t the 1970 conference on “QuantitativeOrganVisualization in Nuclear Medicine,―PJ. Kenny definedquantitative organ visualization (QOV) as “. . . knowledge ofthe absolute amount of activity in each volumeelement in the organ of interest in successive timeintervals . . . “(1). He noted that at that time the capability to determine radionuclide distributions in threedimensions did not exist. Since then, with the development of techniques for single photon emission computed tomography (SPECT), the technology for accomplishing QOV, as Kenny defined it, has become available.

SPECT appears ideally suited for QOV. The tomographic approach permits the separation of structureswhich are superimposed in conventional images. Thisseparation is further enhanced by the intrinsic highcontrast of SPECT images. The reconstruction of contiguous, thin sections permits evaluation of the threedimensional distribution of radioactivity and, whenappropriate corrections are made for attenuation losses,the SPECT images can allow quantification of organsize, shape, and radioactivity.

Several authors have utilized these characteristics to

ReceivedNov. 5, 1985;revision accepted July 17, 1986.For reprints contact: Louis T. Kircos, DDS, MS. PhD, Div. of

Oral Radiology, University of California, San Francisco, SanFrancisco,CA 94143.

quantitate both phantom and in vivo radioactivity distributions using SPECT (2—6)and positron emissioncomputed tomography (7-12). This paper is a reportof our efforts to develop a technique for QOV, basedupon a SPECT system, the Humongotron (13). Wepresent techniques for quantifying organ volumes andthe absolute amount of radioactivity within organs. Inaddition, an analysis of the accuracy of these methodsis presented.

MATERIALS AND METhODS

All studies were performed using the Humongotron, asingle photon emission computed tomography (ECT) systemwhich has been previously described (13). This system, whichincorporates a Searle°Pho/Gamma III HP scintillation camera, was used with either a Searle High Resolution collimatoror a tantalum collimator specially designed for tomography attechnetium-99m (@‘°Tc)energies (14). The images were digitized and stored on disk using a commercially available nuclear medicine computer system,t which was also used forimage reconstruction. A convolution-type reconstruction algorithm wasemployedwith a filter function similar to Budinger's “newfilter―(15,16). Attenuation correction was performed using the “meanexponential technique―of Kay (17)assumingelipsoidalbody marginsand a uniform attenuationcoefficientof u = 0.15/cm.

Serial, contiguous SPECT images of selected organ(s) of

334 Kircos,Carey,andKeyes The Journal of Nuclear Medicine

Quantitative Organ VisualizationUsing SPECTLouis T. Kircos, James E. Carey, Jr., and John W. Keyes, Jr.

Department ofRadiology, University ofCaljfornia, San Francisco, California; Division ofNuclear Medicine, University ofMichigan Medical Center, Ann Arbor, Michigan; andDepartment ofRadiology, Georgetown University, Washington, DC

Quantitative organ visualization (QOV) was performed using single photon emissioncomputed tomography (SPECT). Organ size was calculated from serial, contiguous ECTimagestakenthroughthe organof interestwith imageboundariesdeterminedusingamaximumdirectionalgrad@ntedgefindingtechnIque.Organactivitywas calculatedusingECTcountsboundedby the directionalgradient,imagingsystemefficiency,and imagingtime.Thetechniqueusedto performQOVwas evaluatedusingphantomstudies,in vivocanineliver,spleen,bladder,andkidneystudies,andin vivohumanbladderstudies.It wasdemonstrated that absolute organ activity and organ size could be determined with thissystemandtotal imagingtimerestrictedto <45 mmto an accuracyof about±10%providing(a)theminimumdimensionsoftheorganare> theFWHMoftheimagingsystemand(b)thetotal radioactivitywithinthe organof interestexceeds15 nCi/ccfor dog-sizedtorsos. Inaddition,effectivehalf-livesof@ .5 hr or greatercouldbedetermined.

