Download as pdf or txt
Download as pdf or txt
You are on page 1of 14

Original article

Transmission scanning in emission tomography


Dale L. Bailey
MRC Cyclotron Unit, Hammersmith Hospital, London, UK

&misc:Received: 19 February 1998 / Accepted: 19 March 1998

&p.1:Abstract. Attenuation correction in single-photon were made simultaneously with the emission scan,
(SPET) and positron emission (PET) tomography is now something which is now viewed as highly desirable for
accepted as a vital component for the production of arte- practical usage in a busy clinical department. The radio-
fact-free, quantitative data. The most accurate attenua- nuclides used in these early applications included thuli-
tion correction methods are based on measured transmis- um-170, americium-241, and iodine-125. Anger et al.
sion scans acquired before, during, or after the emission measured transmission on the Mark II Whole Body
scan. Alternative methods use segmented images, as- Scanner [3] and, later, a gamma camera with techne-
sumed attenuation coefficients or consistency criteria to tium-99m flood sources [4, 5]. Sorenson et al. used a
compensate for photon attenuation in reconstructed im- 99mTc point source with a dual-headed rectilinear scan-
ages. This review examines the methods of acquiring ner to acquire simultaneous emission-transmission data
transmission scans in both SPET and PET and the man- [6], and Tothill and Galt used uncollimated point sources
ner in which these data are used. While attenuation cor- of 99mTc, iodine-131 and cesium-137 with a rectilinear
rection gives an exact correction in PET, as opposed to scanner for quantitative planar measurements [7]. All of
an approximate one in SPET, the magnitude of the cor- these applications predated the introduction of single-
rection factors required in PET is far greater than in photon (SPET) and positron emission (PET) tomogra-
SPET. Transmission scans also have a number of other phy. At the time of the development of the first PET sys-
potential applications in emission tomography apart tems in the 1970s, transmission scanning was developed
from attenuation correction, such as scatter correction, for the purposes of attenuation correction using either a
inter-study spatial co-registration and alignment, and ring source of a positron emitter (fluorine-18) [8] or a
motion detection and correction. The ability to acquire point source single-photon emitter such as barium-133
high-quality transmission data in a practical clinical pro- [9]. Various planar attenuation correction methods for
tocol is now an essential part of the practice of nuclear gamma cameras based on the use of transmission scans
medicine. also appeared around this time [10–12]. Today, transmis-
sion tomographic scanning has become routinely accept-
&kwd:Key words: Attenuation correction – Transmission scans ed in both SPET and PET, especially when it is com-
– Single-photon emission tomography – Positron emis- bined with the emission scan in a simultaneous acquisi-
sion tomography tion protocol. The aim of this review is to examine the
development and current and future uses of transmission
Eur J Nucl Med (1998) 25:774–787 scans in emission tomography.

Methods
Introduction
Transmission data acquisition
Transmission scanning with radionuclides was first sug-
gested by Mayneord in 1952 [1] and therefore has exist-
Transmission measurements in SPET appeared in the
ed for as long as nuclear medicine “imaging” studies
late 1970s around the time that single-photon tomogra-
have. The first reference to producing two-dimensional
phy was being developed. Budinger and Gullberg dis-
density distributions of transmitted photons for use in
cussed the use of measured transmission data but point-
conjunction with emission scanning appeared in 1966
ed out the impracticality of having to acquire two sepa-
[2]. Transmission scanning was used to match anatomi-
rate data sets [13]. Jaszczak et al. used a xenon-133 line
cal contours to the radionuclide distribution measured by
source to obtain body contours [14] simultaneously with
a rectilinear scanner. Interestingly, these measurements
the emission data on a dual-head gamma camera for use
Correspondence to: D.L. Bailey, MRC Cyclotron Unit, Hammer- in the iterative attenuation correction algorithm devel-
smith Hospital, DuCane Rd, London. W12 0HS, United Kingdom&/fn-block: oped by his co-worker, Chang [15]. In the same paper,

