Semantic Web, OWL & Protege

Document Sample
Semantic Web, OWL & Protege Powered By Docstoc
					Dalhousie University, Master of Electronic Commerce Program
 The Faculties of Computer Science, Management, and Law




                      Research Project Progress Report II:
            Web Ontology Language and Protégé




                                           by
                                     Anju Sharma
                                      B00398765
                                     anju@dal.ca




   In part fulfillment of the requirements of the Master of Electronic Commerce program
                                 Date Submitted: February 2006
MEC Research Project                                                                                                              Progress Report II


TABLE OF CONTENTS

TABLE OF CONTENTS ............................................................................................................................. 2
INTRODUCTION ........................................................................................................................................ 3
SEMANTIC WEB ........................................................................................................................................ 4
            SEMANTIC WEB: ................................................................................................................................. 4
            ONTOLOGY: ....................................................................................................................................... 4
WEB ONTOLOGY LANGUAGE .............................................................................................................. 6
            WEB ONTOLOGY LANGUAGE: ............................................................................................................. 6
PROTÉGÉ ...................................................................................................................................................10
            PROTÉGÉ:.........................................................................................................................................10
ONTOLOGY IMPLEMENTATION IN PROTÉGÉ ...............................................................................12
            RPG CLINICAL PROBLEM ONTOLOGY:...............................................................................................12
            PATIENT ONTOLOGY: ........................................................................................................................13
            RPG PATIENT-CLINICAL PROBLEM ONTOLOGY: ................................................................................16
CONCLUSION ............................................................................................................................................17
            BENEFITS: ........................................................................................................................................17
            FUTURE APPLICATIONS: ....................................................................................................................17
REFERENCES ............................................................................................................................................18




                                                                                                                                                               2
MEC Research Project                                                                          Progress Report II


INTRODUCTION


     At a surprisingly accelerated rate, the Internet has become the central information station for
individual to consume. The current web architecture contains vast amount of information
resources that can be utilize in numerous beneficial manners. Unfortunately, web information is
rendered solely for humans to translate and share data, rather than computer applications. This
data found on the Internet lacks structure and explicit meaning, creating difficulties for information
retrieval and readability by computers.
                              1
    Therefore, the W3C organization has a “Semantic Web” vision, where the Internet is
transformed into a machine-interpretable network. It would contain “information [which has] well-
defined meaning, better enabling computers and people to work in cooperation” [1]. In this
‘semantic’ project, the Internet would be extended with conceptual metadata that reveals the
intended meaning of Web resources, making them more useful to machines. Subsequently, the
software agents can apply this information to perform advanced tasks that humans may not be
able to perform.

     Ontologies are a central building block of the Semantic Web. Ontologies define domain
concepts and the relationships between them, thus provide an architecture that is meaningful to
both humans and machines. The concepts from these ontologies can be used to annotate Web
resources. The latest Semantic Web technology, Web Ontology Language (OWL), will aid in
publishing and sharing data on the Internet, thus enabling the next generation of web applications
[2].

    This report provides an extensive analysis of the Semantic Web and its latest technologies.
Firstly, descriptions of the Semantic Web are presented. Secondly, an outline of Web Ontology
Language (OWL) will be further discussed. Thirdly, Protégé, an ontology editor tool, will be
examined to understand the process of designing and mapping ontologies. Fourthly, the
implementation of a clinical ontology, based on European Commission’s Referral Guidelines for
Imaging (EU-RPG) [3], in Protégé will be illustrated to emphasize the understanding of the
concepts listed above.




1
  The World Wide Web Consortium (W3C) develops “interoperable technologies (specifications, guidelines, software, and
tools) to lead the Web to its full potential as a forum for information, commerce, communication, and collective
understanding” [1]. Tim Berners-Lee, the original architect of the World Wide Web, founded the W3C in 1994.



                                                                                                                   3
MEC Research Project                                                                 Progress Report II


SEMANTIC WEB

An introduction on Semantic Web and ontologies will be provided in this part of the report.

