spect images
TRANSCRIPT
8/21/2019 spect images
http://slidepdf.com/reader/full/spect-images 1/8
image contrast, anatomic detail and the potential for
g reater d iag no stic accu racy (1 ,2 ) th e m eth od is n ot w itho ut
difficulties. SPECT may actually be viewed as an image
enhancement technique and computer enhancement
methods have been shown to bring increased diagnostic
sensitivity, often in association w ith reduced specificity
3 . The latter may simply relate to the enhancement of
unimportant intensity differences, or may be due to
m ethod related technical factors w hich lead to the produc
tion o f artifactu al im ag e d efects.
P atie nt m otio n d urin g SPE CT im ag e a cq uisitio n is w ell
recognized as an important factor which may adversely
effect the accuracy of 201T l SP EC T m yocardial perfusion
im aging (4,5). S PE CT m yocardial perfusion im aging m ay
present a greater frequency of patient m otion and related
im age artifact than other SPECT im aging protocols since
c ard ia c SPE CT a cq uisitio n u sin g 2 01Tla nd 9 9mT cs estamib i
requires 20—30m m and is often perform ed in elderly and
ill patients who m ay have difficulty lying still. Im aging is
frequently done tw ice in a single day and is often per
formed soon after vigorous dynamic or pharmacologic
stress. Y et, such m otion-related im age artifacts have been
evaluated to only a lim ited extent. DePuey and G arcia (5)
note that “.. it is unknown what type and degree of
motio n w ill ro utin ely c au se sc an a rtifa cts .â €•
In a prelim inary screening of clinical studies, the varia
tion in location of a radioactive marker dot in a dynamic
display of projection images supported the presence of
artifactual v ertical p atient m otio n o f so me ev id en t d eg ree,
in approxim ately one of ten clinical studies. Friedm an et
a . (4) have highlighted the problem ofpatient m otion and
suggest its im portance for quality control of SPECT stud
ie s. T he y h av e a lso n ote d th e o cc urre nc e o f u pw ard c re ep
(6), a gradualchangein intrathoraciccardiacorientation,
likely due to variability in respiratory m otion soon after
ex ercise. T his v ariab le b rin gs a d issoc iatio n o fcard iac from
chest wall related m arker m otion. E isner and coworkers
7 ,8 have p revious ly r epor ted the e ffe cts o f mo tion in
both patient and phantom studies of SPECT myocardial
p erfu sio n im ag e a cq uisitio n. T he y a pp lie d a s op his tic ate d,
cross-correlation com puter m ethod to identify and quan
tita te su dd en â €œno nre tu rn in gâ €•e rtic al p atie nt m otio n a nd
to attem pt correction for m otion postacquisition. N onre
Pa tien t mo ti on du ri ng image acqu is iti on is a f requen t cause
of SPECT perfusionimagearti facts.We soughtto determine
the relationship betw een patient m otion and the resultant
im age a rtifa ct. T he e ffe ct o f p atie nt mo tio n on @°1TIPECT
scintigramswas assessedwithcomputersimulationo create
66 new image sets w ith a rtifa ctu al v ertic al, h oriz on ta l a nd
c om bin ed p atie nt m otio n in tro du ce d o ve r a b ro ad ra ng e in s ix
norma l s tu die s. V is ua l a na ly sis o f re gio na l ra dio ac tiv ity in
these simulatedimages,as wel l as quant itativeanalysisof
th e re su lta nt p ola r c oo rd in ate d is pla y was perfo rmed. The
p re se nc e a nd e xte nt o f â €œmotio nâ €•rtifa cts v arie d w ith th e
n umbe r a nd lo ca tio n o f th e p ro je ctio n im ag es a ffe cte d, a s
wel las the extentof theirdisplacement.Al thoughthe extent
of the defectvariedwi th the f ramesaf fected, they were not
necessar il y more ex tens ive when related to vert ica l d isp lace
ment inthe centerof the orbit.The locationof induceddefects
varied with direction of disp lacem ent and the location of
f rames affec ted . Vert ica l and horizon ta l mot ion c rea ted add i
ti ve de fect s. De fect s ize g rew w ith i nc remen ta l ver ti ca l d is
p la cemen t b ut s ub sequ en tly d ec re as ed w ith y et in cre as in g
displacement.Boththe irregular,“lumpy―istribu tionof rad io
ac tivi ty , o ft en wi th oppos ing “defec ts―,s we ll a s cur vil inea r
extraventncularradioact ivi ty,were visual clues suggest ing
SPECT de fec ts rel at ed to motion art if ac t. A c lin ica l case
review revealedthat approximately25 of studiesdemon
stratesu chmotionduringacquisitionbut only 5 contribute
to visibleimage deteriorat ion.Whi le detectionis important ,
postacqu is it ion a ttempts to correc t such ar ti fac ts are incom
plete and opt imal ly,theymust be prevented.
J NucI Med 1993 ;34 :303-310
tress and redistribution 201T 1 m yocardial perfusion
scintigraphy has been w idely applied to the diagnosis and
evaluation of patients w ith know n or suspected coronary
artery d isease . R ecently , th e ap plic atio n o f sing le-p hoton
em issio n com pu ted to mo gra ph y (S PE CT ) h as p rolifera ted
widely owing to its apparent advantages beyond those of
the planar im aging m ethod. W hile SP ECT provides greater
Rec ei ve d J un . 1 , 1 99 2; r ev is io n a cc ep te d O ct. 7 , 1 99 2.
For correspondenceor reprintscontact: EliasH. Botvinick,MD ,University
ofCalifornia,anFrancisco,505PamassusAvenue,Box0252RoomL340,
San Francisco,CA 94143.0252.
E ffe ct o f P atie nt Motio n on SPECT Images â €¢o tv in ic k e t a l.
