AD-A142 726 SUBTRACTION RADIOGRAPHY FOR THE DIAGNOSIS OF BONE •LESIONS IN DOGS(U) ARMY INST OF DENTAL RESEARCHWASHINGTON DC M P RETHMRN ET AL. 31 MAY 84

UNCLASSIFIED F/G 6/5 NL

mEEEEEmhshEmhIEE0IEEEIIILll-llllll

II1.0 Jim 328Elii'.-

jI.2 ".4 I

MICROCOPY RESOLUTION TEST CHARTNATIOAL BUREAU OF STANDARDS-1963-A

ilg ili il Ji l lll)

SECURITY CLASSIFICATION OF THIS PG We e.Kard

REPORT DOCUMENTATION PAGE FOKCMETNORIREPORT NUMBER 2. VY ACCESSIO6 NO0. 3 RECIPIENT'S CATALOG NUmUeef

I TLC (d SubItil) 6 V P fOF -POT, AP -3.R0 07 ,V~ 1 3 ~lSubtraction Radiography for the oiagnii;s ) t' SIIII,11sk01 o -py-

Rt one. Lesions in Dogs 11,4 -OG. 3(0?NUL

AU THOR~q caraACT 0R GRANT NUMBER,&)

* * __ M.P. Rethman, U.E. Ruttiman, R.B. O'Neal, R.I.% Webber, A.A. Davis, S.C. Woodyard

S PERFORMING ORGANIZATION NAME AND ADORES 10 PROGRAM ELEMENT. PROJECT. TASK

C US Army Institute of Dental Research AE OKUI UBR

Walter Reed Army Medical Center 027?5.A, 3'l,6277A825II Washington. DC__20307______________

CONTROLLING OFFICE NAME AND ADDRESS 12 REPORT DATE

US Army Medical Research & Development CommandHQDA- IS 13 NUMBER OF PAGES

FortDetrick. MD 21701 _______________lb OIORING AGENCY NAME A AOORESS(Ii dlierentfkrm Controlling Office) IS. SECURITY CLASS. (00 this report)

(INC LASS I F I EDISe. DECLASSIFICATION DOWNGRADING

SCHEDULE

16. DISTRIBUTION STATEMENT (of tis Report)

This document has been approved for public release and salc; its distributionis unlimited.

17. DISTRIBUTION STATEMENT (of the abstract onteted in Block 20. iI diII.,.nr from Report)

If, . SUPPLEMENTARY NOTES

P~s Radiographs, computer enhancement JL0r

20 ^SAACT Continue an revere, aide It necessary and Identify by block number)

* Resolution of osseous wounds utilizing conventional radiographic techniques isdependent upon favorable angulation without superimposition of densec anatomic-al structures. A technique of computer subtraction utilizing sequentialradiographs has been demonstrated to enhance visualization of such defects it)dry skulls, but usefulness in live animals had heretofore niot beei demoni-strated. This investigation demonstrated the usefulness of computer subtrac-tion radiography in live animals, both error rate and diagnostic time being

DD I JA 7 1473 EDITION OFt 1 1 V 66IS OBSOLETE NLAS1:1Eii SECURITY CLASSIFICATION OF THIS PAGE (Whein Date Fnrerd)8407 05 0611

q~~~~~~ - % V'. .V.-'. V .. * % *.*%* .

4!

DEPARTMENT OF THE ARMYuweo STATES AuiV mSISIuTE OF MNWIA. R*S& eeesgsoa r -

WALTER A19D AMIV WIEDCAL CEWIlEla

*AS4410" De "0301 XJI Gl &*a fftVft~g foMjy 1. 984DTIC TARfMlay 1 , i9S4 Ukannt eduncd

Just itcat ton

Professional Development Div alst

(D Availabillty Codes%! The Editors Avail ander

Journal of Periodontology Dist Special"4 211 East Chicago Avenue

Room 924Chicago, IL 60611

Dear Sirs:

I m submitting an original research article titled "SubtractionRadiography for the Diagnosis of Bone Lesions in Dogs" solely to theJournal of Periodontology for review and acceptance. In light of yourrequest for transference of copyright rights, I must point out I donot have this right. This research was conducted while I was a memberof the United States Army Dental Corps and as United States governmentresearch, the article cannot be copyrighted. It is freely availableto you for publication without restrictions on your use of it, now orsubsequently.

The United States Army Medical Research and Development ComndJudge Advocate advises me that the following disclaimer should be pub-lished with the article:

The views of the authors do not purport to reflect thepositions of the Department of the Army or the Departmentof Defense. (Pars. 4-3, AR 360-5).