           Semantic Web:

           At the present moment, the World Wide Web is able to display the information with the
       aid of markup languages, such as HTML. However, they neither provide any meaning to this
       data nor help in understanding the content [4] for computers to process. As a result, the true
       potential and effectiveness of the Internet is limited since it is relying on human interpretation
       to use this information. Thus, W3C has decided to pursue the “Semantic Web” project, where
       explicit meaning is given to the data presented online so that it is easily interpreted by the
       machines.
            The vision of the Semantic Web is to enhance today's web via the exploitation of
                                          2
       machine-interpretable metadata . In other words, the documents are ‘marked up’ with
       semantic information so that the content of the document is easily interpreted by the
       computers. In a semantically enabled system, the tags in the document refer to defined
       concepts, and the system can parse the definition of the concept and use that to combine the
       information with other potentially related data [2] in other applications.
            The Semantic Web addresses the shortcomings of the current web by offering a data
       centric markup language, XML, and the descriptive standards, RDF and OWL. eXtensible
       Markup Language (XML) provides a surface syntax for structured documents, but it does not
       provide sufficient data meaning for “efficient sharing of conceptualization” [1]. Resource
       Description Framework (RDF) is a basic ontology language with graphical applications that
       combines XML Syntax and semantics to represent information about resources on the web.
       Resources are described in terms of properties and property values using RDF statements.
       OWL (Web Ontology Language) has “more facilities, [such as additional vocabulary], for
       expressing meaning and semantics than XML, RDF, and RDF Schemas, and thus OWL goes
       beyond these languages in its ability to represent machine interpretable content on the Web”
       [4].
           To facilitate the exchange of data between computer applications, standard vocabularies
       of a domain must be established and captured in an ontology. It is a knowledge
       representation model defined in terms of classes, properties and relationships for individuals
       who need to share information in a domain [4].

           Ontologies compose the primary foundation of OWL, hence it is vital to comprehend
       ontologies and its components prior to elaborating on OWL and its tools.


           Ontology:

            One the main challenges faced in healthcare is the lack of semantic interoperability. The
       exchange of medical knowledge between computers is difficult since each healthcare facility
       has its own medical terminology. Developing an ontology based on a common healthcare
       domain model (such as HL7 RIM), it will add value to a healthcare system in numerous ways
       [4]:

             Ontologies provide semantic-based and common criteria to structure medical
       information.
             Ontologies allow the flow of patient information and other crucial medical data to be
       efficient due to a standard clinical classification model.

2
    Metadata is “data about data”, meaning a set of facts about the content.



                                                                                                       4
MEC Research Project                                                                 Progress Report II


            Ontologies can support and improve the healthcare workflow by transmitting,
       analyzing, re-using and sharing patient and/or other medical data.
            Ontologies are helpful resources in organizing healthcare content on the web and
       extracting semantic content from any documents.
            Ontologies can help build more powerful and more interoperable information systems
       in healthcare.

            Ontologies “use common metadata vocabularies [such as classes, properties and
       relationships] that allow document creators to know how to mark up their documents so that
       agents can use and share the information across various computer applications” [5]. To
       further elaborate the components of an ontology, classes are a collection of individuals.
       Individuals represent objects in the domain. They are related to other objects and to data
                                                                                        3
       values by properties. Properties are used to link individuals in a binary fashion . Relationships
       are formed with properties between individuals or data. An ontology together with a set of
       individual instances of classes constitutes a knowledge base [6].

            The essential steps to consider when engineering an ontology are the following [6]:

                 Determining the domain of the ontology
                 Enumerating important terms of the domain
                 Defining classes in the ontology
                 Arranging the classes in a taxonomic (superclass-subclass) hierarchy
                 Defining the properties of each class and describing their permitted values
                 Creating individual instances

           OWL is W3C’s latest Semantic technology that builds these ontologies to enable agents
       to exchange data across web applications and resources.




3
    Binary signifies relationship between two objects.



                                                                                                      5
MEC Research Project                                                                           Progress Report II


WEB ONTOLOGY LANGUAGE

    This section of the report is dedicated in describing Web Ontology Language and its
framework.