303
A Quantitative Assessment of Patient Motion and
Its E ffect on Myocardial Perfusion SPECT Images
Elias H. Botvinick, YuYing Zhu, William J. O 'Connell and Michael W . Dae
Departments ofRadiology Nuclear Medic ine Section and Medic ine Cardiovascular Div is ion and the Cardiovascular
ResearchInst itute of the Universi ty ofCalzfornia San Francisco
8/21/2019 spect images
http://slidepdf.com/reader/full/spect-images 2/8
turning patient motion was applied in normal studies to
assess th e in cid en ce an d cha ra cteristics o f false-p ositive
studies. Prigent and cow orkers (9) also intentionally dis
placed the SPECT table or manipulated the display of
projection im ages to determ ine the effects of patient m o
tio n o n im ag e fin ding s.
We sought t o i dent if y i n g reat er d et ai l t he speci fi c e ff ec ts
of regional, incremental patient motion on SPECT myo
cardial perfusion im aging w ith 201T l. W e applied a sim ple
com puter sim ulation m ethod to create specific patterns of
artifactual patient m otion varying in location, duration,
m agnitude and direction, in order to assess the effects on
the distribution, extent and density of resultant image
defects and the interpretation of SPECT m yocardial per
fusion scintigrams. Like some prior studies (6, 7) we as
sessed the effects of coordinate motion of the chest wall
a nd h ea rt. We d ev elo pe d so ftw are to c re ate , m ea su re a nd
conversely, correct, motion in a vertical orientation in
referenc e to a low tho racic o r u pp er a bd om in al rad io active
mark er d ot a nd th e c ard ia c s ilh ou ette . Compute r s oftw are
was also developed to im plem ent horizontal rotational
artifac ts an d an aly ze the resultan t im ag e d efects.
MATER IALSAND METHODS
BaselineSPECTImages
S ix norma l 201T1SPECT myocar di al p er fu sio n s cin tig rams
wer e s el ec te d fr om ma le p atie nts w it h a low p re te st lik eli hood o f
coronary disease, according to the criteria ofDiamond et al. (10).
T o assess th e effects o f v ertical m otio n on im ag e find in gs, w e
ap plied o ur stan dard m etho d. H ere, 3 2 p ro jectio n im ag es, o r
fram es, w ere acq uired fo r 4 0 sec e ach, em plo ying th e step an d
sh oo t te ch niq ue , a pp ly in g a c irc ula r o rb it from th e 4 5Â °RA O to
th e 4 5Â °L PO p ro jectio n, u sin g a leap co llim ato r ap plied to a
S iem ens O rbitor (D es P laines, IL ) scintillation cam era. A single
s tu dy , s pe cif ic all y a cquir ed ove r 360°n 64 f rames employed a
se le cte d 3 2 frame s in p ro ce ss in g in o rd er to p ro vid e d ata to te st
th e e ff ec ts o f hor iz on ta l p atie nt mo tio n. In a ll c as es , p ro ce ss in g
employed a Butt erwor th f il te r w it h a 0 .4 Nyqu is t cu to ff f requency
and fifth order roll off. lmages were displayed on a 64 x 64
m atrix in short as well as vertical and horizontal long axis
p ro je ctio ns. G iv en th e fie ld a nd m atrix s iz e, e ac h p ix el re pre
s en te d app roxima te ly 0 .4 cm in e ach d ir ec tio n. A ll image s wer e
te chnic al ly optima l w ithou t e vid en ce o f l oc aliz ed s of t ti ss ue a t
tenuation.
Each s tudy wasacqui red i n assoc ia ti on wit h a norma l max ima l
tr ea dm ill e xe rc is e te st ta ken t o ove r 85 o f maximum p redi cte d
he art rate fo r ag e. In itial im ages w ere read as n orm al b y b oth
v isua l ag reemen t o f two expe ri enced r eader s, and the ir ob ject ive
a na ly si s a nd compari son w ith a la rg e popul atio n o f norma l ma le s
a pp lie d b y th e p ola r c oo rd in ate d is pla y a nd a na ly sis p ro gram
developed by Gar cia a nd cowo rk er s ( 1,2 ,1 1). A 201T lma rk er dot
placed on the skin below the heart and view ed in sum med and
d yn am ic d isp la ys o f th e p ro je ctio n im ag es a nd th e im ag e s in o
gram served to conf irm ini ti al pa ti en t immobil it y and subsequen t
artifa ctu al v ertic al an d ho rizo nta l m otio n in tro duc ed b y alte rin g
t he r el at ive posi ti on and r el at ionsh ip o fp ro ject ion images i n t hese
normal s tudies .
Intervention: Ar ti factual Vert ica l Mot ion
Speci ficsof tware was employed to produce apparent ar ti fac tual
v ertica l p atien t m otio n by sh ifting 1 to 8 co nsecu tive fram es,
f rom 1 t o 20 p ix el s up o r down (F ig . 1 ). Sp ec if ic al ly , t he s of tw ar e
pe rmi ts ve rt ic a l t ransl at ion ofp ro ject ion images i n i nc remen ts o f
5 pixels Each shifted projection image was produced by relo
cating the original unshifted image. The operator shifts a partic
ular projection as m any pixels as desired and then view s the
r esu lt an t c ine d ispl ay of p ro jec ti on images.
D isp lacem en t artifacts w ere pro duc ed o ver a w ide ran ge in
o rde r t o sample i nt erva ls t hough t by prev ious i nvest iga to rs (5—9)
to p rod uce S PE CT a rtifacts and to test th e ch aracteristics an d
r el at ionsh ips o fsuch a rt if ac ts i n i nd iv idual pat ien ts . Chosen were
d im inut iv e d is pla cemen ts , d es ig ned to t es t th e s en sitiv it y o f th e
SPECT method to motion artifact and large deviations, likely
surpass ing those gene ra ll y obse rved c li ni ca ll y, i n o rde r t o gai n an
a pp re cia tio n o f th e e ffe cts o f v ary in g m ag nitu de o f motio n o n
p ro ce sse d SPE CT im ag es. T he n umbe r a nd lo ca tio n o f a lte re d
frame s w as a lso v arie d to d ete rm in e th e m ag nitu de a nd c ha ra c
teristics of resultant SPECT image artifacts. A large selection of
d ispl acemen ts were made bo th supe ri or ly and inf enorl y.
