An Applicationof SoftComputing to RETS: Rheumatic Evaluation and Treatment System by OrientalRheumaticMedicin

An Applicationof SoftComputing to RETS: Rheumatic Evaluation and Treatment System by OrientalRheumaticMedicin

SIIg BMFsA(2oo6-i2-oi-o2)

BiomedicalSoft Cornputing and Human Seiences, VoL.11, No.1,pp.7-13.

Copyright @ 1995 Biomedecal Fuzzy Systems Assoeiatien

Evaluation and Treatment System by OrientalRheumatic Medicine

and Hoang Phuong NGUYEN""""Katsuari

KAMEI"*

*"Dept

clfrer usingjuzay iizferences, RE7S diagnoses the most suitable rheumatic state in which thepatient c\lpears

networks,77'om skitted OMpltysiciansand OM text books, RE7S givesan orientat prescriptionwritten in

suitabie herbswith recxsonabte amozints Finally,we clescribe evaluations and restrictions ofRE7:Sland

1 Introduction

accounts for 15% of all soft tissue diseases The most

rheumatism are physicalrnethods, anti-infiammatory

and oriental medicine (OM) Among them, OM isan

than western medicine and gives good treatment

results Besides, herbal prescriptionsare easy to find

and relatively cheap in comparison with western

OM isabout 50 %.

system such as RETS basedon knowledge from

skiiled OM physicianswiLl he]pmoderate evaluation

in rheumatic diagnoseswhich tend to besub.jective It

maintain and share their profound knowledge with

col]eagues and to assist medica] students or young

rural areas

Shiga-prefecture,

In the last recent decades, the advent of the

computer has greatly stimulated developments of ES

which perfbrms the roles of a specialist or carry out some works requiring specific expertise Therewere

applied, such as medicine [2-4], geology [6],

chemistry [7] and business [81.

symptoms The number of herbs for rheumatic

symptoms on patients, doctors diagnose and classify

states of rheumatism, then give corresponding herba]

1 shows the process ofdiagnosis and prescriptionof

rheumatic treatment by OM doctors Such a process

can be suitably assisted with an ES as shown in Fig 2

explanations, sample prescriptionsand training data.

Fuzzy Inference: Checks rules calculates weights and giyes properrheumatic state Inputs to

symptoms and ruEes, and one output that is themost

serious ofthe theumaticstate that the patient has.

Neural Network: Gives reasonable amounts of

severities of observed symptoms, and outputs from

7

Biomedical Soft Computing and Human SciencesVol.11 No.1.

NN are coefficients ofherbal amounts in [O, 1].

severities from users and shows resutts written in

prescrlptlons.

rheumatism, ES and explains results from RETS.

POi,i'ikE/,:.,i,//l.:[:"TiONS

Pulseexffniination

.Dlagnose

±

Adiseasestate[sdete[mned cjin/ea]expeFiences

"'

l standaTdpFescriptiOn Herbaludiustrnents

'

FINALPRESCRIPTION

ll:i:L//1:'/1/iYgrrMn".

Herbkl.'.1.'. -.gra71

Fig 2 Structureof ES for RETS

symptoms such as high fever,slightly numbed joints,

moderately yellow urine etc These fuzay expressions

of symptorns make itunsuitable for traditional

vague variables using membership functionsrather

than with crisp vaiues, proyed to be one of the most

December2006

They also enable developers to use linguistic

variables and build a friendly user interface OM

with such expressions as `Cthis

typicalsymptoms with these severities, so I prescribe

those herbswith those weights." These expressions

can be represented naturally in IF-THEN fuzzyrules

In addition, fuzzy rules can giye expert-like

explanations, and make it easier for doctors to

understand the ES.

applications inmedicine have been built, including

OM [2i and rheumatic disease with western medicine

[4].