        Web Ontology Language:

        Released in February 2004 by the W3C, Web Ontology Language (OWL) is an ontology
    language that describes the classes, properties and relations between them that are inherent
    in Web documents and resources. Jim Hendler and Guus Schreiber, co-chairs for the Web
    Ontology Working Group, describe OWL as the following:

                  "OWL takes a major step forward in representing and organizing
              knowledge on the World Wide Web. It strikes a sound balance between
              the needs of industry participants for a language which addresses their
              current Web use cases, and the restrictions on developing an ontology
              language that meshed with established scientific principles and research
              experience." [1]

         OWL is used to describe, share and publish the set of terms that are inherent in Web
                                                                                           4
    documents and applications [7]. OWL uses both Unique Resource Identifiers (URI) (e.g.:
    http://www.w3c.org) for naming and the description framework for the Web provided by RDF
    to extend the capabilities of ontologies [7]. OWL is a vocabulary extension of RDF and RDF-
    S by providing an elaborated description of classes, properties, and individuals. This feature
    enhances the machine interpretability of Web content. OWL is derived from two other
    languages, DAML (DARPA Agent Markup Language) & OIL (Ontology Inference Layer).

        OWL has three sublanguages, each with a different level of expressive description of the
    data:

              1) OWL Lite: It is the simplest language for ontologies with simple class hierarchies
                 and constraints. This subset of OWL-DL contains an easier reasoner than the
                 other species.
                                                                 5
              2) OWL-DL: It corresponds to Description Logics , meaning that it has “decidable
                 reasoning”. Thus, it automatically computes the classification hierarchy and
                 checks for inconsistencies. OWL-DL does not allow datatype properties to be
                 transitive, symmetric, or have inverse properties. Therefore, relationships can
                 only be formed between individuals or between an individual and a data value.
              3) OWL Full: It is an extension of RDF with OWL syntax, where it allows for classes
                 as instances. In OWL-Full, classes can be related, but this cannot be reasoned
                 with.

        An OWL ontology is a network of classes, properties, and individuals. OWL has six ways
    to describe classes, illustrated in Table 1:




4
 URI is an address for a resource available in the Web [7].
5
 Description Logic focuses on descriptions to express logic (such as union, intersection and negation) of a domain. It
emphasizes on the use of classification and subsumption reasoning for inference [7].



                                                                                                                    6
  MEC Research Project                                                         Progress Report II



 Core Class       Class              Definition          Symbol   Meaning             OWL Element
                Description
Named Class                    All the individuals are                          owl:Class
                               part of this named
                               class
Intersection   Intersection;   The individuals that              “And”         owl:intersectionOf
Class          Conjunction     are contained in the
                               overlap portion of 2+
                               classes
Union Class    Union;          The individuals                   “Or”          owl:unionOf
               Disjunction     contained in the
                               combination of 2+
                               classes
Complement     Negation        The individuals that        ¬      “Not”         owl:complementOf
Class                          are not in the
                               negated class
Restrictions   Universal       It describes the class            “Only”;       owl:allValuesFrom
                               of individuals that                “All
                               have only one kind of              values
                               relationship along a               from”
                               specified property to
                               an individual that is a
                               member of a
                               specified class
               Existential     It describes the class            “Some         owl:someValuesFrom
                               of individuals that                Value
                               have at least one                  from”; “At
                               kind of relationship               least
                               along a specified                  one”
                               property to an
                               individual that is a
                               member of a
                               specified class
               Absolute        It defines the exact        =      “Exactly      owl:cardinality
               Cardinality     number of                          n”
                               relationships that a
                               class of individuals
                               can have.
               Max             It defines the                    “At most      owl:maxCardinality
               Cardinality     maximum number of                  n”
                               relationships that a
                               class of individuals
                               can have.
               Min             It defines the                    “At least     owl:minCardinality
               Cardinality     minimum number of                  n”
                               relationships that a
                               class of individuals
                               can have.
               hasValue        It specifies the class            “equals       owl:hasValue
                               of individuals that                x”
                               participate in a
                               specified relationship
                               with a specific
                               individual



                                                                                                7
  MEC Research Project                                                           Progress Report II


Enumeration                      It specifies explicitly     { ... }                owl:oneOf
Class                            and exhaustively the
                                 list of the individuals
                                 that are members of
                                 the enumeration
                                 class
                          Table 1: Six core class descriptions and their definitions.

