describes a method of attenuation correction for use innonhomogeneous media that may be applied to cardiacSPECT. The attenuation correction method, termed multiplicative variable attenuation compensation (MVAC),uses tissue contours determined from segmentation of atransmission scan to assign a priori determined attenuationcoefficients to different tissue regions of the transaxialimages.

Information about the anatomy of the patient requiredfor attenuation correction of SPECT images in nonhomogeneous media is not readily available from a standardemission scan. Several methods that use a transmissionscan to reconstruct a map of attenuation coefficientswithin the subject have been developed and studied (3—7).However, these methods are bound to the inaccuracies ofthe determination of attenuation coefficients due to thesignificant Poisson noise, scattering ofphotons and scalingduring the reconstruction process. The use of simultaneousemission and transmission acquisitions with the gammacamera requires the use of two radionucides with sufficiently different energy to prevent the corruption of oneimage by photons from the other radionuclide. In thissituation, to successfully correct for attenuation, the relationship between the absorption properties for the twodifferent energies must be known. Furthermore, transmission scans of sufficient statistical accuracy to calculate theattenuation values directly imply a significant radiationdose to the patient and a longer study time.

These difficulties in determining an accurate map of theattenuation coefficients may be reduced by using the transmission transaxial slices to determine the boundaries between areas of different density. A method first suggestedby Budinger (8) would assign a priori determined attenuation coefficients to the areas identified in the transaxialimages. With this approach, the required statistical accuracy is reduced and a relatively short transmission scanmay be used, reducing the time the patient must remainon the imaging table as well as the patient's radiation dose.This approach has recently been investigated for PET byXu (9), who was able to reduce the transmission scan timefrom 30 to 10 mm while maintaining the quantitativecapability with ‘3N-ammonia.

The quantitative and visual interpretation of SPECT myocardialperfusionimagesis limitedby physicalfactorssuchasphoton attenuation, Compton scatter, and finite resolutioneffects. A method of attenuation correction is described foruse in nonhomogeneousmedia and applied to cardiac SPECTimaging.Thismethod,termedmultiplicativevariableattenuationcompensation(MVAC),usestissuecontoursdeterminedfromsegmentationof a transmissionscan to assign a prioridetermined attenuation coefficients to differenttissue regionsof the transaxial images. An attenuation correction map isthen constructed using a technique inspired by Chang'smethod that includes regionally dependent attenuation withinthe chest cavity and is applied after reconstruction by filteredbackprojection.Scattercorrectionusingthe subtractionof asimultaneously acquired scatter window image enables theuse of narrow beam attenuation coefficients. Experimentalmeasurementsto evaluatethese methodswere conductedfor 201TIand semTc SPECT using a homomorphic cardiacphantom. Finite resolution effects were included in the evaluation of results by computer simulation of the three-dimensional activity distribution. The correction methodology wasshown to substantiallyimproveboth relativeand absolutequantificationof uniformandnonuniformregionsof activityinthe phantom'smyocardialwall.

J NucIMed 1992;33:2232—2237

he interpretation and quantification of single-photonemission computed tomographic (SPECT) images arehampered by the physical factors of photon attenuation,Compton scatter, and finite resolution effects. Compensation for these effects is particularly difficult in cardiacperfusion imaging since the heart is surrounded by otherorgans of varying density. The methods of attenuationcorrection commonly distributed with commercial SPECTsystems (1,2) assume uniform attenuation within the bodyand prove ineffective for cardiac imaging. This report

ReceivedJan. 2, 1992;revisionaccepted Jun. 11, 1992.For reprints contact: James R. Gait, PhD, Department of Radiology, Emory

LhiiversftySchoolof Medians, 1364 CliftonRd. N.E.,Atlanta,GA30322.

