Medical Information Mining for

1

• Medical Information Mining for
• Picture Archiving and Communication Systems PACS
• M. Datcu, R. Srungaru, M. Ungureanu,
• Colapiccioni, A. Zaia, R. Murri, K. Seidel
• German Aerospace Center DLR
• University Politehnica Bucharest
• Adavanced Compure Systems, Rome
• University Camerino
• Swiss Federal Institute of Technology ETH Zurich

2
Postulates Existing volume of unstructured
data prevents any systematic exploitation of its
information content Information extraction
depends critically on the descriptive or
predictive accuracy of the model employed The
augmentation of the data with meaning, e.g. image
understanding, can be interpreted as a coding
task which includes the model of users
conjecture Paradox People have trouble in
caching more than 7 items a time We design
systems to enable people to access 1000 TB
3
IMAGE MINING

• Goals
• find certain images in an archive
• explore the image information content
• discover scene structures or dynamics by
  exploring image collections

4

• EO
• The Image Analyst
• Training
• Optical data 3-4 years
• SAR data 5-7 years
• Medicine
• The Radiologist
• Training 6 - 8 years

5
Medicine Erath Observation
References   Lehmann T.M., Schubert H., Keysers
D., Kohnen M., and Wein B.B. (2003) The IRMA
code for unique classification of medical
images, Proc. SPIE Medical Imaging, San Diego,
California, USA, 18-20 February 2003, 5033, pp.
440 451 Lehmann T.L., Wein B.B., and Greenspan
H. (2003) Integration of Content-Based Image
Retrieval to Picture Archiving and Communication
Systems, Proc. Med. Informatics Europe, IOS
Press, Amsterdam, The Netherlands, 47 May 2003,
CD-ROM Datcu M., et.al. (2003) Information
mining in remote sensing image archives-Part A
system concepts, IEEE Trans. Geosci. Remote
Sensing, 41(12), pp. 1-14 4 Datcu M., Seidel
K. (2005) Human-Centered Concepts for
Exploration and Understanding of Earth
Observation Images, IEEE Trans. Geosci. Remote
Sensing, 43(3), pp. 601-609
6
Parameter retrieval vs. signal analysis Axiom
parameter retrieval is possible only if signal
classes are separable for the assumed physical
models Thus Signal analysis is proposed as a
help for parameter retrieval IIM is proposed as
a help for browsing non visual data exploring
archives prototype classificatns discover
information
7
CONCEPT DE COMMUNICATION AVANCE
Sourcedinformation
Utilisateurs
Extraction de linformation
Représentation sémantique
Inférence du modèle de signal objectif
Inférence du modèle d information subjectif
Modélisation de la conjecture utilisateur
Modèles sémantiques
Modèles syntaxiques
8
KIM concept
9
The image based diagnosis methods are
continuously developing It is necessary to use
databases for these images Queries shall be
instrument independent Queries shall be oriented
to answer the diagnosis needs Solution Picture
Archiving and Communication Systems (PACS) are
available within a hospital allowing a global
access to the shared resources
10
References   1 Lehmann T.M., Schubert H.,
Keysers D., Kohnen M., and Wein B.B. (2003) The
IRMA code for unique classification of medical
images, Proc. SPIE Medical Imaging, San Diego,
California, USA, 18-20 February 2003, 5033, pp.
440 451 2 Lehmann T.L., Wein B.B., and
Greenspan H. (2003) Integration of
Content-Based Image Retrieval to Picture
Archiving and Communication Systems, Proc. Med.
Informatics Europe, IOS Press, Amsterdam, The
Netherlands, 47 May 2003, CD-ROM
11
Advanced communication
Clusters
Semantic labels
Images
Ik
?j Model 1
p(? I )
p(L ? )
?k Model 2
12
Definition of Ontology and Knowledge             
   Ontology and Knowledge Base o Ontology -gt
define the concepts of the system in the
application domain o Knowledge Base -gt contain
the instances of the system concepts and their
inter-relationships          Ontology o
Ontology is an explicit conceptualisation of a
domain
13

•        Knowledge Base (Instance of concepts in
  the ontology)
• o Instances of
• Domains
• Users
• Features
• Files
• Models
• o Value is in the relationships between those
  instances

14
SkinLab system

• Illumination source
• Polarizers
• Digital camera
• Central processing unit

15
Examples of relevant dermoscopic features
16
Hierarchical modelling of image content
17
Classification and semantic labeling
Interactive learning
18
The texture model
19
(No Transcript)
20
Image labeling
21
Training for the pigment network label
22
Probabilistic retrieval for the pigment network
label
23
Cse study 2 Ortodenthal X ray
24
The KIM system
Knowledge based Image Infromation Mining on line
system http//kes.acsys.it/kimv Aknowledgments
- European Space Agency ESA - European
Coordination Group for Image Information Mining
IIMCG