European Journal of Nuclear Medicine


Vol. 25, No. 7, July 1998 – © Springer-Verlag 1998
775

Jaszczak discussed an alternative method for obtaining


body contours from Compton scattered photons for de-
lineating the patient’s outline to use with the same cor-
rection algorithm. Both approaches initially started with
a single, mean attenuation coefficient within the body
outline. Maeda et al. used transmission computed to-
mography for anatomical delineation of the lung in
SPET [16] while others started to investigate the use of
transmission studies quantitatively for attenuation cor-
rection in SPET [17]. Unfortunately, the latter work is
not well known or cited as it was published in Japanese.
The current interest in the use of transmission data
for attenuation correction in SPET commenced with the
work of Malko et al., who used an uncollimated 99mTc
flood source to measure transmission prior to the admin-
istration of the radiopharmaceutical 99mTc-sulphur col-
loid [18]. Greer et al. reported the performance charac-
teristics of an uncollimated 99mTc flood source for trans-
mission tomography and suggested that the data could
be suitable for use in attenuation correction [19]. As Fig. 1. Schematic examples of some of the transmission geome-
both studies used 99mTc as both the transmission and the tries for SPET. From top left, clockwise, they are: transmission
flood source or one moving line source, the profiled multiple line
emission source, the method was impractical for routine
source method where the source size indicates the relative activity
use in producing attenuation-corrected SPET reconstruc- of the source, the fan beam and focussed line source approach on
tions. At the same time in Australia, we had started us- a triple-headed camera, and finally, the dual moving line sources
ing gadolinium-153 as the transmission radionuclide in on a double-headed camera&ig.c:/f
an uncollimated flood source for simultaneous emission-
transmission measurements with 99mTc [20]. 153Gd was
produced for us in the high flux atomic reactor (HIFAR) and a stationary line source fixed at the collimator’s fo-
at the Australian Nuclear Science and Technology Orga- cus to measure simultaneous, narrow-beam transmission
nisation (ANSTO) in an (n,γ) reaction with 152Gd. The and emission [22]. The major disadvantage of the meth-
use of 153Gd had already been documented for bone min- od, though, is severe truncation of the transaxial field of
eral measurements and lung densitometry [21]. With view (Fig. 1) for most applications apart from the head
photon energies of 97 and 103 keV and a long half-life [23]. Various strategies have been devised for dealing
(242 days), 153Gd offered a transmission source that was with this problem such as combining data from conju-
practical to use and could be separated from the 99mTc gate views with an asymmetrically focussed collimator
photons using energy discrimination on the gamma cam- [24, 25] or using a parallel slant hole collimator [26].
era. There were two significant drawbacks to this meth- The alternative approach which emerged at the same
od: (a) the transmission data contained a high proportion time and has found widespread acceptance is the use of a
of Compton-scattered emission photons from 99mTc in scanning line source with parallel hole collimators [27].
the 100 keV±10% energy window which corrupted the The technique employs both strict physical collimation
data, and, (b) as the flood source was uncollimated, of the source and electronic windowing of the detector to
many of the transmission photons accepted were also separate the emission and transmission data. Such a
scattered, producing measures of “effective” or broad- system offers a flexible approach to simultaneous emis-
beam attenuation coefficients (µ) rather than narrow- sion-transmission measurements as the radionuclides
beam values. The former problem was dealt with by esti- used can be of higher, lower or the same energy as each
mating the scatter component from the upper energy other. Additionally, as the system uses parallel hole col-
emission window into the lower energy transmission limation there is no truncation of the transaxial field of
window and subtracting it. The latter problem was not view. The drawbacks of the system are that it requires a
dealt with in this implementation, but the use of broad- slightly more complex mechanical design, it does not
beam µ values partially offset the scatter in the emission work with triple-headed SPET systems due to geometri-
window, which was not corrected explicitly. cal constraints, and there is a slight decrease in emission
At approximately the same time SPET was becoming count rate (~10%) due to the use of the scanning trans-
widely disseminated and the method of choice for myo- mission window. However, the cross-talk between the
cardial perfusion studies. This coincided with the gener- emission and transmission windows is sufficiently low
al availability of multi-head gamma cameras, which af- (typically <5%) and the configuration produces high-
forded more flexibility in developing simultaneous emis- quality, narrow-beam attenuation factors.
sion-transmission scanning protocols. The triple-headed PET measurements have traditionally been corrected
systems, in particular, could use a fan-beam collimator for attenuation using measured transmission data. The