          The OWL vocabulary provides two ‘extreme’ predefined classes, which are used to state
     all or no instances. An empty ontology contains one class representing the set containing all
     individuals has been defined as owl:Thing. New classes created will be the subclasses of
     this root class. The owl:Nothing class is a empty class that has no member individuals.

         The relationships between the classes are defined in two manners:

              1) Subsumption: There is a hierarchical relationship, where ‘subclass’ and
                 ‘superclass’ terms are applied. All subclass members can be members of the
                 superclass.
              2) Disjointness: The classes that do not overlap or do not have any instances in
                 common are known as ‘disjoint’ classes.

         A property provides an association that relates an instance to a value. The following
     represents the categories of properties:

              1) Object properties (owl:ObjectProperty): Object properties link individuals
                 to individuals. They may have an inverse property (e.g the inverse of “worksFor”
                 might be “employs”).
              2) Datatype properties (owl:DatatypeProperty): Datatype properties link
                 individuals to datatype values (e.g. integers, floats, strings).
              3) Annotation properties (owl:AnnotationProperty): Annotation properties
                 describes instances by adding information to classes, individuals, and
                 object/datatypes properties.
              4) Ontology properties (owl:OntologyProperty): Ontology properties relate
                 ontologies to ontologies.

         In OWL, properties can be specified as domain and range. For example, the property
     ‘hasTopping’ will allow the class ‘Pizza’ to select a range of toppings from the class
     ‘PizzaTopping’. Additionally, they may have sub properties leading to a hierarchy of
     properties. Certain property characteristics can be specified as the following:

              1) Functional: The functional property of an individual has a single value. For
                 example, there can be at most one individual that is related to the individual via
                 the property.
              2) Inverse Functional: The inverse of the property is functional. For example, a
                 property links Individual A to Individual B, then its inverse property will link
                 Individual B to Individual A.
              3) Symmetric: Symmetric property can relate property values back to subject
                 resources. For example, if a property links A to B, then it can be inferred that it
                 links B to A.
              4) Transitive: The transitive property allows elements in a chain to be equal. For
                 example, if a property P links A to B and B to C, then it can be inferred that it
                 links A to C via property P. A property that is transitive cannot be functional.

        One of the appealing features of OWL is its reasoning power. Reasoning capabilities,
     such as consistency checking and classification, are used to detect logical inconsistencies



                                                                                                  8
MEC Research Project                                                             Progress Report II


   within the ontology. OWL-DL requires a reasoner to infer information that isn’t explicitly
   represented in an ontology. The reasoner can check whether or not all of the statements and
   definitions in the ontology model are mutually consistent and can also recognize which
   concepts fit under which definitions. Standard ‘reasoning services’ are:

              Subsumption testing
              Equivalence testing
              Instantiation testing
              Consistency testing

        Classification is used to infer specialization relationships between classes from their
   formal definitions. A classifier takes a class hierarchy including the logical expressions, and
   then returns a new class (inferred) hierarchy, which is logically equivalent to the input
   (asserted) hierarchy. Asserted hierarchy would represent the ‘ideal’ tree of disjoint primitives.
   Inferred hierarchy displays the polyhierarchy tangle (several classes with multiple parents),
   the ‘reality’ of the ontology. The reasoning feature adds great value in the domain of
   healthcare since it is consist of nested hierarchies and multi-relationships between various
   medical concepts. Using OWL, ontology developers could just add a new concept by
   describing its logical characteristics, and the classifier would automatically place it in its
   correct position. They can help tremendously in the construction and maintenance of large
   clinical terminologies.

         In OWL, the reasoner applies the Open World Assumption, which means that there exists
   more information than presented. The reasoner cannot assume something doesn’t exist until
   it is explicitly stated that it does not exist. In this case, it will assume ‘the knowledge hasn’t
   been added to the knowledge base’ [7].

       A knowledge-based modeling environment can be used to implement this Semantic Web
   technology. Protégé is currently the most popular ontology editor for OWL. Hence, the next
   chapter describes thoroughly this tool.




