A MEDICAL INFORMATION RETRIEVAL

BASED ON RETRIEVERS’ INTENTIONS

Osamu Takaki

School of Knowledge Science, Japans Advanced Institute of Science and Technology, Nomi, Japan

Koichiro Murata

School of Medicine, Kitasato University Hospital, Sagamihara, Japan

Noriaki Izumi

Social Intelligence Technology Research Laboratory, National Institute of Advanced Industrial Science and Technology

Tokyo, Japan

Koiti Hasida

Social Intelligence Technology Research Laboratory, National Institute of Advanced Industrial Science and Technology

Tokyo, Japan

Keywords: Medical informatics, Ontology-based retrieval system, Ontology mapping, Query expansion, Intention,

Viewpoint.

Abstract: This paper introduces a methodology to retrieve information from medical databases based on intentions

and viewpoints of retrievers, who retrieve some information from databases. The methodology above helps

a retriever to organize his/her intention and viewpoint and to make a proper query based on the viewpoint. It

also helps a retriever to record historical data of retrieving with intentions. This paper introduces a series of

ontologies to organize a relationship between intentions, viewpoints and keywords that are used to make

queries and each of that has interpretations based on viewpoints of retrievers, and explains the methodology

based on the ontologies.

1 INTRODUCTION

1.1 Background

Recent years, medical database systems, that include

ordering systems of medicines and so on, have

become widely used in Japan with digitalization of

medical information (cf. (Ministry of Health Labour

and Welfare of Japan, 2005) and (Monthly New

Medicine of Japan, 2009)). On the other hand, the

kind of users of medical databases has been growing.

Nowadays, various users including patients regularly

utilise medical databases. So, it becomes more

significant to enable many different kinds of people

to easily retrieve medical information from multiple

databases, and thus, it is important to address the

problem of diversity of concepts or interpretations of

keywords in the databases. In the following

paragraph, we explore this problem in more detail.

Health is a common concern to people

throughout the world. In fact, the knowledge related

to medicine is enormous, difficult and growing

significantly not only as scientific knowledge but

also knowledge about institutions, economics or

philosophy. Moreover, the meaning of a concept in

medicine varies from the viewpoint of a person who

considers it. For example, let consider a concept of

``a proper duration of hospital stays for a patient

who gets bowel cancer in an early stage''.

From the scientific viewpoint, it would mean the

duration between the day when he/she gets bowel

cancer and the day when he/she gets cured of the

596

Takaki O., Murata K., Izumi N. and Hasida K..

A MEDICAL INFORMATION RETRIEVAL BASED ON RETRIEVERS’ INTENTIONS .

DOI: 10.5220/0003174805960603

In Proceedings of the International Conference on Health Informatics (HEALTHINF-2011), pages 596-603

ISBN: 978-989-8425-34-8

 2011 SCITEPRESS (Science and Technology Publications, Lda.)

cancer under the assumption that he/she gets proper

treatment. However, the average of real-life

durations of hospital stays for patients who get

bowel cancer in early stages often differs from the

``ideal'' duration above. The reason seems to come

from the fact that the length of real-life duration is

determined by a complex mix of problems of

medical care processes, institutional problems,

problems of management, patient's individual

problems, and so on.

1.2 Objective of this Paper

As noted above, in order to make clear

interpretations of keywords by retrievers, it is often

helpful to know their intentions and viewpoints and

to consider the interpretations of keywords based on

them. Thus, we introduce a methodology to retrieve

information from medical databases based on

viewpoints and intentions of retrievers, where a

``retriever'' denotes one who retrieves some

information from databases with some statement

what data he/she require. We call such a statement a

``query’’. More strictly speaking, we introduce a

methodology to assist a retriever to (i) organize

his/her intention, (ii) find a viewpoint related to the

intention, (iii) make a query based on the intention

and/or the viewpoint, and (iv) record a historical

data of retrieving with the intention. When one

performs retrieval, he/she has a certain intention. So,

the methodology formalizes such a retriever's

intention and utilizes it to retrieve medical

information more appropriately.