Then in the second stage, the fuzzy severities of symptoms are put into a corresponding neural network to getan appropriate prescription.

symptoms, l= 12rheumatic states A rheumatic state

the clinical symptoms will not appear in one rheumatic state, the value of n may yary depending

on each state, fbr example n =: 16 for state 1,n == I3 fbrstatc 2,and so on

symptoms on a patient where S,O･ is a fuzzy

rheumatic states

Let si{J==(s]RJ,s2RJ, s,R,J) be a set of symptoms in the premise of rule RJ-(J' 1,2, ,l)

where R.i is generally described in the fo11owing fbrrn :

IF Sfi'andS2R' and ,andS.it

THEN therheumaticstate is II

r

number

Si'bedefined:

Thang Caoet al.:An Applieationof Seft Cemputing to RETS: Rheumatic Evaluationand Treatment System by Oriental Medieine

R

rheumatic state Hy given by skilled doctors via

survey in advance, wherc:

'H.i,

affecting H,i, and O</tsitj <1 means S,RiafTects

Hi with certainty factor ptsfe,

Forexample, symptom "Afraid

ofwind" is the most

no effect on rheumatic state 3.

rnaximum beliefdegree ofthe premise of ru]e R, is

consider the importance yaiues within this boundary,

avoiding many doctors freely express the importance

of symptoms with theirown view

i LtsR) /.

10.17

Ima

O.25 O.12ee

O.22 s,

symptoms in the premise of a rule In this case,

symptom S7 is more important than the others S?

affects Riwith thc certainty factor pts,ft,= O･25

If an observed symptom S,O is found in the

M'<, ="sf) op ItsitJ' (i rmr l,2,･･･,n) (3)

where op is a t-norm operator, xopy=(xxy) in

RETS.

If symptoms SR' of RJ match observed symptoms SO , conclusion weight of the rule RJ

(denoted by u,R, ) is calculated as:

WI{J =s,

.{s g, ?J.so} MJg,

(4)

where e is a t-conorm operator, this t-conorm

should be compatible with (1).xe}, (x+y) in RETS.

For example, rule Rs has 6 syrnptoms:

observed on the patient: Si, S2, , Sii.Suppose that

fbr Rs as in thecolumns2 and3 on Tab.1 Then we

have theactua] effects (u･tt,,)of each symptom in

weight ofthe rule wR, as in the final cell ofthis tab]e.

Tab.I.ExamleofCalculatinRuleConclusion Symptomlrnportanee

value"s,1jSeveritiesUs,o

5li,si

Zl'.s･f.{fi=1WR5=,s,,.{sORin,go}W.St,

-O.675

Then RETS finds the most serious rheumatic

state H' which has the largestconcLusion weight wR,

among l rheumatic states:

H'-{HklW&-M,aXU{R,} (s)

rules are as shown in Fig 5 In this case, the most

serious rheumatic state is the state 4 with wi{, = O.55

iliseeIE

I"

gge･, ll T //,?/x,,,s,#13,/ee/r,,liliii i i･piii , i･ilii'

ITiil･i l･

t, ¢ e e di $ ee val m esb l - de th･mut ee ua' {"' " emp fi ep ep e ge.

Fig 4 Example of selected symptoms

9

Biomedical Soft Computingand Human SciencesVol.11 No.1.December 2006

with one or more rheumatic states, system finds the

most appropriate rheumatic state H" corresponding

with any rheumatic state, RETS givesan advice about

the closest rheumatic state In this case, patient may

fx

,f N

l 1 ST,XRT 1i

.il

'K

./.7

Fig 7, is used to adjust amounts of herbs in

with 12 rheumatic states, Input data to NN are

state-specified symptoms S,9 (i 1, , n) with severities

ck (k= 1, 2, , , p) in [O,1] In RETS, NN often hasn

INPUT.F,AYER I-IrDDENL,XYER OUTPUTLAYER SeveritiesorspeciricsymptenEin[O,1] Cvefllcieetsofamountsofherbsin[O,11

"s,

Getinputsy!nptoinsSO

ebservedunthepat ÷ ent

mptomshtomatch

withonerhcumatic state?

No

Yes

Findthccloscstrhcumatic

statehavingmaxirnum

Y'it.,fuz4'value

1Giveexplanutionsuboutthc/

closestrheurnat ± cstaLe

arcmatchr

[

E

i"t,zs 'i

:{,:.,,i,:g,g"///:,:,::･:･//:s

to

llS,

･-i

ENDY

,is

-Gctprcscriptienfiomc/iitputsofNN -GeLcxplalnationsabeutthcrhcumaticstatc

/?i.splay results

Prescriptions

analyze, model and make sellse of complex clinical

data across a broad range of medical application [5].