You have my assurance that I retain no rights in the article.

Sincerely,

MIOIAEL P. RETHANMajorU.S. Army Dental Corps

Enclosure

Correspondence regarding this article should be sent to:

MAJ Michael P. RethmanUSA DENTACFt. Monmouth, NJ 07703

4' .Vm? 3. d d .. .J* *

SUBTRACTION RW)IOGRM'WY FMB DUi lilA iM % 017 b*III U"SIO IN 40GS*

H.P. 3eghe a0 U.E. Stuggina". It.$. O'aVI**%~ 1.5. babber""%A.A. Deas . . bodjyard'*~** '

* Commgmiaj mterials and equipmt are ideatified to this report tospecify the investigative Procedure. Such identification does *ot implyreiememilation or amdorsewmt, or that the materials and equipet are

oesawily the best available for the purpose. Furthermore, the viewsof the authors do met marmort to reflect the posit ions of the Departmentof the Aiuy or the Department of Defense."-

"140jor, K,. USA, US Army Dental Activity, Fort Ollinouth, NJ 07703

01"eanior Staff Fellow, Diagnostic Systan Brench, WIGS, KIM, Bethesda,NO 202S

"Liouttmnt. Colenel, K . USA. MAt. Director, US Army Residency ProgrmiA Periodmntics, UBAIDS, NMMC Washington, DC 2030?

~'Oief.Diagnostic Systems Branch, WION, MIN, Bethesda, ND 2020S

***Dapostic Systems Branch, PIOS. WN, Bethesda, 00 2020S

*****Colnel, SC. USA, Director, US Army Residency Program in Perio-dostics9 IMAISS, umMM, Washington, Dc 20307

The authors wish to thank Mr. Walter Franz of the Division of Instruments-ties, fater Seed Army Institute of Research for technical assistance and16. Lowanda Thoan Of U.S. Army Institute of Dental Research fot administra-tive assistance.

% % A4 V

* - - ** * - - * .* -

4~ S

4 54

40

4

USTMCT1 SAWAMW IU .1A13 OP 3~ LBS IONS 3~ ~

- -

studies Imb dry 64*4 injdibIw bAVV 404A~irge'd 1kV*~~ri

of subtaiso reatslgrapr to detcta.e Iff4weJ it0iaine Ow

convestial 01de'bii#d comfrta of see rediogroob.

4;Tbe pups of tht .tsad1 as to eornheorit.te qk~ ed04e06

aS Itwe anissi ed. to a. *&Its dog*. 1#61ast wor. &*"*Ae

w~wir awlckme i. swihbiar alveolar bows as 14 prvdh~mew~d

foiles, 7b* otweli P10-4611111 of a le0wSO pvw..c. of a

psn Icuiar site *a* iML ft*- mad poetopen, at rada06V400

"ems to&"s mith the aid of a (Istomited occveul tmpleoe

60141408 Me 11-8 dft. d &IIimrq 4 rigid SwIWIAce #ttecheWMf

to The S3.tsmw hV@P~a- post .operet itS effpapS *Wer

ismed #0 pars an pro*"~ *ast it eMtiasis tot 0masetioe.

A copten rmimaaed th* efdof of proseetattoo Oad Ffwf04

the aboot to eamin asedicaeted site, solicitms a ISlet.I

graed rwsP~' f0*1iq trm lesion defistio pro Is to

1"u00 deISiteip ebses. Own, 116rcftiet rweepbs Otv

presefted so a ideo stv" dad possible lesi.. site$ meted

by dittles Oft OfA tf IS is uM116 SwqeestQtIit fWSPOWAq

ad Oocisiom times aee lot~r by caW fte. biepstit

Sewrae wans mswfted by feeit e natt gChatctetitic

(MCC) mosipsig. iadiiiwl pooled it Iuts d"Mmssteated

ioewd adeguetic peef"Ofe for the 9"Mrsctio tectmiqw

(F < .001). e'p.eties Owv &ISO lmoweud by the 546te-actu'v1

tecomiu (P .01). PseeOle, an ,.alysi, SkOWed tkat

a - - a a - a - . -

.4'I.

Er#4'

b~ dsgauat ~wh(fR5.J E~hm~~.r a ~ ~i

J

ages are ~

"a

W

Th.erisam of ieMfq v ikai' i-aelit- dt

I~UMU4401 CMVSM*, 101 $0sppb ad~.~ Avs414 ..*6

$bow 5.,iin, #10 3avpsel I Te pvvWWO4 ot 611I~W,0

mage Sb ec wello 1*~ of o~It f oil M0446 tossofV& 0t 4k

00es0 of dsp.i eauee,01u santiketa 14.q.t *.