1.3 Outcomes and Structure of this

Paper

The core component of the methodology of

intention-based medical information retrieval is a

series of ontologies to organize relationships

between retrievers’ intentions, retrievers’ viewpoints

and retrieval keywords that are used as parameters in

queries. Each keyword has a single or multiple

definitions (interpretations) based on viewpoints (or

certain patterns of intentions) of retrievers. The

reader will notice that historical data of retrieving

with intentions are useful to consider not only how

to interpret given keywords but also what keywords

retrievers should select as parameters in queries.

Moreover, as a foundation of the retrieval system,

we introduce an ontology-based retrieval system for

large-scale medical data warehouses. The data

warehouses are real-life, that are in operation in

center hospitals. The retrieval system is developed

based on a developing tool of an ontology called

``Semantic Editor'' (K.Hasida, 2010) and an

ontology-mapping tool called ``D2RQ'' (C.Bizer,

2010).

The reminder of the paper is structured as

follows. In Section 2, we explain related work as a

foundation of the research of this paper. In Section 3,

we briefly explain a process to retrieve information

from medical databases based on retrievers'

intentions and the structure of an abstract intention-

based retrieval system. In Section 4, we explain

ontologies to organize relationship between

intentions and viewpoints of retrievers and several

concepts related to medical information retrieval. In

Section 5, we add supplemental remarks about

utilizing historical data of retrieving to make queries.

In Section 6, we explain an ontology-based retrieval

system for medical data warehouses. We conclude

this paper in Section 7.

2 RELATED WORK

In general, it is not easy to retrieve desired

information from multiple databases in a lump with

an existing keyword-based retrieval system, since

the databases may have structures that differ from

each other. One of the techniques to address this

problem is semantic-based retrieval, in particular,

ontology-based retrieval. A semantic-based retrieval

system denotes a system that retrieves data with

concepts that have the same meanings as given

keywords. Semantic-based retrieval systems for

medical databases also have been investigated

and/or developed actively (cf. (Kementsietsidis et al.,

2009), (Kohler et al., 2003), (Perez-Rey et al.,

2006)).

In technical aspects, a lot of ontology-based

retrieval systems employ two methodologies

``ontology-mapping'' and ``query-expansion (with

ontologies)''. Ontology-mapping denotes a

methodology or a technology that maps between an

ontology and a (relational) data model. Ontology-

mapping integrates data in multiple databases that

may have different structures of schema, by

constructing a proper ontology and assigning the

ontology to the data models of the given databases.

Ontology-mapping also enables one to retrieve data

from multiple databases with different data models,

by queries that are defined at some abstraction level.

In general, this technology has been investigated

actively (cf. (Kalfoglou and Schorlemmer, 2005),

(Konstantinou et al., 2008), (Noy, 2004)).

A MEDICAL INFORMATION RETRIEVAL BASED ON RETRIEVERS' INTENTIONS

597

As a research of ontology-mapping for medical

information retrieving, Aronson et al (Aronson,

2001) developed an algorithm that maps information

in biomedical texts to concepts in the UMLS

Metathesaurus (Bodenreider, 2004). Moreover,

Farfan et al (Farfan et al., 2009) point out several

issues on ontology-based retrieval of documents

from EMR (Electronic Medical Records) and

propose a solution to the issues by using descriptive

logic. They also introduce a retrieval algorithm

based on HL7 (HL7-International, 2010) and

SNOMED-CT (IHTSDO, 2010).

On the other hand, query-expansion denotes a

methodology that aims to make a exhaustive query

by finding keywords having the same meaning as

that of the keywords in a given query based on

thesauruses on medicine and by complementing the

given query with the keywords (cf. (Bhogal et al.,

2007), (Billerbeck and Zobel, 2004)). As researches

query expansion based on medical thesauruses, one

can refer to (D´ıaz-Galiano et al., 2009) and

(Wollersheim and Rahayu, 2005).

Both of the methodologies above often utilize

thesauruses or repositories of medicine or biology

such as ICD (WHO, 2010), SNOMED-CT, GO

(Stevens et al., 2000), UMLS, MeSH (S.J.Nelson

and B.L.Humphreys, 2001), if the ontology-based

retrieval system deals with data in medical domain.