2232 The Journal of Nudear Medicine•Vol.33 •No. 12 •December1992

SPECT Quantification: A Simplified Method ofAttenuation and Scatter Correction forCardiac ImagingJames R. Galt, S. James Cullom and Ernest V. Garcia

Department ofRadiology, Emory University School ofMedicine, Atlanta, Georgia and Veterans AffairsMedical Center, Atlanta, Georgia

by on June 7, 2018. For personal use only. jnm.snmjournals.org Downloaded from

Quantitative imaging also requires correction for theeffects of Compton scatter. In this study, a scatter correction methodology is used which subtracts a fraction of asimultaneously acquired scatter image from the photopeakimage (containing scatter) is suggested by Jaszczak (10).

In the experiments described here, scatter correction isapplied first. The MVAC method is then applied to thescatter corrected transaxial images and evaluated for quantitative cardiac imaging in phantoms with 201T1and 99mTc.Evaluation consists of quantitative comparison of maximal count circumferential profiles (MCCPs) extractedfrom the phantom images with the count distributionexpected in the absence of attenuation and scatter.

METHODS

Scatter Correction MethodologyImplementation ofthe scatter correction methodology requires

that planar imagesbe acquired simultaneouslyover the photopeak and a scatter region of the spectrum. The primary (orphotopeak) window produces the image to be used clinically andprimarily contains photons of energies near the photopeak. Thesecondary (or scatter) window is placed over a lower energy regionof the spectrum and therefore contains a significantly higherpercentage of scatter photons than the primary window. Once

the data have been acquired and the transaxial images reconstructed, the scatter correction is applied as follows:

f@(x,y) = f,@,(x,y) —kf5(x,y),

where f,(x, y) is the scatter window image, f@(x,y) is the primarywindow image, k is the scatter window fraction (SWF) whichmust be determined a priori, and f@(x,y) is the corrected photo

peak image.The use of a constant scatter window fraction for all pixels

assumes a linear relationship between the counts in the scatter

window and the scattered counts in the photopeak window. Thevalidity ofthis assumption has not been addressed directly in thiswork. However, Koral has recently investigated the dependenceof quantitative accuracy with 99mTcSPECT on the value of thescatter window fraction (11). He found that the use of a constantscatter window fraction was a good first approximation for quan

titative accuracy for different source sizes and position within ahomogeneous attenuating media.

The photopeak and scatter windows used in the experimentsdescribed here for 201Tl were 72 keY, 20% and 60 keV, 15%respectivelyand for99mTcwere 140keY,20%and 106keY,32%respectively.The photopeakwindowsfor each isotopeare thosecommonly used in the clinic. The @mTcscatter window is broad(to maximize counts) and covers the Compton plateau. The 20Tlscatter window covers the energy range below the photopeakwindow allowed by the scintillation camera. Earlier experimentswith @mTcline sources (12) gave results consistent with thosereportedby Jaszczak(JO)and indicateda @mTcscatterwindowfraction of 0.5. The geometry of the cardiac phantom used here

is much more complex and thereforedifferent scatter windowfractions were evaluated for both isotopes.

Attenuation Correction MethodologyA transmission scan was acquired for the phantom measure

ments with 30 mCi of@mTc in a collimated fillable flood source

mounted opposite and parallel to the camera surface. Acquisition

parameters included: a 140 keY, 20% energy window, 64 x 64matrix (6 mm pixels), 64 views at 40 sec per view, 360°circularorbit, and low-energy, all-purpose (LEAP) collimation. The transmission scan was acquired prior to the emission scan withoutmoving the phantom or table position to avoid alignment problems. The transmission transaxial slices were reconstructed usingfiltered backprojection after decay and flood correction. Theseimages were then segmented into soft-tissue and lung regions bythresholding.

Attenuation coefficients were determined experimentally byapplying linear regression to the measured count rate as variousthicknesses ofwater(soft tissue)and Styrofoam (lung) were placed

betweena collimatedphoton source and a scintillationcamera(12). Coefficients assigned for 2OVflwere 0.176 cm@ and 0.0 18cm ‘for water and Styrofoam, respectively. Coefficients assignedfor 99mTcwere 0. 154 cm ‘and 0.01 5 cm@ . It was not possible todifferentiate between water and other dense material in the transmission images. The imaging table was visible in the transmission

images and was assigned the same value as the lung. All pixelsoutside of the segmented body boundary and table and in thecentral lung region were assigned a value of zero for air.

