Natural Language Processing

advertisement
Applying Natural Language
Processing in the Clinical Setting
Peter Haug, MD
Homer Warner Center for Informatics Research
Intermountain Healthcare, Salt Lake City, Utah
Applied NLP Research
•
•
Goals:
– Affect Care Delivery
• Extract Clinical Data from Medical
Documents
• Use Extracted Data to Alter Care
– Improve Documentation
Examples:
– Support for Diagnostic Systems
• Screening for Disease
• Assess Risk
• Triggering Orders
• Activating Clinical Protocols
• Identify Necessary Data
– Encode Admit Diagnoses
• Identify Eligible Patients
– Identify Patients for Trauma Registry
– Support Clinical Research
• Data Extraction
• Phenotype Recognition
– Complete Problem List
– Identify Patients for Research
Recruitment
• Research Alerting
– Improve Administrative Data
• Improve Data for Business Planning
• Improve Data for Billing
2
Process Begins with a Document
Document
Parsing Process
•Find Document
Structure
•Extract MetaData
•Plan Further
Parsing
•Find Sentence
Structure
•Determine
Meaning
(Concepts)
•Store/Process
Concepts
Chest Xray Report
History: Cough and fever. Previous history of right-sided
pneumonia.
Exam: PA and Lateral Chest Film.
Observations: Prior films showed confluent opacification of the
RLL. This finding remains in today’s exam. These opacities, seen
in multiple previous films, have spread to the right and left upper
lobes.
Interpretation: Extension of previously diagnosed pneumonia to
right and left upper lobes.
A Sequence of Processing
• Find Document Structure
• Find Section Structure
• Find Sentence Structure
• Determine Meaning (Semantics)
– Map to Concepts
– Build Data Structures
Document
Document
Sections
Find Internal
Structure
Structured
Document
Planner
Output
Output
Document
Parsing Process
Find
Document
Sections
•Find Document
XML Based Output
XML Based
<Body>Prior films showed
confluent opacification
ofOutput
the RLL.
Structure
<Doc>Chest Xray Report</Doc>
Xray Report</Doc>
This finding remains in today’s exam. These<Doc>Chest
opacities,
seen in
<Sec>History</Sec>
<Body>Cough and fever. Previous
history of right-sided pneumonia.</Body>
<Sec>Exam</Sec>
<Body>PA and Lateral Chest Film. </
Body>
<Sec>Observations</Sec>
<Body>Prior films showed
confluent opacification of the RLL.
This finding remains in today’s exam.
These opacities, seen in multiple
previous films, have spread to the right
and left upper lobes. </Body>
<Sec>Interpretation</Sec>
……………………..
</Doc>
<Sec>History</Sec>
<Body>
<Sen>Cough and fever.</Sen>
<Sen>Previous history of right-sided
pneumonia.</Sen>
</Body>
<Sec>Exam</Sec>
<Body>
<Sen>PA and Lateral Chest Film.</Sen>
</Body>
<Sec>Observations</Sec>
<Body>
<Sen>Prior films showed confluent
opacification of the RLL.</Sen>
<Sen>This finding remains in today’s
exam.</Sen>
<Sen>These opacities, seen in
multiple previous films, have spread to the
right and left upper lobes.</Sen>
</Body>
…………..
</Doc>
multiple
•Extract
Meta- previous films, have spread to the right and left
Data upper lobes.</Body>
•Plan Further
Parsing
•Find Sentence
Structure
<Sen>These opacities, seen in multiple previous films, have
spread to the right and left upper lobes.</Sen>
•Determine
Meaning
(Concepts)
•Store/Process
Concepts
Build Structures for:
1) Document Type
2) Sections
3) Paragraphs
4) Sentences
5) Tables
6) Internal Markup
7) Images
8) Lists
9) Images??
Choose Models to Parse Against:
1) Rules to determine which group
of concepts to find based on
Document Type and Sections
Present
Sentences
Structured
Document
Planner
MetaData
Determine
Sentence
Structure
Linked Phrasal
Components
Determine
Semantics
Output
Document
Parsing Process
Includes Document
Type and Section
Information.
XML Based Output
•Find Structures
<Doc>Chest Xray Report</Doc>
…………..