I n t he i ni ti al pa ti en t s tudy, 38 a lt er ed image set swere ana lyzed
spanning a 0.5 to 20-pixel vertical displacem ent over 1 to 8
f rames . A concentr atio n o f s amp li ng wa s p la ce d a t th e le ve l o f 3 -
p ix el d is pla cemen t in 8 f rames s in ce th is was t hought t o y ie ld a n
imp re ss iv e mo tion a rtif ac t w ell s ui te d f or c ha ra ct er iz atio n. I n
o rd er to co nfirm th e o rig in al find in gs in th e in itial in ten sely
evalua ted pa ti en t s tudy and fur ther char ac te ri ze r esul tan t moti on
ar ti fac ts , addit ional a l tered image sets , each wi th 3-pixel di splace
m en t in 8 d iffe re nt a nd w id ely v ary in g frame s, w ere a na ly ze d in
e ac h o f 5 a dd itio na l n orm al stu die s. T his la tte r a na ly sis te ste d
th e e ffe cts o f p os itio n o r tim in g o f motio n d isp la ceme nt o n th e
loc atio n a nd m ag nitu de o f res ultin g artifac tu al d efects. O verall,
56 n ew S PE CT im ag e d ata sets w ere g en erated an d an aly zed .
Spe ci fi c image a lt er atio ns we re p er fo rmed and r el ate d f in di ng s
a re e nume ra te d in T ab le s 1 †”3n d F ig ure 1 .
Arti factual Hor izonta l Mot ion
Hori zon ta l pa ti en t moti on in t he rot at ional o rb it was s imul at ed
b y re pla cin g a p ro je ctio n frame a nd a ll su bse qu en t frame s w ith
V. I
F IGURE1. D efectsize .Sh ownstherelatio nshipetweenhe
visualextent and intensityof the defectgeneratedandthe extent
um berof p ixels),durat ion numberof f rames) ,and t iming spe
c ific frames ) o f u pward image d is pla cemen t. D efe ct s iz e and
in te ns it y in cr eased w ith in cr eased extent and dur atio n o f mo tio n.
2 o
18
16
. 14
No.ofpixels
shifted (up) 12'
0 n orm al
a m in or d e fe ct
moderate defect
•ross defect
†¢
10
8
6
4,
2'
OD U
on U
U
0
U
0 U
0
S
U
0
S
U
02
r...
4688
2
r.@
88
o.offrames
Location
304
The Journal of N uclear M edicine V ol. 34 N o. 2 February 1993
8/21/2019 spect images
http://slidepdf.com/reader/full/spect-images 3/8
TABLE 1
Ef fec ts o f I nc rementa lVer tica l D isp lacement in Fram
on Mo ti on A rt if ac t in a S ing le Pa ti en tes
7—14
PolarDefect
D isplacement Up Down
magnitude(pixels)
L A X L AX
TABLEEffects
of Hor izon ta l, Ro ta tiona l D isp lacement, in +)rCounter
t o †”),he D ir ec ti on o f O rbi t on Mo ti onrtifact,Beginning
w ith F rame 16 in Pa tie nt
Polardefect
No. +€”frames
shifted L A X L A X
Baseline
1—8
Frames
vasculararea L A X L A
Defecton Polarisplay3
pixelsupwardisplacementX7—14
13—20
L R X L AXL18—25X1
3 0 0 7 0
2 4 8 0 10 13―
3 1 0 0 7 13―
4 2 0 0 10 15―
5 6 3 0 11 13―
6 —0
2
0
9
10 5
1 7 24― 2
6 18― 4 —
8 6 0 —
6 29― 2 —
6 5 14― 18― 21―
15― 58
9
15―23―
15 *
25―
20―
23―27
45 *
54―
49―
25―2
0
9
1
3L
= lef tanter iordescending coronary region, A = r ight
* Beyond normal limits, p < 0.05 compared to baseline.coronary
regionand X = leftcircumflexcoronaryregion.
0 3 0 0 3 0 0
2 7 0 3 25― 12― 2
4 14― 8 19― 33― 25― 24―
6 23― 27― 23― 42― 0 28―
8 21― 2 45― 25― 0 0
10 20― 0 21* 9 0 4
16 16― 0 4 11 0 6
L = le ft an terio r d esce ndin g co ro nary re gio n, A = rig ht co ro nary
regionandX = leftcircumflexcoronaryregion.
* Beyond normal limits, p < 0.05 compared to baseline.
0
1
3
5
0 0 0 0 0 0
1 0 0 1 0 0
1 0 0 8 0 0
3 0 0 23― 3 8
= hor iz on ta l r ot at io n in d ir ec ti on o f o rb it —= hor iz on ta l r ot at io n
counter to directionof orb it , L = left anter iordescendingcoronary
region ,A = r igh tcoronaryreg ionand X = lef tc i rcumflexcoronary
region.
* Beyond normal limits, p < 0.05 compared to baseline.
a no th er, im ag ed e arlie r o r la te r in th e a cq uisitio n se qu en ce b y a
g iv en interv al. S uc h a rtifac tu al h orizo nta l ro tatio nal p atie nt m o
tio n c ou ld b e sim ula te d in 5 .9°n creme nts in , o r c ou nte r to , th e
orb ita l d ire ctio n. T hu s, the fra me s eq uen ce ev alu atin g th e e ffec ts
o f hor iz on ta l mo tio n o f 29. 5Â °n th e d ir ec ti on o f r ota ti on b eg in
ningatfram enwouldbe l,2...n,n+5,n+6...37.To
accommoda te t hi s s equence, 64 pro jec ti on f rames were acqu ir ed
ove r 360°n a s ingle s ub je ct a nd th e app ropr ia te 32 f rames wer e
chosen in sequence and used for reconstruction. T his sam e,
specifically acquired norm al study was utilized as w ell to evaluate
hor iz on ta l mo tio n a rtif ac t c ounte r t o th e o rb it al d ir ec tio n wher e
a 29.5Â °shift beginning in fram e n w ould be represented by
projectionrames1,2...n,n —,n —...n, n + 1....27.
