MENTAL CLONING BASE VIRTUAL DIAGNOSTICIAN

SYSTEM

Virtual Medical Doctor System (VDS) Reasoning

Hamido Fujita, Jun Hakura and Masaki Kurematsu

Iwate Prefectural University, Iwate, 020-0193, Japan

Keywords: Emotional mental reasoning, Human user interaction, Intelligent interface, Ontology integration, UMLS.

Abstract: Human computer Interaction based on emotional modelling and physical views, collectively; has been

investigated and reported in this paper. Two types of ontology have been presented to formalize a patient

state: mental ontology reflecting the patient mental behaviour due to certain disorder and physical ontology

reflecting the observed consequences of such disorder. These two types of ontology have been mapped and

aligned for reasoning purposes. We have constructed an integrated computerized model which reflects a

human diagnostician as computer model and through it, an integrated interaction between that model and

the real human user (patient) is utilized for 1

stage diagnosis purposes. The diagnostician knowledge has

been utilized through UMLS for testing, and the integrated mapping of the two views been represented

through OWL framework. The reasoning instantiation is done using Description logic. We have

implemented the system and empirically, examining it, for revision and evaluation.

1 INTRODUCTION

There have been extensive move towards changing

the way health care is delivered, financed and

regulated (Smith, 20009) Medical innovations have

become an important lever inquest of improving

efficiency. The main purpose is to improve the

efficiency so that more patients could receive

treatment more quickly without reducing the quality

of care (Mikkola, 2003). How to cope with a rise in

the need for the elderly care services is a formidable

issue facing all the industrialized countries.

Unfortunately, Japan's health care system has not

been prepared enough to respond to the needs ahead.

Particularly Japan's home care services have heavily

been relying on voluntary labour of family members

with little social services available. The proportion

of the population 65 and over has doubled from 10%

in 1985 to 20% in 2005, and is projected to be 30%

in 2023 (2006, NIPSSR).

In June 2006, the Diet (Japanese Congress)

passed a comprehensive package of reform to make

the delivery system more efficient. First, the average

length of stay in hospitals is to be decreased. To

achieve this goal, the number of long term care

(LTC) hospital beds will be reduced from the 2006

level of 380,000 to 150,000 by the end of fiscal year

2011 and converted to LTC Insurance facility beds

and assisted living (Leflar, 2005). The system

proposed in this paper participates in helping

physicians to manage the diagnosis procedure using

the same knowledge that that physicians have by

copying (mimic) his/her style, mentality, diagnosis

routines and medicine recipes. It is not replacing the

physicians but it would participate to utilize his/her

Figure 1: Simple outline of the VDS.

250

Fujita H., Hakura J. and Kurematsu M. (2010).

MENTAL CLONING BASE VIRTUAL DIAGNOSTICIAN SYSTEM - Virtual Medical Doctor System (VDS) Reasoning.

In Proceedings of the Third International Conference on Health Informatics, pages 250-256

DOI: 10.5220/0002717102500256

 SciTePress

Figure 2: The VDS outline.

knowledge for preliminary diagnosis and health care

services for patient for efficiency purpose.

This paper contributes to present part of our

experimental work on building a virtual system

based or what we called as Virtual doctor System

(VDS) Fig.2, to act as a physical or medicinal doctor

for diagnosis purposes. In other paper we have

presented the outline of the interface, and in this

complementary part we are presenting the inference

engine and the ontological integration as in Fig.5.

1.1 System Conceptual Outline

The system we called here as VDS (Virtual Medical

Doctor) (Fig.2) is to work together with the

corresponding medical doctor. So the system (VDS)

and the MD are working together in comprehensive

coherency; the former is complementary to the latter

but not vice versa. The former is to diagnose

outpatient 1

and classify these diagnosis into

classes. Simple cases classes that the VDS would

take conclusion and set the diagnosis procedure and

accordingly take action (e.g., issue drugs to the

patient). The overall procedure is supervised by the

medical doctor later on in a report. There are other

cases which the system concludes to have the MD to

participate in the final decision. In such cases, the

system sends the diagnosis reports to the MD and

provides an appointment to the patient in the

Figure 3: VD avatar.

Figure 4: VDS experiment style.

hospital queue. The system reads the queue data at

the management centre of the hospital reception.

