Download Anatometabolic Image Fusion: PET/MRI/CT for Locating Increased Activity and more Slides Anatomy in PDF only on Docsity!
uptake of FDG is so high that relatively little soft-tissue
background activity is identified. In such cases, determin
ing where foci of increased FDG uptake are located can be
difficult. This problem is even more apparent with high
contrast parametricimages (12). Fusion of metabolic emis
sion images from positron emission tomography (PET) to
the much higher resolution anatomic informationavailable
from magnetic resonance imaging (Mifi) or computed to
mography (CT) scans seems desirable (13). This approach
has been applied by several groups to fuse PET on single
photon emission computed tomography(SPEC!') emission
images of the brain with MR or C1@anatomic images
(14,19). Indeed, tracer uptake has been mapped onto pa
tient-specffic or atlas-based brain anatomy (14—23).Since
the normalbrainhas reasonably high levels of FDG uptake
throughout it and moderate anatomic information is avail
able, anatomicinformationfrom emission brainimages can
be used in the realignments (1824).
Whereas anatomically defining the location of PET
tracer uptake in the brain is feasible, the situation in the
body, where many normal tissues have much lower levels
of FDG accumulationthan the brain and thus provide less
anatomy, is more challenging. In addition, while the brain
and its contents are fixed, the body can twist and bend. In
the present study, we demonstrate the feasibility of digital
image fusion between body PET and MRI or CT to pro
duce hybrid “anatometabolic―images of the body, which
contain both the exquisite anatomic detail of CF or Mill
andthe uniquephysiological informationfrom PET. This is
achieved using a rigid rotate-translate scale model and a
realignment algorithm based on operator-identification of
external markers present on both emission PET and Cr
Mill as well as, in some instances, unique anatomic points
defined on both transmission PET images and the anatomic
study.
METHODS
Ten patientswere studied.Eachpatientwas > 18yr old and
providedwritteninformedconsentforparticipationin the study.
The PET Studieswere performedas partsof ongoingclinical
Positron emission tomographic (PET) Images of v@ceralcan
cars are commonlyvisualizedas “hOtSpots―ofincreased activity
with relativelylittlenormal anatomy discemable, when 2-[18F]-
fluoro-2-deoxy-D-glucose(FDG)Is used as the tracer. We de
scribe a method by which computed tomography or magnetic
resonance anatomic images can be digitallyfused in three di
mensions, using a dgid rotate-translate scale model with PET
“metabolic―images, to simuftaneouslyd@ay registered ana
tomic and metabolic information. Such “anatometabolic―fusion
images were produced in 10 patients witha vwlety of visceral
cancers. E@ctemaIfiducialmarkers placed during both the ana
tomic and the metabolic study, as well as internal anatomic
fiducials defined from landmarks observed on reconstructed
transmission Images, were used to achieve image fusion. The
mean error magnitude ±sam. offidudal reg@trationInthe nine
patients with successful realignmentswas 5.0 ±0.8 mm. The
mean accuracy In realignment between known anatomic struc
tures seen on boththe anatomicstudy and on the emission PET
scan (butnot used Inrealignment)was 6.3 ±0.8 mm. Localiza
tion of fociof Increased FDG uptake to specificanatomic struc
tures was achieved by this method, which represented an en
hancement over the rudimentaiy anatomy available from the
emission images alone. Anatometabolic fusion images made
using this reasonably simple method should prove useful in the
management of patients withcancer and other diseases.
J NucIMed1993;34:1190-
positron-emitting analog of glucose, 2-['8F]-fluoro-2-
deoxy-D-glucose (FDG), rapidly accumulates in many hu
man neoplasms following intravenous injection (1—li).In
deed, when the FDG accumulation of neoplasms is
compared to that of most normal tissues, a relatively high
target-to-background ratio is often observed (7,8).
Although high target-to-background ratios are useful for
tumor detection, they may also be problematic if tumor
ReceivedOct 19, 1992;revIsionaccepted Feb. 16, 1993. For correspondence or reprints contact Richard L Wahi, MD, DMsbn of
NudearMedIc@ne,UniversityofMichiganMedicalCenter, 1500 E.Med@Center
DrIve,B1G412,AnnAibor,Ml48109-0028.