He re , d ev eloped s of tw a re p ermit te d f rame s el ec tio n s o t ha t c er
tam specific fram es w ere used tw ice at specific locations in the
sim ulated acquisition sequence, w hile others w ere shifted from
th eir a pp ro pria te lo ca tio n, a nd y et o th ers a t th e e nd o f th e c yc le
w ere o mitted . A rtifactual h orizo ntal m otio n of 1 -, 3-, an d 5 -
w ere o mitted . A rtifactu al h orizo ntal m otio n o f 1 -, 3-, and 5 -
frame d isp la ceme nt in a nd c ou nte r to th e d ire ctio n o f th e o rb it
we re evalua ted . I n fou r example s ver ti ca l d isp lacemen t was com
b in ed in v ary in g in cre me nts w ith h oriz onta l m otio n.
Image Analysis
Follow ing i nt roductio n o f t he d es ir ed mo tion a rt if ac t in p ro
jection im ages, each set w as reprocessed as w as the original
unal te red post exe rc is e c li ni ca l images and aga in d ispl ayed in t he
th re e sta nd ard SPE CT a xe s. T he y w ere th en re ad b lin dly b y th e
sam e tw o readers as norm al or abnorm al, w here defects w ere
g raded as fol lows: mi ld— if demons tr at ing a s li gh t, sub tl e de fect
in vo lv in g o nly a p art o f a s in gle c oro na ry v asc ula r a re a; m od er
ate—if dem onstrating an obvious defect involving all or m ost of
a s ingl e c or on ar y v as cu la r r eg ion o r s ev er eâ €” ifd emons tr ati ng a
dense de fect and invo lv ing mult ip le vascu la r a reas . Again, ob jec
t ive quan ti ta ti ve analys is o ft he pe rfus ion abnorma li ty compared
to a n orm al p op ula tio n w as p erfo rm ed u tiliz in g a p ola r c oo rd i
n ate disp lay . H ere, region al de fec ts w ere related to co ro nary
v as cu la r a re as a nd abnorma litie s wer e p re se nt when exc eeding
12 in t he l eft a nte ri or d es ce nd ing (LAD) vas cu la r a re a, 9 in
th e rig ht c oro na ry a rte ry (RCA ) a re a a cc ord in g to e sta blis he d
TABLE2
Eff ec ts o f F rame Loca tion on Mo ti on Art ifa c t
305
ffe ct o f P atie nt M otio n o n SPECT Im ag es Botv in ic k e t a l.
8/21/2019 spect images
http://slidepdf.com/reader/full/spect-images 4/8
c rite ria a nd 1 2 in th e le ft c irc um fle x (L CX ) a re a. F ollowin g
this e va lu atio n o f th e effec ts o f p ro je ctio n im ag e d isp la ce me nt o n
n orm al SPE CT s tu die s, w e re viewe d 16 5 c on se cu tiv e c lin ic al
SPECT studies for the frequency and severity of patient motion
and rela ted ar ti fac t.
Reproducibility
T en SPE CT im ag e d ata s ets d eriv ed from a lte re d p ro je ctio n
im ages spanning a w ide range ofvertical displacem ent, from 1 t o
6 f rames , wer e e a ch p ro ce ss ed 3 time s by th e s ame obs er ve r. T h is
perm itted the evaluation of the reproducibility of defect grade,
exte nt a nd lo ca tio n a ss es se d both v is ua lly a nd w ith r ef er en ce to
t he po la r coordi na te d ispl ay of r eg ional r ad ioac ti vi ty .
Statistics
Reproducibility and differences in quantitative polar map
d efe ct siz e w ere a na ly ze d w ith th e S tu de nt's t-te st. A p v alu e <
0.05 wassignificant.
RESULTS
Ini tial Studies and Reproducib il ity
B oth v isu al an d o bjectiv e p olar coo rd in ate an aly sis w ere
normal in each of the six initial baseline SPECT patient
studies. Reproducibility was excellent and there was no
sig nifican t d ifferenc e in d efect g ra de, size o r d istrib ution
with repeated analysis of ten selected SPECT data sets,
e ach altere d w ith a v arying m ag nitu de o f vertical d isplace
m ent. Sim ilarly, the findings described in depth for analy
sis of a single data set paralleled those seen among more
lim ited sam ples in five other norm al studies.
Eff ec ts o f Vert ica l Motion
D efects identified visually were always confirmed on
o bjective an aly sis o fpo lar co ord in ate m ap s. T he m inim um
m agnitude of vertical displacem ent that was required to
produce an evident visual abnorm ality varied inversely
w ith the num ber offram es affected (Fig. 1). A m ild m otion
related defect was evident in association with 8-pixel
upward m otion affecting only two fram es (7 and 8). H ow
ever, displacement of only 3 pixels produced an image
artifact when affecting four fram es (7—10),and only 2-
pixel m otion produced a sim ilar artifact w hen eight fram es
(7—14)were involved. Displacem ent of a single fram e,
(7) even over 20 pixels, brought no evident significant
image artifact.