And assign the patient to the queue. If the Doctor

found the assignment is appropriate (check mark

OK) then the system learned that the decision is

appropriate, however, by certain feedback from the

doctor the system can learn from the doctor’s

feedback. We provide a window at the doctor office

Bodyweight ・ Temp ・

BloodP

Case and scenario a

cold (92%)

Influenza (83%)

bronchitis (64%)

pneumonia (42%)

A moon-faced woman

cold (38%)

Voice Input

Processing

Image

Processing

Output

Ego gram &

Profile DB

sync

ronous

management

Emotion

Estimate Processor

Emotion

Transmission

Processor:

From Keywords a

related scenario would

Keywor

Feel Heavy

Fever 38.6℃

Keywords

with

annotated

Wei

Emotion Processor

Emotional

Emotion

Scenario Catch Cold

A1：Since when？（92%）

A2：Stomach-ache（83%）

A3：Phlegm？（64%）

Patient

Sensor

Touch Panel

MIC

Camera

Image Processing

Voice Input

Profile Processor

profile management

Mental

Knowledge Base

Action Processor

The decision of the next

behaviour based on the

observed emotion and

Reasonin

Diagnosis Processor

MD Scenario

カ

ルテ

カ

ルテ

Carte

MD-

Personalit

Extraction

MENTAL CLONING BASE VIRTUAL DIAGNOSTICIAN SYSTEM - Virtual Medical Doctor System (VDS) Reasoning

251

Figure 5: System architecture of the inference engine.

to fill a sheet of evaluation to enforce the learning

procedure for the system. Such evaluation sheet

would provide a learning mechanism to increase the

reasoning procedure for the diagnosis. However

such knowledge management would be based (i.e.,

mimic) on Medical Doctor A, therefore, it would be

stored in knowledge management on the top of the

management system. So when another doctor is

doing the outpatient diagnosis then the profile of

decision making related to that Doctor would be

used (i.e., recalled). So there is a general diagnosis

and on top of it there are diagnosis categorized on

physicians actual practices.

The paper is showing the state of art in making a

system that can interact with human user based on

new concept named as mental cloning at (Fujita

2009). The cloning is based on analysis of human

medical doctor (HMD). The analysis is projected

using his/her observed styles as a person and also as

expert in medical diagnosis related practices. So

there are different style of categorized knowledge

reflecting such representation and related reasoning.

As shown in Fig.2 and Fig.3. The system would

create a virtual face (i.e., screen mask) of an actual

doctor that through it the patient communicates with

pre-assigned virtual version of that medical doctor.

Physical doctor face is masked copied and attached

on manikin (as shown in Fig 4). Inside it there is a

projector that reflects the 3 dimensional generated

images on the mask screen that reflects the actual

facial real-time created images and voices of the

medical doctor namely, Doctor A (Fig. 4). These

animated facial image synchronized with a spoken

language in the same manner as the actual physical

doctor is doing diagnosis practices in Japanese

hospital. The style mimics the actual doctor

emotional expression as well his/her diagnosis Case

and scenario a cold (92%) influenza (83%)

bronchitis (64%) pneumonia (42%) a moon-faced

woman cold (38%) style (Fig.2). Also, the MD

would speak in natural accent with emotions based

on the patient mental mode, estimated by the patient

profile (age, gender, ego data), and his/her situation

automatically measured by data resembles (blood

pressure, body weight, body temperature, and

thermal analyzer).

These devices (equipments) are assembled to a

patient desk chair that the patient would sit on, and

automatically these measurements are collected

and transferred through serial connection to the

virtual doctor system. These physical data are all

measured and send online to the VDS together with

the mental status data (situation) of the user

(patient), with estimated ego state retrieved from

the databases.

This system is been built by a support from

Ministry of Interiors affairs and communications of

Japan under SCOPE project.

2 VIRTUAL DOCTOR SYSTEM

REASONING ENGINE

The paper reports here, part of our project outcome

HEALTHINF 2010 - International Conference on Health Informatics

252

that is related to interaction between VDS avatar and

Patient. The voice recognition issues is been also,

discussed in (Kurematsu, 2009). The action scenario

is to create a diagnosis based on the guidelines given

by the Doctor A. Doctor A is a nominated Doctor

which is the object system would mimic to interact

with patient through VDS avatar. Implementing

medical guidelines of Doctor A in active computer-

based decision participates to enhance the best

practices of medical services on behalf of Doctor A.