<Sec>Observations</Sec>
<Body>
<Sen><Coded Index=”1">Prior films
showed confluent opacification of the
RLL.</Coded></Sen>
<Sen><Coded Index=”2">This finding
remains in today’s exam.</Coded></Sen>
<Sen><Coded Index=”3">These
opacities, seen in multiple previous films,
have spread to the right and left upper
lobes.</Coded></Sen>
</Body>
…………..
<Codes>
<code ref=”1 Term=”Snomed”>2345</code>
<code ref=”2 Term=”Snomed”>3456</code>
<code ref=”3a Term=”Snomed”>1234</code>
<code ref=”3b Term=”Snomed”>1235</code>
…………..
</Doc>
•Extract MetaData
•Plan Further
Parsing
Sen><Coded Index=”3">These opacities, seen in multiple previous
films, have spread to the right and left upper lobes.</Coded></Sen>
•Find Sentence
Structure
•Determine
Meaning
(Concepts)
•Store/Process
Concepts
<code ref=”3a” Terminology=”TermX?”>1234</code>
<code ref=”3b” Terminology=”TermX?”>1235</code>
Choose Models to Parse Against:
1) Rules to determine which group
of concepts to find based on
Document Type and Sections
Present
Coded Findings:
Localized Infiltrate-RUL
Localized Infiltrate-LUL
Data Base Storage
Structured Data
Storage
These opacities, seen in multiple views, have spread to the right and left upper lobes.
Semantic
Lexicon: Derived
Sentence
Parsing Process
•Categorize
Words
•Extract
Concepts
Etc.
Phrasal Grammar
Compositional
Grammar
Output
•Test for
Semantic
Congruence
Output
•Find
Relationships
Among Phrases
Output
•Identify Phrasal
Boundaries
Restrictions
from SpecialistAdj.
these=>POS:
Phrasal Grammar
(ME&E Restrictions,
Lexicon??
And
Semantic
Rep: N/A
Conjunctive
semantic KB
Restrictions)
opacities=>POS: Noun
Num; Pleural
Semantic Rep: RadFind.finding.opacities
ID: 1
Identify
Identify
verbSentence/Word Properties
Phrasal Info
Sentence seen=>POS:
Word
Candidate
Num;
N/A
Properties
Phrases
Semantic Rep: N/A
………………………...
right=>POS: adj.
POS: Noun
these=>POS: Adj.
P1: these opacities: pntr- ID:1
Num; Singular
Semantic Rep: N/A
P2: in multiple
Semantic Rep: RadFind.side.right
opacities=>POS: Noun
P3: previous flims
Num; Pleural
P4: In multiple previous films
ID:
2
Semantic Rep: RadFind.finding.opacities
P5: to the right: pntr- ID:2
ID: 1
left=>POS:
adj.
P6: to the left pntr- ID:3
seen=>POS: verb
P7: upper lobes pntr- ID:4,5
POS: Noun
Num; N/A
P8: to the right upper lobes pntrSemantic Rep: N/A
Num;
Singular
ID:2,4,5
………………………...
P9: to the left upper lobes pntrright=>POS: adj.
Semantic Rep: RadFind.side.left
ID:3,4,5
POS: Noun
ID:
3
Num; Singular
Etc.
Semantic Rep: RadFind.side.right
upper=>POS:
adj.
ID: 2
Num; N/A
left=>POS: adj.
POS: Noun
Semantic Rep: RadFind.sup_inf.upper
Num; Singular
Semantic Rep: RadFind.side.left
ID: 4
ID: 3
Etc.
lobes=>POS:
noun
upper=>POS:
adj.
Num; N/A
Num; pleural
Semantic Rep: RadFind.sup_inf.upper
ID: 4
Semantic Rep: RadFind.anat_loc.lobes
lobes=>POS: noun
ID: 5
Num; pleural
Semantic
Rep:
RadFind.anat_loc.lobes
Etc.
ID: 5
these opacities
these opacities
to the right
these opacities
to the left
tothese
the
left upper
opacities
these opacities
lobes
to the right upper
lobes
these opacities
upper lobes
to the left upper
lobes
These opacities, seen in multiple views, have spread to the right and left upper lobes.