D irec tion and Loca ti on (Tim ing ) o f Motion
Sim ilar effects w ere noted w ith vertical displacem ent
inferiorly and superiorly (Fig. 2 and T able 1). H ow ever, in
the sam e patient, defect location varied with direction of
displacement (Fig. 3 and Table 1), and defects due to
inferior displacem ent generally projected to ventricular
regions opposite in location to those related to defects
produced by superior displacement or upward vertical
motion. Sensitivity to motion, as well as the location of
artifacts ofm otion, appeared to depend onthe location of
frames affected (Figs. 3 and 4 and Table 2). A mild defect
was evident in a different location with only 1 pixel of
•1
4
46
8
16
306
The Journal of N uclear M edicine V ol. 34 N o. 2 February 1993
t0
C,
FIGURE2. Effectsofmotion.h ownaremidventncularhort
axis s lic es der iv ed f rom in cr ement al d is pl acement o f F rames 7†”
1 4. A t le ft a re th e in itia l no rm al p osts tre ss (a bo ve ) a nd d ela ye d
images. The e ff ec t o f upwa rd d ispl acement is shown in the cen te r
column , wh il e the e ffect o f downwa rd d isplacemen t is shown
w ith a s ligh tl y g rea te r magn if ic ati on , in t he r igh t column . The
numberof p ixels displacedis indicatedat r ight .N otethe extent ,
locat ionand conf igurat ionof art ifactualdefectsa nd the apparent
reduct ionin defect extent with increaseddisplacement.
up wa rd v ertical d isp lac em ent o ver eigh t fra mes w hen th ose
fram es were 18—25(Fig. 1). Sensitivity to vertical m o
tion appeared sim ilar w hen central fram es 13—20were
affected (T ab le 2 ).
The extent of the artifactual abnorm ality produced m i
tially appeared to vary with the am ount of fram e displace
ment, when the number and location of frames affected
w as held constant (F ig. 1). T here w as an observed biphasic
effect where defect size increased initially, but then ap
p eared to d ecrease w ith ye t in cre asin g d isplacem en t, either
superiorly or inferiorly (Fig. 5 and T able 1). W hen inferior
and superior m otion exceeded the vertical ventricular di
mension, the resultant defects appeared to decrease in
8/21/2019 spect images
http://slidepdf.com/reader/full/spect-images 5/8
objective defects (Fig. 6 and Table 3). W hen five-frame
cou nte r ro ta tio na l d is pla cemen t w a s comb in ed w ith th re e
A pixel upward displacement in central frames 13—20,
H an extensive defect was noted involving all vascular
te rrito rie s (F ig . 6 ).
Internal C lues
T he summed d isp la y o f a ll p ro je ctio n im ag es c ou ld n ot
provide consistent, sensitive evidence of the presence and
e xte nt o f m otio n a rtifa ct. H ow ev er, c are fu l re view o f th e
dynam ic display of projection im ages w as diagnostic in all
cases. A lso, review ofthe im age sinogram provided objec
tiv e id en tificatio n o fth e p resen ce and lo catio n o f ho rizo n
ta@lmo tion a rti fa ct.
C lin ic al R ev iew and E ffo rts a t Corre cti on
The subsequent review of 165 serial clinical SPE CT
perfu sio n s cin tig rams re ve ale d evid en t mo tio n i n app ro x
imately 25 (41/165) of cases. However, motion was
sufficient to produce evident artifacts on resultant SP EC T
im age s in on ly 8 (5 ) stu dies. R ev ersal o fth e d isplacem en t
process w hich produ ced the artifacts form ing th e m ain
body of this study permitted a degree of realignment of
d isp la ce d c lin ic al p ro je ctio n im ag es a nd th e re du ctio n o f
th e su bse qu en t d efe ct o n SPE CT re co nstru ctio n (F ig . 7 ).
O nly ra re ly , h owev er, c ou ld th e a rtifa ct b e fu lly c or re cte d
i f associatedith pureverticalmotion.Here,with vertical
d isp la ceme nt, d efe cts w ere most e vid en t in a nte rio r a nd
inferior w alls as projected in short- and vertical long-axis
projections. One patient in our series demonstrated an
identifiable horizontal shifting m otion during im aging re
suiting in displacement lim ited to the horizontal axis of
rotation (Fig. 8). Here, the SPECT “defect―as m ost
evident in septal and lateral w alls as displayed in the short
a nd h oriz on ta l lo ng -a xis p ro je ctio ns. A pp lic atio n o f th e
software correction algorithm was of no help in reducing
the defect which disappeared com pletely on repeat study
per fo rmed success fu ll y w it hou t pat ient mo ti on .
DISCUSSION
The d ata p re se nte d h ere w ere g en era te d from exte ns iv e
analysis oflim ited S PE CT data sets. O nly general conclu
sions should be drawn in applying the results to other
cases. However, observations made and trends demon
strated suggest principles w hich are w orth extrapolating
with care and may serve as the basis for future work.
S tudi es cho sen f or ana ly si s exc luded con fu si on w ith def ec ts
created by soft tissue attenuation or other extracardiac
artifacts. The observations should apply as well to perfu
sion im ages m ade using the new technetium -based agents.
Motio n a rtifa cts a re amo ng th e most fre qu en t a s w ell a s
t he most d es tr uc ti ve o ft echn ical a rt if ac ts r el at ed to SPECT
m yocardial perfusion im aging (4,5). T hese are intrinsic to
the method and influenced only by patient cooperation
a nd th e a bility o f te ch no lo gist a nd p hy sic ia n to o ptim iz e
them . Prior studies have identified the problem in clinical
B
C
S. £
@.O.
‘ ‘ OO
O000
FIGURE 3. Ef fectsof mot iont iming.Shownarethe resultant
contiguousshor tax is s l icesf rom apex upper lef t) to base lower
r igh t) a fter upward d isp lacemento f two p ixe ls in e igh t f rames .
Pane lA i llus tra tes short ax is s li ces when Frames 7—14were
a ffec ted wh ilepane lB shows s lices resu lt an twhen Frames18—
2 5 w ere a ffe cte d. T he n ormal u nd is tu rb ed ima ge s et is s hown in
C. Notethe differencen defectsizeand location.
magnitude. Displacement of the same frames in the
same direction and to the same magnitude brought
sim ilar defect size in the same or adjacent locations in
d if fe ren t sub ject s.
