 |
Claims  |
|
|
What is claimed as new and desired to be secured by Letters Patent of the
United States is:
1. A knowledge-based system for the retrieval of images based on expert
knowledge and patient information stored in a computer system, the
knowledge-based system comprising:
a first memory means for storing a knowledge-base comprising the expert
knowledge information including structural knowledge on a plurality of
classes, general procedural knowledge, heuristic knowledge stored in the
form of rules, and control knowledge for controlling the functions of
searching hierarchies of the plurality of classes and dynamically creating
objects;
a second memory means for storing a database comprising the patient
information including patient data, examination data and images;
knowledge-base/database interface means for coupling said database to said
knowledge-base for receiving and transmitting information therebetween;
means for storing new patient information in the second memory means;
reasoning means for searching the hierarchies of the plurality of classes
stored in the knowledge-base, for dynamically creating objects and for
selecting rules in response to storing the new patient information in the
second memory means, said searching and selecting being conducted based on
said stored control knowledge, structural knowledge and general procedural
knowledge, and said stored patient information;
retrieving means to search the patient information stored in the database
and to retrieve the examination data indicated by the execution of the
rules selected by said reasoning means;
user interface means for accessing said stored patient information from
said database and outputting the patient information retrieved by said
retrieving means; and
control interface means for coupling said user interface means to said
knowledge-base for receiving and transmitting information and controlling
the flow of information therebetween.
2. A knowledge-based system according to claim 1, wherein said structural
knowledge includes attribute information, identifier information, and
relationship information on said plurality of classes.
3. A knowledge-based system according to claim 2, wherein the relationships
among said classes represented in said relationship information are
association relationships, specialization relationships, and aggregation
relationships.
4. A knowledge-based system according to claim 2, wherein said general
procedural knowledge includes information on a plurality of data
processing procedures performed on said classes.
5. A knowledge-based system according to claim 4, wherein said data
processing procedures includes query and update operations for performing
on said classes and said relationships among said classes.
6. A knowledge-based system according to claim 1, wherein said heuristic
knowledge includes information on reasoning processes for determining the
images to be retrieved.
7. A knowledge-based system according to claim 6, wherein said information
on reasoning processes is stored in the form of a plurality of rules.
8. A knowledge-based system according to claim 7, wherein said stored rules
are in an "IF-THEN" format.
9. A knowledge-based system according to claim 8, wherein each rule
selected by said reasoning means is executed by said retrieving means upon
satisfaction of the "IF" condition of said rule.
10. A knowledge-based system according to claim 1, wherein said control
knowledge comprises rules represented as triggers owned by at least one
respective of the plurality of classes.
11. A knowledge-based system according to claim 1, wherein said user
interface means includes means for displaying the images retrieved.
12. A knowledge-based system which includes a coupled knowledge-base and
database stored in a computer system for operating in a predetermined
application domain, the knowledge-based system comprising:
a first memory for storing a knowledge-base comprising data representative
of application domain expert knowledge including structural knowledge on a
plurality of classes, general procedural knowledge, heuristic knowledge
stored in the form of rules, and control knowledge for controlling the
functions of searching hierarchies of the plurality of classes and
dynamically creating objects;
a second memory for storing a database comprising data which are a
plurality of instances of said structural knowledge;
knowledge-base/database interface means for coupling said database to said
knowledge-base for receiving and transmitting data therebetween;
means for storing new instances of said structural knowledge in said second
memory means;
reasoning means for searching the hierarchies of the plurality of classes
stored in the knowledge-base, for dynamically creating objects and for
selecting rules in response to storing the new instances of said
structural knowledge in said second memory means, said searching and
selecting being conducted based on said stored control knowledge,
structural knowledge and general procedural knowledge, and the data stored
in said database;
retrieving means to search the data stored in the database and to retrieve
the data indicated by the execution of the rules selected by said
reasoning means;
user interface means for accessing the data stored in said database and
outputting the data retrieved by said retrieving means; and
control interface means for coupling said user interface means to said
knowledge-base for receiving and transmitting data and controlling the
flow of data therebetween.