J Nuci Med 2&334-341, 1987

interest were analyzed by the technique described by Kircos(16,18). This technique assumes an isotropic tomographicimaging system response function and is based upon computing a five point, least squares, directionally sensitive firstderivative (LSDSFD) of the image which results in a deconvolution of the imaging system response function from theobject. The processed image shows each organ edge outlinedin a topographic manner representative ofthe gradients of theoriginal image. A profile normal to the organ edge is thesystem response function (Fig. 1) for organ edges which canbe considered step functions in the tomographic plane. Thegeometric edge of the organ is defined by the maximum ofthis deconvolved imaging system response function. It hasbeen shown (16,18) that this technique is mathematicallyexact for sources of radioactivity whose exposed edges can beconsidered either a step or a ramp function in the tomographicplane (Fig. 1). Once the edge had been established, the size ofthe organ of interest was determined by measuring the area(in square pixels) encompassed by the edge of the organ ofinterest (Ak), summing over all the tomographic sectionscontaining the organ (k), multiplying by pixel size (P@2),seetion thickness (DeltaZ), and tissue density (rho) (19):

M(gm) = (rhoXDeltaZXP@2)L-1 (Ak),

where k is the number oftomographic sections through whichthe organ of interest extends in the third dimension. Theextent of the integration in the third dimension is establishedby the observation ofthe point near the periphery ofthe organ

(a) 1.0

I

woFIGURE 1Techniquefor definingorganedgesand extent of radioactivity:(A) Profilethrougha °perleCtorganedge―whichis a convolution of the PSF (B) with an organ edge represented by a step function (C).The organ edge(C) is definedby the maximumof (B)whichis the first derivativeof (A).The numberof counts attributableto the organ extendbeyond the geometric edge (C) and is outlined by theextentof the gradient(B).

where the counts per pixel fall to about 50% of the valueobserved at a slightly more central location (Fig. IA).

The absolute quantity of radioactivity in the organ ofinterest (Q) also was determined by analysis of the deconvolved imaging system response function (Fig. 1). Blurringintroduced by the imaging system tends to spread organ countsbeyond its geometric boundaries in the SPECT image causingdifficulty in determining the radioactivity in the organ ofinterest. To correct for this, the number of counts (Nk) weredetermined within an area (ak) which extends beyond thegeometric border ofthe organ (Figs. 1 and 2) and encompasses

the full extent of the deconvolved imaging system responsefunction (Figs. 1Aand 2D). This was accomplished by flaggingthis extended area manually with a light pen in the firstderivative image (Fig. 2B) and then determining the numberof counts within the same area in the original SPECT image.The integration in the third dimension is extended until thecounts per pixel falls to ‘@-l0%of the value observed in aslightly more central location. Organs distributed in a uniformmedia of radioactivity also required background subtraction:

Nk' Nk (nXak), (2)

(1) where n is equal to the background intensity in counts perpixel squared. This, together with the factors for imagingsystem efficiency (E) and imaging time (Delta t), are sufficientto determine the absolute quantity of radioactivity in theorgan of interest.

FiGURE 2TechnIquefor definingorganedgesand extent of organradioactMty, dinical example. A: Tomographic section ofdogliver.B:Directionallysensitivefirstderivative(gradient)of tomogramin A. Theextent of the gradientdeterminesthe area over which counts are integratedfor activitydeterminations. C: Maximum of the first derivative (gradIent) image, B. This outhne defines the geometric borderof the organand was used to determineorgan size. D:Organbordersuperimposedon the first derivativeimage.Notehow the gradientextendsbeyondthe organborderanddefinesthe extentof countsattributableto the radioactivityintheliver.

wI 23

(c) 1.0

wI 23

Volume28@ Number3 •March1987 335

a@-11]S$I

Q(uCi) = (E)/(Delta t) Lc-.i Nk'. (3)

From this equation, the absolute mean concentration of radioactivity in the organ can be found by dividing by thecalculated mass of the organ (Eq. 1) and multiplying by thedensity. The system efficiency, E, is defined as the reciprocalofcounts per mCi second and empirically derived from regression analysis of Figure 9.

Phantomand ClinicalStudiesA variety of phantom and animal studies were performed

to evaluatethesetechniques.The choiceof studieswasbasedon the objectiveof providingrelativelyinclusivecoverageofthe imaging situations that might be encountered in performing QOV. Technetium-99m in various chemical forms wasused in most studies with the exception ofthallium-201 (@°‘Tl)which was used for the kidney studies@Imaging times variedfrom 10to 45 mm.