European Journal of Nuclear Medicine Vol. 25, No. 7, July 1998


776

to window the transmission data so that events were only


accepted when the line of response recorded between
two detectors was collinear with the known location of
the rod. This has two effects: (a) reduced scatter in the
transmission measurements, and more importantly, (b)
provision of a means of separating emission and trans-
mission data to allow either post-injection transmission
[31–33] or even simultaneous emission-transmission
measurements [34, 35]. Windowed rotating rod sources
remain the transmission source geometry of choice in
PET at present. PET transmission measurements have
also been used in their own right and not just as a meth-
od for generating factors which corrected for photon at-
tenuation, such as in estimating lung density [36] in PET
studies.
Until very recently, nearly all PET systems manufac-
tured since 1990 had been designed to operate in both
2D and 3D modes [37], and transmission measurements
were usually acquired with the septa in place in 2D
Fig. 2. Schematic examples of some of the transmission geome- mode with rotating rod sources. However, PET systems
tries for PET. From top left, clockwise, they are: ring sources of a are now being constructed which operate exclusively in
positron emitter measuring transmission (coincidence mode), ro- 3D and a reappraisal of the optimal transmission acquisi-
tating rod positron sources (coincidence mode), rotating single-
tion system for these “open” volumetric systems has
photon source where a “coincidence” is made between the source
position and photons detected on the opposing side of the ring,
been required. In an early paper on PET system design
and finally, fixed rod sources on a rotating, partial ring tomograph and performance, Derenzo et al. suggested the use of a
(coincidence mode)&ig.c:/f single-photon emitting point source of 133Ba [9]. This
approach has recently been implemented on a number of
commercial tomographs [38–40]. When using a single-
reasons that this has been the case for PET, and that this photon source in a modern PET system, a mechanism is
has only come more recently in SPET, is that in PET the required to form a “coincidence” event between the de-
correction for attenuation is exact, as attenuation is inde- tector which records a valid single photon and the detec-
pendent of the origin of the annihilation along a line of tor on the opposing side of the tomograph which is clos-
response, and because PET started out in the research est to the current location of the rotating point source. In
domain where there was greater emphasis on accurate the whole body 3D system in our laboratory, the ECAT
quantitative measurements and less emphasis on practi- EXACT 3D [41, 42], there is a 48-turn helix of stainless
cal aspects of data acquisition. The transmission ring steel filled with water, through which a point source of
source approach (Fig. 2), which is still in widespread use 137Cs (half-life =30.2 years, E =0.662 MeV) is pumped
γ
today, uses radioactive rings [8], usually containing the at a velocity of approximately 1 m·s–1. As this is a total-
positron emitting radionuclides gallium-68/germanium- ly uncollimated system, a high proportion of the events
68 (half-lives 68 min and 270.8 days respectively [28]) that are recorded are scattered. However, as this is a sin-
which co-exist in transient equilibrium. The rings are gle-photon counting system, more radioactivity can be
mechanically moved in front of the detectors and coinci- used and consequently count rates are increased by ap-
dence events are recorded by the detector near the anni- proximately an order of magnitude, to produce highly
hilation event and the detector in the opposing fan, precise data which are biased due to the scattered pho-
which records the other photon of the pair after it has tons. Strategies for dealing with this will be examined in
passed through the subject. The energy of the transmis- the following section.
sion photons and emission photons is the same in this
case. There is usually one transmission ring per ring of
detectors in discrete systems such as those based on the Transmission data processing
“block” design [29]. In a manner similar to the way the
2D flood source in SPET transmission measurements The calculation of the transmission factors through the
was “collapsed” in one dimension to give a 1D line object of interest is the same for SPET and PET. The
source which could be electronically windowed, the con- transmitted count rate through the subject is compared
tinuous ring sources in PET were collapsed radially into on a pixelwise basis with a transmission scan without
“rod” sources which were continuously rotated about the the subject in place, often referred to as a “blank” scan.
subject [30]. A major limitation of this approach is the The standard exponential attenuation equation is used:
local dead time on the detector block that is near the rod
source. However, in favour of the method was the ability Nx=N0 e–∫µ(x)dx, (1)