1190 The Journalof NudearMedicine•Vol.34 •No. 7 •July
‘‘Anatometabolic' ‘Tumor Imaging : Fusion of
FDG PET with CT or MRI to Localize Foci of
Increased Activity
Richard L. Wahi, Leslie E. Quint, Richard D. Cieslak, Alex M. Aisen, Robert A. Koeppe and
Charles R. Meyer
Departments ofRadiology and InternalMedicine, Division ofNuclear Medicine, University ofMichwan Medical Center, Ann Arbor, Michigan
MeanMean
ofNumber erroraccuracy
ofmagnitudereallgnmenttPt.
no. Cancer MOdality fiduclals(e,l)*(mm)(mm)
@ *e numberofexternalfidudals;I- numberofInternalfidudalsusedInrealignment.
tNU@Ik@eI@In parentheses Is number of points assessed for accuracy of fusion. @TechnlcalIyunsatisfactory due to patient motion during emission PET study and was not used In calculating mean error magnitude or accuracy.
TABLE I
Accuracyof Fusion InTen Patients withAnatometabolicTumor Imaging
IBreastMR4(1,3)4.66.2(3)2TestIsMR8(8,0)4.08.4(3)3TestIsMR8(8,0)5.69.3(4)4ProstateMR5(3,2)5.57.3(3)5BladderMR5(5,0)5.14.4(2)6LungCT4(3,1)2.36.6(4)7BladderMR*
(1)8LungCT5(4,1)3.61.0(2)9LungCT5 (5,0)10.424.V*
.07.4(3)10LungCT4 (1,4)1 1 (3)Mean± (1,3)3.76.
s.e.m.5.04 ±0.826.3 ±0.
because of their improved visibility, their commercial availability and their ability to define the geometric center of the markers. Crscanswereobtainedinfourpatientsinthestandardfashion
with or withoutintravenouscontrast,dependingon the clinical
situation. A GE 9800 CT (General Electric Medical Systems,
Milwaukee,WI)scannerwas usedwith a 2-secscan time. Images
were displayed using a 512 x 512 pixel image matrix. Small
cylindricallead markers, roughly2 mm in diameter x 1.0 cm in
length,were placedover the same markedskin sites used for the
PETstudy,orientedperpendicularto the scanplane.Generally,
i-cm thick contiguous tomographic sections slices were obtained.
InsomecasestheanatomicstudyprecededthePETscan.Ineach
case, however, the PET and anatomic studies were performed on
thesameday.Duringeachimagingprocedure,effortsweremade
to positionthe patientstraightand fiat on theirbacks on the
scanning table but no external frames were used.
Imag Fusion
All cross-sectional image sets were reconstructed initially on
each acquisitionsystem usingstandardvendor-suppliedrecon
struction algorithms.Image fusionswere performedon a Sun 4
workstationrunningSunOSUnixandthenativeMITX-windows
graphicaluser interface.Reconstructeddatasetswere transferred
to the workstationusingthe Internetfile transferprotocol.To
reducethe spatialgranularityof z-axis identificationof marker
location,all imagesets were first interpolatedin the z-axisdirec
tion using a fourth-order,Hamming-weightedsinc functionto
yieldthree evenlyseparatedinterpolatedimagesforeveryoriginal
cross-sectionalimage.The sinc functionwas chosen because of
its particular suitability for reconstructing sample data sets (25).
Thez-axisinterpolationmaintainedanisotropicvoxels, i.e., “cu
beriles―were not made.Note thatthe interpolateddatasetwas
used only for determininggeometrictransformation.Identifica
tion of corresponding three-dimensional markers (homologous
point pairs) in each dataset were selected by the operator using
mouse-driven,pairedslider-scrolledimagesandpointer.Thecen
ter of the fiducial markers was used in the realignments. After
point pairs were identifiedin the datasets, a rigid rotate-scale
protocols to evaluate the diagnostic accuracy of PET in a variety
of cancers.Thestudypopulationis brieflydescribedin Table1.