13. A knowledge-based system according to claim 12, wherein said structural
knowledge includes attribute information, identifier information, and
relationship information on said plurality of classes.
14. A knowledge-based system according to claim 13, wherein the
relationships among said classes represented in said relationship
information are association relationships, specialization relationships,
and aggregation relationships.
15. A knowledge-based system according to claim 14, wherein said general
procedural knowledge includes information on a plurality of data
processing procedures performed on said classes.
16. A knowledge-based system according to claim 15, wherein said data
processing procedures includes query and update operations for performing
on said classes and said relationships among said classes.
17. A knowledge-based system according to claim 12, wherein said heuristic
knowledge includes information on reasoning processes for determining the
data to be retrieved.
18. A knowledge-based system according to claim 17, wherein said
information on reasoning processes is stored in the form of a plurality of
rules.
19. A knowledge-based system according to claim 18, wherein said stored
rules are in an "IF-THEN" format.
20. A knowledge-based system according to claim 19, wherein each rule
selected by said reasoning means is executed by said retrieving means upon
satisfaction of the "IF" condition of said rule.
21. A knowledge-based system according to claim 12, wherein said control
knowledge comprises rules represented as triggers owned by at least one
respective of the plurality of classes.
22. A knowledge-based system according to claim 21, wherein said user
interface means includes means for displaying the data retrieved.
23. A method for retrieving images based on expert knowledge and patient
information stored in a computer system, the method comprising the steps
of:
storing in a knowledge-base the expert knowledge information including
structural knowledge on a plurality of classes, general procedural
knowledge, heuristic knowledge stored in the form of a plurality of rules
in an "IF-THEN" format, and control knowledge for controlling the
functions of searching hierarchies of the plurality of classes and
dynamically creating objects;
storing in a database patient information including patient data,
examination data and images;
receiving and transmitting information between the knowledge-base and
database;
storing new patient information in the database;
searching the hierarchies of the plurality of classes stored in the
knowledge base, creating objects dynamically and selecting rules in
response to storing the new patient information in the database, said
searching and selecting being conducted based on said stored control
knowledge, structural knowledge and general procedural knowledge, and said
stored patient information;
executing each rule selected from searching the plurality of classes upon
satisfaction of the "IF" condition of said rule;
searching the patient information stored in the database and retrieving the
patient information indicated by the rules executed; and
outputting the patient information retrieved.
24. A method for retrieving images according to claim 23, wherein the step
of storing structural knowledge in the knowledge-base includes storing
attribute information, identifier information, and relationship
information on said plurality of classes.
25. A method for retrieving images according to claim 24, wherein the
relationships among said classes represented in said stored relationship
information are association relationships, specialization relationships,
and aggregation relationships.
26. A method for retrieving images according to claim 23, wherein the step
of storing the general procedural knowledge in the knowledge-base includes
storing information on a plurality of data processing procedures performed
on said classes.
27. A method for retrieving images according to claim 23, wherein the step
of storing the heuristic knowledge in the knowledge-base includes storing
information on reasoning processes for determining the images to be
retrieved.
28. A method for retrieving images according to claim 27, wherein the step
of storing the control knowledge in the knowledge-base includes storing
the control knowledge in the form of rules represented as triggers owned
by at least one respective of the plurality of classes.
29. A knowledge-based system which includes a coupled knowledge-base and
database for operating in a predetermined application domain, the
knowledge-based system comprising:
a memory for storing data representative of domain expert knowledge
including structural knowledge on attributes and identifiers of a
plurality of classes and relationships between the classes, general
procedural knowledge on a plurality of data processing procedures
performed on the classes, heuristic knowledge on reasoning processes for
performing predetermined functions in the application domain, and domain
dependent control knowledge for determining the reasoning process specific
to the application domain by controlling the functions of searching
hierarchies of the plurality of classes and dynamically creating objects;
and
control means coupled to said memory for controlling the storage of new
data in said memory by searching the hierarchies of the plurality of
classes and dynamically creating objects in response to storing the new
data in said memory.
