Title: Semantic Engineering Notation
1With Semantic Analysis from Noun Phrases to
SNOMED CT and Classification Codes Hans Rudolf
Straub Semfinder AG, Switzerland
2 Reality Instances
? Free Text ? Terminologies ? Classifications ? Gr
oups / Statistics
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
3 InformationContent per Case (in Bits)
Reality Instances
? Free Text ? Terminologies ? Classifications ? Gr
oups / Statistics
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
? ?
.
.
.
.
.
. .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
(512k) 19 100
.
.
.
(16k) 14 28
(512) 9
4 Reality Instances
? Free Text ? Terminologies ? Classifications ? Gr
oups / Statistics
.
.
.
.
.
.
.
- Information
- Reduction is
- Enormous- Inevitable- Ubiquitous
- Necessary
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
5 Reality Instances
.
.
.
.
.
.
.
- Information
- Reduction is
- Enormous- Inevitable- Ubiquitous
- Necessary
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. .
Simplification is necessary and useful
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Overviews Statistics Comparisons Basis for Actions
6 Reality Instances
.
.
.
.
.
.
.
- Information
- Reduction is
- Enormous- Inevitable- Ubiquitous
- Necessary
- Deducible?
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. .
Simplification is necessary and useful
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Overviews Statistics Comparisons Basis for Actions
7Reality Instances
Deducible?
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. .
Simplification is necessary and useful
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Overviews Statistics Comparisons Basis for Actions
Hierarchy?
8When is a Hierarchy a True Hierarchy?
1) Disjunctivity ? Species
Barrier2) Unidirectionality ? Time Arrow
Deducible!
Hierarchy
9Forms any Ontology of Diseases a True Hierarchy?
1) Disjunctivity ? Species Barrier
for Diseases? 2) Unidirectionality ? Natural
Time Arrow?
Deducible?
10 Problems with Hierarchies for Diseases 1)
Inherent Multidimensionality (Combinatorial
Explosion) 2) Sum/Summands Problem
11ICD-10 Several Dimensions in 1 Hierarchy
organic complications
causes
causes
I10 primary hypertension
I11 hypertensive cardiopathy
I12 hypertensive nephropathy
I13 hypertensive cardiop. nephrop.
I15 secondary hypertension
3-digits
I10 I15 arterial hypertension
group
12ICD-10 Several Dimensions in 1 Hierarchy
5-digits
I11.00without crisis
I11.01with crisis
I15.20without crisis
I15.21with crisis
...
...
...
I11.0with cardiac insufficiency
I11.9without card. Insuff.
I12.0with renal insufficiency
I12.9without ren. insuff.
.. H
.. N
I13.2 HN
.. -
4-digits
I15.0reno- vasc.
I15.1reno- paren.
I15.2 endo- crinous
I10.0benign
I10.1malignant
I10 primary hypertension
I11 hypertensive cardiopathy
I12 hypertensive nephropathy
I13 hypertensive cardiop. nephrop.
I15 secondary hypertension
3-digits
I10 I15 arterial hypertension
group
13Representation of 2 Axes ( 2 Hierarchies)
Germ
Bacterium
Germ
Virus
Organ
2 Choices 2 Degrees of Freedom
Respiratory Organ
Digestive Organ
Organ
2 Semantic Axes 2 Degrees of Freedom
1412 Hierarchies
Germ
Certainty
Localisation
Bakterium
Progression
Germ
Virus
Organ
Respiratory Organ
Digestive Organ
Organ
12 Hierarchies cannot be Represented
Simultaneously
15Encapsulated Axes
Pathogenesis
Localisation
Organ
Certainty
Root Focus
Pathology
Focus
e.g. Fractur
29 Axes (8 Active)
16Concept Molecule
StructuredRelationship Group (Postcoordinated)
17Concept Molecules From Texts to Semantics
18Sum/Summands Problem
Diagnosis of its Own
I11.0with cardiac insufficiency
I11.9without card. Insuff.
I12.0with renal insufficiency
I12.9without ren. insuff.
.. H
.. N
I13.2 HN
.. -
4-digits
I15.0reno- vasc.
I15.1reno- paren.
I15.2 endo- crinous
I10.0benign
I10.1malignant
I10 primary hypertension
I11 hypertensive cardiopathy
I12 hypertensive nephropathy
I13 hypertensive cardiop. nephrop.
I15 secondary hypertension
3-digits
I10 I15 arterial hypertension
group
Attribute of a Diagnosis
19Sum/Summands Problem
- Autom. Included?
- Additional Item?
- Sequence?
Diagnosis of its Own
? I50.00 I50.19
I11.0with cardiac insufficiency
I11.9without card. Insuff.