PETimageswereacquiredwitha SiemensCli 931i08-12PET
scanner (Siemens Medical, Iselin, NJ). The method for image
acquisition has previously been described and 15 slices of ‘—6.& mm thickness were obtained per scanninglevel(810,11). In brief,
transmissionPET datasetswere firstobtainedover specificareas
ofthe body suspected ofcontaining cancer for a periodof ‘—10min per set of 15 slices (with axial coverage of —10cm) using the
retractableasGe/@Garingtransmissionsources of the PET scan
ncr. Transmissionimageswere reconstructedby filteredbackpro
jection and generallycontained5—10milliondetectedcoincidence
events per image. Following the performance of one or more transmissionscans, radioactivefiducialmarkerswere taped onto
ink-markedskin sites on the patient that were expected to be
within the field of view of the scanner. These fiducialmarkers
contained 0.5-3 iCi of FDG adsorbed onto —-2-5-mmdiameter
polystyrene beads. Markers were placed so that the suspected
tumor region was thoroughly surroundedby markers (generally covering 15—20cm in the z-axis). Up to eight markerswere placed per patient. Emission images reconstructed using filtered back
projectionwere obtainedfollowingthe intravenousinjectionof
approximately 10 mCi of FDG. The two 10-mis duration images
from50—70miii postinjectionwere used for fusions.Generally
—20cm of the patient was imaged in the z-axis, which represented
two contiguous 10-cmacquisitions.A 128 x 128image matrix
displaywas used to displayemissionimagesof approximately
mm resolution.
Six patients had MR scans obtained with a Picker Vista 0.
Tesla or GE Signa 1.5 Tesla Mill device. StandardTi-weighted,
T2-weightedand/or proton density spin-echo images were ob
tamed, which were 5—10mm in slice thickness, generallywith a
i-mm gap between images. Images were displayed using a 256 x 256 image matrix. Vitamin E capsules (iOOI.U. vitamin E, Rorer
Inc., Ft. Washington,PA)weretapedoverthesamemarkedskin
sites used forthe PETimagingstudyandwere orientedperpen
dicularto the axialscan plane. Anisotropicmarkerswere chosen
AnatometabolicTumor Imaging•Wahi at at. (^) I 191
@ I p@@GroROTATE-SCALE- Physicianassessfor I TRANSLATE I ALGORiTHM+patientmotion/
_____________H
@ I@ie@o@ma@
Ideletequestionable
@@ I FusionAnatoetbo' I
I imagesproducedover I I 20ariz-a,ds I
Flow Chart of Registration Process Between CT/MR and Emission PET to Form the “AnametabolicImage Set@'
Technologist Identifies unique anatomic points on recon structed transmission scan (ifany) and applies to registered emission PET
Adequate
FiGURE 1. Row chart detalhingmethod of anatometabolicImageformation(after
Ref.18).
a representationof stable scarringafter aggressive chemo
therapy(7,27).
The potential of the method in the care of a patientwith
newly diagnosedlung canceris illustrated in Figure 3A. In
this patient, CF shows volume loss in the rightlower lung
and a localized right pleural effusion. The corresponding
PET image shows intense central FDG uptake with a small
focus of modest FDG uptakein the anteriorleft chest, with
only minimal FDG uptake elsewhere. The fusion image
clearly demonstrates intense FDG uptake in the central
portion of the rightlower lobe collapse. This appearanceis
most consistent with central FDG-avid lung cancer,
whereas the peripheral area represents postobstructive
lung volume loss with little FDG uptake. A lower set of
images (Fig. 3B) shows that the anterior left chest FDG
activity is myocardial in origin. This fusion over multiple
image planes illustrates that the fusions are true fusions in
three dimensions,not just realignmentswithin a single
imaging plane.
Theabilityof our algorithmto realignvolumesof tissue
is also clearly shown in a patient with metastatic breast
cancer (Fig. 4). Images high in the thorax (Fig. 4A), in the
mid-thorax (Fig. 4B) and in the mediastinum (Fig. 4C)
show excellent alignment. Three foci of increased FDG
uptakewere difficultto localize on emission PET alone due
to the high target-to-background ratios. The fusion image
demonstrated that there was focal tumor in an anterior
right rib and not in a lymph node as was suspected without
the fusion images (Fig. 4B). An unexpected observation
was two foci of intense FDG uptake in the mediastinum
(Fig. 4C). The fusion image (Fig. 4C) showed that the FDG
uptake was into normal-sized lymph nodes. On follow-up
studies several months later, this patient had progressive
systemic and mediastinal involvement with breast cancer,
which implied that PET detected cancer in normal-sized
lymph nodes. Such cases where FDG activity localizes to
apparentlynormal structures on anatomic imagingstudies
would seem to represent a situation in which anatometa
bolicfusionimagingwouldbe of greatvalue. Fusion imaging was also possible in the abdomen,
although external fiducial markers were more necessary
than in the chest because relatively little internalanatomic
informationwas provided by transmission images of the
abdomen. In our small series, there were no quantitative
Matometabolic Tumor Imaging•Wahi at at. 1193
) ‘T'
___________ luniqueanatomic I
.. Ipoints from tram- I _________*—ImiuionscanontheI [MR/cs I
differences in the quality of realignmentsin comparingthe
chest and abdomen (Table 1). For example, in a patient
beingevaluatedasa partof anongoingstudyof prostatic carcinoma, excellent image realignment was achieved
(Fig. 5). This individual had an abnormal bone scan and
Ti-weighted MR image with metastatic prostatic card
noma to the acetabulum/hipregion. Emission PET demon
stratedtwo foci of FDG activity, one of moderate intensity
in the prostate itself(he had residualprostaticcarcinomain
the gland), and a more intense triangularfocus to the right
of the prostate. Fusion images showed the right-most ac
tivity to be in the right acetabulum (Fig. 5, bottom).