WMi W# 06#9 *Ift910004 401*0 #6 OVOWI4. 1"ht #4f4,#4

the *10labti ift of fodgeepmhst te.w 40~ otlb'. *if ~f

$0110dI of 11M. Ml*~ wtft 6" brom 4fr.tmme to*1 1%

woto mp owbefuiim tuirIqe Ws low 41#0006#6 of

let~ a- Ms o e l eion is8 a3I** sime. Swar to **-

oNMI ~Wfia 4100MIKI ii~t t t*#WAI..c CWV.qd eItw

u~w ~ ~.md S * .e~3 e~se-q6

bi0*0 *"4 *wo M44*

41SIOO W *bt40"tM 4 OIlkf446,4W AA*JJ~

330 IM I# 1p W*"WX &@AWV* ~*4dtJ 0fI1

we iVlas allc to*l is*f O£ nvt 0~~s40 O 4 IahItoo we4

insrKV* hSO iliM ~&.M .R f#1144 ftff~ *S 1Mw 1sa*

sa. Wi ai tsn s0iabslau pftw.m*s oS PSaftti fI* I 1f~4

v.e.e gtw tewu srl0t 10w OKble (wild 00 1006 *d#0PWe

tea~ r Ibs I*@ I a m oflWM #a *v.ol too 4~ 1 %

loseE~ Ibs s toer b.edt zf p WCW# f 4miab tIftifi*

lb ,esi tiu. teos a &*a cat*o ofs th. wteitao ofw

* 1."O eftsea b oetivst ee,.,oI&e~,e I.

00 Otb*Pstle; t~b. CII CO, fd f*#4 Wis 19tVtl

bi Caea fo& Colp.b kmbs" W~tt. 166I~(

. .

p

"o r~ to buR~ lowI~ o"aflo aF~IOiA'OsiaI 4 ttkdkImi

IC3116*as %WOW &i**)W~hJ *1lb * 341* "WW #" *'bw A W1J. 14

000444g ot $**lv.awu **a tbmS, ~bO lkS **Iw

~# 'WU IPWO&StVO 60 f% *UIP 141 I&0 OSbg* Oe.M"CU

coo ~~t*$ p5lw~i o* $i0vtpso' 3~f*0 MA, dig. *Aes. altow".IS O9tiI (t*WW** 0" SW#1 $t*flolW~ 190 44W **l4 44019

4411610 4*00 S*6e10044iA.I* **t. fouI6 *

W# 10 IWOAi *on 1. us.ff pew~tu, tf* u** 4g$t*

Ct '~JiI~e .ltrfes ** **A#I1e4 oo cmtflo 4,l 0#*1t lbow t,vesp~pm w, twm*q I to ; qk;

41 ~1 "Op6Wovit aitO fit"- lb. vmiii4 o,..f "ofwI. f4iAt'Iit

owblef. so v~f *v* testfma~ e t II pwe...V.* dlgilt#*let pffte us d#I.ii* * *sp&IIO taei ofj~ 0JW"L##*j

~i~et., t 0s ff"4I~W~ lb. #"-~tPIO'm *f t'* 1

td.toi 00i wwwI~e is # mII oe " is v 4 of 11* 'w

wil #ft*msiprv*t oke swtli~ I** . l. it.t a*

*~~~~ol lob~ asps.w.e . . %imIailed to Vlsie

tie* of IN* t.,m F*vo w*e 41&*i f%1vtlit fivvo *

* tmtsat **Vvwl.. Aaue-f It* to tok. tmv sc iowt

u4kha.rl p 4CCOW40J a" i 0Ai0 "s~~~ sopll a C*-4"Ntow

a*No to go* "Vmo~i*j a s eWtsmo 4 o A

-m *wisp *Ui4i1 # pIb*N S. podrw h.*t cwl iwm 6ptsiaaVtSo ft lb.AV pu.* O~o pObuelf tow dbI.Sf dot~~~

of Oe twa#06040 - *W"*EOd 604, l ~ w toI $sft*et" "' 'btim ISls0"pw*449M4tf

"fi 4((p lb. #fw* -* 4#tsv.oo * W*440 444 Owt

took* lb.W14 #n IM4ul1106 **4iw. & (ItOw lows ts 4

Wt wqe *~uwi* too *41#I~ 0Sws *-I tot fl 1w*I.

ftIte fe'S01104 to Oww.owl towfiev es ~ wS tt,k**

* lim~to mo#$* JO to MA~ **wiqV t" tw lbm ma..wi..l t*wpWs.

