service
WebDAV
servlet
REST-‐interface
LOBCDER
service
backend
Core
component
host
(vph.cyfronet.pl)
GUI-‐based
access
Resource
factory
Storage
driver
Storage
driver
EncrypGon
Resource
keys
(SWIFT)
catalogue
Atomic
Service
Instance
(10.100.x.x)
Mounted
on
local
FS
(e.g.
via
davfs2)
Amazon
S3
Storage
backend
•
•
•
SWIFT
storage
backend
Generic
WebDAV
client
Master
Interface
component
Data
Manager
Portlet
(VPH-‐Share
Master
Interface
component)
Service
payload
(VPH-‐Share
applicaGon
component)
External
host
VPH-‐Share
federated
data
storage
module
(LOBCDER)
enables
data
sharing
in
the
context
of
VPH-‐
Share
applicaGons
The
module
is
capable
of
interfacing
various
types
of
storage
resources
and
supports
SWIFT
cloud
storage
(support
for
Amazon
S3
is
under
development)
LOBCDER
exposes
a
WebDAV
interface
and
can
be
accessed
by
any
DAV-‐compliant
client.
It
can
also
be
mounted
as
a
component
of
the
local
client
filesystem
using
any
DAV-‐to-‐FS
driver
(such
as
davfs2).
#OIA4
Paris,
France
04-‐Dec-‐13
38
40. Elephant in the Room
• Security & Privacy
• Legislation & Ethics
• Training
•
Long Tail of
Physicians and
Care-takers
• P4 Medicine Journey
•
Predictive
• Preventative
• Personalised
• Participatory
#OIA4
Paris,
France
04-‐Dec-‐13
40
42. <Thank
You!>
Dr Susheel Varma <susheel.varma@sheffield.ac.uk>
VPH-Share - Scientific Workflows Coordinator
Department of Cardiovascular Science
The Medical School, The University of Sheffield
Beech Hill Road, Sheffield S10 2RX UK
T: +44 (0)114 271 2863
#OIA4
04-‐Dec-‐13
42
43. @neurIST Simulation Workflow
•
•
•
•
•
@neurIST- Integrated Biomedical Informatics for the Management of Cerebral
Aneurysms
(http://www.aneurist.org)
IST project funded within the EU FP6
Duration: 2006-2010
Budget: 17M€
Participants: 28 institutions
–
–
–
Public and private,
Industry, hospitals, academia,
12 European countries
•
External collaborators: from USA,
New Zealand, Japan)
•
@neurIST main objective:
@neurIST will transform the management
of cerebral aneurysms by providing
new insights, personalized risk
assessment, and methods for the
design of improved medical devices
and treatment protocols.
45. Two Workflows from @neurIST
Selec(ng
Boundary
Condi(ons
Volume
Rendering
GAR
Segmenta(on
Skeletoniza(on
Input:
3D
image
Output:
3D
image
DescripGon:
aneurysm
and
vessels
VisualisaGon
Input:
Image,ROI
Output:
surface
mesh
DescripGon:
vessels
and
aneurysm
extracGon
CFD
preprocessor
Flow
Simula(on
Input:
xml,
surface
mesh
Output:
surface
mesh,
ccl
DescripGon:
Defines
hemodynamic
model
Neck
Selec(on
Input:
DICOM
Output:
3D
image
DescripGon:
Converts
a
DICOM
image
to
VTK
image
Input:
wall
shear
stress
Output:
.csv
file
DescripGon:
computes
hemodynamic
descriptors
Input:
surface
mesh.
Output:
skeleton.
DescripGon:
necessary
to
set
the
boundary
condiGons
DICOM
Input:
surface
mesh
Output:
volumetric
mesh
DescripGon:
creates
a
volumetric
mesh
of
the
selected
geometry
Input:
surface,
1D
model
Output:
xml,
vtk
DescripGon:
boundary
condiGons
for
CFD
Flow
Simula(on
post-‐processing
Volumetric
Mesh
Input:
volumetric
mesh,
ccl
Output:
wall
shear
stress
map
DescripGon:
solves
flow
equaGons
Input:
surface
mesh
Output:
surface
mesh
DescripGon:
aneurysm
neck
surface
and
dome
selecGon
Bounding
Box
Input:
3D
image
Output:
ROI
DescripGon:
volume
selecGon
Mesh
Edi(ng
Input:
surface
mesh
Output:
surface
mesh
DescripGon:
clipping
vessels,
cleaning
surface
(cell
removal,
closing
holes,
smoothing…)
Morphology
Descriptors
Aneurysm
isola(on
Input:
surface
mesh
Output:
surface
mesh
DescripGon:
aneurysm
isolaGon
Input:
surface
mesh
Output:
xml,
vtk
DescripGon:
surface,
depth…
and
ZMI
calculaGon
Morphological
analysis
GIMIAS
Hemodynamic
analysis
ANSYS
(ICEM)
Common
opera(ons
@neuFuse
ANSYS
(CFX)
Manual
interacGon
46. Morphological Workflow
• Morphological, hemodynamic and structural analyses have been
linked to aneurysm genesis, growth and rupture.