Although realignments with our current algorithm are sometimes imperfect (such as in Patient 7) and more so phisticated warping algorithms might be useful, it is clear
that our relatively simple approach, which can align to
within 5—6mm, has practical utility.
DISCUSSION
Our study demonstrates that it is possible to obtain rea
sonably accurate anatometabolic fusion images between
Cr orMRandFDG/PETscansbyusingarelativelysimple
rotate-translate-scalerigidmodel and manualidentification
of fiducial points from the CT, MR and transmission and
emissionPET images.Theseanatometabolicimagesap
FiGURE2. Patientwitha historyof germcellcaranomaof the
testis with muWptestable residual pulmonaiy nodules followingag
gres@vechemotherapy.TheupperlefthandimageisaTl-walghted
cm667,TE20)spin-echoMRimageshowinga retrocardiacnod
ule,a smallnoduleatthetightcosto-phrenicangleandthreevitamin
E markers positioned over the anterior chest. The FDG PET emis slon image at 60 mm postinjecdon (upper nght) clearly shows the three FDGfiducialmarkersatthe same skinlocationsas the vitamin E markers, myocard@IFDG uptake and a small photopenic area in
the retrocardiacarea.Thefusionimage(bottom)shows excellent
registrationbetweenthe FDGand vitaminE markers,demonstrates
FDG uptake in the cardiac wall and shows no FDG uptake in either pulmonarynodule. Modest FDG uptake projects in the expected
locationfortheliver.Follow-upofthsspatientshowedstabilityofthe
pulmonarynodulesmostconsistentwithscarnngand a lackof FDG
accumulation.
FiGURE3. (A)CTshowsmarkedvolumeloss Inthe tightlower
lung and a lOcalIZedrightpleuraleffusion in a patientwith newly
diagnosed lung cancer. The correspondingtransverse emission
PETimageobtainedat 60 mmpostinjectionshows intensecentral
FDG uptake in the right chest modest focal FDG uptake in the
anteriorleftchest, but onlyminimalFDGuptake in the bicodpool
andthe collapsedrightlowerlobe/effusionarea.Thefusionimage
(bottom)dearly demonstratesintense FDG uptake in the central
portionofthe collapsedrightlowerlungmostconsistentwithuptake
inthe centraltumor,whereasthe remainderof the rightlunghas
actMtymostconsletentwithpostobstrucffvevolumeloss.MRI per
formedelsewheresupportedthisinterpretationandthe largecentral
turnorwastreatednonsurgically.(B)Amorecaudalseriesofimages
from the same patient as in (A) demonstrates that the left anterior
chest activityseen in (A)is myocardialInorigin,withthe leftventric
ular wall clearly delineated by PET, despite its poor delineation on
the CTstudy.Thisis bestseen onthefusionimage.
pear to be particularlyuseful in defining the anatomic lo
cation of intense foci of FDG uptake that are otherwise
difficult to localize precisely due to limited background
activity.