Image Pa tt ern
A simple v isu al re view o f S PE CT im ag es d eriv ed from
altered p ro jection im ag es in m ultip le pro jectio ns re vealed
typical artifacts of patient motion (Figs. 2 and 3). These
in clu de d a n irre gu la r o r lumpy d is trib utio n o f ra dio ac tiv ity
with often opposing defects between the lum ps. This w as
m ost exaggerated as defect size peaked and overlap of
initial and displaced ventricles m axim ized. A lso evident
was a cu rv il in ea r ex tr aven tr icu la r r eg ion o f r ad io act iv it y,
likely related to the projection of aspects of the displaced
ve ntricle ad jace nt to th e n on displaced v en tricle . H ow ev er,
artifactual defects appeared to triangulate in the m ultiple
SPECT pro jec ti on s and o therwi se suggest ed tha t t hey were
tru ly m yoca rd ial in o rig in .
Effec t o f Horizontal Mot ion
L im ite d sa mplin g o fh orizo ntal m otio n artifact rev ealed
g re ate r se nsitiv ity to p atie nt m otio n c ou nte r to th e d ire c
tion of rotation. H ere, defects on the polar m ap were only
a pp re cia te d in a sso cia tio n w ith d isp la ceme nt o f fiv e p ro
jectio n fram es. L im ited defects w ere p ro du ced in an terio r
septal an d lateral areas. H ow ever, displacem ent of five
f rames in th e d ire cti on o f r ota tio n b ro ug ht n o d is ce rn ib le
307
ffe ct o f P atie nt Mo tio n on SPECT Image s Botv in ic ke t a l.
8/21/2019 spect images
http://slidepdf.com/reader/full/spect-images 6/8
FIGURE6. Combinedffectsof vert ica landhorizontalis
placement .Shown are polard isplaysof resul tantSPECTimages
fo llowing computers imulationof hor izonta lrotationalmot ion of
5- framemagnitude,beginningwith Frame16, in the direct ion of
o rb it al rotat ion ,upper lef t and counter to the d irec tion o f o rbi ta l
rotation,upper r ight in Pat ient6 seeTable3 ).A significantdefect
shown in v io let ,b lueor black,herein the left anter iordescending
dis tr ibut ion,was only evidentwith ar ti fact oppositeto the direc
t io n o f o rb it al r ota tio n. S hown b el ow a re p ola r c oo rd in at e d is pla ys
of resu lt an tSPECT images fo llowing computer s imu la tiono f 3 -
pixelvert icalupward motionin Frames13—20n the same Patient
6 (Table2),lowerleft,andw ithaddedhorizontalotationalmotion
of 5 -f rame magn itudebeg inn ingwi th Frame 16, counter to the
d ire ctio n o f o rb ita l ro ta tio n, lo we r rig ht. N ote th e d rama tic in
creaseindefectsizewhenverticalandhorizontalmo tionartifac ts
werecombined.
function to characterize and quantitate continuous or
“nonreturning―e rtic al p atie nt mo tio n in itia te d a t F rame
16. H ow ever, they did not fully determ ine the sensitivity
o f th e m eth od to v ertic al d is pla ceme nt, n or d id th ey fu lly
e va lu at e th e r ela tio ns hi p b etween mo tio n d ire ctio n, mag
nitude o r tim ing, and resultant defe ct size and location.
This could only be done by applying a computer simula
t ion of “retu rn ing―a tie nt mo tio n in d es ig na te d lo ca tio ns
a nd d ire ctio ns th ro ug h th e d ata se t.
F urth er, a lth ou gh su ch re tu rn in g motio n w ou ld a pp ea r
u nc ommon , b oth c lin ic al e xp erie nc e a nd th e re ce nt lite r
ature (7—9)support the variable nature and tim ing of
p atie nt mo tio n d ur in g SPECT acquis iti on . E is ne r e t a l. (7 )
note that m ultiple episodes of patient m otion rather than
a sq ua re w av e n on re tu rn in g motio n is th e â €œ moree ne ra l
case.― Additionally, incom plete application of the de
s cr ib ed compu te r a lgor it hm to cor rect appa rent non re tu rn
in g v er tic al mo tio n d is pla cemen t c ou ld re su lt in remnant
re tu rn in g mo tio n. S im ila rly , th e e ffe cts o f h oriz on ta l mo
tion a rtifact on resultant S PE CT perfu sion im ages hav e
no t been p reviou sly ana ly zed .
Th is s tudy app li ed a s imp le , s peci al ly desi gned compu te r
a lg orithm t o in tro du ce app are nt mo tio n d urin g p ro ce ss in g
FIGURE4. Ef fectsfmot ioniming—polaroo rdin ateisp lay.
Shown a re r esult an t pola r d is pla ys fo r t he s tudie s i llu st ra ted her e
wherethe ef fectof “mot ion―n Frames7—14s shown aboveand
the e ffec t due to a lt erat ions in Frames 18—25s shown be low.
Here d if fe rences in de fec t s ize and loca tion are apparen t.The
co lo r po la r d isp lays a t lef t r epresen t the relat ive d if fe rence in
count distribution betw een altered stress and delayed SPE CT
imag esets.Hereredo rd arkgreyrep resentspeaka ctivity.Those
at r ight indicateregionalabnormal it iesin al teredpoststressand
delayedimagescomp aredo a normalpopulation.
evaluations of the relationship betw een m otion displace
m ent and resultant SPEC T artifacts (7,8). O thers per
fo rm ed lim ite d a na ly sis o f a rtifa cts p ro du ce d b y motio n
in ten tio nally in tro duced d urin g acq uisitio n o r p rocessing
(7,9) . Eisneret al . (7) useda complexcross-correlation
F IGURE5. Exte ntofdisplacement.hownstherelationship
be tween the exten t o f d isp lacement abscis sa ) and resul tan t
de fec t s ize ord ina te )quan ti ta ted on po la r d isp lay , f o r upward
and downwa rd d is pla cemen t i n th e s ame p ro je ctio n image s
Frames7—1).N otethe similarpatternsand defectsize, and the
reduct ionin defect s ize with increased“mot ion― .
308
T he Journal of N uclea r M edicine V ol. 34 N o. 2 F eb ruary 1993
8/21/2019 spect images
http://slidepdf.com/reader/full/spect-images 7/8
a le rt to th eir o cc urre nc e a nd h elp u s u nd ersta nd th e lim its
C and effects of motion. This could aid image interpretation.
