Transcript AMIA_Image_Nov06 - Buffalo Ontology Site

Doing Ontology Over Images
Barry Smith
1
What ontologies are for
what molecular function ?
what disease process ?
need for semantic annotation of data
3
need for semantic annotation of data
through labels (nouns, noun phrases)
which are algorithmically processable
4
natural language labels
to make the data cognitively
accessible to human beings
5
compare: legends for maps
6
compare: legends for cartoons
7
ontologies are legends for data
9
ontologies are legends for images
10
what lesion ?
what brain function ?
11
ontologies are legends for
mathematical equations
xi = vector of measurements of gene i
k = the state of the gene ( as “on” or “off”)
θi = set of parameters of the Gaussian model
...
...
12
The OBO Foundry Idea
GlyProt
MouseEcotope
sphingolipid
transporter
activity
DiabetInGene
GluChem
13
annotation using common ontologies
yields integration of databases
GlyProt
MouseEcotope
Holliday junction
helicase complex
DiabetInGene
GluChem
14
annotation using common ontologies
can yield integration of image data
15
annotation using common ontologies
can support comparison of image data
16
truth
17
simple representations can be true
18
there are true cartoons
19
a cartoon can be a veridical
representation of reality
20
Cartographic Projection
21
maps may be correct by reflecting
topology, rather than geometry
22
an image can be a veridical
representation of reality
a fully labeled image can be an
even more veridical representation
of reality
23
24
25
cartoons, like maps, always have a
certain threshold of granularity
26
grain resolution
27
grain resolution serves cognitive
accessibility
we transform true images
into true cartoons
28
there are also true cartoon
sequences
29
30
Pathway diagrams are annotated
dynamic cartoons
31
pathways can be represented at
different levels of granularity
32
the jaw
33
34
Joint capsule
Netter
35
Mandible and condyle movement
36
Condyle position in fossa wrt location of disc
37
TMJ in jaw open and closed positions
38
Holes and Parts
Parts
• 1 head of condyle F
• 2 neck of condyle F
• 3 disc B
• 4 retrodiscal tissue B
• 7 articular eminence F
• 8 zygomatic arch F
• 10 upper head of lateral pterygoid
muscle F
• 11 lower head of lateral pterygoid
muscle F
Holes
• 5 lower joint compartment B
• 6 upper joint compartment B
39
Temporomandibular Joint (TMJ)
ANTERIOR
from Thomas Bittner and Louis Goldberg, KR-MED 2006
40
adjacency relations
No connectedness
Only (temporary)
adjacency
Adjacency
relations
Connectedness
adjacency
graph
41
Frames of reference
Rigid = do not
change shape
(bones)
B
C
D
E
F
A
The extension of the axis of the condyle
intersects the fossa in region D
42
instances vs. types
43
two kinds of annotations
44
names of instances
45
names of types
46
pathway maps are representations
of complexes of types
47
molecular images and
radiographic images are
representations of instances
48
MIAKT system
49
50
51
52
53
Patient #47920
54
55
Mammography #31667
56
Medical-Image #44922
Mammography #31667
57
Patient #47920
Medical-Image #44922
Breast #1388
MRI-Exam #32388
Mammography #31667
Abnormality #86023
58