30. A knowledge-based system according to claim 29, wherein the structural
knowledge stored in said memory includes attribute information, identifier
information, and relationship information on said plurality of classes,
said attribute information, identifier information, and relationship
information on each class being represented in the form
Attribute-name: data-type=default-value
Identifier: Attribute-names
Operation-name (arguments): return-type.
31. A knowledge-based system according to claim 30, wherein the
relationships among said classes represented in said relationship
information are association relationships, specialization relationships,
and aggregation relationships.
32. A knowledge-based system according to claim 31, wherein the general
procedural knowledge stored in said memory includes information on a
plurality of data processing procedures performed on said classes, said
data processing procedures performed on each class being represented in
the form
Procedure-name (arguments):
SELECT output-specification
WHERE constraint-specification.
33. A knowledge-based system according to claim 32, wherein said data
processing procedures includes query and update operations for performing
on said classes and said relationships among said classes.
34. A knowledge-based system according to claim 33, wherein the heuristic
knowledge stored in said memory includes information on reasoning
processes for performing predetermined functions in the application
domain, said information on reasoning processes being stored in the form
of a plurality of rules encapsulated in said plurality of classes, each
rule encapsulated in each class being represented in the form
Rule-name:
IF condition-specification
THEN action-specification.
35. A knowledge-based system according to claim 34, wherein the domain
dependent control knowledge stored in said memory comprises rules
represented in the form
Trigger-name (arguments):
IF: activated-condition-specification
THEN: action-specification
owned by at least one respective of the plurality of classes. |
|
 |
|
 |
Claims |
|
|
|
Description |
|
|
NOTICE REGARDING COPYRIGHTED MATERIAL
A portion of the disclosure of this patent document contains material which
is subject to copyright protection. The copyright owner has no objection
to the facsimile reproduction by anyone of the patent document or the
patent disclosure, as it appears in the Patent and Trademark Office patent
file or records, but otherwise reserves all copyright rights whatsoever.
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to a method for designing a coupled
knowledge-base/database using a synthesized object-oriented
entity-relationship model and, more particularly to a knowledge-based
system and method for the retrieval of images using a coupled
knowledge-base/database designed using a synthesized object-oriented
entity-relationship model.
2. Discussion of the Background
Over the past two decades, databases have evolved into the central
component of organizational information systems. A great many information
systems are operated in environments requiring that a large amount of
information or data be managed and processed. Further, the users of such
systems often require high performance operation in information retrieval
and processing functions.
Database and knowledge-base technologies have been widely employed by
organizations to meet their information needs. Knowledge processing is
widely accepted in business, industry, and engineering as valuable for
handling information and has made possible advances in the capabilities of
data processing and information systems.
Typically, a database maintains well-structured data representing the facts
that traditionally are essential to data sharing and processing
activities, while a knowledge-base contains less precise, more abstract,
and possible subjective knowledge used mainly for decision and planning
support. Because data come in larger quantities and are more dynamic than
knowledge, database technologies are most often concerned with efficient
update and retrieval operations on large sets of data whereas
knowledge-base technologies address small-size knowledge-base processing
and infrequent knowledge-base maintenance. Knowledge-based technologies
have been applied to database systems to enhance data retrieval functions
by providing decision and planning support.
One such effort to provide knowledge-based capabilities in a large, complex
information system has been made in the area of image retrieval, and more
particularly in a system for the retrieval of radiographic images for use
in a hospital environment.
Advances in computer, communication and digital radiographic imaging
technologies have encouraged many organizations to launch efforts to
realize totally digitized radiology services. Extensive research and
development is taking place in the areas of picture archiving and
communication systems (PACS) that handle the creation, storage, retrieval,
transmission and display of digital patient radiographic images and
pertinent information for radiology-related services.
One of the core components of a PACS is a database system that stores and
manages images and pertinent textual data. Massive amounts of data are
generated every day (e.g., 3.8 gigabits in a 500 bed hospital) as new
images in digital form are stored. In operation, the digitized images are
retrieved from the PACS database for use by radiologists at viewing
workstations. In order to meet the performance requirements of
radiologists, a response time of less than two seconds is required. In an
effort to satisfactorily handle the large amount of information and meet
the performance goals required, some designs for PACS database systems
have adopted a multi-level storage architecture and a distributed database
approach.