Ot Of .b WsuiOke' .t~iI.* a"d ***I #*o~wq6lm O0YV #5*6*.

"P#1*001 #60 lw*0*q*,t .1 *A* lSt Imkom .'"d lb. Ppifle-

Sfo f~j~ tWO SsW ~ I0 WoV* ifc lb '490itWJ SW*i"'1IfiidS..*twet ii"c~ 1W*~. b wilI *" qiteOe Atju. .te mod #Pk* lql dSi#*4 siiftwogik4 ptwwo 4%0'S#f

lb. twW100 04 lb. fs. v o(e6W6,"

1%0 twdai* Wits' Me8S.tme~~ibi wigW

pmqwt" A fof Vlot. ib, o~IpeRW4 ovmw' 40 Ji. eir sa0 lb 54e I fwi -

tot is* "ae1 06f sb. pv# -ei f* 8W 'w4CA 004 4 110hIhC0ft fe*#l. lb, . -*k jp*4e S *qn. u"qtISj' 160 4dsjW*.

of lb pvw~atsvw~#w~it1~. * lb #fwb -I.'t %

7e ba eob" ne vwp*404 b tg lpa a &wuAw r (#A* Av a . . u

the x"044 rogss b46 coaindeftv 3A4l of tow #644,51,04

- C tel deiiii*f~ lleSj. of te~ resa

Ikq tf*POO 9010#00ld lhu Pftfem RQ l'%*qiq"ao.V cisR* *iftb*f

IN*llmoft t 0*. viw % Ie tveupedq' *cOwe' Op" O

OM. -w41 ftf lb jq185* *lot *-Oiisci, I kil~p JP1 h

C" ~M * ltk* dltrt*v*-fk* ftOileet.To *4s,Iv4 -44. ~V-q

4t. 0,VA

correrpwading Iwm.i rLitisre.J sit aji I t Jmiklni the were'.°

recorded ituttta .alb- .

mlied ate (eedbacL on each dc+ision us proviJed in order

to sustain the reders' interest and to help then j4 hieve aid

aiintain stable performance. The computer progrja respond Jd

to each entry as follows:

decision rating response

lesion present lesion absent

I 00you are correct" "there is no lesion"2 s"you are zorract" "there is no lesion"3 "there is a lesion" "there is no lesion"4 "there is a lesion" "you are correct"S "there tS a lesion" "you are correct"

4

MATA AR ISIS

The diagnostic performance of the readers uslng both

4 of the two radiographic modalities was evaluated by ROC (receiver

,peat.iml characteristic) analysis. M0C analysis provides an

Itias of diagnostic acuracy that 1i independent of extra-image

decisiom factors and prior probbillity of lesion occurrence.

"' Specifically. for a diagnostic system with given discriminatory

capnlity the, u rve shows the Ituding relationship between

the pe ttive,% of triue-positive (TP) and false-po'stive (FP)

resposs, as the dtimion criteria to call the findings positive=

or megatitV it varied %ystqmticall. In our particular

applicatlion, this graph can be asseseed from the loci or points

p, deseribiag the relative tP and FP Jecisions that would be side

by cosidetiig etch hiowsdary between the five esaminer choices

a different dtisim criterion. the above procedute provided

'Or

-a -

7

four jwssible points that are located on a conceptually smooth

curve characterizing the discrimination capacity of a particular

modility. A commercially available computer program (RSCORE)8

was ud to fit an ROC curve through the four empirically

obtained data points. The theoretical curve is based on the

a.oueption that the distributions of the psychologically perceived

sitnal strengths in the presence or absence of a lesion are

normal. 7 Consistent with this assumption, the data points

can be plotted on double probability (binormal) coordinates***

&dd fitted by a straight line. The computer program provided

meawrs of goodness-of-fit of that line and a maximum-likelihood

eilieste of an ROC index of diagnostic accuracy, AZ , as well

&s its corresponding sampling variance. The index A reflectsz

the location of the entire ROC curve rather than any particular

etv*ting point thereon. A is defined by the area beneathZ

the fitted ROC curve, and ranges from a minimum of 0.5 for chance

pwffpotnce to a maximum of 1.0 for perfect discrimination

tel4pbi I ityr.

In order to summarize the performances achieved with each

"41lity. the accuracy indices, A , estimated from each of

0e easinrers' responses were either pooled or averaged.