The phantoms includeda cylindricalabsorption phantom17.2 cm in diameter in which focal volumes of radioactivitycouldbe placedin a radioactiveor nonradioactivemilieu(Fig.3)—aresolution bar phantom consisting of various slotsmilled in a block oflucite which could be filled with radioactive fluid(Fig.4); a 15.0cm diameter cylindricalpolystyrenequality control phantom in which line sources of radioactivitycould be accuratelyplaced (14); and an anatomic phantom,ELRAD (Figs. 5 and 6), which consists of a hollow plastictorso in which a variety of hollow, simulated organs can beplaced containing varying amounts of radioactivity.

In vivo studies were performed on mongrel dogs ranging inweight from 14.0 kg to 25.0 kg. Liver and spleen studies wereperformed using [@“Tc]sulfurcolloid, bladder studies with[@“TcJpertechnetateor pyrophosphate,and a kidney studywith @°‘Tl.The dogs were anesthetized with sodium pentobarbital, scanned, killed, and the organs ofinterest were removed.For bladderstudies,the urine wasremovedfrom the bladder,measured volumetrically, then assayed in a Capintec CRC 2Nionization chamber dose calibrator.* For liver-spleen studies,tissue samples for the organ ofinterest were removed, weighed,and counted in a Packard 5320 well-type scintillation counter.*

FIGURE 3AbSOrptiOn phantom. This is a SPECT image of the absorplion phantom wfth the @rgecompailment and onetube filled with low and high concentrations of activityrespectively.Thisphantomis 17.2cm in diameterand25cm long. It hasa volumeof 5,800cm@.Differentimagingsituationsareproducedby the inclusionof up to three19mmdiametertubes locatedon a 4.5 cm radiusfrom thecenter of the phantom.

FIGURE 4Resolutionphantom.The imageon the left is a planarimage of the resolution phantom Obtained by placing thephantomincontactwith the collimatorface.Theimageonthe nght is a SPECTof the samephantom.Thisphantomis made of lucite and has slots milled in it in which variousquantitiesof radioactivitycanbe placedandmeasures15x 15 x 5.1 cm thick. The slots measure 6.5 x 1.6; 5.0 x1.3;6.0 X 1.0;and6.0 x 0.7 cm, respectively.

In addition,a calibrationcurvewasobtained(r@>0.99)foreach study by counting a number of serial dilutions of aknown amount of [@“Tc@pertechnetatein the scintillationcounter. From this calibration curve the absolute concentra

tion of radioactivityin @zCi/gwascomputed for the organ ofinterest. The weight ofthe organ of interest was recorded.

RESULTS

Representative phantom data are presented in Table1 and illustrated in Figure 7. Data to determine theaccuracy of this technique for half-life measurements

are presented in Figure 8. These time lapse studiesinclude data obtained from the absorption phantomfilled with a uniform distribution of radioactivity(Curve 1, Fig. 8) and also the same phantom containinga focal source of radioactivity (Curve 2, Fig. 8) withina more dilute uniform distribution ofactivity (Curve 3,Fig. 8).

The in vivo results(Table 2) substantiate the accuracyofthe technique as demonstrated in the pha.ntom studies (Table 1) with the exception ofthe spleen data. Thiserror results from a technical problem due to the rapidloss ofsplenic blood volume following killing and beforethe spleen could be removed and weighed.

In addition to the phantom and canine studies, an invivo human bladder study was performed using [@mTc]pyrophosphate. Although the Humongotron is not wellsuited to studies of the human torso because of itslimited field-of-view, the actual voided urine volume of250 cc with a radioactivity of 950 @Cicompared rcasonably well with the computed volume of 299 cc andactivity of 1,010 MCi.The difference may be a reflectionof the technical conduct of the study including limitedfield-of-view, nonideal attenuation correction, etc., orpatient management, i.e., residual urine or movementduring the study.

336 Kircos,Carey,andKeyes TheJournalof NuciearMedicine

.,@@@ .r@.