Critical to the design of the PACS database system is the observation that
in a radiological examination reading, radiologists usually compare a
newly generated examination (image) with previous examinations (images) of
the same patient. Based on this observation, the retrieval of old images
is a critical design requirement of PACS.
In current film-based radiology systems, such retrieval is initially
performed by nurses or assistants, who hang the most recent images on
alternators for reading. During reading, radiologists may dig into film
jackets and fan through other old images for additional relevant images.
To effectively support digital primary reading using a PACS, it is
essential to identify relevant previous patient images that can be either
pre-fetched off-line or retrieved on-line to arrive for current diagnosis
and to reduce the significant delay caused when access is obtained through
a slow and remote storage device in a hierarchical (multi-level) and
distributed PACS database system.
Effective patient image retrieval depends upon the radiologists' expert
knowledge, which enables them to select images for comparison based on
information about the new images (to be diagnosed), the patient and
previous images. The implementation of the PACS system using a
knowledge-based approach based on the expert knowledge of radiologists
would provide the advantages of (1) reducing system response time, (2)
reducing radiologists' time in selecting images for review, (3) minimizing
the turnaround time of the exam interpretation function, and (4) improving
the diagnostic effectiveness and quality by providing relevant and
sufficient images automatically.
A precursor to the present invention is described in Sheng, Ovitt, Wang,
and Garcia, Image Retrieval Expert Systems, Proc. SCAR 90 (Computer
Applications to Assist Radiology), edited by Arenson and Friedenberg, pp.
198-204, and Sheng, Wang, and Garcia, IRES--Image Retrieval Expert System,
Proc. SPIE Medical Imaging IV Conference, Feb. 4-9, 1990, pp. 832-841. The
knowledge-based system for the retrieval of images, Image Retrieval Expert
System (IRES), described was prototyped using three major components:
databases, a procedural control algorithm, and a rule base. Although the
capability of knowledge-based processing was added to the PACS, the design
of IRES using a flat rule base resulted in problems with maintainability
and extensibility of the system due to the lack of defined relationships
among rules and the redundancy inherent in rule based systems. Further,
the lack of a structured rule organization in the IRES rule base according
to natural rule characterization and relationships negatively impacted
upon the efficient design of knowledge inferencing procedures and severely
hampered the extendibility of the knowledge-base to include a larger set
of rules.
It is a known fact that as the structures and manipulations of databases
become more complex and the size of knowledge-bases increase, as, for
example, in the rule based IRES discussed above, existing separate
database and knowledge-base design technologies become inadequate. In an
effort to provide information systems having large, complex databases with
knowledge processing capabilities, attempts have been made to couple
knowledge-base and database design. Coupling knowledge-base and database
design provides the advantages of (1) improving data management by using
knowledge-base technologies to manage complex relationships among data and
to perform deductive data processing, and (2) improving knowledge
management by using database techniques to maintain the factual data
imbedded in knowledge, thereby reducing the size and improving the
extensibility and maintainability of knowledge-bases.
Modeling or representation of data and knowledge relationships is critical
to the design of coupled knowledge-base/database systems. Modeling any
less than all of the knowledge on data and knowledge interactions for a
given application domain (1) severely restricts the maintainability and
extensibility of the system, (2) limits the advantages realized by
knowledge-base/database coupling, (3) greatly increases the burden on
system designers/developers during the design and development stages, and
(4) substantially increases the likelihood of errors in implementation.
Attempts to model data and knowledge relationships in designing coupled
knowledge-base/database systems are characterized, generally, by the
application of (1) semantic data modeling techniques or (2)
object-oriented techniques to represent the knowledge/data relationships.