• Evidence indicating differences in morphology and flow between
ruptured and unruptured aneurysms have been shown for reduced
patient cohorts.
• Structural wall mechanics has been used to justify the growth and
remodelling happening at the aneurysm level.
+
images
Morphological
analysis
Morphological
descriptors
+
BC,
material
Haemodynamic
analysis
Hemodynamic
descriptors
+
BC,
material
Structural
analysis
Structural
descriptors
Confidence
in
physical
measures
+
Direct
diagnos6c
power
PracGcally,
morphological
characterizaGons
might
currently
have
the
highest
predic(ve
capabili(es
with
respect
to
the
other
analyses.
48. GAR Segmentation
A
surface
mesh
represenGng
the
vascular
geometry
is
required
to
perform
the
@neurIST
morphological
and
hemodynamic
analyses
•
•
An automatic segmentation method based on Geodesic
Active Regions (GAR) and an image standardization
technique is used
The method:
– eliminates most of the dependency on the operator, and on
the specific imaging protocols and equipment used.
– is able to segment (extract the surface mesh) a region of
interest with a size of 2563 voxels in 17+4 min (avg+std dev)
on a PC (Intel quad-core, 2.4 GHz, 4GB memory).
Medical
image
from
imaging
equipment
Surface
mesh
a<er
segmenta6on
Hernandez,
M.
et
al.
2007
Non-‐parametric
geodesic
acGve
regions:
method
and
evaluaGon
for
cerebral
aneurysms
segmentaGon
in
3DRA
and
CTA.
Med.
Image
Anal.
11,
224–241.
Bogunovic,
H.
et
al.
2011
Automated
segmentaGon
of
cerebral
vasculature
with
aneurysms
in
3DRA
and
TOF
MRA
using
geodesic
acGve
regions:
an
evaluaGon
study.
Med.
Phys.
38,
210–222.
49. Mesh Editing
A
surface
mesh
represenGng
the
vascular
geometry
is
required
to
perform
the
@neurIST
morphological
and
hemodynamic
analyses
• The surface mesh obtained after the GAR
segmentation needs to be manually manipulated by
an operator to either remove or correct:
– some of the artifacts not belonging to the cerebral vasculature
– those parts of the geometry not relevant for the subsequent
analyses (morphological or hemodynamic).
P
O
Kissing
vessels
Surface
mesh
a<er
segmenta6on
Remove
cells
Close
holes
Surface
mesh
a<er
correc6on
and
aneurysm
isola6on
50. Several
morphological
measurements
are
based
on
the
aneurysm
sac,
to
idenGfy
it,
the
aneurysm
neck
is
required
• User is asked to manually delineate the neck
• Unfortunately, automatic methods are not an option
because there are:
– unacceptable differences in Neck
DelineaGon
cases among
a large number of
methods and manual selection of experts
– too complex vascular topologies where there is not even an
agreement among experts about where the aneurysm neck is
A
u
t
o
m
a
t
ic
vs
Manual
delinea6on
Surface
mesh
a<er
correc6on
and
aneurysm
isola6on
Too
large
differences
in
performance
and
lack
of
consensus
Too
complex
vascular
topologies
Manual
aneurysm
neck
selec6on
51. Morphological Descriptors
Morphological
descriptors
of
various
complexity
are
automaGcally
extracted
and
stored
for
their
subsequent
analysis
• Among the wide variety of existing morphological
descriptors, @neurIST chose to compute:
– Basic size indices describing the aneurysm sac: aspect ratio,
non-sphericity index, aneurysm volume and surface area.
– Complex indices describing the sac and a portion of the
surrounding vasculature: Zernike moment invariants (volume
and surface-based).
depth
Basic
size
indices
describing
aneurysm
sac
neck
Complex
indices
(Zernike
moment
invariants)
Manual
aneurysm
neck
selec6on
@neurIST
morphological
descriptors
Ujiie,
H.
et
al.
1999
Effects
of
size
and
shape
(aspect
raGo).
Neurosurgery
45,
119–130.
/
Ma,
B.,
Harbaugh,
R.
E.
&
Raghavan,
M.
L.
2004
Three-‐dimensional
geometrical
characterizaGon
of
cerebral
aneurysms.
Ann.
Biomed.
Eng.
32,
264–273.
/
Raghavan,
M.
L.,
Ma,
B.
&
Harbaugh,
R.
E.
2005
QuanGfied
aneurysm
shape
and
rupture
risk.
J.
Neurosurg.
102,
355–362.
Pozo,
J.
M.
et
al.
2011
Efficient
3D
Geometric
and
Zernike
moments
computaGon
from
unstructured
surface
meshes.
IEEE
Trans.
Pagern
Anal.
Machine
Intell.
33,
471–484.
52. Morphological Analysis Workflow
depth
Basic
size
indices
describing
aneurysm
sac
Medical
image
from
imaging
equipment
neck
@neurIST
morphological
descriptors
Complex
indices
(Zernike
moment
invariants)