Phrasal
Grammar
Document
Parsing Process
From Words to Coded Findings:
Localized Infiltrate-RUL
Localized Infiltrate-LUL
Structured
Data
Storage
Compositional
these
opacities
Grammar
•Categorize
Words
Output
•Identify Phrasal
Boundaries
these opacities
•Find
Relationships
Among Phrases
s_o bservatio ns
these opacities
to the left upper
lobes
to the right
to the left
c pneumothorax
c enlargement of the heart
c abnormality
c ambiguity
c lclzd prnchyml bnrmlty ns
c generalized infiltrate
c localized infiltrate
c bilateral pleural effusion
c nlrgmnt f th plmnry vssls
c generalized volume reducti...
c parenchymal abnormality no
c pulmonary nodular lesion
c mediastinal widening
c pleural thickening
c generic infiltrate nos
c pleural effusion ns
c bony abnormality
c consolidation nos
c localized volume reduction
c vlm rfrnc t prvs rcrd
other-
s_to pic_co ncept
c enlargement
12.7
c air density
11.5
c parenchymal abnormality ns 10.6
c prly mrgntd pcty nfltrt
9.17
c abnormality
8.73
c fluid density
7.94
c volume reduction
4.78
c see previous event
4.19
c consolidation
3.99
c smll wll mrgntd pcty ndl
3.56
other22.8
to pic_mo difier
null
nodular adj
accumulation n
increased
focal adj
hazy adj
patchy adj
active adj
degenerative adj
linear adj
other-
79.3
1.98
1.66
1.58
1.30
0.67
0.59
0.51
0.51
0.51
11.4
9.76
9.48
5.97
5.37
5.18
4.98
4.35
3.87
3.24
2.77
2.73
2.73
2.69
2.61
2.33
2.21
2.13
2.09
1.94
1.86
21.7
s_anato mic_co ncept
c bilateral lung
c heart
c interpleural space ns
c see previous event
c bilateral interpleural space
c null
c mediastinum
c left lower lobe
c pulmonary vessels
c pleura
c left upper lobe
c right lung
c left interpleural space
c right lower lobe
c all thoracic locations
c left lung
c bilateral lower lobes
c right upper lobe
c right interpleural space
c aorta
other-
16.0
9.56
8.93
5.49
4.74
4.50
4.46
4.35
3.00
2.84
2.73
2.57
2.57
2.49
2.25
2.21
1.90
1.78
1.62
1.58
14.4
anato mic_lo catio n_mo difier
null
otherwise adv
structures n
generalized adj
remainder adj
other-
22.3
10.3
6.32
5.85
4.03
3.64
3.32
2.65
2.25
1.94
37.4
93.0
0.71
0.40
0.36
0.28
5.22
s_o bservatio ns
16.2
15.7
8.95
8.31
7.12
5.77
5.42
4.03
3.08
2.46
23.0
s_to pic_co ncept
c enlargement
12.7
c air density
11.5
c parenchymal abnormality ns 10.6
c prly mrgntd pcty nfltrt
9.17
c abnormality
8.73
c fluid density
7.94
c volume reduction
4.78
c see previous event
4.19
c consolidation
3.99
c smll wll mrgntd pcty ndl
3.56
other22.8
s_anato mic_link_co ncept
c involving
c null
c throughout
c adjacent to
c projecting on
other-
null
in
of
at
throughout
other-
73.3
11.7
8.02
1.26
1.22
4.51
s_supero r_inferio r
s_side
86.5
8.34
4.07
0.75
0.12
0.20
c null
81.5
c upper 7.35
c middle 1.86
c lower 9.25
c null
61.4
c left
13.8
c right
10.6
c bilateral 11.3
c either 2.77
to pic_mo difier
null
left
right
bilateral
both
other-
null
nodular adj
accumulation n
increased
focal adj
hazy adj
patchy adj
active adj
degenerative adj
linear adj
other-
anato mic_mo difier_sup_inf_
anato mic_mo difier_side_
null
lower
upper
base
apical
other-
61.4
13.6
10.5
7.82
2.81
3.83
c pneumothorax
c enlargement of the heart
c abnormality
c ambiguity
c lclzd prnchyml bnrmlty ns
c generalized infiltrate
c localized infiltrate
c bilateral pleural effusion
c nlrgmnt f th plmnry vssls
c generalized volume reducti...