FIGURE 5ELRAD.The pictureon the left is an anteriorview demonstratingthe liver in solid arrows and the spleenin hollowarrows. The picture on the right is a left lateral view. Note on these views the close resemblance of human anatomyandthe actualinclusionof a humanskeleton.

The lower limit to the usefulness of this techniquefor organ sizing appears to be limited by the resolutionof the imaging system (Fig. 9). A Gaussian spreadfunction with a full width half maximum (FWHM) of16.0 mm (approximately equal to the response functionof the Humongotron in these studies) was convolvedwith rectangular step functions of various widths. Theresulting functions were sized by the derivative technique already described and the results plotted. Thecurve is linear down to approximately the FWHM valueof 16.0 cm, below which it rapidly flattens out and noobjective size information can be obtained. Also shown

in Figure 9 are data points from measurements obtainedon phantoms which correspond to the theoretical predictions.

DISCUSSION

The phantom and canine studies produced QOVresults which agree relatively closely with measuredorgan size and activity concentration, with the exception of the spleen data as already noted. These resultsindicate that, with the techniques described, it is possible to determine the size and absolute quantity of

FIGURE 6The image at left is an AP planarimage of ELRAD's liver and spleen.Theimageat right isa SPECTimagethroughthebottomthirdof ELRAD'sliver demonstrating a 1-in. tub 10-catedin the liverwhichwas not filledwith activity. The inferior tip ofELRADsspleencan be appreciatedin the SPECTimagein the posteriorright,adjacentto the liver.

337Volume 28 •Number 3 •March 1987

PhantomLocationActualactivity(pCi)Computed

activity(@C@Actual

volume(cc)COmputed

volume(cc)Size

error(%)Activity

error(%)1

. ELRADUver780.0750.01 ,280.O1,247.O332.ELRADSpleen63.065.0105.096.08123.ELRADBladder42.840.54.304.901454.ELRADBladder39.438.54.304.901425.

AbsorptionTube13.323.962.171.9311196.AbSOrptiOnTube22.022.312.171.9111147.AbsorptionTube31.02s2.17ss•8.AbsorptionVolume54.456.9176.0184.0159.

AbsorptionTube151.054.42.172.055710.AbsorptionTube228.331.52.172.0081111. AbSOrptionVolume120.0127.0176.0178.01612.

ReSOlUtiOnSmallbars56.060.03.4*•713.ResolutionMedium

bars62.067.03.84.617814.ResolutionBigbars96.0104.05.86.513815.OCLine

source98.0100.0<0.1ss2*

Cannot resolve.

TABLE IPhantomStudies

radioactivity in organs to an accuracy of ‘@-‘±10% forrelatively small uniformly distributed sources. The radioactivity determination is dependent on a number offactors including linearity of the imaging and recon

150xlO4

C

U)

C

0C)

C.,

(I)C0C.)w

Micro Curies

FIGURE 7Relationshipbetweenthetotalcountsinthe reconstructedimagesto the actualradioactivityin the variousphantomstudies, i.e., calibrationcurves. The different slopes ofthese two curves reflect the different sensitivities of thehigh resolution and the specially designed tantalum collimator. (•)TOMO collimator; (A) Searle HR collimator.

struction processes as well as an accurate technique forattenuation compensation (20—31).The techniques employed here worked well for dog-size torsos and uniformattenuators. Presumably, these errors would increasefor larger, human torsos.

Unexpectedly, however, the human phantom studiesusing ELRAD and the sole in vivo bladder study produced relatively accurate estimations ofthe activity andthe size of the radioactive source. In these cases, thetheoretical tendency of the attenuation compensationroutine to underestimate Nk for large objects may have

FIGURE 8Calculatedradioactivitiesin variousgeometriesof the absorption phantom (Fig. 3) versus time. These curves wereused to evaluate the usefullness of these techniques forhalf life measurements and give a measure of the limitdetectability of this SPECT system and these techniques.

14A

100

: 102

C.,

0

;

50 100 120

Time(see)xl02

338 Kircos,Carey,andKeyes TheJournalof NuclearMedicine

SubjectACtUalCOmputedErrorCOmputedError(kg)Organ sizesize(%) Actualactivity activity(%)

299cc 20 10

TABLE2In VivoStudies

cause the major errors in quantification that wouldseem apparent from casual consideration.