Our system reported here briefly, participates to

provide cognitive interaction between real patient

and specialized doctor A (avatar) through computer

interpretable guidelines (CIGs). We have used

UMLS Metathesaurus® for testing purposes (Fig.5).

The contextualization of the two ontologies

specified namely, by the mental ontology and

physical ontology annotated mapping alignment,

that produce annotated search profiling in the

medical knowledge base Unified Medical Language

System (e.g.,UMLS) semantic net. We think this is

compatible with vocabularies and classification used

in patient diagnosis. The combined usage of the two

ontologies is intended (Fig. 5) to derive semantically

appropriate reference to the patient’s status and

correspondingly the appropriate concluded aligned

key words for relative diagnosis search. The

diagnosis retrieval from the knowledge base (i.e.,

UMLS) would be related to the specialization of the

patient case through the combined ontology. This in

turn would lead to derive the validated

correspondence related to patient condition.

We define the structure of classes that are to be used

in the hospital Class patient: as the class that specify

all types of patient in hospital class and can be

attributed by their: Unique assigned number; The

number can have a structure representing a form that

store the profile of that patient and retrieve it when

necessary, and to update it is content. This code

represents the hierarchy of the data structure of

patient ID.

The patient ID file would have information related

to the, personal information, like name, gender, age,

employment type, tallness, address, marital status,

special information and other type of personal

related medical information. This also would

represent the patient module that would have related

information to the mental state of that patient,

his/her physiological state definition (Fujita, 2009).

This would to establish a template that also stored as

part of the patient module structure. Templates

related to mental cloning issues. Like the user ego

gram, his/her universal templates, and six basic

templates of Ekman emotion and the neutral state

(no emotion recognized state) (Fujita, 2008). As part

of the language issues we have defined other modes

that are needed during the diagnosis procedure.

These types of emotional modes are related to the

effect produced due to a certain combination that the

patient would have as a result of disorder or sickness

of different cause or nature. These states can be

specified as pain(x), x would be a percentage values

that to be extracted from the combination of other

observed cognitive state. We have six primitives’

cognitive states. {Happy, sad, disgust, fear, surprise,

anger}. Each state can be attributed with a value

represents the grades among high, low medium.

These are primitive’s states, due to their

characteristic nature to express human emotion,

universally. Albeit, the degree of exposing these

states are different among people due to several

factors. These factors can be predicated and

estimated (i.e., computed) based on specialization

and symmetrical projection through people

characteristic, observed through other disciplinary,

like Type_Age, Type_Gender, Type_Ego-gram and

so (Hakura, 2009). In this paper, we call these as

stereo data as typed meta data related to complex

representation for emotional states. [Such

presentation would be useful to use object orientated

technology in the implementation.] For example:

TypeFear:(Type_Age:20th, Type_Gender:male,

Meta_Type_perosnality). This is a stereo type (i.e.,

class definition) on Class:fear be characterized. For

sickness related issues, the pain (user) can be a

stereo type of a combination of {disgust (medium),

(sad (low)), neutral}. The combinations of the

emotional related states can be extracted in real-time

from the frame video collected from user images

labeled on the spoken sentences that he/she

articulate to present his/her condition. This

articulation representation would be to express user

physical status or/and mental status using informal

representation in spoken language. This time the

spoken language is Japanese. As a spoken language

it would be arbitrary set of sound words (spoken

words). The sound is a stereotype collecting the

emotional feature of the patient (i.e., user).

(Kurematsu, 2009). The facial and related situational

information collected by the system would reflect

the status of the patient. Here, is the stereotype of

the emotion of the user. So the user status is a

combination of the pair: {StereoType_Voice(pitch,

power), Stereotype_ (face(happy, sad,..), situation

(gender(Boolean), temperature(integer), blood

pressure(integer,integer), BMI(integer))}. These two

different ontology stereotype information represent

the mental cloning of the user for reasoning purpose.

The wording (key word extraction by the mapping

alignment between the two view schemata related to

MENTAL CLONING BASE VIRTUAL DIAGNOSTICIAN SYSTEM - Virtual Medical Doctor System (VDS) Reasoning

253

the two ontologies), are to be concluded by the

system. These wording would construct the situation

related abstraction useful to be articulated to

construct the schema needed to establish the

conceptual view of the user diagnostic situation.

Words collected from user (i.e., patient) and

conceptualized. For example Headache; as a concept

is conceptualized into mental view, and physical

view. Each view would have a set of condition and

assertion to be fulfilled collectively as assertions.