A Previousorkersuggestedhatmotionslitt leasa
cen tim eter co uld cause a significant im age artifact ( 7 ).
Wh ile th is may be t h e c as ewhen many frames a re a ffe cte d,
t he s itu atio n i s c le arly mo re comp lex.
Ô The presence and extent of SPECT vertical motion
artifacts vary with num ber and location of the projection
im ages affected as w ell as the direction and extent of their
B displacement.pecificrtifactualffectsf motionon
0 SPECTmagingereelatedohextentfdisplacem
th e duration of d isplacem en t, the num ber of fram es an d
th e s pe cific frames a ffe cte d ( Fig s. 1 †”4n d T ab le s 1 and 2 ).
W hile the SPEC T m ethod seem ed quite sensitive to dis
pla cem ent tow ard the ce nter of the orbit, a u nique sensi
C tivityof thecentralregioncouldnotbedemonstrated.
The extent and duration of vertical displacem ent ap
peared additive in their influence on im age artifacts. T his
‘ s upports the pre lim inary findings ofP rigent and coworkers
(9 ) who notedthat the productof extentand durat ionof
m otio n, th e p ix el-fram e †œarea ,†•eterm ine s th e p resenc e
a nd e xte nt o f re su ltin g im ag e a rtifa ct. H ow ev er, S PE CT
per fu si on images a re r el at iv ely tol er an t t o i so la ted mo ti on
a ff ec ti ng f ew f rames. Wh i le even ext en si ve mo ti on b rought
little change when affecting one or two frames, minor
d isplacem en t b ro ugh t ob vio us d efects w hen affectin g eig ht
frames . T he se fin din gs agre e g en era lly w ith o bs er va tio ns
of others (7†”9) nd carry them further. T he loc ation of
S PE CT m otion a rtifacts relates to th e d irectio n o f d isp la ce
men t a s w ell a s th e lo ca tio n o f frame s a ffe cte d.
S om e o fth e e ffe cts o fp atie nt m otio n o n SPE CT im ag es
appear obv ious , wh il e o th er s, s uch as t he b iphasi c r el at ion
ship betw een m otion displacem ent and S PE CT defect size,
w ere surprisin g (F igs. 3 and 6, T able 1). T hese findin gs
suggest an effect of the backprojection technique on the
an aly sis o f disp laced an d n on displaced pro jection im ag es;
th ese m ay o verlap as d isp lacem en t app ro ach es th e v entric
F IGURE 8 . C lin ic al s tu dy †”h oriz on ta l
mo tio n. Shown a re e xample s o f SPECT
shor t-ax is left column)ver tica l long axis
cen te r co lumn) and horizon ta l long-ax is
rightcolum n)slicesacquiredpo stexercise
i n a c li ni ca l s tudy associ at ed wit h obvious
horizontaldisplacem ent above ).Gross ar
t if ac tua l dupl icat ion of shi ft ed vent ri cu la r
wallsisapparent inshort-axisandhorizon
te l lon g-ax is slices. T hese cou ld n ot b e
s oftwa re c orre cte d but we re no lo nger
ev ident on subsequen t r e- study (below) .
Rota tion in or coun te r to the orb it ald i rec
ti on was eva lua ted in compu te r s tud ies.
H ow ev er, in th is clin ic al c as e th ere w as
eviden t ho ri zon ta l mo ti on in additi on to
rotation.
.
:@ :b
e b
A
. . O O
0 00
FIGURE7. Clinicalstudy—vert icalotion postexercise.
Show naccordingo the formatof Figure3, are immediatepost
stress(A)an d4-hr delay(C)co ntiguou short-axisSPE CTs lices
f rom apex upper le ft ) t o base lower r ight )acqui red in a c linica l
s tudy w ith g ross ver tic al pa tien t mo tion du ri ng the immediate
pos tstress cqu isition.ere,obviousdefectswereim provedafter
applyingthe mot ion cor rectionalgor ithmon the same immediate
poststre ssdata .Thisp artiallyco rrec tedver sionof imagesshown
i n A) i s shown in B). Postp rocess ingcou ld no t cor rect mo tion
ar ti facts in most cases.
0
in otherw ise norm al SPE CT studies in order to determ ine
the sp ecific effects o f v ertical p atien t m otio n alon g a sin gle
axis on resultant SPECT images. In the course of this
analysis, tolerance ofthe SPE CT m ethod to patient m otion
w as assessed an d the in fluence o f the ex tent, direc tion,
duration and location of motion, which determine such
tolerance, were identified and quantitated. W hile the find
ings here w ill not prevent artifacts, they w ill m ake us m ore
309
ff ec to f Pa tien tMotionon SPECT Images Botv in icke t a l.
8/21/2019 spect images
http://slidepdf.com/reader/full/spect-images 8/8
u la r d imens ion. Here , p ro ject ed myocar di al wa ll s i ni ti al ly
m ove apart, then overlap with reinforcem ent of radioac
tiv ity in su pe rim po se d s tru ctu re s, m ax im iz in g th e re su ltin g
SPECT defect, and then move apart again to minimize
t he def ec t. I nc remen ta l d is pl acement b rought f ir st i nc reas
ing, then decreasing SPECT defects related to the degree
o fim ag e ov erlap an d the v arying reinfo rcem en t o f reg io nal
ray sums. The nature of image motion artifacts relates to
the interaction of image data in ray sums and can be less
severe and more lim ited in relation to greater displace
m en t. W hile o f u nlik ely practical clin ical sig nifican ce, th e
analysis of the effects of such extensive m otion provides
in sig ht in to th e m etho d.