The correction map derived from the attenuation coefficientmap is similar in formulation to the first order (no iterations)correction described by Chang (2). The attenuation at any givenpixel location within the transaxial image is approximated by the

average attenuation experienced by photons originating in that

voxel for all measured projections. A significant departure fromChang's method as it is commonly applied is the incorporationof varying linear attenuation coefficients. The correction factorsat each point are computed using the assigned attenuation coefficient values from the segmented transmission image with the

equation:

11 M N@C(x, y) = L@ @([Je@@'@)j

where the correction factor C for each point (x, y) in the transaxialimage is the average of the attenuation factors for projections atM angles (O@)over 360°around the point. Each attenuation factoris the product of the attenuation due to each of the N voxelsdefined by the radial distance from the point (ri) along theprojection angle (@Ljis the attenuation coefficient for thejth voxel,li is the lengthof the projectionthrough this voxel).

A simplified version of Sorenson's method of attenuationcorrection ( 1) which was commercially available was also usedfor comparison of our results. This version assumes uniformattenuation (effective attenuation coefficient value of 0. 12 cmand that the activity is distributed uniformly within the medium.

Convolution ModelTo evaluate the effects of finite resolution on quantification, a

three-dimensional convolution was performed on a computer

simulation of the phantom configurations. The point spreadfunction (PSF) used in the model was a spherically symmetric

Gaussian function with a half-width-at-half-maximum (HWHM)of 9 mm and 8 mm for 201Tl and @mTcrespectively and wasassumed to be invariant over the region of the distribution. The

resulting image is that expected under ideal conditions in theabsence of attenuation, scatter and noise (13). MCCPs from thesimulated images were used as the baseline for evaluation of thecorrection methodology after calibration.

The assumptions made about the PSF in the simulation are

Cardiac Attenuationand Scatter Correction •GaItet al 2233

by on June 7, 2018. For personal use only. jnm.snmjournals.org Downloaded from

only valid ifthe experimental PSF varies only slightly in the areaoccupied by the heart. This approximation is supported by theuse of 360°camera orbits and the fact that noise reducing filterswere applied equally in all three dimensions. Application of themethods presented here to 180°orbits will require compensationfor spatially varying resolution and will be addressed in the laterreport. Even with these assumptions attenuation and scatter willcause the shape of the PSF to vary, particularly near attenuationboundaries. These variations are largely corrected by the methodsproposed here.

PhantomExperimentsThe phantom used to evaluate the correction methodology is

a Data Spectrum Cardiac Insert (model 7070) mounted in a DataSpectrum Elliptical Phantom (model 2230) (Fig. 1).Lung inserts,made of Styrofoam, and a spine insert, made of packed bonemeal, were developed in-house. Fillable defect chambers, withdifferent activity concentrations, were placed around the midventricular circumference ofthe phantom in a ring configuration.This ring of defects was 2 cm in length along the long axis andplaced in the center of the myocardial wall so that there wereareas of normal myocardium both above and below the ring ofdefects. Three fillable defects were used, spanning 180°,90°,and450, with tracer concentrations of 50%, 75%, and 25% of the

normal wall concentration respectively. The remaining 45°wasopen to allow for even mixing of activity through the normalwall. This distribution ofactivity is shown in Figure 2 along withthe image predicted by the convolution model for @mTc(HWHMof 8 mm). The activity concentrations in the normal wall were15 MCi/cc for 20Tl and 16.5 @Ci/cc for @mTcwith a normal

myocardial wall volume of approximately 90 ml. Backgroundactivity in the ventricular chamber and phantom body was 5%of the normal myocardial wall concentration.

Sixty-four views were acquired using a 64 x 64 matrix over a360°orbit, a LEAP collimator and a circular orbit which contoured the phantom as close as possible. The time per view was40 sec for 201Tland 60 sec for 99mTcThe photopeak and scatterenergy windows were as described above. Before transaxial recon

struction, both the photopeak and scatter window projections

were filtered with a two-dimensional symmetric Hann filter witha cutoff frequency ofO.833 cycles/cm.