It enables intelligentsystems to learn from

experience, examp]es and clinica] records, improving

symptoms and herbal amounts based on typical

sample prescriptionscollected from experienced

symptoms observed on the patient,

data are norrnalized as:

ck=nj/ifS (6)

vaVP=ek ×ur" (7)

where nj is the arnount of herb k in training data,

nj' is actual amount of herb k in the prescriptive

results, and M* is the maximurn amount ofa herb in

''i

== 20or 60 grams.

from experienced doctors in Thaibinh OM Col]ege,

we haye assessed irnportant fuzzy va]ues of

symptoms in rheumatic states, chosen standard

and equiyalent herbs,selected specific symptoms that affected herbal adjustments, then generated 12,OOO

severities of the state-specified 'symptoms using

ranges of herbaladjustments Training data for NN

rheumatic prescriptions from theexperienced doctors.

survey, we built RI]TS in Visual C# running on MS

adopted a sigmoid activated function Adaptive

Thang Cao et al./An Applieationof Soft Computing to RETS: Rheumatie Evaluationand Treatment System by Oriental Medieine

andexplanations,respectively

adjustments in the training data.rangesof

herbal

reasoning and explaining capabilities forrheumatic

treatments.

nonlinear relations (real prescriptions and rules of

accuracy of 1o-2 mean-square error for both training

then shows the advised prescriptionwith appropriate

amounts ofherbs by NN Most of these prescriptions

are comp]etely compatible with the rea] prescriptions,

We built RETS: Rheumatic Evaluation and

system by fuzzy inferencesand herbaE prescription

system by NN Then we could confirm that RETS has

narrow domain of expertise, RETS is developed for

other diseases besides rheumatism, doctorscannot

solely rely on thissystem since they do not have

evidence to contro] potential effects of the herbal

remedies on theother concurrent diseases Hence, it is

recommended that the system be used on]y for

other concurrent diseases.

system in therea] patients and comparing system's results with the doctors' diagnoses

'

e.e#.neeeed/4tm'utliDsl AbEu.l

: /: : " : X :

{,S-fie"-1rnftt.":le: , i･'seeeas i'lt1111I･l･･/t

pFtilo,,wh.t/A･g,･･

/;g.ex1

t .

-/1･;ur :ffdeCaoThengFrm.pde

ft di

S]idctbrsettin!theseveritiesof

theselectedsymptom from thclist

.･t･ ti de wn

･ '･ g 'gisi is.,2･･ ,,

"es･m.i.e.c;{.t/.gg.vaggil::;i;$e

s di w tr w' w W.vr.T :.: .tm.mm

"' /2 m'me ss :

e e p pt' t 'mpethth' tsl:it

rtsikgtiedi"-th-thXe v ･-ny-ee･ di ec '.'Fiva e ･rejt･

rm.e m e e t pt wt s ww wpt 'tajiSfecgniede.e.i?sg&ieempvrwwrsemeeeee1Wwee ptps"i".-.± ididw

; 1:'

1-11.1:it

i

le'

/. 25Revengi ･ ･ -･･?E teihaagvimev' # FyeTHEregyswttwg' de '

di/lg mame'ee de 'eewh 'dei'ee ir. th･ S.

1e, Jqlntsvrti beU]loddtcd Ev)olie'npt op en W'ge.

･ ･･ -tsu naan su･･

'

9 Jo/nls:d/'l'ls;vo'ltm.-/hot/thoL"t-lomi-min ･

,Sl'!i,',,,ny',i'hs

"kn/iliir,s",coFdi.',ghgi,II'i'l,,

･igee mTouNGuE.patpAIIoNs'nerwcptxeasvaeeiiM '1-14.HandtEhd'fe'etatecol(tiaggasS\iilg

1 ･IE L'frkjd/o/･1-wl'id' ･ - . e ew pt ts' 'retff

･le"'nl:efaz-,･ mOSS ' '.

17Wti.rteasttc(S t?-nple mo,sg g gw::' ff'.