European Journal of Nuclear Medicine Vol. 25, No. 7, July 1998


777

where Nx and N0 are the attenuated and unattenuated


count rates (units: counts·s–1) respectively, µ is the linear
attenuation coefficient (units: cm–1), and x is the distance
(units: cm) from the source to the detector. The attenua-
tion coefficients for the body range from air values
(close to zero) in the cranial sinuses, through very low
values in the lungs, to fat, tissue and cortical bone,
which has the highest value. The attenuation coefficient
depends on a number of factors including photon energy,
scattering cross-section of the material, and electron Fig. 3. The concept of broad-beam and narrow-beam transmis-
density, and has typical values for water, which is almost sion, shown for a simple single channel detector. In the narrow-
the same as tissue, of 0.15 cm–1 for 140-keV photons beam case (top) the majority of photons that are scattered will not
and 0.096 cm–1 for 511-keV photons. The linear attenua- be counted. This is due to the collimation of both source and de-
tion coefficient is a measure of the probability of the tector. The difference between the count rates with and without
photon being absorbed per unit length in the object. the object in place is due to both total attenuation within the ob-
Equation 1 can be rearranged to give the integral of at- ject and scattering out of the line of sight of the detector. In the
broad-beam case (lower) the source is uncollimated and this al-
tenuation coefficients for any projection in this manner:
lows photons with original trajectories that would not have been

z µ ( x ) dx = log FH NN IK ,
counted to be scattered back into the acceptance angle of the colli-
0 (2)
e mated detector. A similar number of photons would be expected
x
to be scattered out of the line of sight and attenuated as the nar-
which, as the left-hand side of the equation is now just a row-beam case, but the count rate will be higher in the broad-
simple summation of a single variable, is amenable to beam case due to the acceptance of scattered events. This lowers
reconstruction with standard back projection methods. the effective attenuation coefficient value&ig.c:/f
Attenuation coefficients are often referred to as “narrow-
beam” or “broad-beam” depending on whether they con-
tain scattered transmission photons or not. Figure 3 illus-
trates the simplest example of this with a single detector
measuring uncollimated and collimated sources. Ideally,
narrow-beam transmission measurements are required
for attenuation correction, and, as in Fig. 3, this is deter-
mined by the geometry of the transmission and acquisi-
tion system.
In applying the transmission factors in emission to-
mography, we will deal with the PET case first, as it is
the simplest. Figure 4 shows the attenuation paths for
single-photon and annihilation (coincidence) radiation Fig. 4. Attenuation for a point source in SPET is measured by dis-
detection. For coincidence detection, Eqn. 1 is modified tance a as single photons only are counted. In the PET case both
as both photons pass through the full thickness of the photons are counted and therefore the attenuation for the event is
object for each angle and we therefore have an attenua- a+b, the total thickness through the object at the projection angle
shown. The attenuation for the projection is independent of the
tion equation of:
position along the chord for the projection in the PET case, and
Nx=N0 e–∫µ(a+b)dx, (3) provides an exact correction. In SPET the attenuation varies with
the position of the point source along the chord, as the position is
where the integral now is a constant, independent of the usually not known&ig.c:/f
point source location along the chord, for any projection.
This is not the case in SPET as the attenuation to the de- for any particular line of response. The attenuation cor-
tector varies with depth in the object as the photons are rection factor (ACF) for each line of response is there-
only recorded in one direction. For a distributed source, fore just 1/α. The correction is an exact one in PET, and
with multiple emission points along any projection, the is the reason that PET attenuation correction is regarded
emission count rate recorded is: as accurate. In SPET, the attenuation term in brackets
varies with the depth of the point source in the subject,
Cx=∫C0(x) · (e–∫µ(x))dx, (4)
and therefore the correction term cannot be moved out-
where Cx is the integrated count rate from each of the at- side the integral.
tenuated point sources C0(x). In PET, the term in brack- In practice, PET transmission and blank scans are
ets is a constant and therefore we can rewrite the equa- measured and the ratio calculated, often after some spa-
tion as: tial filtering (smoothing) to reduce statistical variations
in the recorded count rates. The blank is usually ac-
Cx=α∫C0 (x)dx (5) quired for a longer duration than the transmission scan