FIGURE 1. Phantomthorax. The Data Spectrumellipticalphantomis shown with the Data Spectrumcardiac insert. Lunginsertsand a spineinsertwere developedin-house.

FIGURE 2. Distributionof activityinthecardiacphantom.Theleft panelshows the distribution of activity (achievedwith threefillable chambers) used in the myocardial wall of the cardiacphantom.The 2-cm long chambersspanned180°,45°and 90°withtracer concentrationsof 50%, 25% and 75% of the “normal―wallconcentration,respectively.The right panelshows the shortaxis imageexpected in the absenceof attenuation and scatter(predictedby the convolutionmodel).

RESULTS

The correction methodology is illustrated in Figure 3for the 201T1phantom experiment. Panel A shows thetransmission transaxial slice of the phantom thorax with acardiac insert. The cardiac insert contained activity at thetime of acquisition and can be seen in this image. Panel B

FIGURE 3. Thallium-201phantom study, attenuationandscattercorrections.(A)Transmissiontransaxialsliceof the phantornthorax. Thecardiacinsertcontainedactivitywhenthis imagewasmadeandis visible.(B)Mapof assignedattenuationcoeffidents. (C) Attenuation correction map. (D) Correspondingphotopeak emissiontransaxialslice through the phantom's myocardial wall chamber. (E) Scatter emission transaxial slice (scaled toshow relationshipto D). (F) Scatter and attenuation correctedemission transaxial tomogram.

0 0

E. F.

2234 The Journal of Nuclear Medicine•Vol.33 •No. 12 •December1992

CardiacPhantomActivityDistribution

TracerConcentration

Image Pred@tedbyConvolutionModel

CorrectionsforAttenuationandScatter

T@2O1PhantomStudy

by on June 7, 2018. For personal use only. jnm.snmjournals.org Downloaded from

A 1.00

0@ 0.75

C

00

0.50

0@ 0.25

— Simulation

— ActivityConcentrationRaw Counts..-.......SWF-O.O

shows how regions have been defined and a priori determined attenuation coefficients assigned. Panel C showsthe attenuation correction map constructed from Panel B.Panels D and E show transaxial slices through the cardiacinsert acquired with the photopeak and scatter windows,respectively. These images have been scaled to show therelative magnitude of counts in the two windows. Theimage in Panel F shows the same transaxial slice withscatter and attenuation correction applied. Importantitems to note in the corrected image are the increaseduniformity of the apical and basal regions ofthe phantom(which had the same 201Tlconcentration), the increasedcontrast of the defect in the septal wall, and the increasedcontrast of the ventricular chamber.

Relative Evaluation of the CorrectionsThree short-axis slices of the phantom were chosen to

evaluate the correction methodology on a relative scale:(1) a uniform basal slice, (2) a nonuniform slice throughthe center of the ring of defects, and (3) a uniform apicalslice. These slices were extracted from the experimentalimages and simulated using the convolution simulation.Maximal count profiles through these slices were thennormalized to the mean of the two uniform profiles togive a basis for comparison.

Figure 4 shows relative MCCPs for 201―flobtained forthe ring of defects and normalized to the mean of thenormal apical and basal profiles (uniform slices above andbelow the ring of defects). The simulation curve gives theresponse expected from the activity distribution with thesystem resolution (no attenuation or scatter) and formsthe gold standard for our measurements (Fig. 4A). Thedefect with the lowest concentration can not be discernedin the raw count (uncorrected profile). Attenuation correction alone (SWF = 0.0) overestimates the true count valuesbecause the added counts due to scatter are neglected.Application ofattenuation correction after scatter subtraction (Fig. 4B) clearly improves the correlation of correctedand predicted values over the range of defects. The normalized mean square error (NMSE) between the experimental and simulated profiles indicated best results with ascatter window fraction of 1.5.