Semantic data modeling uses semantic models for representing structurally
complex interrelationships among data. The primary components of semantic
models are the explicit representation of objects, attributes of and
relationships among objects, type constructors for building complex types,
IS-A relationships, and derived schema components. An example of a
semantic data model is the Entity-Relationship (ER) model. The ER model
and other semantic data models and modeling techniques of the
above-described type are discussed in Hull and King, Semantic Database
Modeling: Survey, Applications, and Research Issues, ACM Computing
Surveys, Vol. 19, No. 3, pp. 201-60, September 1987; and Peckham and
Maryanski, Semantic Data Models, ACM Computing Surveys, Vol. 20, No. 3,
pp. 153-89, September 1988. Although semantic data modeling provides a
technique for effectively modeling the structural aspects of objects, no
construct or mechanism is provided for representing the behavioral aspects
of objects exhibited during knowledge and data interactions. This inherent
limitation in the semantic data model precludes modeling of a substantial
part of the knowledge exhibited during data and knowledge interactions.
Further, the available semantic data modeling techniques are limited in
the types of relationships among objects that can be represented and
provide no mechanism for sharing of properties and methods among objects
by inheritance.
Object-oriented modeling techniques have been applied to database designs
to represent data relationships and their behavior. The Object-Oriented
Entity-Relationship Model (OOERM), discussed in Gorman and Choobineh, An
Overview of the Object-Oriented Entity Relationship Model (OOERM),
Proceedings of the Twenty-Third Annual Hawaii International Conference on
System Sciences, 1990, pp. 336-345, extended the Entity-Relationship (ER)
model by using object-oriented constructs to model the operational
properties of entities or objects for the purpose of database design.
Because the ER model has been frequently used in database design methods,
its object-oriented extension in the OOERM permits application of existing
ER-based design concepts, while adding object-oriented principles to
dictate entity behavior (procedures, rules or operation). The Object
Modeling Technique (OMT), discussed in Blaha, Premerlani, and Rumbaugh,
Relational Database Design Using An Object-Oriented Methodology,
Communications of ACM, Vol. 31, No. 4, pp. 414-27, April 1988, similarly
incorporated the main concepts of the ER model into an object-oriented
model and associated design methods that model both the static (passive)
and behavioral (active) properties of entities for software system and
relational database designs. The application of object-oriented modeling
techniques for database design provides the advantage of natural
abstraction representation, data/behavior encapsulation and
superclass-subclass inheritance features. However, these techniques (OOERM
and OMT) cannot be effectively used to model coupled
knowledge-base/database systems due to the limited types of entity or
object behavior represented. This limitation of the OOERM and OMT
precludes modeling of a crucial part of the knowledge exhibited during
data and knowledge interactions.
Another object-oriented design method, the Structured Object Model (SOM),
discussed in Higa, Morrision, Morrison, and Sheng, An Object-Oriented
Methodology for Knowledge Base/Database Coupling, Working Paper,
University of Arizona, 1990, and Morrison, Morrison, and Sheng, A
Hierarchical Object-Oriented Knowledge-Based Architecture for Coupled
Knowledge-Base/Database Systems, University of Arizona, Working Paper
Series, 1990, has been used for modeling coupled knowledge-base/database
systems. In the SOM design method, data semantics are represented using
objects, attributes, and two types of relationships (aspect and
specialization). Although this model provides the advantages associated
with the object-oriented modeling method, the modeling constructs and the
design procedures for the knowledge-base components are incomplete and
imprecise, and cannot effectively represent the domain expert knowledge
and data and knowledge relationships. Further, the problem solving control
knowledge for performing the reasoning process in an object-oriented
coupled knowledge-base/database system is not defined and no method for
modeling such knowledge is described, as the systems were implemented
using expert systems shells coupled with database management systems
(DBMS).
Notwithstanding the available coupled knowledge-base/database design
methods, there is a need for a method of designing a coupled
knowledge-base/database system that provides (1) a mechanism for modeling
all of the knowledge on data and knowledge interactions for a given
application domain, (2) a defined construct for modeling all the
knowledge, (3) a mechanism for sharing of properties and methods among
objects by inheritance, and (4) a schema for the coupled database.
Further, there is a need for a knowledge-based system and method for the
retrieval of images that uses a coupled knowledge-base/database system
having a knowledge-base storing all of the knowledge on data and knowledge
interactions for the image retrieval process and a coupled database having
a schema derived from the knowledge-base.