c parenchymal abnormality no
c pulmonary nodular lesion
c mediastinal widening
c pleural thickening
c generic infiltrate nos
c pleural effusion ns
c bony abnormality
c consolidation nos
c localized volume reduction
c vlm rfrnc t prvs rcrd
other-
anato mic_lo catio n_term
null
lung n
lobe n
pleural adj
interpleural adj
heart n
lungs n
cardiac adj
mediastinal adj
chest n
other-
to pic_term
null
air n
opacities n
opacity n
consolidation n
fluid n
infiltrate n
infiltrates n
effusion n
changes n
other-
anato mic_link_term
81.5
6.48
5.22
1.66
1.19
3.91
79.3
1.98
1.66
1.58
1.30
0.67
0.59
0.51
0.51
0.51
11.4
9.76
9.48
5.97
5.37
5.18
4.98
4.35
3.87
3.24
2.77
2.73
2.73
2.69
2.61
2.33
2.21
2.13
2.09
1.94
1.86
21.7
s_anato mic_co ncept
c bilateral lung
c heart
c interpleural space ns
c see previous event
c bilateral interpleural space
c null
c mediastinum
c left lower lobe
c pulmonary vessels
c pleura
c left upper lobe
c right lung
c left interpleural space
c right lower lobe
c all thoracic locations
c left lung
c bilateral lower lobes
c right upper lobe
c right interpleural space
c aorta
other-
•Test for
Semantic
Congruence
16.0
9.56
8.93
5.49
4.74
4.50
4.46
4.35
3.00
2.84
2.73
2.57
2.57
2.49
2.25
2.21
1.90
1.78
1.62
1.58
14.4
anato mic_lo catio n_mo difier
null
otherwise adv
structures n
generalized adj
remainder adj
other-
22.3
10.3
6.32
5.85
4.03
3.64
3.32
2.65
2.25
1.94
37.4
93.0
0.71
0.40
0.36
0.28
5.22
anato mic_lo catio n_term
null
lung n
lobe n
pleural adj
interpleural adj
heart n
lungs n
cardiac adj
mediastinal adj
chest n
other-
to pic_term
null
air n
opacities n
opacity n
consolidation n
fluid n
infiltrate n
infiltrates n
effusion n
changes n
other-
16.2
15.7
8.95
8.31
7.12
5.77
5.42
4.03
3.08
2.46
23.0
s_anato mic_link_co ncept
c involving
c null
c throughout
c adjacent to
c projecting on
other-
86.5
8.34
4.07
0.75
0.12
0.20
anato mic_link_term
these opacities
•Extract
Concepts
Concept
Extraction
Linked Phrasal Components
null
in
of
at
throughout
other-
73.3
11.7
8.02
1.26
1.22
4.51
s_supero r_inferio r
s_side
c null
81.5
c upper 7.35
c middle 1.86
c lower 9.25
c null
61.4
c left
13.8
c right
10.6
c bilateral 11.3
c either 2.77
anato mic_mo difier_sup_inf_
anato mic_mo difier_side_
null
left
right
bilateral
both
other-
null
lower
upper
base
apical
other-
61.4
13.6
10.5
7.82
2.81
3.83
81.5
6.48
5.22
1.66
1.19
3.91
these opacities
s_o bservatio ns
c pneumothorax
c enlargement of the heart
c abnormality
c ambiguity
c lclzd prnchyml bnrmlty ns
c generalized infiltrate
c localized infiltrate
c bilateral pleural effusion
c nlrgmnt f th plmnry vssls
c generalized volume reducti...