Ie ;viso ted standard errors were obtained from the sampling

vria.~e. of the maximum-likelihood estimates given by the

eltwper program. The statistical significance of the observed

Chart Y4231, Codex Book Company, Norwood, MA 02062

". " "- "- - - • " -'- -'w *" € '- -'' e '" .''''-' ." -" " - ," '.- -" ." , f

Hi ' ~~~~~~. .-. -' . *-.' . . "- -",*t-.*,"" ."', * ," - ,N " " " ", - "'

M-4 41r A . .= - s .. 0'. ir--- -.7. 9 - . -

di fference in A between the two modal it ies, and between groups

of lesion sites with comparable anatomic obscuration was

tested by a paired comparison. This was possible because eachi

reader participated in the evaluation of both modalities.

A non-parametric test (sign test) was preferred in view of

the limited range of A zand the small number of readers which

renders the normality assumption questionable.

The time intervals required in making decisions were

averaged over all readers and all lesion sites, or groups of

lesion sites with presumed similar detection difficulty.

The observed averages were compared by the t-test for statistically

significant differences.

RESULTS

F~igure 2 shows a representative example of corresponding

pre- and postoperative radiographs, and the ensuing subtraction

image. The superimposed circles appeared one at a time in

a random sequence over each potential lesion site. While it

is nearly impossible to detect all lesions by comparing the

postoperative (upper right) versus the preoperative (upper

left) radiograph, the lesions are easily detected in the

subtraction image (below) as dark blotchy areas. In this

particular example from a right mandible, lesions were induced

at sites 1, 3, 4, and S. The bright disk-shaped artifacts

in the radiographs are projections of spherical radiopaque

markers serving as reference points to monitor the reproducibility

iv%

-V ,V -7 -Y

9

of the radiographic projection geometry.

The diagnostic performance attained with each modality

is shown in Table 2. A clear superiority of the subtraction

technique over the conventional method of comparing radiographs

is evident. At the outset of the investigation, lesion sites

were grouped as shown in Table I based upon the presumption

that members of each group would be subject to comparable

obscuration due to anatomic overlay. For the conventional

technique, the results indicate a definite decrease of A

for the detection of interproximal lesions as compared to the

*interradicular and radicular groups (P < .01). No significant

differences existed between groups for the subtraction technique,

consistent with the premise that the source of anatomical

obscuration is cancelled by subtraction.

The data also show that pooling the 11 readers' raw data,

i.e., treating them as one reader by merging their rating responses,

leads to a small depression of the accuracy index as compared

to the average taken over the individual's indices. This is

to be expected theoretically,8 however, the small difference

observed between the two summary measures attests to the relative

uniformity of the decision criteria used among the different

readers.

Figure 3 shows the detection performance evaluated for

the total set of lesion sites. Every reader achieved a higher

accuracy using subtraction images (P - .001). Also evident

is the more uniform performance among the readers for the

subtraction as compared to the conventional technique.

Vi . . . . & . . * . " ." . " . " -" -, ."-'--. - - °- ' . .' I.- ' '.

- *8 " ' ''wl~i """""""" "-'""" """' "" ""' ""* :

- 's........................................ . .

Figt re 4 displays the I W(I ,I; l )It ll .- ,olit'lilled h) pooll l.

tile I(CSOnses of each reader. \I- - )htt .nJ'e t le best-litted

lines for" each modality p1lott 'd on dotible probability coordinates.

The ROC for the subtraction technique is seen to be consistently

above that for the conventional technique, with respective

values of A of 0.98 and 0.83.z

A comparison of the time intervals required to decide

whether at an indicated site a lesion was present or absent

is shown in Table 3. In general, for each of the groupings,

as well as the total pool of lesion sites, the time differences

between the two modalities were highly significant (P < .001).

The average response times observed for the conventional technique

were almost four times longer and displayed approximately twice

the standard errors as compared to the subtraction technique.

Furthermore, the relative difficulty of detecting lesions

at different sites was somewhat reflected in the times recorded

for the conventional technique. The average response time for

lesion sites 3 and 6, where presumably the least amount of

obscuration existed, was the shortest, and was statistically

different (P < .02) from that obtained for sites I and 2.

In contrast, the times required in making the decisions using

the subtraction technique were homogeneous among the lesion

sites (analysis of variance, P > .75).

i #6g' i

T* V

DISCUSSION

Previous studies with skull phantoms have indicated that

subtraction radiography can improve diagnostic accuracy when

compared with the conventional radiographic technique.5

This investigation has confirmed these results in a live

animal model. The clear superiority of the subtraction technique,

as demonstrated in this and other studies, is critically

dependent on the ability to limit geometric and densitometric

variation between radiographs to be compared. However, despite

the authors' best efforts these variations were. at times, quite

evident and dictated the two-film packet technique. The two-

V film packet technique allowed the authors to continue making

radiographs until an empirical on-site visual confirmation

of geometric standardization could be made. Two radiographs per

site at each observation interval was usually sufficient.