SUMMARY OF THE INVENTION
The primary object of the present invention is to overcome the deficiencies
of the prior art described above by providing a method for designing a
coupled knowledge-base/database system for use in a knowledge-based system
for the retrieval of images that can effectively provide knowledge
processing capabilities in information systems having large complex
knowledge-bases and databases.
Another key object of the present invention is to provide a coupled
knowledge-base/database design method and a system for image retrieval
using a coupled knowledge-base/database that can be effectively and
efficiently maintained and extended.
Still another key object of the invention is to provide a coupled
knowledge-base/database design method and a system for image retrieval
using a coupled knowledge-base/database that substantially reduces the
burden on system designers/developers in the design, development,
implementation, maintainability, and extendibility of knowledge-based
information systems.
Yet another key object of the present invention is to provide a coupled
knowledge-base/database design method and a system for image retrieval
using a coupled knowledge-base/database that provides a mechanism for
modeling all of the knowledge on data and knowledge interactions including
knowledge involved in data processing, knowledge-based problem solving and
object-oriented reasoning.
Another object of the invention is to provide a coupled
knowledge-base/database design method and a system for image retrieval
using a coupled knowledge-base/database having improved data and knowledge
management and the capability of performing deductive data processing.
Still another object of the present invention is to provide a coupled
knowledge-base/database design method and a system for image retrieval
using a coupled knowledge-base/database having the capability of
minimizing knowledge-base and database sizes through sharing of properties
and methods among objects by inheritance.
Another object of the present invention is to provide a coupled
knowledge-base/database design method and a system for image retrieval
using a coupled knowledge-base/database that is not limited in the types
of entity or object behavior or in the relationships among entities or
objects that can be represented.
Yet another object of the present invention is to provide a coupled
knowledge-base/database design method and a system for image retrieval
using a coupled knowledge-base/database that provides a structured
approach and a well-defined construct for modeling the domain expert
knowledge on data processing, knowledge-based problem solving and
object-oriented reasoning.
Yet another object of the present invention is to provide a coupled
knowledge-base/database design method and a system for image retrieval
using a coupled knowledge-base/database that can effectively bind data and
knowledge interactions.
Still another object of the present invention is to facilitate the method
of design of coupled knowledge-base/databases and to reduce errors in the
implementation of coupling knowledge-base and database by providing a
modeled knowledge-base structure from which a schema for the coupled
database can be derived.
Another object of the present invention is to provide a method for
designing a coupled knowledge-base/database and a system for image
retrieval using a coupled knowledge-base/database representing structural
knowledge as object classes having relationships and encapsulating general
procedural, heuristic, and control knowledge within the object classes.
Yet another object of the present invention is to provide a method for
designing a coupled knowledge-base/database and a system for image
retrieval using a coupled knowledge-base/database, the knowledge-base
storing expert knowledge information including structural knowledge on a
plurality of classes, general procedural knowledge, heuristic knowledge
stored in the form of rules, and control knowledge.
Still another object of the invention is to provide a coupled
knowledge-base/database design method and a system for image retrieval
using a coupled knowledge-base/database having the capability of
effectively representing and using, during processing, domain expert
knowledge on the relationships among objects including specialization,
aggregation, and association relationships.
The present invention achieves these objects and others by providing a
method for designing a coupled knowledge-base/database and system for the
retrieval of images using a coupled knowledge-base/database, the method
comprising the steps of modeling structural knowledge by identifying
classes and attributes of classes, specifying an identifier for each
class, determining relationships among the classes, and defining
operations for each class; modeling heuristic and general procedural
knowledge by acquiring heuristic rules for each class dependent on the
application domain, specifying data processing procedures required by the
heuristic rules acquired, representing the procedures specified on the
classes and on the relationships among the classes, and representing the
heuristic rules in an "IF-THEN" format and using the procedures and
operations defined; modeling control knowledge by specifying
intra-class-hierarchy searching paths, specifying inter-class-hierarchy
searching paths, and representing the specified searching paths in
triggers for each class; and deriving a schema for the coupled database
from the structural knowledge. The method for designing a coupled
knowledge-base/database system of the presen | | |