I de ntif ie d wer e v is ua l c ue s s ugge sti ng a rtif ac tu al mo tio n
related SPECT defects. The irregular, lumpy nature of
d efe cts, th e o fte n o pp os in g d is trib utio n o f motio n-re la te d
artifacts an d the p resen ce o f c urv ilinear extra ven tric ula r
ra dio ac tiv ity lik ely re la te to th e re la tio ns hip o f b ac kp ro je c
tion ray sum s of displaced and nondisplaced ventricular
radioactivity (Fig. 3). T hese should be carefully sought on
SPE CT in te rp re ta tio n. H ow ev er, o nly a re view o f p ro je c
tion images can reliably provide supporting evidence of
p ati en t mo tio n.
W hile horizontal m otion w as less intensely analyzed, it
appeared as though the SP ECT m ethod w as m ore sensitive
to motion counter to the rotational axis and that defects
related to vertical and horizontal m otion w ere additive.
Some have suggested the possibility of correcting for
patient motion during SPECT processing ( 7,12). The al
gorithm s applied here to test the effects of patient m otion
artifacts could be applied to correct such artifacts by the
proper adjustm ent and alignm ent of affected projection
images. Of course, such correction could only succeed if
the artifact was produced by a pure vertical or horizontal
motion in specific identified frames. As we noted in our
clin ical su bgro up, this oc curs rarely an d o ur ab ility to fu lly
correct m otion artifacts is lim ited (Figs. 7 and 8). Further,
this method can do nothing to correct for artifact due to
motion in other planes. The clinical patient example of
h oriz on ta l m otio n a rtifa ct (F ig . 8 ) w as fu rth er c omplic ate d
by evident horizontal displacem ent. In addition to proba
ble horizontal rotation, the latter alone could likely be
analyzed and corrected w ith the algorithm presented here.
T he d ou ble c ard ia c im ag e p re se nte d in th e c lin ic al stu dy
likely results from this horizontal displacement but was
not evident in association w ith sim ulated horizontal, ro
tational m otion created here by the com puter algorithm .
O n th e other hand, applica tion of the m ethod, seeking
correction of apparent motion artifact with partial but
obvious improvement in resulting SPECT defects, may
provide tangible support to the origin of SPECT image
defects in patient m otion. As noted by D ePuey and Garcia
5 , motion is best detected by observation ofthe dynamic
disp lay o f th e p ro jection im ag es. It is o bviou sly im po rtan t
to recognize the presence of m otion and the observations
here help to relate this observation to SPECT im age find
ings. However, it is clear from this study and a wealth of
clinical material that the best way to minimize a SPECT
m otion artifact is to p rev ent its o ccu rren ce.
ACKNOWLEDGMENTS
T he authors thank C arm en B reslin and C laudia Fuller for
helping to put our thoughts dow n on paper. W e also thank our
te chno logis ts , who hel ped u s avo id th e p itf alls th at we ana ly zed
her e. T hi s wo rk was s uppo rt ed in p ar t b y a g ra nt f rom The Fannie
E . R ippe l Founda tio n, Annandal e, N ew J er se y.
REFERENCES
1. G arcia EV , van Train K , M addahi J, et al. Q uantification of rotational
thall ium-20l myocardia ltomography.J Nuc/Med 1985;26:17—26.
2 . M ad da hi J , v an T ra in K F, P rig en t F , et a t. Q ua ntifica tio n o f T l-2 0l
myocardi al s in gl e pho to n emis si on comput er iz ed rot at io na l t omogr aphy :
developm ent,alidationandp rospectivevaluationof an optimizedcom
put er iz ed me thod . J NuciMed l 986; 27 :8 99 †”907 .
3 . Mas ticB, Bo tv in ickEH, Arnold S. e t a . Ef fec to f cont ra s tenhancement
on the sensitivity and specificity of T l-201 scintigraphy. A m H eart I
198l;l02:37—44.
4 . Fr iedmanJD, BermanDS,vanTra inK, e ta l . Pa tien tmo t ion in thal lium
201 m yoca rd ial S PE CT im agin g: a n easily id en tified frequ ent sou rce of
ar ti fac tua l defec t. C/ in Nuc/Med l988 ;13 :321—324 .
5 . DePuey EG, Garc ia EV . Op timal s pe ic fi ci ty o f t ha ll ium-20l SPECT
throughrecognitionofimaging artifacts.I Nuc/Med 1989;30:441-449.
6. EisnerRL, NoeverT, NorwakD, et al.U seof crosscorrelat ionunct ion
t o de te ct pa ti en t mot ion during SPECT imaging . JNuc Med l987 ;28 :97—
101.
7 . E is ne rR , Chur chwel lA, Noev er T , e t a l. Quan ti ta ti ve ana ly si s of t he
t omog ra ph ic th al li um -2 0l my oc ar di al b ul l's -e ye d is pl ay : C rit ic al r ole o f
cor re ct in g for p a ti en t mo ti on . I Nuc/Med 198829 :9 l—97 .
8 . F ri edman J , v an T r ai n K , Madd ah i J, e t a . â €œUpwa rdreep―fthe heart:
a frequentsource of false-positivereversibledefectsduring thallium-20l
stress-redistributionSPECT .I NuciM ed 1989;30:1718—1722.
9. Pr igentF, Bel li lD , BermanD, eta l . E f fecto f mot ionon Tl -20 l SPECT
s tud ies . Ci rcu la t ion l989 ;80 (supp l I I) :l I-366 .
10. D iamond GA , Fo rr es te ri S. An al ys is o fp robabi li ty a s an a id i n t he c li ni ca l
d iagnos is o fcoron a ry a rt er y d is ea se . N Eng JMed 1979; 300: 1350 †”1354 .
11. DePasquale EE, Nody AC , DePuey EG , et a . Q uantitative rotational
thallium-20l tomographyfor identificationand localizingcoronaryartery
disease. Circulat ion l988;27:999—1004.
1 2. F lem in g i S. A te ch niq ue f or mo ti on c or re ct io n in d yn am ic s cin tig ra ph y.
Eur J Nuc iM ed 1 98 4; 9: 39 7â €”40 7.
310 The Journal of N uclear M edicine Vol. 34 N o. 2 February 1993