MarcCUGGIA,SaharBayat,DelphineRossille,PatricePoulain, - PowerPoint PPT Presentation

Marc
CUGGIA,
Sahar
Bayat,
Delphine
Rossille,
Patrice
Poulain,

 Patrick
Pladys,
Régis
Duvauferrier
 Inserm
U936

 Conceptual
Modelling
of
Biomedical
Knowledge
 Faculté
de
Médecine
–
Rennes
‐
France


IntroducKon
 Seman&c
Interoperability
 Knowledge
 Models
 InformaKon
 Models
 InformaKon
 InformaKon
 System
B 
 System
A 


IntroducKon
 Terminologies
 Ontologies
 ICD,
RxNorm,
 • Terms
et
concepts
of
the
domain
 Knowledge
 Snomed,
 • DefiniKons
:
ontologies
 Models
 LOINC…
 Interoperability
 standards
 InformaKon
 • InformaKon
structure
 HL7 ,
EN13606,
 Models
 • 
messages
or
documents
 OpenEHR ,
DICOM,
 HPRIM…


IntroducKon
 Terminology/ontolology
 Knowledge
 Models
 RepresentaKon
 Interface
 formalism
?
 InformaKon
 Data
element
 Models
 Rector,
A.
L.
(2001).
"The
Interface
between
 Informa;on,
Terminology,
and
Inference
 Models."
STUDIES
IN
HEALTH
TECHNOLOGY
 AND
INFORMATICS:
246‐250.


ObjecKves 
 To
Compare
two
relevant
informaKon
standards


 HL7
 and 
Open
EHR
 according
to
 ‐
The
informa&on
representa&on
formalism

 ‐
The
interface
with
the
knowledge
models


Materiel
and
Method 
 Virginia
Apgar
 • Apgar
score
 :
To
assess
the
health
of
newborns
just
 a\er
childbirth
 • Score
range
0
to
10
 • Measured
at
1,
5,
10
minutes
 Score
value
 Pulse
rate
 Breathing
 Muscle
Tone
 Reflex
irritability
 Skin
color
 No
response
to
 0
 <80 
 absent 
 none 
 cyanosis 
 sKmulaKon 
 Weak
or
 Grimace
cry
 1
 80
‐
100 
 Some
flexion 
 acrocyanosis 
 irregular 
 when
sKmulated 
 AcKve
 2
 >
100 
 strong 
 vive 
 pink 
 movement 


HL7
v3
 Act Role Link Reference
 Relationship InformaKon
 Model
 (RIM)
 Act Entity Role Participation Domain
 Message
 InformaKon
 DMIM
 refinement
 Model
 (DMIM)
 Messages
:
 Documents
structures
 Business
process
 (CDA
&
Template)


OpenEHR
 Reference
 Model
 Refinement
 Archetypes
 ECG
 APGAR
Score
 Weight
 MenstruaKon
 report
 Pulse
rate
=  
 Breathing=  
 Birth
 Adult
 Menstrual
 Muscle
Tone
 =  
 Pain
 Reflex=
  
 weight
 weight
 Cycle
 Skin
Color=  
 inheritance
 composiKon


Material
and
Method 
 Archetype
Apgar
score
(2008)
 published
by
(2008)
 
«
APGAR
score
»
part
 
of
the
 perinatal
DMIM
published
by
 Goosen
(2005)


Method
 • EvaluaKon
criteria:
 – What
is
the
degree
of
formalism
of
each
model
?
 – Is
the
context
expressed
in
the
model
?
 – Could
we
use
the
model
in
another
context
?
 – How
is
the
binding
with
knowledge
models


Birth
 Weight
 APGAR
 Birth
data
 SCORE
 Composi&on 


No
birth
or
perinatal
 context
in
the
OpenEHR
 Taxonomy


No
link
between
 pulse
rate
and
apgar
 score
archetype


Header
:
archetype
metadata
:

 purpose,
use,
authors
(free
text)
 Data
DefiniKon
 Temporality
 No
hierarchy
between
the
 data
elements
(e.g
TOTAL)


Study
of
the
InformaKon
Models

 
 ComparaKve
analyses
of
the
representaKons 

 Apgar
Score
 HL7
V3
 OpenEHR
 • 
lack
of
organizaKon
in
the
 • 
Formal
structure
 Formalism
 archetypes
hierarchy
 • 
Few
defined
relaKonships
 degree
 • 
Complexity
 between
the
data
elements
 • 
limited
 Reusability
 • 

Modularity
like
Lego
Bricks
 • 
Possibility
of
modularizaKon
 (CMET
:

generic
model
of
score)
 Context

 • 

Context
is
“embedded”
 • 
No
relaKonship
to
the
context
 Binding
 • 2
data
linked
to
LOINC
 • 
2
data
linked
to
LOINC


The
terminology
binding
 Knowledge
 LOINC
 Models
 Muscle
Tonus
 Informa&on
 Muscle
tone=
<
 Model
 Act Role Link Relationsh ip RIM
 Entity Role Act Participatio n

How
to
handle
the
binding
to
 different
Knowledge
Models?
 Knowledge
 LOINC
 SNOMED
 Models
 Observable
 Finding
 Muscle
Tone
 Muscle
Tone
 MuscleTone
 Informa&on
 Muscle
tone=
<
 Model
 Act Role Link Relationsh ip RIM
 Entity Role Act Participatio n

How
to
handle
the
binding
to
 different
Knowledge
Models?
 Knowledge
 LOINC
 SNOMED
 Models
 Observable
 Finding
 Muscle
Tone
 Muscle
Tone
 MuscleTone
 GREY
ZONE
 Informa&on
 Muscle
tone=
<
 Model
 Act Role Link Relationsh ip RIM
 Entity Role Act Participatio n

Discussion
&
conclusion
 • Limits
:
1
example
but
which
illustrates
generic
 problems
 • Gap
between
informaKon
and
knowledge
model
:
 – 
QuanKtaKve
aspect
  
to
increase
the
binding
 – QualitaKve
aspect
:
 • “Easy”
to
solve
if
terminology
is
considered
as
a
flat
 model
(TermInfo
IniKaKve)
 • Need
to
re‐think
the
modeling
process
of
both
models
 in
a
same
Kme 
 • More
reusable
structures
of
informaKon
and
a
beoer
 organizaKon
of
data
elements.