                                                                                                   9
MEC Research Project                                                                             Progress Report II


PROTÉGÉ

         Protégé:

         Standford University’s Medical Informatics team created Protégé, a Java open-source
     software, to provide a development environment for ontologies and knowledge-based
     systems. Protégé is a widely-used tool that allows users to construct domain ontologies,
     customize data entry forms, and enter data. This Java framework can be extended to include
     graphical components such as graphs and tables, media such as sound, images, and video,
     and various storage formats such as OWL, RDF, XML, and HTML.

           Protégé’s has a competitive lead on other ontology editors since it has been “built for
     industry-strength and has a plug-in architecture which makes Protégé a most versatile tool”
     [8]. In addition, according to Ribière et al., this software editor is “well structured, scalable and
     it is very easy to extend the tool to fulfill different goals. It is also easy to implement a new
     knowledge representation language and it generates automatically the documentation on
     ontology concepts” [9]. It is renowned for being a successful knowledge-modeling platform.

          This software contains an OWL plug-in as an extension to support Web Ontology
     Language [10]. The OWL plug-in a medium to load and save OWL files in various formats,
     edit OWL ontologies, and provide access to reasoning based on description logic [2].

         Protégé cannot support OWL Full since their ontologies can state anything about
     anything (Open World Assumption), and OWL Full uses meta-classes which cannot be
     classified. Hence, OWL Lite and OWL-DL can be applied in Protégé projects.

         The developers of Protégé has recently released its beta version 3.2 which has evolved
     significantly in numerous aspects, specifically for its support for SPARQL and its integration
     with Jena [11]. Jena is a Java Application Programming Interface (API) for RDF and OWL. It
     provides “services for model representation, parsing, database persistence, querying and
     visualization tools” [11]. W3C is working towards a standard query language for the
     Semantic Web, known as SPARQL. It has been added in Protégé Beta 3.2 in order to query
     an RDF Schema or OWL model to filter out individuals with specific characteristics. Queries,
     which are executed by the Jena SPARQL engine, can in principle be used for constraint
     checking [11].

          Another significant component that enhances this Protégé tool is its support for
     reasoners. As mentioned before, reasoner software verifies the logical consistency of the
     ontology and infers information that is not explicitly contained within the knowledge
     representation model. It can also be used at runtime in applications as a querying
     mechanism. The OWL plug-in can interact with any reasoner that supports the standard
                                                    6
     Description Logic Implementers Group (DIG) interface since it typically sets up a service
     running locally or on a remote server. Additionally, Protégé has a feature where it displays
     the relationships of classes inferred by the reasoner. The recommended reasoner is Pellet
     version 1.3 reasoner since “it is robust and scalable, and is available under an open-source
     license” [12].

          One aspect that needs to be mentioned is that Protégé has interpreted the superclass-
     subclass relationships between sets of individuals with two types of conditions. Classes with
     “necessary and sufficient” conditions are called ‘defined’ classes. This signifies that if an
     individual is a member of a specific class, then it must satisfy the conditions. If an individual
     satisfies the conditions, then it must be a member of the class. As for classes with only

6
  The DIG interface is an emerging standard for providing access to description logic reasoning via HTTP-based interface
to a separate reasoning process [12].



                                                                                                                     10
MEC Research Project                                                          Progress Report II


   “necessary” conditions, which are also called primitive classes, means that an individual of a
   class must satisfy the conditions, and not vice versa.

       Protégé OWL Plug-in provides a promising platform for healthcare ontology and
   Semantic Web projects. With its strengths in reasoning capabilities and others, OWL
   ontologies are “more compact, less error-prone, and easier to maintain” [2]. Hence, this
   robust and reusable model will play a key role in the evolution and sharing healthcare
   knowledge and wisdom.




                                                                                              11
MEC Research Project                                                           Progress Report II


ONTOLOGY IMPLEMENTATION IN PROTÉGÉ

This chapter of the report demonstrates a simple development and implementation of an ontology
in Protégé. Furthermore, it will demonstrate the utility of OWL in a medical application.

In order to understand the basics of the Protégé tools and functions, the Agfa team has
suggested a small project where various types of patients are classified. Derived on the referral
radiation imaging guidelines by the European Commission (EU-RPG) [13], the patient types have
been created based on randomly selected and obvious clinical problems from the imaging
guidelines due to limited time and medical knowledge for this exercise. Nonetheless, it can be
used in conjunction with another ontology by importing it and creating relationships among them.
Two ontologies, ‘ClinicalProblemOntology’ and ‘RPGPatientOntology’, have been developed in
order to enhance their re-usability in other ontologies.