./ l, /// l･i //Iitlg.tlli,3111:,li/

l' i' l l i l･ i' ll･ I-1i' l･,A･

l .2i EaSte:ng.hgrd'pui'se''/ : : : ':

1/-22,Hartimlilse ., t･ts su ･sdi･isma･

in.m x

ma de･-,/･

in

ti-di as ere･vemveeew

.t.lat.g ut･t.

t.t

las

t'ti de/ge ･ve･wpm

vafimin

e/eeeeeeeveeva

.sS fffiva:a / ;1 ;Fentvaru#,vb:; :

f}1 asnueveteswa e

iF-eeeeew gee

1/,WWwwwv ,- va sltgzalime we w

l el

v.

ma digegagesg geempssg : , . g ff

li,i'

ll/kell'ili'

iX･l,:l･/i

g}･g･g,pg･x/:･ggs･k-gii-,T'IS･,

[llllll

i

i"'u"

-ag'k'

i,tts,ut,T,1/iov i,' ,,ci'

ttlj･kTi il}il･i.i･ll '

l "

l'

l'

ll･l

Selectedsyptomsandtheirseveritiesonthepatient

v, za fi tu Ide de in it･ diedegeg,ffva va.'

BrLeflvlafitsgttrtspasdisg -tgngwaeetwg "}

ts::･::.ww.eyw- th wb.

/:

1:di1e･

,,tkeX,e･/9,/･.

lliri'/ec,as,.i,,i lllifi/k/ifa

.ee/

i

i

'j"

fii?'/l,i'

li/iC

i'/i'

Sli

ffre

tvaeeee weeee･eeepm

fg ggg $ ww.i.ma

,

･wny ww

eeeetaClinicalSymptoms

i:Me-im-th-ee-wwtw-ew-w-ww-ma-/:;"wh nt es de leth diea ¢wdieecaW/

v de"th de skia ee e.g#ggsgvg,

e ip p･ de inineIM ifeijxe re W e･WW'

filiiialmgi :, x, gileeit

ee/es es ee ww - .

::gsiei{

11

BiomedicalSoftComputingandHuman Sciences Vol.11 No.1.December 2006

1' esISiillilllg$#leei maee /wyes/veiiv//',ei i asi#l:ll!ISigeSIFiliiiuei#lliliiililge;IIi

･g#･ggeefigff,g'

ei'

;;;E

mesen

il･IA･:-!,:#k 'e･wii･vee

g,#/,,it･:,z/kts ･

iz.:,･is:w･,a

g･

"e

,S4rwhrtetsneue, g

wrVSsuSM"meeunw

easSoUtieme pulw- su w nyeive de di th n et va - nv

$lkgv e[g/.ff]t./ /xw.

gellee ee

eee di th nt e t p ve tu

vaffSkS4m aj "e2ema vgeWeedidi didi aj di} ane wva w ve ee w va et ca otk4miut

deee va va vavaecretawpmaas

mpgeewveveeeeve mame ee wts ee W va su eneeen

tm Rheurnaticstate-specifiedsymptorns

andtheirseverities

re eeB ee l

}"vvalet

ttva

/as

gang]1v

esme

ane}

asvi:diei ･siee2･

ii'ire

# mp

ili･

I

'ifgi'

i

'

tiliSifi

e･hg/ik･I'g/gii/i･i//i/

Mg･J/

di tv"

l/1ee

utwtptee

ssff

tme

tt

-BriefexplanaLionsabotittherhcumaticstate.m.i

Treannentherbswithrcasenab]eainounlsl'ortherheumaticstatc

{eeeeeereeets

e va va th as

labr,erwwoTke

speurtiEvat" de "

ee} ee tuas s'

rimplww gedi, rtdi

illl r

-il'x･

.illi

･

;F'

E,h illigile,/w,,\･lg･,/g,lr/T,iia･l,T,li･ il, iiit tsiiesti1?'

ge

gte,$s,;,Eee,xes sst ･

it.th

ffefieRag

tg".de S.""gn,,,egah,mamopeeqI#Errorgraphintrajningprogress

ii,iiizi s.gk,i,,?iy.i.,ee-w

,i3jz･,e･g･ft,slg

"

:enlar/,as

[weights scee za

:.:-.x-.rv :x･ter:･ne.#a,-gssg-.Gg:

,Wee eewassggsl:

:･ws

be de

ft di thma

asts-

i,mp1

ve /va'es imp

ng

ewwwuavaN'W'va

Mlffdenttata"-ri-i

ft {' thewts me mt de w t tu

di tE"e eefi#e 4.