c parenchymal abnormality no
c pulmonary nodular lesion
c mediastinal widening
c pleural thickening
c generic infiltrate nos
c pleural effusion ns
c bony abnormality
c consolidation nos
c localized volume reduction
c vlm rfrnc t prvs rcrd
other-
s_to pic_co ncept
to the right upper
lobes
c enlargement
12.7
c air density
11.5
c parenchymal abnormality ns 10.6
c prly mrgntd pcty nfltrt
9.17
c abnormality
8.73
c fluid density
7.94
c volume reduction
4.78
c see previous event
4.19
c consolidation
3.99
c smll wll mrgntd pcty ndl
3.56
other22.8
to the left upper
lobes
to pic_mo difier
null
nodular adj
accumulation n
increased
focal adj
hazy adj
patchy adj
active adj
degenerative adj
linear adj
other-
79.3
1.98
1.66
1.58
1.30
0.67
0.59
0.51
0.51
0.51
11.4
9.76
9.48
5.97
5.37
5.18
4.98
4.35
3.87
3.24
2.77
2.73
2.73
2.69
2.61
2.33
2.21
2.13
2.09
1.94
1.86
21.7
16.0
9.56
8.93
5.49
4.74
4.50
4.46
4.35
3.00
2.84
2.73
2.57
2.57
2.49
2.25
2.21
1.90
1.78
1.62
1.58
14.4
22.3
10.3
6.32
5.85
4.03
3.64
3.32
2.65
2.25
1.94
37.4
s_o bservatio ns
anato mic_lo catio n_mo difier
null
otherwise adv
structures n
generalized adj
remainder adj
other-
16.2
15.7
8.95
8.31
7.12
5.77
5.42
4.03
3.08
2.46
23.0
s_to pic_co ncept
c enlargement
12.7
c air density
11.5
c parenchymal abnormality ns 10.6
c prly mrgntd pcty nfltrt
9.17
c abnormality
8.73
c fluid density
7.94
c volume reduction
4.78
c see previous event
4.19
c consolidation
3.99
c smll wll mrgntd pcty ndl
3.56
other22.8
s_anato mic_link_co ncept
c involving
c null
c throughout
c adjacent to
c projecting on
other-
86.5
8.34
4.07
0.75
0.12
0.20
anato mic_link_term
73.3
11.7
8.02
1.26
1.22
4.51
s_side
c null
61.4
c left
13.8
c right
10.6
c bilateral 11.3
c either 2.77
anato mic_mo difier_side_
null
left
right
bilateral
both
other-
61.4
13.6
10.5
7.82
2.81
3.83
c pneumothorax
c enlargement of the heart
c abnormality
c ambiguity
c lclzd prnchyml bnrmlty ns
c generalized infiltrate
c localized infiltrate
c bilateral pleural effusion
c nlrgmnt f th plmnry vssls
c generalized volume reducti...
c parenchymal abnormality no
c pulmonary nodular lesion
c mediastinal widening
c pleural thickening
c generic infiltrate nos
c pleural effusion ns
c bony abnormality
c consolidation nos
c localized volume reduction
c vlm rfrnc t prvs rcrd
other-
93.0
0.71
0.40
0.36
0.28
5.22
anato mic_lo catio n_term
null
lung n
lobe n
pleural adj
interpleural adj
heart n
lungs n
cardiac adj
mediastinal adj
chest n
other-
to pic_term
null
air n
opacities n
opacity n
consolidation n