       RPG Clinical Problem Ontology:

        The ‘RPGClinicalProblemOntology’ classifies all the clinical problems mentioned in EU-
    RPG. The clinical problems have been grouped from Table A to Table M in the guidelines, as
    the following Table 2 demonstrates:

                    EU-RPG Tables              Type of Clinical Problems
                    Clinical Problem A         Head
                    Clinical Problem B         Neck
                    Clinical Problem C         Spine
                    Clinical Problem D         Musculoskeletal
                    Clinical Problem E         Cardiovascular Systems
                    Clinical Problem F         Thoracic System
                    Clinical Problem G         Gastrointestinal System
                    Clinical Problem H         Urological, Adrenal &
                                               Genitor-Urinary System
                    Clinical Problem I         Obstetrics & Gynaecology
                    Clinical Problem J         Breast Disease
                    Clinical Problem K         Trauma
                    Clinical Problem L         Cancer
                    Clinical Problem M         Pediatrics
                           Table 2: The types of Clinical Problems [13].

        Hence, each of these clinical problems (ex: ClinicalProblemA) is a subclass of the main
    class, ‘ClinicalProblem’, and is a superclass for specific clinical situation (ex:
    ClinicalProblemA1 ClinicalProblemA2), illustrated in Figure 1. The feature of disjointness has
    been applied to these specified classes since these medical clinical issues are distinct.




                                                                                               12
MEC Research Project                                                             Progress Report II




                Figure 1: Illustration of subclass-superclass of clinical problems.

        When the reasoner was activated, the inferred model was identical to the original model.
    Instances have been created for all the clinical problems, following the naming convention
    where the clinical problem number as first followed by the name of the problem, like this:
    “A1_Congenital_Disorders”.

       Patient Ontology:

In the ‘PatientOntology’, three main disjoint classes have been defined: ‘Age’, ‘Gender’ and
‘Patient’. The ‘Gender’ class has the ‘Male’ and ‘Female’ subclasses. The ‘hasGender’ property is
a functional relationship with a cardinality of 1 between the ‘Patient’ and ‘Gender’ class. The
‘Patient’ class also has a functional object property called ‘hasAge’ with a cardinality of 1 that
links with the ‘Age’ class. This class contains several subclasses with age categories, such as
‘InfantAge’, ‘YoungAdultAge’, ‘OldAdultAge’, etc., that are applied particularly for the
‘AdultPatient’ and ‘NonAdultPatient’ classes. For example, the ‘NonAdultPatient’ class has a
restriction that is comprised of three age categories: “hasAge only (InfantAge or ChildAge or
AdolescentAge)”.

Under the ‘Patient’ class, the following subclasses have been classified, based on obvious EU-
RPG clinical problems:
            1) ‘AdultPatient’ class: Patients’ age range is from 18 (inclusive) to 125 years old.
                Disjoint from ‘NonAdultPatient’ class.
            2) ‘NonAdultPatient’ class: Patients’ age range is from 0 to 17 (inclusive) years
                old. Disjoint from ‘AdultPatient’ class.
            3) ‘MalePatient’ class: Patients’ gender is only ‘Male’. Disjoint from
                ‘FemalePatient’ class.
            4) ‘FemalePatient’ class: Patients’ gender is only ‘Female’. Disjoint from
                ‘MalePatient’ class.
            5) ‘PregnantPatient’ class: Patients’ gender is only ‘Female’. Clinical Problems
                are from I1, I2, I3, and I4. Disjoint from ‘MalePatient’ class.