This is a temporal order related situational reasoning

based on integrated views of reasoning. It is sequent

cascade incremental reasoning, based on

regenerative schema at the functional

correspondence mapping. It is cyclic iterative

reasoning based on the same schema’s structure for

diagnosis. The 1

schema structure would be used to

collect the best response, and accordingly the related

diagnosis is fetch and customized with diagnostician

template to readout, to the patient with emotion

mimicking the MD namely A. The response from

the patient would lead to another incremental

schema that is semantically consistent with the

previous schema with similar partial set semantically

consistent fragmented diagnosis outcome, with extra

key word, extracted from the patient response

derived from patient performance due to the readout,

role-act diagnosis initiated by the schema of the

alignment of the two ontologies. The incremental

schema generation from the mapping alignment

constructed through the patient and computed by

system due to the VDS generated scenarios derived

from the semantic net (UMLS), and shown in Fig. 5,

is cascading and incremented schema type

generation, in nature. This would provide structured

interoperability through such integration.

Medical diagnosis process is built by merging a set

of fragments instances of the mental views instances

with the corresponding physical views instances.

These mappings are expressed with various

relationships between classes in the two different

ontology fragments, for machine executable medical

diagnosis purposes.

The VDS final conclusion on the diagnosing

situational reasoning process would also be part of

the readout to validate the patient performance if

he/she is satisfied with the outcome. This validation

process is essential to help the patient be more

interacted with the VDS on basis on collective

engagement for best practices. All diagnosis reports

are also sent to the MD for conformance checking.

Above we have expressed the stereotype

emotion, and below we integrate it with the

stereotype voice. The stereotypes voice is also used

by the VDS to express the question and related

responses for information extraction from the patient

user. The VDS doctor would express these

synchronized stereo types using avatar which is 3D

generated graphics and synchronized voice sounds

as shown in Fig. 2 and Fig.3. The 1

response would

be presented through the avatar using the

representation of the stereotypes (emotion and

voice) mentioned above. Then the collected

information for the patient user is also represented in

the next cycle of diagnosis hierarchy, for another

round of reasoning in the search engine. The

information knowledge is represented as semantic

net and based on the stereotype representation. The

domain knowledge is specified by automatic

retrieval in establishing a link between the ontology

and the patient database.

The word selection is specified through the

keywords. The schemata outline (as above) would

construct the conceptual schema that would be used

to do search in the semantic net based on the

situational abstract articulated on the tow views,

namely the mental and physical views.

The semantic net is constructed based on Object

Web ontology (OWL). The data base is constructed

such that the diagnosis would be articulated on

structural hierarchy. For example pain, with fever,

specify the diagnosis of pain fever class hierarchy,

then pain would be specified as location by the user

answer.

We establish a process of finding the reasoning

of the system using these two types of ontology.

Also this problem is called as ontology mapping

(Kalfoglou,2003), to establish the semantical

relation between procedures and entities among the

mapped ontologies based on the previously

explained stereotype views. We align the properties

of physical diagnosis specified by the medical doctor

with those related to cognitive reasoning based on

the patient mental cloning articulation. Features

expressed by stereotype Meta data definition

mentioned above are used to make such alignment.

The mapping is an abstraction that encapsulate the

features (properties) related to certain abstraction in

an ontology and reflect or map that feature into other

ontology such that to qualify the related features

through such mappings. This semantic level

mapping based on conceptual schema related to each

ontology. The mapping feature would provide

coordination among different schema such that to

establish semantic correspondence for reasoning

purposes, in semantic web bases reasoning fashion.

This would enhance and smooth the interoperability

on service through different schema reflected on

different type ontologies. As shown in Fig.2 and

Fig.5., This mapping is incremental process as the

HEALTHINF 2010 - International Conference on Health Informatics

254

patient produce new schema and also the VDS

consume this schema to produce another request by

which the patient user would correspond to produce

another schemata to be correspondingly, consumed

again by the diagnostician (the system). This

procedure is recalled based on mutual induction

mapping discovery through conceptual schemata.

This is incremental indexing schema because the

process is qualitative driven by the diagnostician. It

represents incremental diagnostic refinement by

including new collected purified information.

2.1 Implementation Aspects

Ontologies used in semantic web consist of

hierarchical description of concepts and their

properties in a domain (nest of concepts organized

due to the nature of the situation instantiated from

the integrated ontologies). Fig. 2 shows the details of

the system implementation.