In this paper, we utilize ontology-mapping and

query-expansion as a fundamental methodology.

Moreover, we newly consider a series of ontologies

(see Section 4) to organize retrievers' viewpoints and

intentions to realize retrieval based on retrievers'

intention and/or viewpoints.

3 CONSTRUCTION

OF THE METHODOLOGY

In this section, we show a process to retrieve data

from medical databases based on a retriever's

intention. For practical use, we will use some

ontology-based retrieval system plus several

ontologies defined in Section 4, keywords that are

used to make queries and that are organized with the

ontologies, and historical data of retrieving (see Fig.

1). The set of ontologies above is called ``Intention-

based Medical Retrieval Ontologies (IMRO)”. Until

Section 6, we regard the ontology-based medical

retrieval system above as an abstract one.

Historical data of retrieving denote a series of

data related to retrieving that retrievers have

performed. The data include not only queries but

also intentions (purposes) to make the queries. These

intentions are organised and recorded as well as

viewpoints based on IMRO.

Figure 1: Process to retrieve medical information based on

retrievers’ intentions with a retrieval system.

Here we show a process to retrieve data in the

environment that is described in Fig. 1, as follows.

i. The retriever first login into the system (cf. Step

1 in the left side of Fig. 1). Then, the system checks

his/her fundamental information such as the type of

a job and historical data that he/she has achieved

until now. After that, according to the fundamental

information and the historical data, the system

selects one or several candidates of viewpoints,

patterns of intentions, and intentions themselves, and

shows them.

ii. Then, the retriever selects a viewpoint, a pattern

of intentions and an intention among candidates that

the system lays out at (i) above. He/she also inputs a

new intention when he/she cannot find a desired one

among candidates (cf. Step 2). Then, the system

selects one or several candidates of templates of a

query.

iii. Then, the retriever makes a query by selecting a

template among candidates that the system lays out

at (ii) above and by inputting keywords (parameters)

that the system keeps up (cf. Step 3). Moreover, for

each keyword, he/she checks interpretations of the

keyword and selects the best one. Note that for each

keyword the system lists up and orders

interpretations of the keyword according to the data

about the intention in (i) and (ii) above and historical

data of retrieving, which are organized with IMRO.

The retriever also can refer to queries in the

historical data of retrieving, by checking intentions

and/or viewpoints that are attached to the queries.

iv. After the retriever establishes a query and

submits it to the system, according to the query and

additional data including the retriever's intention, it

retrieves information from databases that are con-

HEALTHINF 2011 - International Conference on Health Informatics

598

nected with the system.

v. Then, the retriever checks the result of the

retrieval in (iv), and if he/she can confirm the result

to be a desired one, he/she goes to the activity (vii)

below (cf. Step 4). If the retriever is not satisfied

with the result, he/she can undo the previous activity

(iii) or (ii). Moreover, the retriever also improves the

query with checking the definitions of keyword in

the query, when it is necessary.

vi. When the query is completed, the system

transforms the query to some query in the SQL-

language the given medical database use by mapping

keywords in the query to indexes and by mapping

the query in an ontology-based language to another

one in SQL. Note that, from a query that the

retriever makes, the system may generates multiple

queries when the system connects to multiple

databases that have different schema (and/or

different indexes) and/or different query-languages.

vii. If it is necessary, the retriever input

complementary information into the historical data

of the query (cf. Step 5). Remark that the system

also records data how to make the queries that may

include not only the queries themselves but also how

the queries were constructed, in particular, new

intentions and/or keywords the system required.