V The amount of empirically observed geometric variation over

the eight week period during which radiographs were gathered

appeared constant. Even a rigid registration method may,

over times longer than thos, used in this investigation,

present geometric variation problems due to normal minute

4 changes in tooth position which may occur over time in some

animals. Other researchers have used a non-rigid occlusal

**.registration with some success, 9although in any subject

under general anesthesia, as Well aS any animal, the use

of a non-rigid occlusal registration would likely add additional

undesirable geometric variation. The method utilized in this

investigation to limit densitometric varlatlon (ell bhwrt of

the authors' goal of virtual elimination. IndeeJ. (requentl)

the films processed in a hand developer immediately following

exposure showed less densitometric variation than did the

duplicate films stored and processed under more carefully

controlled conditions. Fortunately the program used for subtrac-

tion radiography can rompensate for densitometric variation.10

The choice of 901Kvp exposures was made in order to parallel

clinical practice in our area. 60-70 Kvp would have produced

more contrast in the radiographic films used for the conventional

technique, but likely would have had little effect on the

results produced after subtraction because the contrast under

the latter conditions can be manipulated electronically.

The third and fourth mandibular premolar area was selected

for this investigation because, in the dog, this region has

sufficient lingual vestibule depth for parallel film placement

and there is no interproximal contact or overlap of the third

premolar with the adjacent teeth. Potential lesion sites were

chosen to reflect incipient interproximal periodontal lesions

without cortical plate penetration (Figure 1, Lesions I and 2),

as well as a variety of overlaying anatomical structures for

those lesions designed to penetrate the cortical plate (Figure

1, Lesions 3, 4, S, 6, and 7). The results shown in Table 2

indicate that the diagnostic accuracy in detecting interproximal

lesions (sites I and 2) by the conventional technique was

substantially reduced as compared to the other lesion sites

(A8 X .77 versus Aa .87). This finding is in agreement

- - * * e ~** ~ lii~ .i. iili4* d .. %%*.-*%~ *~ ....-. . .. . i - '

with other rese4rih cg'UtVning thatl |0m* hot ino|ling the

cortical plate are *Orr JIfficult to delt.' (in ComwetutsMS1

radiographs) than those lesions Wilth corical plate involvmeat.

Contrasting with this. the corresponding data from subtraction

9- :9 radiography do not show a specific association of accuracy with

lesion type. Such a result of constant detection performance

irrespective of anatomical context should be espected from a

technique that is effective in suppressing structured noise.

In a clinial situation, a diagsnticias frequently must

make a decision utilizing less than conclusive evidence. In

'S thes situations a clinician Is likely to skew his decision

Atowards a diagnosis, which once ade, imposes the least harm

to the patient if the diagnosis is later determined to be in-

correct. In am investigation as this, there was no danger

to a patient in the case of an incorrect diagnosis, and the

decisions were presumbly based solely on the knoledge of

the prior probability of lesion occurrence and the information

derived frn the imaSes. The diagnostician was not restricted

by clinical pressures and thus was free to express his confidence

in each diagnosis by the rating scale provided. Hence, this

technique permtted estimating selective points on the MOC

curve from the proportion of TP anf FP decisions that would

be made by choosing, in turn, each of the possible rating

* levels as decsision thresholds between accepting or rejecting

the presence of a lesion.

*~~~ % % .

14

Uder clisical co a ..s, the decision hraslwd4s or

opereting pot"s on O W1C curve chae by a diagstictas

are ually m a beosame &My vory with ths purticular

diaposeti. sek the valus Judgments amds about morb'idity and

fiamsi pacts anm eatmotes of disease provs lme or a

4 peartlaw paImo. MUrmev , the estimted pIphe swei iu

P1 4 still apply in th eliaical contm becase thoy

scrin the possible twedo-offe that can be mi betweem

wre t and ima'rect de ilsmna. Sbpresed diftoratly,

oway operaings palm that ma be adopt" by a diagnstician

t loem t pprg gete m Mer a give. radiographic

:eaioi. Pe ompl it a flso-positive (M) rate of .10

is climically soeaptable Is a particular situation, subtractie.