OWL (web ontology language), is knowledge

representation scheme in semantic web exploiting

web standards (XML, RDF). OWL is suitable to

specify medical knowledge. We think OWL is more

suitable for us than OBO as the later is mostly

suitable for biological application, and its format

does not include all the feature of OWL.

Deductive inference is used based on explicit

given knowledge, represented by description logic,

like RACER www.racer-systems.com/ We envisage

using the OWL axioms and class constructors as

descriptive logic for explicit inference from medical

knowledge base.

For example, there is a correspondence relation

between the medical process integration “drugs” and

“medical decisions” domains. Consequently, there is

semantics alignment integration among the instances

semantic selection between, process decision and

process of selecting the best match instances of drug

property that fit to the decision to be expressed to

the patient. (as shown in Fig.2).

We envisage templates that can be used by the

diagnostician for matching diagnosis use cases to the

drug prescription databases impeded in the

knowledge base. This would be a sub ontology

derived from the aligned two ontologies, (mental

and physical). We envisage using a sequence

diagram with association rules to examine

interactions arising in various patient scenarios to

establish the validity of diagnosis procedure as

approved by the diagnostician.

Discovering association rules is an important data

mining problem. For example patient who has

running nose can be either allergenic side effect or a

side effect of cold sickness. Therefore, the main

purpose of implementing relationships in the

knowledge base is by analyzing the data as reference

during decision making.

Protégé UMLS plugin supports browsing and be

use to query the UMLS knowledge base for medical

keywords and retrieve them in terminological format.

We construct domain ontologies by entering

diagnosis related data and storing them in formats

standards like OWL. The formal of OWL is

provided by description logics (DLs). DLs are used

to describe structured objects whose parts are

interconnected in complex ways, such as medical

diagnosis. OWL facilitates greater machine

interpretability of Web style content than that

supported by XML, RDF, and RDF Schema (RDF-

S) by providing additional vocabulary along with a

formal semantics. Instance data pertaining to the two

view mapped ontology is making the assertion in

terms of DL using Ontomat. OntoMat

http://annotation.semanticweb.org/ontomat Public

annotation tool is used to build OWL instances, and

attribute relations, participate in the logical part of

DL.

MD doctor diagnosis routines, as well as the MD

personality (mental view) are added to the action

scenario related to diagnosis (Fi.g.2). All these

knowledge based are to be represented as a concept

in DL in structural formal way using the stereotype

based views mentioned above. The reasoning would

be reflected through what we called as diagnosis

map: reflects the clustering of different knowledge

map articulated through the mapping of the two

previously mentioned views (mental and physical)

shown in Fig.5.

Also, there is a correspondence to the patient

pattern: behavioural pattern that is reflected by the

mental view ontology along with the attributed

values of the physical view. Recall that the physical

view is values reflecting by the physical status of the

patient. This all resembled in profile processor

shown in Fig.2. The strategy of changing

diagnostician (i.e., MD) routines would be affected

by specified instances of patient observation, these

related observation would be categorized in a region

that to be confirmed by actual MD, like blood or else

physical collected data analysis. In most cases MD

chose from limited evoked set of drugs which comes

up in their minds, given a certain health problem of a

patient. This evoked service is influenced by MD

mental states, background and other local

parameters. MD usually does not consider all

possible treatment options, but chose approximately

among two to 5 different options reflected to his/her

experiences. This is resembled by the Emotion

Estimate processor in Fig.2.

The knowledge base diagnoses scenarios can be

revised and updated based on new diagnosis

MENTAL CLONING BASE VIRTUAL DIAGNOSTICIAN SYSTEM - Virtual Medical Doctor System (VDS) Reasoning

255

scenarios and participate to revise this efficiency

related issue. However, it is still be approved by the

MD in order to be used by the system. We currently

are collecting these action scenarios from a hospital

case study in Iwate region, in Japan. We would

report on this in another paper.