4 ONTOLOGIES FOR

INTENTION-BASED

RETRIEVAL

Not only in retrieving medical information, but also

in all kinds of tasks, one performs his/her tasks

based on his/her own intentions, even though

retrievers themselves are sometimes not fully aware

of their intentions to retrieve information. While a

retriever's intention may differ from one retrieving to

another, in many cases, he/she has a tendency of

his/her intentions to retrieve information. In many

cases, such a tendency can be regard as his/her

viewpoint. Moreover, a retriever's viewpoint

strongly depends on which type of an actor he/she is

or which kinds of jobs he/she does. Let consider a

doctor as an example. The typical viewpoints that a

doctor tends to have when he/she try to retrieve data

from medical databases would be the following

viewpoints.

i. Researches of medical knowledge such as

diseases or medical care based on statistical data.

ii. Pursuit of highly productive medical service

based on statistical data.

iii. Solution of a problem of a special patient or a

special group of patients based on statistical data

about the (group of) patient(s).

iv. Improvement of medical management based on

statistical data.

While the first viewpoint is that of a doctor as a

researcher of medical science, the second one is that

of a staff who tries to improve a medical setting. On

the other hand, while the third viewpoints is that of a

professional of a medical care who tries to solve a

problem of a (group of) particular patient(s) own,

the fourth one is that of a manager of a hospital who

tries to restore his hospital.

Take an intention to research relationships

between incidence rate of bowel cancer and age of

patients as an example. Then, this intention can be

regarded as an element or an instance of the

viewpoint (i). On the other hand, another intention to

investigate monthly cost of medical care for patients

of cancer in the medical expense system with DPC

can be regarded as an element of the viewpoint (iv).

Here, we introduce ontologies to organize

relationships between intentions and viewpoints of

retrievers, types of jobs of them, and interpretations

of keywords that are defined based on their

viewpoints. We call these ontologies ``Intention-

based Medical Retrieval Ontologies (IMRO)''.

In this section, we only show the ontology that

organizes the relationship between intentions and

viewpoints (Fig. 2) and another one that organizes

the relationship between keywords and viewpoints

(Fig. 3).

Figure 2: Ontology of intentions and viewpoints.

Figure 3: Ontology of retrieval keywords.

A MEDICAL INFORMATION RETRIEVAL BASED ON RETRIEVERS' INTENTIONS

599

In this paper, we consider a viewpoint as well as a

pattern of intentions as a set of intentions. The

ontology in Fig. 2 indicates that there are patterns of

intentions and that each pattern has a unique

viewpoint. We consider each pattern to have a

unique viewpoint that is a super set of intentions.

We define concrete viewpoints and patterns of

intentions in Table 1, which are formalized as

instances of the concepts in Fig. 2.

Table 1: Viewpoints and typical patterns of intentions.

On the other hand, the ontology in Fig. 3

indicates the relationship between keywords, the

definition (interpretation) of keywords and

viewpoints. Each keyword has at least one definition,

and each definition has just one concept in Fig. 3 as

the content of the definition and at least one

background. Each background of a definition

corresponds to a pair of a single viewpoint and a

single strength value that we will explain later. So,

we can assign to each keyword several definitions

and to each definition several viewpoints and

strength values.

As an example, we show two definitions of a

keyword ``first visit’’ in Fig. 4. The first

definition ``first visit: Def.1’’ in Fig. 4

indicates that ``first visit'' denotes a visit by a

specified patient to a specified department in a

specified hospital and that the visit had not existed

for a specified duration to the relevant period but it

existed at the period. On the other hand, the second

definition ``first visit: Def.2’’ in Fig. 4

indicates that ``first visit'' denotes a visit by a

specified patient to a specified hospital when the

patient first gets an ID-number of the patient at the

hospital.

Figure 4: Two definitions of the keyword ``first

visit’’.

Figure 5: Information obtained from a potencially grouped

historical data of queries sharing V.

The first definition is universal and we assign to

it all viewpoints in Fig. 5. On the other hand, the

second one is a particular one and we assign to it a

viewpoint ``Viewpoint of Medical Business’’ in

Table 1. Note that, in ``first visit: Def.1’’ in

Fig. 4, the strengths of the support to the definition

by each viewpoint is assigned to the same value as

any other’s one. Here, the strength of the support to

the definition by a viewpoint denotes a value how

much the viewpoint supports the definition. This

value will be dynamically changed by the historical

data of queries. More strictly speaking, if a retriever

make a query based on a viewpoint and he/she

HEALTHINF 2011 - International Conference on Health Informatics

600

employs a definition of a keyword in the query, the

strength value of the definition by the viewpoint

increases. On the other hand, when the retrieval

system assists a retriever to keyword in the order of

the value of the definition by the viewpoint of the

retriever. We will detail this point in the next section.