"wnvvvva eeM t ts di

ut mu-wh e e e ma

va eweee re ts, dipt endi Mesascaas"seede.'tgve egepm,evemsee

ll/y･,ede

va"

,la,#ill･lli･.;i･o l,g 'mp

i i ll' Sl

-iiiiii " t///i,,tk/iii#,ssi,l//･I･lil･llt･lsii,g'

i, i･,yl,ill･i ' X

N

1111'l1

''i

severitiesobserved onthepatient v'bggyfpmecnee km estu

su th ni di tigna-"

:ai,i:/xig･ ,wy.',,x,ixl{/{,i,gr,l/g'tspV.,

'

iilS'tl'

-･ epe::::i

:, iigggifT/ gl :g;'vaffggge :: ::

tvatin"de wn-t

i$asPSies-M.ma.-#ewwwme,.,.:

1an det tde"al.th dethtt

:is･,e:,: ge f

iwkwadie:Lct,.X:ggg#6:g:;,.

i #tgXi$s as" ep, de.

m gEgee,de.

xsi :

l #eggesijdiyGz g ee # #g g ::ii g

in mdithstwhregfiff e veee-dieeety lasema ma in ve metsWptptsu -'MfftiedidiTenmse

su e es e.Z"m s w 'W{4 th th w

'Ii,k[:,:･tTs?ltr!111g,igl,ges･sts'ikkasig,t,'gt･."･i,,i･,kgi･S#,k,g,ggi'

ess･}e- l-,ww:･/:g,･ii'l,e,.i,g"cgeij/gptw-.:ift:},gl/isgilgeglei･v.gli,!g,k'i

･eq gSgg･gg9g･9fY/di ℃ g'S{eSeeby"7re.2exl:`z"!i .:: g2 gevny･"

ts,l'/eu,/k･/stg,S g'ESfijin i.,ce,ISS.ggst,gs'la,srg,g･as,ngg･g'g.':rp.,,,':t:,gg,4,i"

g",es"i.

'

lki,/wyik･# ,//it･,elilie ･

,g

'

,gk,

v

.twlgi,,lii:,i,ilis

i

s･,wilo

,g,klitiii'

wwwumma/W en ma en en th me ma m. ,

imlmenwggxve.eeiS#lg

'\ig#igee.i,gffl

v

ffedere

fi

.kfo･geaza : ; ggge# g gi, gigdien : :gx#2::g$x:

gdiaststsas.dieees as ee twgee ee eesi es

va{wwwmp tmwigweewfive va wtW

WI.igPrtutnepimJcrIVSwde mpen?wwu n su mptmp en rm e va ww rpweW op dn

pt1{etrEhtselele) - - ve-o-m ve

sulfiMtha -t-ur dith th .

i#･:fi #,ss･ ,ks- g gg #:s g,,i

eegge eeeiaseeeeff va

ee ee utIee mpeua eentvaeee uaem eua me mel ee

eeg ee va,ee va ee as ee ma me #Ik

w{ vam wee vage eeeeme,w

Advisedtreatmentherbs '1:

!

1di withreasonableamoullts

-lt

-'ajopdiWeeutde,mmumeffvaeeeWe,va

i g･ x issec mei

m'veengdedeeut,deptpm-mdemamefi.,

References

2001

[31

utilisation of fuzzy tcchnology in Medicine and

331-349E.H.

North-Thang Cao et al./AnApplieation of Soft Computing to RETS: Rheumatie EvaluationandTreatmentSystem byOriental Medieine

Holland, New York, 1976

261, 1993

1994

Limited 2002, pp 12

Lectures in Oriental Medicine, Medicine Pub.

***

Thang CAO

1994/BachelorofElectronics.

Hanoi Universit},ofTechnology 2005/ MS Ritsumeikan Univcrsity

enas

bee ifde

'

.l

Eric W COOPER 1998/ MS, Computer Science,

Rits-ncikanUniversity

Univcrsity2003/

Jeined COE prograrnas

post-doctoralrescarcher

2006/ Associate Profossor, RitsurneikanUniversis,

YukinobullOSIItNO

2002/ Doetorof Engineering, Ritsumeikani Universlty 2002/ Assistant Professor, RitsumeikanIJniversity 2006/ Associate Professer, Kochi L)niversityofTechnology

KatsuariKAMEI 1983/ Assistant 1'rofessor.

Ritsumeikan L]nivcrsity ]993/ Associate Professor, RitsumeikanUniversity

University

HeangPhuongNGUYEN l979/MSc,TashkentState

Universits'

Scicnccs Teehnical University of

Vienna.Austria.

200S/Directoroflnformation Tech Center MinistryofIIeallth

ofVietnam

***

13