fluid n
infiltrate n
infiltrates n
effusion n
changes n
other-
null
in
of
at
throughout
other-
these opacities
s_anato mic_co ncept
c bilateral lung
c heart
c interpleural space ns
c see previous event
c bilateral interpleural space
c null
c mediastinum
c left lower lobe
c pulmonary vessels
c pleura
c left upper lobe
c right lung
c left interpleural space
c right lower lobe
c all thoracic locations
c left lung
c bilateral lower lobes
c right upper lobe
c right interpleural space
c aorta
other-
s_supero r_inferio r
9.76
9.48
5.97
5.37
5.18
4.98
4.35
3.87
3.24
2.77
2.73
2.73
2.69
2.61
2.33
2.21
2.13
2.09
1.94
1.86
21.7
s_anato mic_co ncept
c bilateral lung
c heart
c interpleural space ns
c see previous event
c bilateral interpleural space
c null
c mediastinum
c left lower lobe
c pulmonary vessels
c pleura
c left upper lobe
c right lung
c left interpleural space
c right lower lobe
c all thoracic locations
c left lung
c bilateral lower lobes
c right upper lobe
c right interpleural space
c aorta
other-
16.0
9.56
8.93
5.49
4.74
4.50
4.46
4.35
3.00
2.84
2.73
2.57
2.57
2.49
2.25
2.21
1.90
1.78
1.62
1.58
14.4
anato mic_lo catio n_mo difier
null
otherwise adv
structures n
generalized adj
remainder adj
other-
anato mic_lo catio n_term
null
lung n
lobe n
pleural adj
interpleural adj
heart n
lungs n
cardiac adj
mediastinal adj
chest n
other-
c null
81.5
c upper 7.35
c middle 1.86
c lower 9.25
to pic_mo difier
anato mic_mo difier_sup_inf_
null
lower
upper
base
apical
other-
81.5
6.48
5.22
1.66
1.19
3.91
null
nodular adj
accumulation n
increased
focal adj
hazy adj
patchy adj
active adj
degenerative adj
linear adj
other-
79.3
1.98
1.66
1.58
1.30
0.67
0.59
0.51
0.51
0.51
11.4
to pic_term
null
air n
opacities n
opacity n
consolidation n
fluid n
infiltrate n
infiltrates n
effusion n
changes n
other-
22.3
10.3
6.32
5.85
4.03
3.64
3.32
2.65
2.25
1.94
37.4
Data Base
Storage
93.0
0.71
0.40
0.36
0.28
5.22
16.2
15.7
8.95
8.31
7.12
5.77
5.42
4.03
3.08
2.46
23.0
s_anato mic_link_co ncept
c involving
c null
c throughout
c adjacent to
c projecting on
other-
86.5
8.34
4.07
0.75
0.12
0.20
anato mic_link_term
null
in
of
at
throughout
other-
73.3
11.7
8.02
1.26
1.22
4.51
s_side
c null
61.4
c left
13.8
c right
10.6
c bilateral 11.3
c either 2.77
anato mic_mo difier_side_
null
left
right
bilateral
both
other-
61.4
13.6
10.5
7.82
2.81
3.83
s_supero r_inferio r
c null
81.5
c upper 7.35
c middle 1.86
c lower 9.25
anato mic_mo difier_sup_inf_
null
lower
upper
base
apical
other-
81.5
6.48
5.22
1.66
1.19
3.91
Etc.