The method we describe, using both external fiducial
markers and available internalpoints of anatomy (derived
from transmissionimages), is a unique approachto realign
ment. The use of the transmission data in addition to the
1194 The Joumat of Nudear Medicine•Vol.34 •No. 7 •July 1993
Another approach to realigning MR and PET images of
the braindoes not requireexternal markersor transmission
images. This involves defining the interhemispheric plane
on both MR and emission PET, with subsequent operator
interactions to produce realignments in the other dimen
sions (18). Although this approach is said to require less
than an hour of processing, it can requireuser definitionof
points on the emission PET scan, including the border
between the “graymatter and CSE―This can be difficult
due to the poor resolution of PET versus CT or Mill and
because white matter mainly adjoins the CSF. This algo
rithm is also reported to be less successful if the interhemi spheric fissure is nonplanar (18).
The preceding approaches for brain image registration
reported “errormagnitudes―in the 3—4mm range. Our
error magnitude of 5—6mm compares well with these,
particularly considering that the body is much more subject
to deformation than the brain. There is, however, some
uncertainty in our assessment of accuracy of realignment
as it is sometimes challenging to accurately define the lo
cation ofvessels on emission PET, and with some vessels,
to uniquely define their z-axis location is not possible.
While it may become feasible in the future to eliminate
the need for skin surface markers in the realignmentalgo
rithm, a trained technologist can prepare and apply mark ersin a few minutes.In mostcases,onetechnologistap
plied the ‘8Fmarkersand another applied the vitamin E or
lead markers, so a “devoted―technologist was not essen
tial. Careful attention to positioning the patient “flat―on
the imagingtable was employed.
A trained technologist was able to perform the computer realignment procedure in less than 2 hr, often without
assistance from a radiologist. This seems an advantage
over methods requiringa greater familiaritywith emission
PET and Mill anatomy than the modest knowledge of
anatomy needed to define points on the transmission im
ages with our method (18). In our realignment algorithm,
anatomic input from a physician is to assess the subjective
quality and quantitative accuracy of realignment and to
determine if patient motion occurred.
A rapid reliable fusion method for body imaging that
does not require external fiducials would be preferable to
our method. By incorporating anatomic data from the
transmission images, we moved in this direction and were
able to achieve successful fusions using as few as one
external fiducial marker in three of our patients. Clearly,
themethodweappliedislesscumbersomethanhavingthe
patient rigidlyimmobilized in a mold or frame, as has been
done in brain imaging studies (22,29). It should be noted
that fusion images and/or extraction of region of interest
information also have shown promise in better localizing
abnormalities seen on SPECT imaging using @“Fc-red
blood cells for hemangioma detection and with SPECT of
radiolabeledmonoclonal antibodies (30—32).This suggests
a more general applicabilityof such fusion methods.
In summary, we describe a method which allows for the
fusion of CT or MR anatomic images in three dimensions
S
FiGURE5. Ti-weightedMRI(TA600,TE20)demonstratesab
normal signal intensity in the region of the right acetabukim In thus patient with metastatic prostatic carcinoma to bone. Emission PET demonstrated two foolof FDG activity,one inthe midlineand one to
the rightof midline.FusionimagesdemOnStratedthatthe focal
uptake in the midlinewas in the region of the prostategland (the patient had residual prostate carcinoma inthe gland), and the signal from the right of midlinefused into the area of abnormal MR signal (metastasis) bested in the right acetabulum. Fidelityof the fusion was further substantiated by the fact that the two falnt fool of FDG
uptakelocatedanteriorandto eithersideoftheprostatewereseen
at multiple levels on PET (i.e., were tUbUlar)and fused into the
locationofthe externaliliacbloodvessels.
5. Surface anatomy (and thus markerlocation) is variant
between the two studies, such as in very obese pa
tients where a change in the appearance of the pannus
of fat (and markers on it) might not reflect internal
anatomy.
These issues are partly addressed since we have been
careful to perform the anatomic and physiologic studies on
the same day with the patient laid carefully onto his/her
back, in a similarstate of hydrationor feeding. Obviously,
more sophisticated algorithmswhich include the ability to
“warp―the three-dimensional image datasets between the
two studies would be ideal for more precise realignment.
Our approach to realignment and registration of body
PET studies with CT or MIII is related to several of the
approaches described for fusing brain PET or SPECT im
ages with MR or Cl' images (14—2Z24).The realignment
algorithms for brain anatomic and emission studies have
taken several foims, but the situationin the brainis simpler
than in the body because the brain and skull do not sub
stantially deform between studies. For the brain, realign
ment algorithms have been described that eliminate the
need for surface markers. One such algorithm uses the
transmission-scan defined surface of the skull and the sur face of the scalp on Mill or CT to perform realigmnents (14,15,17).Thisso-called“headin thehat―algorithm searches for minima in the quality of surface realignments,
but is not totally operator-independent.This approach re
portedly requires —2—3hr/patient to complete (17).