The corresponding data for the @mTcexperiment areshown in Figure 5. The best correlation in MCCPs ofmeasured and predicted counts for the ring of defects wasobtained with a scatter window fraction of 0.5. Theseimprovements are consistent with the results obtained for201T1 As the figure shows, the commercially availablemethod overestimated the activity in the ring of defectsand did not substantially improve contrast in the area ofleast activity.Absolute Evaluation of the Corrections

Absolute evaluation of the corrections requires an external calibration. This calibration was accomplished bytomographically imaging a 90°(5.5 ml) fillable chamberboth in air and mounted in the lateral wall near the baseof the filled cardiac phantom. An attenuation factor was

90 180

Angle (Degrees)

Sinwlation

B

•1

C

00

0

0

0

90 180 270

Angle (Degrees)360

FIGURE4. Thallium-201cardiacphantom,relativecountMCCPs. Maximal count circumferential profiles through the ringof defectsnormalizedto themeanof twouniformslices(oneaboveandone belowthe defects).Errorbars representstatisticalnoise(estimatedwithSPECTsimulationas two standarddeviations). Error bars at 180°show mis deviationfrom uniformityofa uniform slice processed in the same manner during a separateexperimentandthusrepresenttheeffectsof allsourcesoferror.(A)Thisplot showsthe raw (uncorrected)counts and attenuationcorrection alone. (B) Similar to (A) but with scatter correctionusingdifferentscatterwindowfractions(SWF).Thebestmatchwas found with a scatter window fractionof 1.5.

derived as the ratio of counts extracted from the chambermounted in the phantom (corrected for scatter) and countsextracted from the chamber in air (corrected for self attenuation). From this attenuation factor and the counts inthe lateral wall of a basal slice of the phantom, a factorwas found to scale the profiles predicted by the convolutionmodel for comparison to absolute counts.

Evaluation of the corrected image profiles in terms ofabsolute counts compared the NMSEs between the countsfrom the experimental profiles and the scaled results ofthe simulation model. Best results were found with scatterwindow fractions of 2.0 and 1.0 for 201Tland @mTcrespectively.

Quantitative Evaluation of the CorrectionsThe results of the relative and absolute evaluation of

the corrections are summarized in Table 1 for two uniform

Cardiac Attenuationand Scatter Correction •GaItet al 2235

by on June 7, 2018. For personal use only. jnm.snmjournals.org Downloaded from

IsotopekUniformapical1k@iformbasalRing

ofdefectsRelative

%NMSEAbsolute%NMSERelative%NMSEAbsolute%NMSERelative%NMSEAbsolute%NMSE@°1Tl

9SmTCRaw

0.01.01.52.0

Raw0.00.51.0

Corn7.31

1.031.101.161.264.000.630.691.791.4436.9

29.412.5

7.12.2

29.225.916.5

8.821.44.41

0.270.300.330.392.290.100.120.140.5752.6

11.03.01.00.1

44.615.6

9.14.04.90.61

2.030.600.250.710.300.580.281.150.9345.4

34.59.02.50.5

36.322.9

9.31.5

16.6

short-axis slices (apical and basal) and through the ring ofdefects. Attenuation correction alone improved the uniformity of uniform slices, giving the greatest change in%NMSE. The addition of scatter correction degraded thisuniformity to a small extent in the relative evaluation butwhen absolute counts were considered produced significantly improved results. The scatter window fraction thatgave the best results for each radionuclide in the absoluteevaluation was higher than that found for the relativeevaluation and may reflect inaccuracies in the scaling ofthe results of the simulation to estimate absolute counts.

DISCUSSION

The limiting effects of Compton scatter, photon attenuation and finite resolution of the imaging system onmyocardial perfusion imaging and quantification are wellknown. The nonhomogeneous distribution of attenuationwithin the thoracic cavity complicates the interpretationof these images and precludes the application of moresimple methods of scatter and attenuation correction developed for homogeneous media. The latter methods generally require only knowledge of the external boundary ofthe object and the constant attenuation coefficient in theregion. The method proposed here, which uses a transmission scan to provide information about the spatial variation in tissue attenuation coefficient and an interactivesegmentation method for tissue boundary determination,avoids many ofthe problems ofabsolute determination ofthe attenuation coefficient by employing narrow beamvalues for correction of scatter subtracted images.