                                                                                                13
MEC Research Project                                                          Progress Report II


           6) ‘TraumaPatient’ class: Clinical problems associated with Table K of EU-RPG.
               Subclasses include various injury-related cause such as ‘ArmInjuryPatient’,
               ‘FootInjuryPatient’, etc.
           7) ‘CancerDiseasePatient’ class: These patients would have clinical problems
               primarily from the Tables L, but also Tables D, G, J, and F. Examples of
               subclasses are ‘BoneDiseasePatient’ and ‘LiverCancerPatient’. Certain
               subclasses are applied to specific gender such as ‘TesticleCancerPatient’ class
               contain only male patients. Each of these subclasses is also disjoint.
           8) ‘FollowUpPatient’ class: These patients are following up on their clinical
               problems such as anthropathy, heart disease, pneumonia, etc. A subclass
               named ‘CancerFollowUpPatient’ has been defined to group all the cancer patient
               followup from L28, L31, and L39 from EU-RPG.
           9) ‘DisorderPatient’ class: Patients with clinical problems associated with Tables
               A, B, C, E, F, K & M. Subclasses include ‘CardiovascularDisorderPatient’,
               ‘CongenitalDisorderPatient’, ‘HeadDisorderPatient’ and ‘NeckDisorderPatient’.
           10) ‘PainSymptomaticPatient’ class: This class has patients that are encountering
               pain in their bodies, such as ‘BackPainPatient’, ‘ChestPainPatient’,
               ‘KneePainPatient’, etc.

        With the aid of OWLViz tool, the following image, Figure 2, displays the asserted model
   of the ontology:




                                                                                              14
MEC Research Project                                                         Progress Report II




                             Figure 2: Asserted model of PatientOntology.

      Instances for each patient type described above have been created, where values have
   been added to each slot. Queries have been developed and ran to retrieve instances
   associated with them. For example, if a clinical physician wanted to know the trauma patients
   who are male, the following results appear in the right hand box of the Figure 3:




                                                                                             15
MEC Research Project                                                          Progress Report II




                       Figure 3: Example of a query in the RPGPatientOntology.

       When the reasoner was applied, certain classes were re-classified under different
   classes. For example, under the ‘FemalePatient’ class, ‘PregnantPatient’, ‘OvaryPatient’,
   ‘UterusCancerPatient’ and ‘BreastDiseasePatient’ classes were found since the restriction of
   ‘hasGender only female’ was applied to them.
       Instances for each patient type with age and gender have been defined in the instance
   tab of Protégé.


      RPG Patient-Clinical Problem Ontology:

        For this ontology, the ‘RPGClinicalProblem’ ontology has been imported and combined
   with     the    ‘PatientOntology’,     and     renamed     to    ‘RPGPatientOntology’.    The
   ‘RPGClinicalProblemOntology’         contains    a   ‘CP’   prefix.   The    object  property,
   ‘hasClinicalProblem’, has a domain of ‘Patient’ and a range of ‘CP:ClinicalProblem’. This
   property established a relationship between the two ontologies. The cardinality is minimum of
   zero relationships between the patients and clinical problems since a patient may not have a
   clinical problem listed in this EU-RPG.
        For each type of patients, restrictions have been used to define the patients’ possible
   clinical problems. An example of restriction is that an ‘OvaryCancerPatient’ class will have
   clinical problems L29 (Ovary Cancer: Diagnosis), L30 (Ovary Cancer: Staging), and L31
   (Ovary Cancer: Follow-up).
        Instances for each type of patients have been created with their gender, age and clinical
   problems defined. For example, Jessica is a breast cancer patient who is female, is 48 years
   old, and has breast inflammation as a clinical problem.
        Unfortunately, the Pellet reasoner has detected an error that seems irresolvable by the
   Protégé team. The error message appeared in the command states that certain classes are
   forced to be in the Patient class and its complement. Due to limited time on the current
   research, this error will be further analyzed once the main project has been completed.




                                                                                              16
MEC Research Project                                                           Progress Report II


CONCLUSION

     This progressive report is aim to deliver thorough descriptions of Semantic Web, its
technologies and tools. These concepts provide the robust foundation for the further development
of the overall HL7 RIM-OWL research project.

      Benefits:

        Once the primary goal of this research project of mapping the HL7 RIM to OWL is
   achieved, the medical data will have greater meaning so that computer applications, rather
   than humans, can interpret and use it effectively. This translation mechanism will permit the
   exchange of information and data integration between various healthcare machines,
   achieving interoperability between heterogeneous medical applications. Thus, knowledge
   sharing will aid in providing tools for further advancements, and efficiencies in healthcare
   facilities will improve immensely.