1196 The Journalof NuclearMedicine•Vol.34 •No. 7 •July
@ ii
10. WahIRL, HarneyJ, HutchinsG, GrossmanMB.Imagingof renalcancer using FDG PET: pilot animal and human studies.I Urol 1991;146:1470- **1474.
- Harneyiv, wai@iRL, LiebertM, Ctal. Uptakeof 2-decxy,2-(18F)fluoro** d-glucoscin bladdercancer: animallocalizationand initialpatient positron emissiontomography.J Urol 1991;145:279-283. 12.WeldRL,ZasadnyKR,GreenoughR, KoeppeR. Parametricimagedis plays to enhanceVisUaliZatiOnof cancers usingFDG/PET:influxconstant **andtemporalsubtractionimaging[Abstractj.Radiology1991;181:152.
- WeldRL,,MeyerC, KoeppeR, MacnA, QuintLE. Whereis the tumor** imagedon FDG PET scannrngas a “hotspot―?Methods for metabolic/ anatomicregistration[Abstract).Radiology1991;181:151.
- Levis DN, PelizzariCA, den OTY,Oen C-T,CooperMD.Retrospective geometriccorrelationof MR, Cl', and PET lmages Radiology 1988;169: 817—823. 15. PelizzariCA, Chen GTY, SpelbringDR, WeichselbaumRR, Chen C-T. Accurate three-dimensionalregistrationof CIT,PET and/or MR imagesof the brain.I Cony@AssLctTomogr1989;13:@-26. 16. AlpertNM, BradahawJF, KennedyD, CorralsJA. The principalaxes trandormation—amethodforlmageregistratlon.JNuclMed 199031:1717-
17. HolmanBL, ZimmermanRE, JohnsonKA, et al. Computer-assistedsu perimpositionof magneticresonanceand high-resolutiontechnetium-99m- HMPAO and thallium-@1 SPECT images of the brain. INuci Med 1991; 32:1478—1484. 18. KapouleasI, Alavi A, Alves WM, OurRE, Weiss DW. Registrationof three-dimensionalMR and PET imagesof the humanbrain without mark era. Radiology1991;181:731—739. 19. Fox PT, PerlinutterJS, RaichieME. A stereotacticmethodof anatomical localizationfor positronemissiontomography.JCony,Assist Tmsogr1985; 9:141—153.
- BohmC, OreitzT, KingsleyD, BerggrenBM, OlasonL Adjustablecorn puterizedstereotaxicbralnatlasfor transmissionand emissiontomography. Am JNeUTOTO4IOI1983;4:731-733. 21.MazziottaJC,PelizzariCC,QienGT,BOOksteInFL,ValentinoD.Region of interest issues: the relationshipbetween structure and function in the brain.I CerebBloodflow Metab 1991;11(suppl1)-A51-A56.
- Evans AC, BellC, Marrett S, ThompsonCJ, HaldznA. Anatornical-func tionalcorreistionusinganadjustableMRI-basedregionof Interestatlaswlth positronemissiontomography.JCereb B1OOdFIOWMeab1988;8:513-530.
- VannierMW,GayouDE. Automatedregistrationofmultimodalityimages. Radiology 1988;16@860-861.
- FristonKS,PassinghamRE, Nutt JO, Heather JD, SawleGV, Frackowiak RSJ.Localizationof PETimages:Directfittingof theintercommiuural (AC-PC)line.I Cesrb BloodFlowMetab 1989,9:690-695.
- PapoulisA. TheFowierbuegmiandits acatio@as New Yoric McGraw Hill; 1962:52.
- NobleB, DanielJW.Appliedlineara@geb,mEnglewood @1Iffs,NJ: Prentice Hall; 1988:346—350.
- Wahi RL, Harney JV, HutchinsGD, GrossmanHB. PET imagingof GU neoplasms:initial evaluation. Presented at 76th ScientIficMaembly and AnnualMeeting,RSNA, 1990.
- BookstelnFL. Principlewarps: thinplate splinesand the decompositionof deformation.IEEE Tiw PauemAnoi Med Intel! 1989;11:567-585.