Scatter subtraction, which removes the broad scatterresponse superimposed on the photopeak PSF, can improve quantification by reducing the influence of scatteron surrounding structures. However, the reduced numberof counts in the corrected image can result in increasedimage noise and a loss of uniformity over the normalmyocardium. This method also results in a significantincrease in contrast and therefore the possibility of im

180

Angle (Degrees)

SimulatonActivityConcentration

:::::@Me@7J@

90 180 270 360

Angle (Degrees)

B 1.00

0.75

FIGURE 5. Technetium-99mcardiacphantom,relativecountMCCPs.Maximalcount circumferentialprofilesthrough the ringof defects normalizedto the mean of two uniform slices. Errorbars representstatisticalnoise(estimatedwith SPECTsimulationas two standarddeviations).Error bars at 180°show rms deviation from uniformityof a uniformslice processed in the samemanner. (A) This plot shows the raw (uncorrected) counts andattenuation correctionalone. (B)Similarto (A)but with scattercorrectionusingdifferentscatterwindowfractions(SWF).Thebest matchwas found with a scatterwindow fractionof 0.5. Thecommercialmethod is that suppliedby the manufacturerand isa simplificationof Sorenson'smethod.

TABLE IAttenuation and Scatter Correction Evaluation

2236 The Journal of Nuclear Medicine•Vol.33 •No. 12 •December1992

by on June 7, 2018. For personal use only. jnm.snmjournals.org Downloaded from

tial observations of@mTc@sestamibiscatter images indicatethat it may be possible to estimate the tissue contourswithout the use of a transmission scan.

CONCLUSIONS

The MVAC method applied to scatter subtracted transaxial images of the myocardium offers promise to providea significant improvement in relative and absolute quantification for both 201Tland @mTcmyocardial perfusionimaging. This method can be implemented clinically withrelatively simple modifications to conventional SPECTsystems.

ACKNOWLEDGMENTS

This work was supported in part by the National Institutes ofHealth(NHLBI)undergrants ROl Hl4l628 and ROl H142052,and by the Veterans Administration Research Advisory Group(RAG) under grant 3022-0001. The authors thank David Rex

Hamilton for his assistance with the phantom experiments anddevelopment ofthe software. We are grateful to the reviewers fortheir helpful comments and suggestions.

REFERENCES

1. Sorenson JA. Quantitative measurement of radioactivity in vivo by wholebody counting. In: Instrumentation in nuclear medicine, volume 2. NewYork:AcademicPress;1974:311—348.

2. Chang LT. A method for attenuation correction in radionuclide computedtomography. IEEE Trans Nuci Sci 197825:638—643.

3. Bailey DL, Hutton BF, Walker PJ. Improved SPECT using simultaneousemission and transmission tomography. I NuclMed 1987;28:844—85l.

4. Manglos SH, Jaszczak Ri, Floyd CE, Hahn U, Greer KL, Coleman RE.Nonisotropic attenuation in SPEC!': phantom tests of quantitative effectsand compensationtechniques.JNuclMed 1987;28:1584—1591.

5. Ljunberg M, Strand SV. Attenuation and scatter correction in SPECT forsources in a nonhomogeneous object: a Monte Carlo study. J NucI Med1991;32:1278—1284.

6. Ljunberg M, Strand SV. Attenuation correction in SPECT based ontransmission studies and Monte Carlo simulations of build up functions. JNuclMed 1990;31:493—500.

7. Manglos 5H, Bassano DA, Duxbury CE, Capone RB. Attenuation mapsfor SPECT determined using cone beam transmission computed tomography. IEEE Trans NuclSci 1990;37:600—608.