The time lapse studies introduce the possibility ofdetermining effective half-lives in vivo. The physicalhalf-life of 99mTc as determined from the slope of thecurves shown in Figure 8, is within 0.5% of its knownvalue with r2 >0.99. Assuming a 20-mm scan time anda 10% error in the determination of the absolute concentration of radioactivity, effective half-life determination of ‘-1.5hr are possible. For effective half-lives<1 .5 hr, the error in the estimation becomes greater.

Figure 9 shows that there is a systematic overestimation of organ dimensions which are less than theFWHM ofthe imaging system and that, below FWHM/2, no useful size information can be obtained. Figure 9also shows that there can be size determination errorsof about half a pixel even for perfect data. Thus, providing a smaller image matrix slightly improves theaccuracy of size determinations, especially for smallorgans. Organs whose minimum dimension in anydirection is smaller than the FWHM of the imagingsystem cannot be accurately sized without the knowledge of the absolute concentration of radioactivity inthe organ of interest. When the absolute radioactivityand concentration of radioactivity within the organ ofinterest are known, the volume of the organ is simplythe activity divided by the concentration of radioactivity. Many organ sizing techniques, especially thresholding, rely on a predictable concentration of radioactivityin the organ of interest for accurate results (3,5,32).Note, however, that accurate quantification of theamount of radioactivity in the organ still can be obtamed, independent of size.

The limit of detection of these techniques was established with studies of the absorption phantom (Table 2and Fig. 8). The limit of detection is dependent on thegeometry and intensity of the sources in the tomo

graphic plane (and near planes) as well as the geometryofattenuators in the tomographic plane. There is diffi

70

Human(80)kBladder250 cc3.80 ,@Ci/cc3.41@Ci/ocDog(20) kLiver628 g651 g49.33 @zCi/g8.60@Ci/g8Dog(25) kLiver874 g942 g84.33 ,@Ci/g3.90,@Ci/g10Dog(16)kLiver632g640g33.90@iCi/g2.93,@Ci/g5Dog(16)kSpleen61

g137gs13.2iCi/g6.86,@Ci/gsDog(25) kSpleen127 g317 g*4.15 MCi/g3.81@Ci/g•Dog(20)kBladder260 cc258 cc13.96 pCi/cc3.81zCi/cc4Dog(16)kBladder116cc110cc647.7@Ci/cc44.8@zCi/CC5Dog

(25) kBladder290 cc303 cc46.48 MCi/cc6.44,@Ci/cc1Dog(18) kKidney106 g107 g11 .82 @iCi/g1 .63 @zCi/g10

S Splenic volumes show large apparent error due to postmortem decrease in spleen volume that occurred before the organ could

beremovedandwsighed.

been partially compensated for by an increase in Nkdue to Compton scatter (23,24,28). Although the useof a uniform attenuation coefficient for the dog studiesand the study of ELRAD's liver appears reasonable, theuse of a uniform attenuation coefficient for ELRAD'sbladder and the human bladder study is somewhatquestionable because of the increase in attenuationfrom portions of the pelvis and the coccyx/sacrum.However, for this limited study the results are satisfactory and CT ofthe lower pelvis indicates the volume ofbone attenuating the photons may be reasonably smallin comparison to the volume ofsoft tissue and may not

FIGURE 9Relationshipbetweenactualsize and computedsize forcomputer generated data. The derivative techniques descnbed here were used to size various width steps andplotted for two levels of matrix digitization, 64 and 128.The numbers1 through5 appeanngbetweenthe curvesare actual phantom measurements using the imaging systernandtechniquesdescribedhere.