I12.0with renal insufficiency
I12.9without ren. insuff.
.. H
.. N
I13.2 HN
.. -
4-digits
I15.0reno- vasc.
I15.1reno- paren.
I15.2 endo- crinous
I10.0benign
I10.1malignant
I10 primary hypertension
I11 hypertensive cardiopathy
I12 hypertensive nephropathy
I13 hypertensive cardiop. nephrop.
I15 secondary hypertension
3-digits
I10 I15 arterial hypertension
group
20Sum/Summands Problem
? Very Frequent!
Diagnosis of its Own
I11.0with cardiac insufficiency
I11.9without card. Insuff.
I12.0with renal insufficiency
I12.9without ren. insuff.
.. H
.. N
I13.2 HN
.. -
4-digits
I15.0reno- vasc.
I15.1reno- paren.
I15.2 endo- crinous
I10.0benign
I10.1malignant
I10 primary hypertension
I11 hypertensive cardiopathy
I12 hypertensive nephropathy
I13 hypertensive cardiop. nephrop.
I15 secondary hypertension
3-digits
I10 I15 arterial hypertension
group
21Sum/Summands Problem
Severel Diseases can Coexist and Interrelate in
the Same Patient
Which of them Defines the Medical Case? Is
there a "Hat" Disease? ? Classification
22Sum/Summands Problem
I10
N28.9
I12
Give Sum
23Sum/Summands Problem
I12
N19
N19 I12
Give Summands
24Sum/Summands Problem
I12 Hypertensive renal disease Includes
any condition in N18.- (renal failure)
with any condition in I10 arteriosclerosis of
kidney, Excludes secondary hypertension
(I15.-)I12.0 Hypertensive renal disease with
renal failure hypertensive renal
failureI12.1 Hypertens. renal disease without
renal failure hypertensive renal disease NOS
I12 Hypertensive renal disease Includes
any condition in N18.- (renal failure)
with any condition in I10 arteriosclerosis of
kidney, Excludes secondary hypertension
(I15.-)I12.0 Hypertensive renal disease with
renal failure hypertensive renal
failureI12.1 Hypertens. renal disease without
renal failure hypertensive renal disease NOS
I12 Hypertensive renal disease Includes
any condition in N18.- (renal failure)
with any condition in I10 arteriosclerosis of
kidney, Excludes secondary hypertension
(I15.-)I12.0 Hypertensive renal disease with
renal failure hypertensive renal
failureI12.1 Hypertens. renal disease without
renal failure hypertensive renal disease NOS
25Which Summand is Stronger?
Fine Granularity (Summands)
Deductionspossible?
Coarse Granularity (Sum)
Classification Funnel
Diagnoses Terminologies Classifications
(Hierarchy?)
26Which Summand is Stronger?
hypertension
nephropathy
Summands
Classification Funnel
27Which Summand is Stronger?
hypertension
nephropathy
Summands
diabetic Nethie
athero- Nepthie
Sum with Full Information
hypertensive nephropathy
Classification Funnel
28Which Summand is Stronger?
hypertension
nephropathy
Summands
diabetic Nethie
athero- Nepthie
Sum with Full Information
hypertensive nephropathy
Sum with Reduced Information
nephropathy
Classification Funnel
N
29Which Summand is Stronger?
hypertension
nephropathy
Summands
Hypertonie mit Kardiopathie
Hypertonie mit Kardiopathie
Sum with Full Information
hypertension with nephropathy
Sum with Reduced Information
Classification Funnel
30Which Summand is Stronger?
hypertension
nephropathy
Summands
Hypertonie mit Kardiopathie
Hypertonie mit Kardiopathie
Sum with Full Information
hypertension with nephropathy
Sum with Reduced Information
hypertension
Classification Funnel
I
31Which Summand is Stronger?
hypertension
nephropathy
Summands
diabetic Nethie
Hypertonie mit Kardiopathie
athero- Nepthie
Hypertonie mit Kardiopathie
Sum with Full Information
hypertensive nephropathy
hypertension with nephropathy
Sum with Reduced Information
hypertension
nephropathy
Classification Funnel
I
N
Disjunction
32Sum/Summands Normal part_of Relation?
Finger part_of Hand ? Never on Same
Level - Natural Hierarchy
Hypertension part_of (Hypertensive)
Cardiopathy ? Possibly on Same Level! -
Hypertension Can Stand Alone - No Natural
Hierarchy
33Sum/Summands Normal part_of Relation?
Finger part_of Hand ? Never on Same
Level - Natural Hierarchy
Hypertension part_of (Hypertensive)
Cardiopathy ? Possibly on Same Level! -
Hypertension Can Stand Alone - No Natural
Hierarchy
34Sum/Summands Normal part_of Relation?