5 MAKING A QUERY BASED ON

RETRIEVERS’ INTENTION

In this paper, we do not explain concrete language of

queries, though we actually employ a query-

language SPARQL (W3C, 2008) (cf. Section 7).

However, we briefly explain how to utilize a

retriever's intention or viewpoint formalized by

IMRO in Section 4, when he/she make a query.

In Section 4 we explained that each keyword was

assigned a single or multiple definition(s) according

to viewpoints of retrievers. However, when a

retriever make a query, he/she can consider not only

how to interpret keywords but also what keywords

to be used, by using historical data of retrieving with

others' viewpoints or intentions. We explain this

with two examples of making queries. As the first

example, let consider a retrieval of the lengths of

durations of hospital stays from an intention to

increase efficiency of hospital beds. From the

viewpoints of economics, it would be desirable to

analyze the result with classifying the set of the

lengths in terms of costs of beds or profits from

patients, since these factors strongly affect the

hospital's earning and expense. In fact, from the set

of the lengths of the durations only, one cannot

analyze the result since he/she cannot know the

meaning of each length. On the other hand, let

consider another retrieval of the same values from

an intention to analyse the problem of social

hospitalization. Then, the retriever needs to classify

the set of the lengths in terms of not only economic

aspects but also patient-specific aspects including

patients' ages and their family structures. This

means that retrievers' intentions or their viewpoints

are often related to how to classify or how to order

the retrieval results.

Thus, it is useful to consider concepts to organize

relationships between intentions, viewpoints and a

certain set of keywords. Although we avoid

explaining concrete sets of keywords that would be

related to certain kinds of viewpoints, we briefly

explain how to make and reuse a set of keywords

that are deeply related to a given viewpoint, as

follows.

1. Let consider a case to make a query Q

based on a

viewpoint V.

2. The system has a list L that consists of keywords

and their definitions organized with IMRO. Then,

the system extracts keywords (for example, K1 and

K3 in Fig. 5 above) in L that have definitions (for

example, D12, D31 and D34 in Fig. 5) in L, and

recommends the retriever of Q

to use the keywords

and definitions to make Q

, more strictly speaking, to

make conditions of results data, to make conditions

of patients and to classify the result data.

3. The priority of recommendations of certain

keywords and their definitions is decided according

to the value of strength that is assigned to definitions

with V.

4. If the retriever employs existing keywords and/or

existing definitions, then the values of strengths

related to the keywords and/or existing definitions

and V will be increased.

5. If he/she consider new keywords or new

definitions to make Q

, then the system add the

keywords or definitions above to make L as an

information related to V.

Remark that the retriever can utilize not only his/her

historical data of retrieving but also historical data of

others' retrieving. This means that a retriever with

viewpoint V can refer to information obtained from a

potentially grouped historical data of queries sharing

V. Therefore, this method is especially useful for

retrievers who have clear intentions or clear

viewpoints but who are not familiar with medical

knowledge. Typical types of such retriever would be

patients.

6 PROTOTYPE OF AN

INTENTION-BASED MEDICAL

RETRIEVAL SYSTEM

In Section 3, we describe an ontology-based

retrieval system, which joins IMRO and historical

data of retrieving. In this section, as a foundation of

the retrieval system, we introduce a prototype of an

ontology-based medical retrieval system for large-

scale medical data warehouses, which is currently

under development.

Though the development of the prototype

originally began with the intention of retrieving data

relevant to patients of cancer, it will be improved to

cover all patients.

A MEDICAL INFORMATION RETRIEVAL BASED ON RETRIEVERS' INTENTIONS

601

Figure 6: Prototype system plus IMRO.

The prototype consists of the following

components (cf. Fig. 6).

i. Data warehouses that are developed with

RDBMS.

ii. A series of ontologies of patients, medicines,

tumor markers and exceptional rules.

iii. Mapping components between schema in the

given data warehouses in (i) and concepts in the

ontologies above.

iv. An engine that assists retrievers to make queries

based on the ontologies in Fig. 5 and retrieves data

from the data warehouses in (i).