70

60

50

40

30

EE

wNJU)

wI—

0C)

Of

Of

Of

Of

Of

OfOf

Of

Of

DIGITIZATION

+ 28 x 1280 64 x 64

Of

TOf4

03?L+1

10

00 10 2030405060

ACTUALSIZE (mm)

Volume 28 •Number 3 •March 1987 339

culty in generalizing the results ofthese studies becauseof the infinite variability of the size, shape, intensity,and relative location of sources and attenuators encountered in clinical imaging situations, all of whichhave an effect on their ultimate detectability. However,studies ofthe absorption phantom indicate that, for thisgeometry, an absolute radioactivity of about 5.5 @tCi(1 5 nCi/cc for the large compartment, uniform distribution) and about 1 @iCi(250 nCi/cc) for the tubularinsert (focal area) in the absorption phantom produceSPECT images close to the lower limit of detectabilityby these techniques. These results were dependent onthe duration of the study and the geometry of theattenuator, and they reflect a 40-mm study time andattenuators ofabout the size ofa dog's torso. In general,the detectability of multiple sources is dependent ontheir location, size and shape, absolute and relativeconcentration of radioactivity, geometry and size ofattenuator, and the duration of the study. These areonly samples of a complex imaging problem and represent an area where further study should be directedfor resolution.

In general, the accuracy of the techniques employedare limited by the characteristics ofthe imaging system.The volume determinations are limited by the FWHM,image matrix size, noise, and the assumption ofa linearand isotropic imaging system. Radioactivity determinations are limited by attenuation correction, noise,and related factors including scatter, uniformity, energywindow, etc. (20—31),and are independent of the volume determinations.

Although the sensitivity of this technique is less thanthat obtained by counting tissue samples in a well-typescintillation counter (which often produces errors>±10%), it does introduce the possibility of humandosimetry using in vivo human data. Using this technique, human data on organ size, radionuclide concentration, and effective half-life could be obtained.

Currently, radiopharmaceuticals are assessed anddose calculations performed on the basis of animalstudies (33—36). The major difficulty in performingMedical Internal Radiation Dose Committee dosimetryis the determination of A, the accumulated activitywhich is based on the knowledge of the time course ofradioactivity in the organ of interest (37). This is nowdone by performing tissue distribution studies of theradiopharmaceutical of interest in animals and extrapolating the data to humans. The basis of extrapolationofanimal data to humans is equivocal (38—40)and thereliability of dosage calculations based on such datamay be in error of up to an order of magnitude (41).Potentially, QOV can determine the parameters associated with radiopharmaceuticals in vivo such as uptake(kg % dose/g, target to nontarget ratio (TINT), and thebiologic half-life in humans. There is a current need formeasurementsof theseparametersin normal patients,

in pediatric patients, and in patients with altered physiology and anatomy (42). Dosage calculations are inherently more accurate when based on radiopharmaceutical parameters determined in vivo in man becausethe behavior of a radiopharmaceutical in vivo is thereflection of human anatomy and physiology which noanimal can suitably mimic.

The results are based on a Searle Pho Gamma IIIHP, a relatively old camera. Although newer camerashave increased intrinsic resolution, this would not translate into dramatic increases in overall resolution forSPECT systems. The resolution of SPECT systems isdominated by the characteristics of the collimator andthe radius of rotation so that improvements in intrinsicresolution have only a minor effect on SPEC!' resolution (14). However, preprocessing appears to providesmall improvements in resolution (43—44) and improvements in camera uniformity and uniformity corrections appear to provide some improvement in quantitation (27—31).

Although the results demonstrated are for a specificSPECT system, they should apply to most SPED'systems, providing that the requirements for accurateattenuation correction and for a linear, isotropic imaging system are met. Since the sensitivity and resolutionofour present system are representative ofmost SPED'systems currently available it appears that EDT techniques will permit the accomplishment of QOV withinuseful limits.

NOTES

S Siemens Medical Systems, Islen, NJ.

t Medical Data Systems, Ann Arbor, MI (MDS “Trinary―).

I Capintec, Inc., Ramsey, NJ.

§Packard Instrument Co., Downers Grove, IL.

REFERENCES

1. Kenny PJ. What is quantitative organ visualization innuclear medicine? In: Kenny PJ, Smith EM, eds.Quantitative organ visualization in nuclear medicine.Coral Gables, FL: University of Miami Press, 197 1:1—13.