      A subgroup of the W3C organization, known as Healthcare and Life Sciences, best
   describes the benefits of this research pursuit:

                “The implementation of new informatics models that will unite many
           forms of medical information across all institutions, through the encoding
           of meaning into the data and their interpretations. By focusing on the
           semantics of information, researchers will have more access to the
           knowledge required to effectively find cures to diseases, while doctors
           will have better tools for individualized clinical management of patients.”
           [1]

      Future Applications:

        Insights for future emerging investigations can be based on this HL7 RIM ontology to
   create automated systems that generate specific rules, clinical practice guidelines, and/or
   workflows. These actions would ensure reliability, quality assurance and safety for future
   uses of this RIM-OWL mapping, and thus will be a well-built foundation for other architectural
   projects. An agent technology can be developed that will utilize this medical mapping to
   produce and share ‘healthcare wisdom’ across the clinical applications. As a result, the
   medical professionals can proactively act and innovate further advanced technologies with
   this wisdom. This will dispense and lead to high healthcare services and effective patient
   care, and an overall trust based system.

       There can be potential use cases of this developed ontology such as mapping to a web
   services (such as ICD10), creating decision support for clinical practice, semantic assistance
   for data processing tools, or sharing of healthcare imaging data. A specific Web-based
   application that would interest the Agfa team is to publish this OWL file connected with other
   OWL ontologies based on imaging guidelines (such as dimensions, resolution, contents) on
   the Web so that resources, agents, and services can link to this file and use these ontologies’
   concepts. For example, digital images can be classified, queried and retrieve automatically
   through various repositories, with the aid of intelligent agents [2].

        The HL7 RIM-OWL ontology can be a key instrument in developing electronic health
   records. Since the ontology is based on widely-used standard medical terminologies, then it
   will seamlessly integrate with the decision support system with patient records and patient
   information systems which use this HL7 RIM standard.

       These notions are only the beginning of a new era of Semantic Web and healthcare. It is
   the time to embrace and pursue them with all the efforts available.



                                                                                               17
MEC Research Project                                                           Progress Report II


REFERENCES

   [1] World Wide Web Consortium: www.w3.org

   [2] Holger Knublauch et al. Weaving the Biomedical Semantic Web with the Protégé OWL
   Plugin: http://protege.stanford.edu/plugins/owl/publications/KRMed2004-protege-owl.pdf

   [3] European Commission. Radiation Protection 118 – Referral Guidelines for Imaging. 2000:
   http://europa.eu.int/comm/environment/radprot/118/rp-118-en.pdf

   [4] Robin Cover. OWL Web Ontology Language. Cover Pages hosted by OASIS:
   http://xml.coverpages.org/owl.html

   [5]     Open    Clinical     –     Knowledge       Management         for    Medical      Care:
   http://www.openclinical.org/ontologies.html

   [6] Natalya F. Noy and Deborah L. McGuinness. Ontology Development 101: A Guide to
   Creating                       Your                      First                 Ontology:
   http://protege.stanford.edu/publications/ontology_development/ontology101-noy-
   mcguinness.html

   [7] Matthew Horridge et al. A Practical Guide To Building OWL Ontologies Using The
   Protégé-OWL Plugin and CO-ODE Tools (Edition 1.0). 2004


   [8] Some Ontology Tools: http://dannyayers.com/archives/002138.html

   [9] Myriam Ribiere et al. Ontology Overview: http://www.fipa.org/docs/input/f-in-00045/f-in-
   00045.pdf

   [10] Protégé OWL - Ontology Editor for the Semantic Web:
   http://protege.stanford.edu/plugins/owl/

   [11] Standard University’s Protégé: http://protege.stanford.edu/

   [12] How to use Jena and DIG reasoners: http://jena.sourceforge.net/how-to/dig-
   reasoner.html

   [13] European Commission. Radiation Protection 118 – Referral Guidelines for Imaging.
   2001: http://europa.eu.int/comm/environment/radprot/118/rp-118-en.pdf




                                                                                                  18

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:20
posted:7/7/2012
language:English
pages:18