European Journal of Nuclear Medicine Vol. 25, No. 7, July 1998


778

to reduce the statistical noise. It can be seen from Eq. 1 the 137Cs point source. Segmentation and reclassification
that the ratio of the blank count rate (N0) divided by the deals with both the difference in photon energy and the
transmission count rate (Nx) gives the attenuation correc- inclusion of scattered photons in one step. The method is
tion factor directly for the line of response. The PET automated and rapid to apply, adding 3–5 min to the pro-
transmission scan is ideally acquired before the radio- cessing time on a Sparc 5 workstation for the segmenta-
pharmaceutical is administered, but this often leads to tion and classification of 95 planes of data. Alternative
long clinical protocols which are not ideal. Post-injec- methods that have been proposed for dealing with the
tion transmission using electronically windowed rotating scatter in single-photon transmission measurements in
rod sources of 68Ge helps by allowing the transmission PET include a dual-energy window scatter correction
scan to be acquired immediately before or after the [40] and reclassification using artificial neural networks
emission scan. Short duration transmission scans [49].
(~2–3 min), often interleaved with the emission data in a In SPET, the attenuation correction term cannot be
whole-body acquisition protocol, have been employed separated from the emission count rate recorded and
but often further processing is required such as spatial therefore all SPET methods for attenuation correction
filtering or image processing classification tools such as are approximations. The majority of procedures for at-
tissue segmentation [43–45]. The use of tissue classifica- tenuation correction using measured transmission data in
tion techniques requires that the natural logarithm of the SPET are applied either during or after the emission re-
ratio of blank-to-transmission count rates are recon- construction. One well-known approach is the modified
structed first, as the segmentation techniques are applied version of the Chang method [15]. The original method
to the reconstructed volume. Once the data have been in this paper used a body contour which contained a sin-
adjusted to the “correct” attenuation coefficients, the at- gle attenuation factor, which was modified in an iterative
tenuation correction factors are derived by forward pro- manner. The emission data are reconstructed without
jection at the appropriate angles to form attenuation cor- correction, and an average attenuation correction factor
rection projections which match the emission data. Si- based on the calculation of the sum of attenuation coeffi-
multaneous emission-transmission scanning aids this cients from each point in the object to the detector is de-
further, but care must be taken by using rod sources with rived and multiplied with the uncorrected emission re-
reasonably low activities as the scatter and random coin- construction. The correction term for each point in the
cidences from the rod sources are recorded in the emis- image [A(x,y)] is found from
sion window, and therefore degrade the quality of the −1
primary data [46] as well as causing count losses due to LM M
A ( x , y ) = 1 ∑ e − µ a ( x , y, θ i )
OP , (6)
dead time. N
M i =1 Q
When using a single-photon point source for PET where M is the number of projection angles over 360°.
transmission measurements, post-processing of the data The emission data are then forward projected, with at-
is required. Firstly, the point source most frequently em- tenuation from the assumed body contour, and compared
ployed at present is 137Cs, which has a higher energy with the acquired data. Differences are calculated for
photon than the 511 keV from the emission annihilation each projection element and these are back projected in
radiation; therefore the measured transmission must be the conventional manner and corrected for attenuation.
modified to account for the difference. This is relatively The reconstructed “difference” image is then used to
straightforward and a simple exponent can be used [40]. modify the first estimate of the corrected emission re-
The larger issue is the high amount of scattered trans- construction. This updated emission reconstruction is
mission photons acquired in 3D systems, especially with again forward projected, differences calculated and back
the bismuth germanate (BGO) detectors used in the ma- projected, and so on, iteratively. In the modified version,
jority of PET devices. The energy resolution of BGO in the reconstructed attenuation image is substituted for the
a scanner is around 15%–25% full-width at half-maxi- assumed body contour and single µ value [20, 50]. An
mum at 511 keV, and therefore energy discrimination is example of an attenuation reconstruction and the corre-
not particularly helpful for excluding scattered photons. sponding attenuation correction map is shown in Fig. 6.
This is less of a problem for NaI(Tl)-based devices as An alternative is to use the ratio of the projections, rath-
they have superior energy resolution. Another system er than the difference, to calculate the error term before
which produces high-quality scatter-free data is the con- back projection [17].
tinuously rotating partial ring tomograph (ECAT ART) Iterative reconstruction techniques for emission
[47], which uses collimated point sources of 137Cs [48]. tomographic reconstruction are becoming more routine-
One significant advantage of single-photon transmission ly used clinically due to improvements in algorithms and
in PET is the very high photon flux recorded, and this hardware. In particular, the maximum likelihood, esti-
leads to high-quality scans compared with coincidence mation maximisation (ML-EM) algorithm [51, 52] and
measurements. This, in principle, makes the tissue clas- the optimally partitioned, block-iterative version known
sification and segmentation task easier and more robust. as ordered subset EM (OSEM) [53] have found wide-
At present, we use the local threshold segmentation spread clinical acceptance in both SPET and PET. In ap-
method [45] on the EXACT 3D system with data from plying these techniques in PET, the data are usually cor-