-11 epploy would attaft a tr-oeitive CT prvpert m of

q 0.5 as compared to the emnvetiomal toshaiqus with a TF

pf 0.40. Or if I r some roes the false-eetive

- ) dision rate mt be kept ill. ay below 0.01. It

ca be So rm the isls to the loft m at the top of the

diape. that sobts radio phi could provide a true-

asptisve (TN) detes rme of about 0.60, ceopwd to a

moeup-ool rte of 0.06 attainble with te ao"vetiomal

te-.iquo.

In prece. sottling, diMoticiams my simply uitmold

a definite respons in equivocal situstio. sad request further

dispstie evidmee. Usually it Is desirable to maiatain

both the probabilities of PP ad FN reospass below a certain

SI-',. ., . -,.. - . . ., - .r." ,,.,-"- .""'',' / ' ' , . ''', ; '

l i* * -i.. . -. ..

:S

vatliaunder thee mst vaints ca be ott jeted (no. the

appoprateWC wvm. This bits". ft tbe fact thu the

p ishiUes of the possile, reposes that sa be Simueto

PC 1.. * I VT eit also fale tht preboilittesof the possible. responses am say be gives to site withew a

4%lesin wl so t . as will I.e.. p(M ) .P(W arv p * 1.e.

With theaid of nows, 6tmsnbe leew han Fit" 4 that th

uesroat" toinhelpe WSWd msitala, both F(W) adp(M)

loss than or equel to S.060 "al derialtely setting all losion

sites In@. poet"e or meow.. Nerom to ematain p(MP)

P(H) ±1. the mmoelal todbiqu amrid produe F(T).6.0 md IP (MW a 6.4 so tal to dLagne 38% of the sit".

howve Ioela" me 45% of the Blte vit~ lesione. ows,

* w equal p 6-6 IfllU. of lamb.e pesence or oeeeoo. tOe

mus~m)tesheiw woud ram"a equivocal as abov 113

of the Issei altos Preented. ThiS analysi.s asm it Cleow

theathes oberved dIffameee of 6.1 is %~ bet-eM the twe

.4 eibp'iphlc Modalitie is a s.ubsaa Practical diffeneaco.

SM46 app ears large onsw to euevei* aw value ) Jup ts

tha OW be ssigned to set and imncec iagpoic decisiomn.

S The tiee intervals requited is ambles the iagnostic

Aslaw pewlded inathe loep kint msssmet of the

volathwe iagnstic vtllty of the tw techmiques. For all

Is"a "pase thes lauowals wore sipitipIcmly sherter gad

.ineiMS SPOed. k~te ib praeUMe tq~l Of SIU

w we be is lq~tm, is "me""a t" a s" of itapauc

wellm M-0 a .t sonw bls Is Is" tawly to Of

*Witk asu iM-peies of sieub s Ims dm with omesamal

Amw doe o11W of d" WA~p is 4 f) t off t 4

-. sh evsop .1 twawq%- berns em two eppw

dm. lit"!SW -tw OW. a"W tt ftfd 49 ebwmwnsonu "a team ow be "~iMat or ~IIe to &"a*si

dsenmi om leow neiaedlb bee1.1 "W soresu"r

im is "t dst by "A enw . maif .mtieu

er ublm m@ "m w bm to be htr sqwer

se eiinml someqi &up$ far Rulies deteseim &a a

live Mmi mso". Misost1 smounu on slpntly

IWEiW with momi sip" (F q .UI), "d doeb.'UIs vewlhsi fr alsoms urn sp ly meisi A s well

(P ~ ~ ~~W t 63. rhnuelysis ibmi Ona do diaotkc

Val"e of ,s@11e111 em be otmtalp ISVSeSsed by d&ita1

~iinsI iIitiag is m eut~ed usts% of oeiwc"l

disomme iseisern dm Mus Ime we New. Mai

*00 *

IAM 4010"~: SMaHdgRWW9 6~9W C~t

"alt *

*Amos**~~ I'm &"ot ofs AeN

(1) save. SL: k.1.u pql.: ofu$s f.~~~i~

4MWWO 1913.W$U-M 1"

(1)eo4meUGe,, mef k. w a u

(2Jt. U.S. Ma'em~ ej t gh bumg

-( J Pwt @ 4233 19--11.

- - - P2 -. - - -

9

(6) ~.w.. J. * PicA.ii * S.: f.~'~~aui~ ~'f ~~auP 4.u.~..

9. ad 1. UP ~ 11J4)J. ~ 1ff. Academic Pmi. 1661.