? mandatory part_of ?
contingent part_of
Finger part_of Hand (Always) Finger without
Hand (Not Naturally Possible)
Hypertension part_of Hypert. Cardiopathy
(Sometimes) Hypertension without Hypert.
Cardiopathy (Well Possible)
35Sum/Summands Normal part_of Relation?
? mandatory part_of ?
contingent part_of
hand
palm
thumb
...
deductions possible
hypertensive cardiopathy
hypertension
cardiopathy
deductions NOT possible
36The sum/summands problem is frequent
- duodenal ulcer with gastrointestinal
bleeding - diabetes with complications -
diabetic nephropathy - diabetic foot - cardiac
lung oedema ( symptom cardiac insufficiency) - ce
rebral aneurysm with subarachnoidal bleeding and
coma- monteggia-fracture ( fracture of ulna und
luxation of radius) - Weber fracture (tibia
fibula ligaments) - unhappy triad (meniscus 2
ligaments) - appendicitis with perforation and
peritonitis
37 Reality Instances
Sum/Summands Problem
? Free Text ? Terminologies ? Classifications ? Gr
oups / Statistics
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. .
.
.
.
.
.
Many Points
.
.
.
.
.
.
.
.
.
.
.
.
?
.
.
One Point
Simplification is Intended!
38 Reality Instances
? Free Text ? Terminologies ? Classifications ? Gr
oups / Statistics
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Severel Classifications of One Case are Possible
.
.
.
.
.
.
39From Noun Phrases to Terms and ICD-Codes
Automated Coding System
Free TextDiagnoses
Structured Representationof Semantics
ICD-10Codes
(In Daily Routine in 180 Hospitals)
40From Noun Phrases to Terms and ICD-Codes
Automated Coding System
Free TextDiagnoses
Understanding Texts
Structured Representationof Semantics
Understanding Codesand Coding Guidelines
ICD-10Codes
(In Daily Routine in 180 Hospitals)
41From Noun Phrases to Terms and ICD-Codes
Automated Coding System
Free TextDiagnoses
Completion Queries
Understanding Texts
Structured Representationof Semantics
Understanding Codesand Coding Guidelines
ICD-10Codes
(In Daily Routine in 180 Hospitals)
42From Noun Phrases to Terms and ICD-Codes
Automated Coding System
Free TextDiagnoses
Completion Queries
Understanding Texts
Structured Representationof Semantics
Understanding Codesand Coding Guidelines
ICD-10Codes
(In Daily Routine in 180 Hospitals)
43From Noun Phrases to Terms and ICD-Codes
SNOMED CT
Automated Coding System
Free TextDiagnoses
Completion Queries
Data Entry Tools
Precoord. Terms?Comb. Explosion?Vagueness? Compl
eteness?
Understanding Texts
Structured Representationof Semantics
StandardTerms
Mapping Tables
Understanding Codesand Coding Guidelines
ICD-10Codes
Unambiguous? Many-To-One?Many-To-Many?
(In Daily Routine in 180 Hospitals)
44Are 2 Systems of Data Entry Necessary?
SNOMED CT
Automated Coding System
Free TextDiagnoses
Completion Queries
Understanding Texts
Structured Representationof Semantics
StandardTerms
StandardTerms
Data Exchange
Understanding Codesand Coding Guidelines
ICD-10Codes
ICD-10 Coding will continue
45Are 2 Systems of Data Entry Necessary?
SNOMED CT
Automated Coding System
Free TextDiagnoses
Completion Queries
Understanding Texts
Structured Representationof Semantics
StandardTerms
StandardTerms
Free
Data Exchange
Understanding Codesand Coding Guidelines
ICD-10Codes
ICD-10 Coding will continue
46From Noun Phrases to Terms and ICD-Codes
SNOMED CT
Automated Coding System
Free TextDiagnoses
Completion Queries
Understanding Texts
Structured Representationof Semantics
StandardTerms
StandardTerms
Free
Data Exchange
Understanding Codesand Coding Guidelines
ICD-10Codes
Fixed Standard for Data Exchange
Open Method
47Summary
A description of a coarser granularity cannot
automatically be obtained from a description of
finer granularity. Diagnoses resist to be
arranged in form of a hierarchy due to the
inherent multidimensionality and to the
sum/summands-problem. Simplifying, grouping and
classification of medical diagnosis will always
be necessary in practice. When automatically
assigning ICD-10 codes to medical cases, their
SNOMED clinical terms can be assigned
simultaneously.
48(No Transcript)