Etc.
upper lobes
Semantic Tests
1) Model determines Semantic consistency of phrasal combinations.
2) “Best” combinations trigger concept extraction and storage.
Output of a Semantic Parse
PARSE: A hazy opacity is seen in the right upper lobe.
• Instantiated Event:
• 1001
*Overall Concept : *localized infiltrate (0.998669)
• 1002
*State Concept : *present (0.780993)
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
1003
1004
1005
1006
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
Presence Term : null (0.779583)
*Topic Concept : *poorly-marginated opacity (infiltrate) (1.0)
Topic Term : opacity~n (1.0)
Topic Modifier Term: hazy~adj. (1.0)
Topographic Location Term : null (0.588844)
*Severity Concept : *null (0.969009)
Severity Term : null (0.962739)
*Link Concept : involving (0.686011)
Topic Location Link Term : in (1.0)
*Anatomic Concept : *right upper lobe (1.0)
Anatomic Location Mod : null (0.9375)
Anatomic Location : lobe~n (1.0)
Anatomic Location Mod1 : right (1.0)
Anatomic Location Mod2 : upper (1.0)
Anatomic Location Mod3 : null (1.0)
Anatomic Location Mod4 : null (1.0)
Anatomic Location Mod5 : null (1.0)
Probabilistic Semantics
GlobalObservation
abnormality
0+
localized infiltrate
58.4
alveolar infiltrate nos 14.8
generic infiltrate nos 26.8
State
absent
28.8
present
64.2
possible
6.62
otherwise absent 0.39
state_term
null
62.3
no
18.7
no evidence
4.67
with no
4.28
or
2.33
particularly
1.56
without
0.78
possibly
0.78
and or
0.78
most likely
0.78
other3.12
topic_term
abcess
.002
abnormalities
.002
consolidation
31.1
infiltrates
18.3
opacity
10.5
visualization
0.39
infiltrate
30.7
opacification
0.39
opacities
2.73
densities
0.39
consolidated
0.39
air bronchogram
0.39
kerley lines
0.39
septal lines
0.39
changes
1.17
lesion
0.39
null
0.39
air
0.39
change
0.39
density
1.17
Observation
localized infiltrate
53.3
infiltrate nos
43.2
generalized infiltrate 1.95
interstitial infiltrate
0.78
inflammatory lesion 0.39
other0.39
topic_modifier
null
69.6
focal
6.61
superimposed
0.39
hazy
3.50
mildly coarse pattern
0.39
confluent
0.78
infiltrative
3.11
mixed pattern
0.39
dense
1.56
patchy
1.95
areas of
0.78
area
0.39
reticular
0.39
nodular
0.78
inflammatory
0.38
ill defined
0.78
indistinct
1.17
poorly marginated
1.17
slightly coarse
0.39
bronchogram
0.39
underlying
0.39
fluffy
1.95
ill defined
1.95
poorly defined
0.39
vague
0.39
Location
bilateral lung
26.5
left lower lobe
14.0
null
10.1
bilateral lower lobes
8.17
left upper lobe
7.39
other33.9
anatomic_location_term
pulmonary
3.51
null
37.7
lung
15.5
lungs
3.89
hemithorax
0.78
lobes
3.89
perivascular
0.40
lobe
28.4
lung fields
0.40
chest
0.40
location
0.40
infrahilar
0.39
side
0.40
heart
1.17
retrocardiac
0.78
lingula
1.95
Modifier_Side
null
52.9
left
21.8
right
16.3
bilateral
8.57
either
0.40
anatomic_modifier_side_
null
52.5
either
0.40
right
16.3
bilaterally
0.78
left
22.2
bilateral
3.50
both
4.28
Modifier_Sup_Inf
null
58.7
ascending
.007
descending
.007
superior
11.7
middle
2.73
inferior
26.8
anatomic_modifier_sup_inf_
null
58.7
lower
22.6
upper
10.9
base
1.95
apex
0.39
cranial
0.39
bases
1.56
mid
1.17
basilar
0.78
middle
1.56
Modifier_Cntrl_Prph_
null
97.3
central
1.95
peripheral
0.79
anatomic_mdfr_cntrl_prph_
null
97.3
peripheral
0.39
central
1.95
peripherally
0.39
Example: NLP in Pneumonia
(a computer-based intervention)
• Goal:
– Identify Pneumonia Patients in the ED Rapidly
– Assess Risk
– Suggest Intervention
• Approach:
– Use Probabilistic System to Identify Patients
– Suggest Enrollment in Pneumonia Protocol
– Provide Therapeutic Suggestions
• Requires Data Extracted from the X-ray Report
Care Delivery Framework
Example: Community-Acquired Pneumonia
Computable Medical
Knowledge Reposotory
Chest Xray
Reports
Chest Xray Report
Processing
(Structured Data
Extraction)
Data Supporting
Pneumonia
Assessment
Pneumonia
Screening Tool
Clinical Data
Repository
Pneumonia
Guideline
Enrollment
Pneumonia
Treatment
Protocol
Care Delivery Framework
Example: Community-Acquired Pneumonia
Computable Medical
Knowledge Reposotory
Chest Xray
Reports
Chest Xray Report
Processing
(Structured Data
Extraction)
Data Supporting
Pneumonia
Assessment
Does the patient
have pneumonia?
Pneumonia
Screening Tool
Clinical Data
Repository
Pneumonia
Guideline
Enrollment
Pneumonia
Treatment
Protocol
Care Delivery Framework
Example: Community-Acquired Pneumonia
Computable Medical
Knowledge Reposotory
Chest Xray
Reports
Chest Xray Report
Processing
(Structured Data
Extraction)
Data Supporting
Pneumonia
Assessment
Does the patient
have pneumonia?
Pneumonia
Screening Tool
Clinical Data
Repository
Should we used the
guideline?
Pneumonia
Guideline
Enrollment
Pneumonia
Treatment
Protocol
Care Delivery Framework
Example: Community-Acquired Pneumonia
Computable Medical
Knowledge Reposotory
Chest Xray
Reports
Chest Xray Report
Processing
(Structured Data
Extraction)
Data Supporting
Pneumonia
Assessment
Does the patient
have pneumonia?
Pneumonia
Screening Tool
Should we used the
guideline?
Pneumonia
Guideline
Enrollment
Pneumonia
Treatment
Protocol
Apply Pneumonia
Care Protocol.