3 CONCLUSIONS

This paper is reporting a progress status of our

project related to mental cloning based concept on

how to reason and represent human emotion in

scientific way and use that emotion to reason with

human user. We articulate such realization to

establish a virtual medical doctor for diagnosis

purpose. The MD is a real person that based on

interviews, we extract her/his personality that is to

be used into the system and act on his/her behalf on

mental basis using her/his routine diagnosis

procedure (knowledge and scenarios). Using this

with other related information we created a system

that interacts with the patient user based on

Transaction analysis protocol. The system would

be examined in Beta space at a hospital where that

MD is working. We have represented the patient

mental view and physical view. We have aligned

and mapped these two views to discover the best

integrated correspondence that resulted in a set of

key words that would be used in searching the best

action scenario relative to patient case. The

discovery is incremental and cascade. The

implementation outline of our system is presented

in this paper. All diagnosis knowledge of MD is

stored in the knowledge base as semantic net and in

OWL. For testing purpose we have used UMLS for

knowledge based, as it is based on semantic net.

The mapping is based on DL. The system is under

construction and to be installed in a hospital in

Iwate region in Japan by 2010.

ACKNOWLEDGEMENTS

This research is supported by the Ministry of

Internal Affairs and Communications of Japan under

the Strategic Information and Communications R&D

Promotion Programme (SCOPE).

REFERENCES

Annette ten Teije (ed). 2008 Computer-Based Medical

Guidelines and Protocols: A Primer and Current Trends

(Studies in Health Technology and Informatics, IOS press,

ISBN: 978-1-58603-873-1

Fujita, H., Hakura, J, Kurematsu, M., 2009 “Intelligent

human interface based on mental cloning-based

software” International Journal on Knowledge-Based

Systems, Elsevier, 22 (3), pp. 216-234, April.

Fujita, H., Hakura, J. Kurematsu, M. Chida, S. and

Arakawa, Y.:2008 “Empirical based Techniques for

Human Cognitive Interaction Analysis: Universal

Template Design”，the 7th New Trends in Software

Methodologies, tools and Techniques (Proceedings of

SoMeT_08), pp.257-277. IOS press, ISBN: 978-1-

158603-916-5

Hakura, J., Kurematsu, M., Fujita, H., 2008 An

Exploration toward Emotion Estimation from Facial

Expressions for Systems with Quasi-Personality,

INTERNATIONAL JOURNAL of CIRCUITS,

SYSTEMS and SIGNAL PROCESSING, Vol. 1, No. 2,

137-144.

Hakura, J., Kurematsu, M., Fujita, H 2009 «Facial

Expression Invariants for Estimating Mental States of

Person» Frontiers in Artificial Intelligence and

application series, Volume 199, New Trends in

Software Methodologies, tools and Techniques

(SoMeT_09),. IOS press, ISBN: 978-1-60750-049-0.

Kurematsu, M., Ohashi, M. Kinoshita,O. Hakura, J. And

Fujita, H., 2009: “An Approach to implement

Listeners Estimate Emotion in Speech” Frontiers in

Artificial Intelligence and application series, Volume

199, New Trends in Software Methodologies, tools and

Techniques (SoMeT_09),. IOS press, ISBN: 978-1-

60750-049-0.

Leflar,2005: Leflar, R.B and F. Iwata, “Medical Error as

Reportable Event, as Tort, as Crime: A Transpacific

Comparison,” Widener Law Review 189, no. 25

(2005)

Mikkola, 2003, Hospital Pricing reform in the public

health care system- an empirical case study from

Finland, International journal of Health Care Finance

and Economics, 3 (4) 267-286.

NIPSSR, 2006; National Institute of Population and Social

Security Research, “Population Statistics of Japan,”

2006; Tokyo: NIPSSR.

Horowitz: 1988 Introduction to Psychodynamics. A New

Synthesis: By Mardi J. Horowitz. New York: Basic

Books, Inc.

Kalfoglou, Y., and Scholermmer, 2003 “Ontology

Mapping: the state of the art,” The knowledge

Engineering review, 2003.

Tanaka, M., Noguchi, M 2000: Concept Retrieval of

Medical Text using UMLS, Japan Journal of Medical

Informatics, Vol.20, pp.934-935.

Smith, 2000, Reforming Markets in Health Care-An

Economic Perspective, Open University Press,

Buckingham, 2000.]

Trautmann, R. L. & Erskine, R. G. 1981: Ego state

analysis: A comparative view Transactional Analysis

Journal, 11, 178-185.

Weizenbaum, J., 1966: "ELIZA - A Computer Program

for the Study of Natural Language Communication

Between Man and Machine", Communications of the

Association for Computing Machinery, Vol.9, pp.36-

45.

HEALTHINF 2010 - International Conference on Health Informatics

256