2. Kuhl DE, Edwards RQ, Ricci AR, et al. The Mark IVsystem for radionuclide computed tomography of thebrain. Radiology 1976; 121:405—413.

3. Tauxe WN, Soussaline F, Todd-Pokropek A, et al.Determination of organ volume by single-photonemission tomography. J NuciMed 1982; 23:984—986.

4. Macey Di, DeNardo GL, Denardo SJ,et al. A calibration phantom for absolute quantitation of radionuclide uptake by SPED' [Abstract]. J Nuci Med1984; 25:P 105.

5. Malko JA, Eisner RL, Engdahl JC, et al. A thresholdmethod of volume determination from tomographicreconstructions [Abstract]. JNuclMed 1983; 24:Pl9.

6. Caldwell JH, Williams DL, Hamilton GW, Ctal. Regional distribution ofmyocardial blood flow measuredby single-photon emission tomography: comparisonwith in vitro counting. J NuciMed 1982; 23:490—495.

7. BrownellGL,BurnhamCA,CheslerDA,etal.Trans

340 Kircos,Carey,andKeyes TheJournalof NuclearMedicine

verse section imaging of radionuclide distributions inheart, lung and brain. In: Ter-Pogossian MM, PhelpsME, BrownellGL, et al., eds. Reconstruction tomography in diagnostic radiology and nuclear medicine.Baltimore: University Park Press, 1977:293—307.

8. Phelps ME, Hoffman El, Huang S, et al. ECAT: anew computerized tomographic imaging system forpositron-emitting radiopharmaceuticals. J Nuci Med1978: 19:635—647.

9. KuhI DE, Phelps ME, Hoffman El, et al. Initial dinical experience with (l8)F-2-deoxy-D-glucose for determination of local cerebral glucose utilization byemission computed tomography. In: Ingvar DH, Lassen NA, eds. Cerebral function, metabolism and circulation. Copenhagen: Munksgaard, 1977:192—193.

10. Weiss ES, Ahmed SA, Welch Mi, et al. Quantificationof infarction in cross sections of canine myocardiumin vivo with positron emission transaxial tomographyand (ll)C-palmitate. Circulation 1977; 55:66—73.

11. Budinger TF, Derenzo SE, Huesman RH, et al. Roleof tomography in providing radionuclide distributionand kinetic data. mt J Nuc Med 1983; 10:39.

12. Hoffman El, Huang SC, Phelps ME. Quantitation inpositron emission computed tomography: I. Effect ofobject size. J Comput Assist Tomogr 1979; 3:299—308.

13. Keyes JW Jr, Orlandea N, Heetderks WJ, et al. TheHumongotron—a scintillation camera transaxial tomography. JNuclMed 1977; 18:381—387.

14. Kircos LT, Leonard PF, Keyes JW Jr. An optimizedcollimator for single photon computed tomographywith a scintillation camera. JNuclMed 1978; 19:322—323.

15. Budinger TF, Gullberg GT. Transverse section reconstruction of gamma-ray emitting radionuclides in patients. In: Ter-Pogossian MM, Phelps ME, BrownellGL, et al., eds. Reconstruction tomography in diagnostic radiology and nuclear medicine. Baltimore:University Park Press, 1977:315—342.

16. Kircos LT. Quantitative organ visualization. [Dissertation]. University of Michigan, 1978.

17. Kay DB, Keyes JW Jr. First order corrections forabsorption and resolution compensation in radionuclide Fourier tomography. J Nuci Med 1975;16:540—541.

18. Kircos LT, Keyes JW Jr, Carey JE. Edge findingtechniques in emission computed tomography. MedPhys 1978; 5:352.

19. Geise R, McCullough E. The use of CT scanners inmegavoltage photon-beam therapy planning. Radio!ogy 1977; 124:133.

20. Walters TE, Simon W, Chesler DA, et al. Attenuationcorrection in gamma emission computed tomography.J Comput Assist Tomogr 1981; 5:89—94.

21. Gullberg GT, Budinger TF. The use of filtering methods to compensate for constant attenuation in singlephoton emission computed tomography. IEEE Transactions on Biomedical Engineering 1981; BME28(2):142—157.