Clinical Data
Repository
ChiefComplaint
RESPIRATORY COMPLAINT 32.4
FEVER
6.96
ABD PAIN
6.05
ORTHO INJURY
4.26
CHEST PAIN
4.12
NEURO COMPLAINT
3.69
FALL
3.62
TRAFFIC INJURY
3.50
ABD PROBLEMS
3.45
CHEST PRESSURE
3.10
BACK PAIN
2.82
WEAKNESS
2.79
SYNCOPE
2.28
ENT PROBLEM
2.19
BODY ACHES
1.88
CV COMPLAINTS
1.88
HEADACHE
1.83
DIZZY
1.77
FLANK PAIN
1.43
CV PROBLEMS
0.92
ASSAULT RAPE
0.87
PSYCHIATRIC
0.86
CHEST HEAVINESS
0.82
SKIN COMPLAINT
0.78
SPECIFIC DIAGNOSIS
0.51
DIABETIC
0.44
PAIN CHEST
0.37
HEART RACE
0.33
TRAUMA
0.31
GENITOURINARY PROBLEM 0.31
PALPITATIONS
0.31
HEART IRR
0.30
ALLERGIES
0.29
HIGH BP
0.28
FLUID NUTRITION
0.27
CONVULSIONS
0.25
INFECTION
0.20
RAPID HR
0.19
IRR HEARTBEAT
0.16
LACERATION
0.16
INGESTION
0.16
BP HIGH
0.13
UNCONSCIOUSNESS
0.11
VAGINAL BLEEDING
.098
MED REFILL
.091
UNKNOWN
.087
LOW BP
.064
CARDIAC ARREST
.059
EYE PROBLEM
.055
BP LOW
.054
other0.18
BPSystolic
< 121.5
29.4
121.5 to 148.5 44.6
>= 148.5
26.0
134 ± 22
HeartRate
< 85.5
44.5
85.5 to 99.5
24.7
99.5 to 110.5 13.0
>= 110.5
17.8
92.1 ± 15
BPDiastolic
< 69.5
28.3
69.5 to 82.5 36.2
>= 82.5
35.5
76.9 ± 11
RespRate
< 19.5
52.3
19.5 to 21.5 24.9
21.5 to 27.5 16.1
>= 27.5
6.72
20.8 ± 3.5
MeanBP
< 85.5
23.0
85.5 to 99.5 35.4
>= 99.5
41.7
95.1 ± 12
TempC
< 36.75
62.7
36.75 to 37.45 23.8
37.45 to 38.05 6.04
>= 38.05
7.46
36.79 ± 0.63
Implimented Using:
NLP_FINDING
Positive
25.9
Negative
74.1
Age
< 15.5
8.06
15.5 to 45.5 45.6
>= 45.5
46.4
42 ± 21
• Web Services Infrastructure
Sodium
< 137.5
25.7
137.5 to 140.5 41.8
>= 140.5
32.6
139.2 ± 2.4
• A Bayesian Network
< 13.5
>= 13.5
Chloride
< 103.5
42.1
103.5 to 105.5 25.1
>= 105.5
32.9
104.3 ± 1.8
Creatinine
< 0.405
3.90
>= 0.405 96.1
PNEUMONIA
• Supported by a Production
Rules System
Absent
94.9
Present
WBC
< 11.85
86.1
11.85 to 18.75 12.4
>= 18.75
1.45
9.46 ± 3.4
5.09
(DROOLS)
• Using an NLP System
Yes
No
BS_CLEAR
44.0
56.0
BS_STRIDOR
Yes .083
No
99.9
BS_CRACKLES
Yes 0.72
No
99.3
Yes
No
BS_RALES
0.11
99.9
Yes
No
BS_ABSENT
.030
100
– Sentence Isolation
BS_NO_COUGH
Yes
0+
No
100
BS_WHEEZES
Yes 2.84
No
97.2
BS_INSPIRATION
Yes 0.79
No
99.2
BS_CONGESTION
Yes 0.53
No
99.5
BS_RHONCHI
Yes 0.43
No
99.6
BS_ABNORMAL
Yes 3.87
No
96.1
BS_DECREASED
Yes 2.29
No
97.7
BS_CLEARING_SECREA...
Yes
0.45
No
99.6
BS_FINE_CRACK...
Yes
0.31
No
99.7
BS_MODERATE
Yes 1.36
No
98.6
BS_NON_PRODUCTIVE_CO...
Yes
1.74
No
98.3
– Random Forests-Based Semantics
BS_EXPIRATION
Yes 0.90
No
99.1
BS_NOT_CLEARING_SECREA...
Yes
0.10
No
99.9
BS_FREQUENT
Yes 1.19
No
98.8
SpO2
< 92.1
10.2
92.1 to 95.3 23.6
95.3 to 98.4 44.9
>= 98.4
21.3
96.1 ± 3
BS_TUBULAR
Yes .024
No
100
BS_COURSE
Yes 0.90
No
99.1
Yes
No
BS_WEAK
0.16
99.8
BUN
45.1
54.9
BS_PRODUCTIVE_CO...
Yes
1.81
No
98.2
BS_INFREQUENT
Yes 0.62
No
99.4
BS_STRONG
Yes 0.76
No
99.2
Patient Tracking Board
Patient Tracking Board
Conclusion
• Natural Language Processing Does Play A
Role In Patient Care
• Useful Applications Will Blend NLP-Derived
Data With Structured Data From The EHR
• Radiology Reports Are A Data-rich Target
For NLP
Questions???
23
Download