The data warehouses in (i) above are real-life, that

are in operation in center hospitals.

Patients ontology in (ii) consists of fundamental

concepts relevant to patients such as names, genders,

ages, and so forth.

Diseases ontology consists of concepts of disease

that are organized according to ICD 10 (World-

Health-Organization, 2007). Each concept of disease

has attributes of the name, the code in X10 and the

type of the classification according to X10.

Medicines ontology consists of concepts of

anticancer drugs. Each concept of an anticancer drug

has attributes of the code that is defined by Ministry

of Health, Labour and Welfare of Japan, the names,

the class item, diseases the drug is effective against

and exceptions to the applications of the drug.

Tumor markers ontology consists of concepts of

tumor markers. Each concept of a tumor marker has

attributes of the code, the name, diseases that the

marker is effective against and exceptions to the

applications of the marker.

Exceptional rules ontology consists of exceptions

to the rules of prescription or examination by tumor

markers.

A mapping component in (iii) assigns each

concept in one of the ontologies in (ii) to some

index(es) in one of the data warehouses in (i). The

mapping components are implemented with D2RQ

(C.Bizer, 2010).

D2RQ is a platform that consists of a mapping

language, a RDF-based retrieval engine and an

interface (a server). The mapping language are used

to transform concepts (classes and properties)

described in RDF into schema and attributes in the

schema. On the other hand, the RDF-based retrieval

engine transforms queries described in RDF-directed

graphs into SQL of relational databases, and vice

versa. By D2RQ-platform, one can retrieve data

from RDBMS with a RDF-query language called

SPARQL.

The retrieval engine in (iv) is implemented with

D2RQ and a platform to develop ontologies, which

is called “Semantic Editor (SE)”.

Since this system has a unique mapping file for

each medical database, the prototype can retrieve

medical data from multiple databases in a lump. In

fact, we are trying developing the prototype system,

by which one can retrieve data from three real life

databases in center hospitals. We are supposed to

develop an intention-based medical information

retrieval system by combining the prototype system

with IMRO.

7 CONCLUSIONS

This paper introduces a methodology to retrieve

information from medical databases based on

intention of retrievers. Information about intentions

of the retriever helps he/she to make more

appropriate queries or to check the quality of them

by explicitly checking meanings (definitions) of

given keywords in the queries. It also enhances

reusability of historical data of queries. The

methodology of this paper helps a retriever to

organize his/her intention and to find a viewpoint

related to the intention, and to make a query based

on the viewpoint. It also helps a retriever to record a

historical data of retrieving with intentions.

In Section 4, Intention-based Medical Retrieval

Ontology (IMRO) is introduced. IMRO organizes

the relationship between intentions, patterns of

intentions, viewpoints and definitions of retrieval

keywords. By virtue of IMRO, one can organize

his/her intention or viewpoint and select keywords

based on the viewpoints.

In Section 6, this paper introduces a prototype

system of an ontology-based retrieval system for

HEALTHINF 2011 - International Conference on Health Informatics

602

large-scale medical data warehouses, which are in

operation in center hospitals. The prototype system

is being developed with an ontology-developing tool

“Semantic Editor” and an ontology-mapping tool

“D2RQ”, and it enables a user to retrieve data

relevant to patients from multiple medical databases

in one lump. We aim to connect the prototype with

IMRO to develop an intention-based medical

retrieval system.

REFERENCES

Bhogal, J., Macfarlane, A., and Smith, P. (2007). A review

of ontology based query expansion. Inf. Process.

Manage, 43(4):866–886.

Billerbeck, B. and Zobel, J. (2004). Techniques for

efficient query expansion. In Proc. String Processing

and Information Retrieval Symp, pages 30–42.

Springer-Verlag.

Bodenreider, O. (2004). The unified medical language

system (umls): Integrating biomedical terminology.

Nucleic Acids Research, 32(Database issue):D267–

D270.