8. Budinger TF. Physical attributes ofsingle-photon tomography. JNucl Med1980;21:579—592.

9. Xu EZ, Mullani NA, Gould KL, Anderson WA. A segmented attenuationcorrection for PET. JNuclMed 1991;32:l61—l65.

10. Jaszczak Ri, Greer KL, floyd CE, Harris CG, Coleman RE. ImprovedSPECT using compensation for scattered photons. J NucI Medl984;25:893—900.

11. Koral KF, Swailem FM, Buchbinder S, Clithorne NH, Rogers WL, TsuiBMW.SPECTdual-energy-windowComptoncorrection:scattermultiplierrequired for quantification. I NuclMed 1990;3l:90—98.

12. Galt JR. Reconstruction of the absolute radionucide distribution in ascattering medium from scintillation camera projections. PhD Dissertation,Emory University, Department of Physics, 1988; University Microfilm,Ann Arbor,MI.

13. Gait JR, Garcia EV, Robbins WL. Effects of myocardial wall thickness onSPECT quantification. IEEE Trans Med Imaging 1990;9: 144—150.

14. King MA, Hademenos GJ, Glick SJ. A dual-photopeak window methodfor scattercorrection.J NuclMed l992;33:605—6l2.

I

FIGURE 6. Technetium-99m-sestamibipatient study, attenuationand scattercorrections.(A)Transmissiontransaxialof thepatient's chest. (B)Mapof assigned attenuationcoefficients.(C)Attenuationcorrectionmap.(D)Correspondingphotopeakemissiontransaxialslicethroughthephantom'smyocardialwallchamher.(E)Scatteremissiontransaxialslice(scaledto showrelationship to D). (F) Scatter and attenuation corrected emission transaxialtomogram.

proved diagnostic value. By acquiring a second scatterwindow, the dependence of the scatter component of thephotopeak image on surrounding activity is measured ona patient specific basis, reducing the number of assumptions required in the correction methodology. The dependence of scatter on the inhomogeneity of the media is alsomaintained. It should be noted that the scatter correctionsapplied here are independent ofthe MVAC method. Otherpromising methods of scatter correction such as that recently suggested by King (14) are not precluded.

We have evaluated the protocol described here clinically(Fig. 6) and found that it can be implemented with relatively minor modifications to a conventional single headedgamma camera gantry. The transmission image in Figure6A was acquired in 11 mm using a 30 mCi @mTccollimated flood source. The most evident difficulty for usingthis methodology clinically is the required precautions forthe registration of transmission and emission images.

This methodology is not restricted to the radionuclides(and photon energies) described here. In particular, the useof assigned rather than calculated attenuation coefficientscircumvents the need for the transmission map to beadjusted for photon energy. The assignment of attenuationcoefficients should reduce the statistical requirements ofthe study allowing for a quicker scan that saves the technologist time and reduces patient radiation exposure. Ini

2237Cardiac Attenuationand Scatter Correction •GaItet al

. . CorrectionsforAttenuationandScatter

Tc-99mSestamibi Patient Stu@

@:@ E. F.

by on June 7, 2018. For personal use only. jnm.snmjournals.org Downloaded from

1992;33:2232-2237.J Nucl Med.   James R. Galt, S. James Cullom and Ernest V. Garcia  Cardiac ImagingSPECT Quantification: A Simplified Method of Attenuation and Scatter Correction for

http://jnm.snmjournals.org/content/33/12/2232This article and updated information are available at:

  http://jnm.snmjournals.org/site/subscriptions/online.xhtml

Information about subscriptions to JNM can be found at:  

http://jnm.snmjournals.org/site/misc/permission.xhtmlInformation about reproducing figures, tables, or other portions of this article can be found online at:

(Print ISSN: 0161-5505, Online ISSN: 2159-662X)1850 Samuel Morse Drive, Reston, VA 20190.SNMMI | Society of Nuclear Medicine and Molecular Imaging

is published monthly.The Journal of Nuclear Medicine

© Copyright 1992 SNMMI; all rights reserved.

by on June 7, 2018. For personal use only. jnm.snmjournals.org Downloaded from