.9

(6) bdemnb. J.. ~~'* P.. badeem. J.. Sserem*~. S.: A0

.4 ~Shd j~P ~ ~frIe md t~eft~gv~e 5u~d(meeS.. er- ~ S ~ .~A a .m a~e

W5~D~. * UWTRO DUU~ w.

0

- . ~wu~ Ni. ~uinei Cfferm. a. ~-- j'be-S

~y1e. Cin.iies~am .j' Ipr.5g 5t~ewe (q Desist~oinetm bEdqu~tg. UMI PSI 514:171, 1663.

'p

II ~t4.e. 3M I Peu'tadmsaes Ussi Burnt 3:6. 3663.I

'4

4

*

'a

- - t d.~-. ~ ~~%-~q ~ ~ .~*.*'%~*.~* -. ~-. - ~ ~ .5- -. 9--.. ~

'l~ Table I: Distribtution of Lesions

"* Site (twos Pigmw. l) Type n

1. 2 imterproxiaal 1S3* 6 ialnardicular IS

I.4,, S* 7 radicular 22

totsi 52

4

I,

4.

CI

'.4J4.

i 4" 6,'' ; ' 6l'ii " ' :' ' " ''"' " ''" '"'*" '" "' '

~ ea4ised Iftosures of Performance Az

Cmventional Subtraction

0° .76 .9811• .s,,u, ."' .98 (.03)

.t VMS .a6 .98.sqp .8 (.10) .99 (.02)

6. ~ psS .86.98• , w .86 (.07) .98 (.02)

pa u *,3 .98!.teuee .44 (.OS) .98 (.02)

l% ho9 h 4tfev~t (P * .01) from either site groups (3, 6) or (4, 5, 7).W N . POGNOW64 uwr t standard error of the mean.

V.

I.

Table 3: Time Required to Perform Lesion Detection Task (sec.)

Site Conventional Subtraction

1, 2 10.19 (.40)-- 2.78 (.17)

3, 6 8.69* (.37) 2.41 (.14)

*.4, 5, 7 9.91 (.46) 2.80 (.19)

all 9.21 (.46) 2.66 (.20)

*Significantly different (P < .02) from sites 1 and 2.

*Numbers in parentheses represent standard error of the mean.

C)

.0

CN CU

r_

~0

4)4

-4

CV) 0

L.4

k *o

0

0 F4 0

***~Q 0a Cu u,

C $.4 *0

V)

0n 04

0 bo9: 0.0

0I 0 P

*64 r4 -

*6~~ 9 0 $4t

0 0-4

~41b *4) 1

9.90 0s

16 .9

04*to ra.

-4J

IV CIL '

Fd - -* * *o- -. .

M U

K>-4 a) U

4)

0'U

0S.

0 as

0 0 u

0 us

0 c)

0.0

tInbo t

1L

S.z

04 a

0 14

..- ~4 .J

3I 5

a~~ - #

* Slo

0 sS = a .

A -. I

go 0 V4 -4l

z. ". dVi 4a a 6$. t 0 600

-0 -- 034 -4 C9 0

50 0m4 4. k v~ 0 ~

a.MOWO&C~

II0-4 0I ' 0 u 4 r.

I j :O' 0'4S 41

* Io -0 0p

t ~~ C~ :lb. ~ S*0m 0 EM0449 m

"a 0

I~U 0 .a.

S

(d.) as.. 6. o N

C3 d~vie

44 a

-~~~~~~~~~~~~~~ z,. . . . .~.. -.... se** 4 . .0, o 094

ut ~ ~. '~ i A V 4 ~ *1 ~

* ~ - j,. *rv

t . .

* U * . . . S p

* V** '

U.

' .. ,*

i' .,,

I 9

4 '~ \V 1 1~ F)

~ i~.q4 i~ I* *1

..~ -* . U~4 ~ .~ A -. * -

~~1%

y* ~(. *k~W~.NA..$ ~ S , A .~& ~W~P4~ 1,... 4 ~'*

.. , ~. . .

'.1 .~.4~ 4* . t

9~Eq~S

5

* . . *4 .. F. ~ ,*. ,*.'~ ~ *. . p

.5

4

C *

~. I .. - S

1'

0.A.

~*,41t. * --. -... , -4

S.

j ,j~! ~ -'I ~ , * S .-* ~ .0.*-.

. r~.-- -

N' *- I / '4

0-'1.

* . "A -wC,. '

'

'S. .4.4'.*

* *I:1;~g~ * ~ *i.......-. ** - * *................

A. .* **.%~' S. % ~ . * * - . . . . S... * *