C.Bizer (2010). The D2RQ Plattform - Treating Non-RDF

Databases as Virtual RDF Graphs. http://www4.

wiwiss.fu-berlin.de/bizer/d2rq.

D´ıaz-Galiano, M. C., Mart´ın-Valdivia, M., and Ure na-

L´opez, L. A. (2009). Query expansion with a medical

ontology to improve a multimodal information

retrieval system. Comput. Biol. Med., 39(4):396–403.

Farfan, F., Hristidis, V., Ranganathan, A., and Weiner, M.

(2009). Xontorank: Ontology-aware search of

electronic medical records. In ICDE ’09: Proceedings

of the 2009 IEEE International Conference on Data

Engineering, pages 820–831. IEEE Computer Society.

HL7 -International (2010). Health Level Seven (HL7).

HL7 International. http://www.hl7.org.

K. Hasida (2010). Introduction to Semantic Editor (in

Japanese). http://i-content.org/semauth/intro/index.

html.

Kalfoglou, Y. and Schorlemmer, M. (2005). Ontology

mapping: The state of the art. In Kalfoglou, Y.,

Schorlemmer, M., Sheth, A., Staab, S., and Uschold,

M., editors, Semantic Interoperability and Integration.

Internationales Begegnungs- und Forschungszentrum

für Informatik (IBFI), Schloss Dagstuhl, Germany.

Kementsietsidis, A., Lim, L., andWang, M. (2009).

Profilebased retrieval of records in medical databases.

AMIA Annu Symp Proc, 2009:312–6.

Köhler, J., Philippi, S., and Lange, M. (2003). Semeda:

ontology based semantic integration of biological

databases. Bioinformatics, 19(18):2420–2427.

Konstantinou, N., Spanos, D.-E., and Mitrou, N. (2008).

Ontology and database mapping: A survey of current

implementations and future directions. J. Web Eng.,

7(1):1–24.

Ministry-Of-Health-Labour-And-Welfare-Of-Japan (2005).

Survey of Medical Institutions and Hospital Report (in

Japanese). Ministry of Health, Labour andWelfare of

Japan. http://www.mhlw.go.jp/toukei/saikin/hw/iryosd

/05/index.html. Monthly-New-Medicine-Of-Japan

(2009). White Paper on

Electronic Health Record & Picture Archiving and

Communication Systems 2009-2010 (in Japanese).

The ME Association.

Noy, N. F. (2004). Semantic integration: A survey of

ontology-based approaches. SIGMOD Record,

33(4):65–70.

Perez-Rey, D., Maojo, V., Garcia-Remesal, M., Alonso-

Calvo, R., Billhardt, H., Martin-Sanchez, F., and

Sousa, A. (2006). Ontofusion: ontology-based

integration of genomic and clinical databases. Comput

Biol Med, 36(7-8):712–30.

S. J. Nelson, W. and B. L. Humphreys (2001).

Relationships in medical subject headings (MeSH),

pages 171–184. Stevens, R., Goble, C. A., and

Bechhofer, S. (2000). Ontology-based Knowledge

Representation for Bioinformatics, volume 1, pages

748–756.

The-International-Health-Terminology-Standards-Develop

ment-Organisation (2010). SNOMED-CT. The

International Health Terminology Standards

Development Organisation. http://www.ihtsdo.org/

snomed-ct.

The-World-Wide-Web-Consortium (2008). SPARQL

Query Language for RDF. The World Wide Web

Consortium (W3C). http://www.w3.org/TR/rdf-sparql-

query/.

Wollersheim, D. and Rahayu, J. W. (2005). Ontology

based query expansion framework for use in medical

information systems. IJWIS, 1(2):101–115.

World-Health-Organization (2007). International

Statistical Classification of Diseases and Related

Health Problems 10th Revision Version for 2007.

World Health Organization. http://apps.who.int/classi

fications/apps/icd/icd10online/.

World-Health-Organization (2010). International

Classification of Diseases (ICD). World Health

Organization. http://www.who.int/classifications/icd.

A MEDICAL INFORMATION RETRIEVAL BASED ON RETRIEVERS' INTENTIONS

603