22. Lewis MH, Willerson JT, Lewis SE, et al. Attenuationcompensation in single-photon emission tomography:a comparative evaluation. J NuclMed 1982; 23: 1121—1127.

23. Oppenheim BE. Scatter correction for SPED'. J NuclMed 1984; 25:928—929.

24. Jaszczak Ri, Greer KL, Gloyd CE, et al. ImprovedSPED' quantitation using compensation for scatteredphotons. J Nucl Med 1984; 25:893—900.

25. Siegal JA. The buildup factor a general scheme forattenuation correction [Abstract]. J Nucl Med 1985;26:P54.

26. Harris CC, Greer KL, Floyd CE, et al. Attenuationcoefficients for Tc-99m photons in water-filled phantoms, determined with gamma camera: variation withwindow energy [Abstract]. J Nuc! Med 1984; 25:P22.

27. Graham LS, LaFontaine RL, Stein MA. Effects ofasymmetric photopeak windows on flood uniformityand spatial resolution of scintillation cameras. J Nuc!Med 1986; 27:706—713.

28. Mukai T, Torizuka K, Douglas KH, et al. Improvement in quantitation in SPED': attenuation and scatter correction using non-uniform attenuation data[Abstract]. JNuclMed 1985; 26:P91.

29. Bixais YJ, Rowe RW, Zubal IG, et al. Linearity andenergy corrections for scintillation cameras used inquantitative single photon emission computed tomography [Abstract]. JNuclMed 1984; 25:P87.

30. Gregory RD. Savala J, Zendrt M, et al. Assessment ofdetector sensitivity and uniformity during SPED' rotation[Abstract].JNuclMed 1984;25:Pl03.

31. Rogers WL, Clinthorne NH, Koral KF, et al. Fielduniformity considerations for rotating gamma cameraED' [Abstract].JNuclMed 1981;22:P2l.

32. Keyes JW, Leonard PF, Brody SL, et al. Myocardialinfarct quantification in the dog by single photonemission computed tomography. Circulation 1978;58:227—232.

33. Richter CW, Anderson DW. In vivo dosimetry of(99m)Tc lung scanning agent. Phys Med Biol 1974;19:73—78.

34. Jung A, Russell RGG, Bisag SP. Fate of intravenouslyinjected pyrophosphate (32)p in dogs. Am J Physiol1970;218:1757—1764.

35. IAEA. Measurement of radioactivity in body organs.mtiApp!RadialIsotopes1971;22:385—398.

36. Stevenson JS et al. The toxicity of Sn pyrophosphate:clinical manifestations prior to acute (LD)50. J Nuc!Med 1974; 15:252—256.

37. Johnston RE. Internal dosimetry I. In: The physics ofclinical nuclear medicine. AAPM Annual SummerSchool, Lexington: University of Kentucky, July 23—29, 1977:273—280.

38. Smith EM, Brownell GL, Ellett WH. Principles ofnuclear medicine. In: Wagner HN Jr, ed. Radiationdosimetry. Philadelphia: WB Saunders, 1968:742.

39. Blau M. Radiation dosimetry of(13l)iodocholesterol:the pitfalls of using tissue concentration data. J Nuc!Med 1975;16:248—249.

40. Kirschner AS, Ice RD. The author's reply [Letter]. JNuclMed 1975;16:248—249.

41. Loevinger R. Distributed radionuclide sources. In:Attix FH, Tochlin E, eds. Radiation dosimetry. 2nded. New York: Academic Press, 1969:51—88.

42. Feller PA. Internal dosimetry II. In: The physics ofclinical nuclear medicine. AAPM Annual SummerSchool, Lexington: University of Kentucky, July 23—29, 1977:282—295.

43. Ortendahl DA, Hattner RS, Kaufman L, et al. ABayesean algorithm for resolution recovery in clinicalnuclear medicine. In: Deconick F, ed. Informationprocessingin nuclear medicine. The Hague: MartinusNijhoff, 1984:392.

44. Madsen MT. Park CP. Enhancement of SPED' images by fourier filtering the projection image set. JNuclMed 1985; 26:395—402.

Volume28 •Number3 •March1987 341