European Journal of Nuclear Medicine Vol. 25, No. 7, July 1998


779

rected for attenuation prior to reconstruction due to the photon and annihilation coincidence attenuation correc-
exact properties of the correction, but in SPET, the data tion factors. The aim is to give some idea of the attenua-
can be incorporated into the reconstruction process by tion correction factors required, rather than to produce
inclusion in the transition matrix, along with other fac- exact corrections. Figures 5 and 6 demonstrate the
tors which model physical aspects of the data collection Chang attenuation correction calculation using measured
process such as scatter, depth-dependent resolution and transmission data for SPET. By modifying the limits of
anatomical boundaries. the summation in Eq. 6 to calculate the total attenuation
for any ray in both directions (a+b in Fig. 5), an average
attenuation to each point in the section can be calculated
Comparison of attenuation correction in PET for PET, analogous to the SPET case. This has been
and SPET done for the following imaging situations: (a) thallium-
201 in SPET, (b) 99mTc in SPET, (c) 18F in SPET, and (d)
There is growing interest in the clinical use of PET in 18F in PET coincidence mode. The transmission scan
nuclear medicine. It is possible to use a modified dual- shown in Fig. 5 was scaled so that the tissue attenuation
headed gamma camera in coincidence mode to record coefficient values were appropriate for each of the pho-
annihilation radiation [54]. It will be useful, therefore, to ton energies corresponding to the three cases above (i.e.,
compare the impact that attenuation correction has on 70 keV, 140 keV and 511 keV), namely µ=0.18 cm–1,
coincidence imaging, compared with SPET, and to as- 0.15 cm–1 and 0.095 cm–1 respectively. The transmission
sess the need for transmission scans for attenuation cor- scan used was originally recorded in single-photon mode
rection. on the EXACT 3D scanner using 137Cs as the source.
For this study, modified versions of the method of The first three cases used the single-photon calculation
Chang [15] have been used for calculating both single- of Eq. 6 and the final case for coincidence mode used
the modified equation. It should be stressed that this is
not how PET attenuation correction factors are applied
in practice, but has merely been done to give an indica-
tion of the average attenuation to each point in the body,
and to allow comparison with the SPET situation.

Scatter correction

While the primary motivation for acquiring transmission


data has been for accurate attenuation correction, the da-
ta have found application in other areas. One area that
has received considerable attention has been the use for
Fig. 5. Examples of correction schemes for SPET and PET. In transmission data in scatter correction. This is perhaps
SPET, the method of Chang calculates the attenuation factor e–µa an obvious application as attenuation and scatter are so
for a number of angles and takes the average for each point within
closely related. The use here in SPET predates that in
the object as an approximate average attenuation to that point.
This can be modified iteratively. In PET, the attenuation is given
PET. Mukai et al. [55] used attenuation data to modify a
directly by the factor eµ(a+b) for each projection. Following the convolution kernel which was applied to the attenuation
method of Chang, this can be used to calculate average attenua- corrected projections to derive an estimate of scatter.
tion within the object to compare with the SPET case, although They used CT scans to provide the attenuation data.
this is not the method used for attenuation correction in PET&ig.c:/f Ljungberg and Strand used measured attenuation (“den-
sity”) maps with Monte Carlo simulation to study the
distribution of scatter, but this was not practical for rou-
tine correction purposes [56, 57]. Once there was ready
access to transmission data from simultaneous emission-
transmission scans in SPET, we developed a method for
incorporating this into scatter correction by modifying
the estimates of scatter fraction for each projection ele-
ment [58]. The scatter model used was a convolution of
the photopeak data with a predetermined “scatter kernel”
(k) [59, 60]. The transmission projection data were con-
verted to attenuation projections using Eq. 2. Work from
Siegal and Wu aimed at measuring the “buildup” of pho-
Fig. 6. An example of an attenuation image and the corresponding tons between the chest wall and the centre of the left
calculated attenuation factors using the Chang method. Attenua- ventricle to estimate left ventricular volume from planar
tion is greatest towards the centre of the body, as expected&ig.c:/f gamma camera measurements had demonstrated that if