- ZhangJ, Levesque MF, WilsonCL, et aL MultirnOdalityimagingof brain structuresfor stereotacticsurgety.Radiology1990;175:435-441.
- BirnbauinBA, Nor ME, Chapnick,et al. Hepatichemangiornas:Diagnosis with fusion MR, CT, and Tc-99m-labeledred blood cell SPEC1@images. Radio'ogy 1991;181:469—474. 31. KramerEL, Nor ME, SangerJJ,MegibowAJ, MaguireGO.CF-SPECF fusiontocorrelate radiolabeledmonoclonalantibodyuptakewithabdominal CT findings. Radiology 1989;172:861—865. 32. KramerEL, Nor ME.CF-SPECI@fusionforanalysisof radiolabeledanti. bodies:applicationsin gastrointestinaland lungcarcinoma.IntlRad App! Jrtstnim(BJ1991;18:27—42.
1197
with PET metabolic images to produce a series of
anatometabolic images. The method involves the use of
external fiducial markers for all imaging modalities and is
further refined through the use of internal landmarks from
foci in the body that are uniquely defined on both the
transmission PET images and the CF or Mill study. Such
anatometabolic information should prove useful in
anatometabolic directed biopsies of areas most likely to
contain active cancer that may appear normal by other
imagingtests (29). While additional experience and simplification will be
necessary to routinely apply our method to clinical prac
tice, we believe that our relatively simple approach will
prove practical in those clinical situations in which reason
ably exact anatomic delineation of the origin of the ‘8F(or
other PET tracer) signal is necessary. Indeed, as tech
niques to produce image fusions become simpler, such
anatometabolic images may well become the standard for PET image display in patients with cancer and other ill
nesses.
ACKNOWLEDGMENTS
The authorsthankthe technologistsandchemistsof the Urn
versity of Michigan PET Center for their assistance and Ms. Manette London, Annette Betley, Barbara Burton, and Joan Fog arty for their contributions. Supported by NIH grants 5 ROl
CA53172,1 R55CA52880,CA and MOl RR00042,as well as the
“Hi-techfundinginitiative―fromtheUniversityofMichiganHos
pital.
REFERENCES
1. SomP, Atkins}IL, BandoypadhyayD, et aLA fluorinatedglucoseanalog, 2-fluoro-2-deoxy-Dglucose(F.18):nontoxictracerforrapidtumordelco tion. JNucl Med 1980;21:670—675. 2. DiQiiro0, LeLaPazRL,BrooksRB,etal.Glucoseutilizationofcerebral gliomasmeasuredby ‘Ffluorodecayglucoseand positronemissiontomog raphy. Neurology 198232:1323—1329. 3. YonekuraY, BenuaPS,BTIIIAB,etal.Increasedaccumulationof2-deoxy 2-[1@F]fluoro-D-g1ucoscIn liver metastases from colon carcinoma.I Nuci Med 1982;23:1133—1137. 4. PaulR, AhonenA, RoedaD, NordmanE. Imagingof hepatomawith 18F-fluorodcoxyglucose(Letterl. Lancet 1985;1:50-51. 5. StraussLO, aonue JH, SchiagP, Ctal. Recurrenceof colorectaltumors: EFF evaluation.Radiology1989;170:329-332.
- MinnH, SorniI. Fluorine-18fluorodeoxyglucosesCintigraphyin diagnosis and follow up of treatment in advanced breast cancer. Eu, I Nuci Med 1989;15:61—66. 7. Wahl RI., Hutchins GD, Buchsbaum DJ, UCbeTt M, Grossman HB, Fisher S. fluedne.18-2.decxy-2-fluoro-D-glucose(FDG)uptakeintohumantumor xenografts:feasibilitystudies for cancer imagb@gwith PET. Cancer 1991; 67:1544—1550. 8. WahiRI, CodyRL, HutchinsGD, MUdgetiEE. Prhnaiyandmetastatic breast carcinoma: initial dilnical evaluation with PET with the radiolabeled glucose analogue 2-(F-18J.fluoro-2.decxy.D-glucoscRadiology 1991;179: 765—770. 9. GriffethLK, DehdashtiF, McGuireAN, McGuireDJ,PerryDJ,Moerlein SM. PETevaluationof soft-tissuemasseswithfiuonne-18flnoro-2.deoxy. D-glucose.Radiology1992;182:185-194.
MatometabollcTumor Imaging•WahI atat.