European Journal of Nuclear Medicine Vol. 25, No. 7, July 1998


780

the transmission factors were known the buildup, or and is particularly useful for dynamic studies. Similar
scatter, could be estimated from the equation: methods could be used in SPET.
Nx
= 1 − (1 − e − µ d )n , (7) Definition of anatomical regions of interest. &p.1:As the at-
N0
tenuation images display density differences, it is possi-
where d is the depth in tissue and n is the buildup factor ble to use the data to define regions of interest based on
at infinite depth, B(∞) [61–63]. Narrow-beam transmis- anatomical landmarks. The data have limited density
sion factors were essential for this method. We used this resolution for discerning differences between soft tissues
approach in SPET and developed a transmission-depen- (as opposed to X-ray CT), but can be used for the tho-
dent scatter correction [64]. The transmission data re- rax, in particular. This is a return to the original use of
placed the term e–1µd in Eq. 7, and the only factor re- transmission scanning in nuclear medicine, but in tomo-
quired was B(∞), which was determined a priori with ex- graphic mode. More recently, transmission scans have
perimental measurements. This method has been im- been used to define the lung regions for application to
proved by using a Gaussian basis function for the esti- functional images of ventilation and lung clearance in
mation of the scatter component [65]. An alternative both planar and SPET measurements [71, 72]. It is con-
method developed by Welch et al. used the reconstructed ceivable that using the high photon flux and resolution
attenuation map to calculate the expected scatter distri- of single-photon transmission measurements in 3D PET
bution in the projections [66]. This method is computa- it will be possible to separate bone and soft tissue re-
tionally intensive; however, it is well suited to inclusion gions from attenuation images to define further regions
in the EM-ML algorithm. of interest.
In 3D PET, the reconstructed attenuation data have
also been used for calculation of the predicted scatter Intrinsic correction for limited spatial resolution. &p.1:Trans-
distribution [67, 68]. The method tends to use coarse mission scans have been used in PET studies to correct
matrices for the calculation to reduce the computation for cardiac movement and partial volume effects due to
required. This approach has been found to be accurate, the limited spatial resolution of the system. Iida devel-
although it remains susceptible to the influence of scat- oped the concept of the myocardial tissue fraction to
ter from outside the coincidence field of view, in areas overcome the limitations in measuring tracer concentra-
which are not measured by either the transmission or tion in the myocardium due to motion and the small
emission scans. transmural wall thickness. This was originally done by
compartmental modelling and estimating the volume of
distribution from oxygen-15 labelled water studies,
Further uses of transmission data which is proportional to tissue fraction [73]. Subse-
quently, a transmission-based technique was developed
Motion detection and correction. &p.1:Subject and organ which involved the subtraction of a C-[15O] blood vol-
movement degrade the image quality obtainable in both ume scan from the attenuation image, which produced
SPET and PET, unless it can be corrected for. Cardiac an image of the extravascular density [74]. The ratio of
gating is an example where organ movement is mea- these two parameters gave a measure of the perfusable
sured and compensation made, by recording in small tissue fraction which was independent of the partial vol-
cardiac temporal sub-intervals. For gross subject move- ume effect. In effect, the limited resolution in each mea-
ment, transmission scanning offers a potential method surement was used to cancel out the limitation of finite
for tracking and correcting this motion which should be resolution in the final measurement.
potentially more robust than using the emission data, as
the emission distribution may change with time. We Spatial co-registration and anatomically guided recon-
have used this approach to record dynamic planar emis- struction. &p.1:Spatial co-registration of an individual’s re-
sion-transmission scans with a scanning line source to constructed data with other image data is becoming in-
measure the clearance of inhaled radioaerosols from the creasingly utilised. The simplest example in nuclear
lung [69]. medicine is the comparison of myocardial perfusion pro-
Movement can also lead to misalignment between the files with normal data bases. Co-registration of paired
transmission and emission scans if they are recorded se- scans in an individual acquired at different times is also
quentially. This will result in an inaccurate attenuation utilised [75]. These techniques match a single study to
correction being applied. A method has been described another, or possibly to a group of others, in a standard,
in PET which derives the translation and rotation after defined spatial co-ordinate system. The problem with us-
movement based on emission data, and then applies ing functional images to perform this is that they often
these transforms to the attenuation reconstruction to re- have patterns of distribution which differ from normals
orient it into the corresponding location [70]. The data or other individuals with the same disease. Anatomical
are forward projected in this new space to provide the at- data are often likely to match more closely, and are less
tenuation correction factors. This ensures that the emis- likely to change over a short period of time, unlike, for
sion data have the correct attenuation factors applied, example, stress and redistribution myocardial perfusion

European Journal of Nuclear Medicine Vol. 25, No. 7, July 1998


781

images. One possible use of transmission data acquired with transmission-dependent attenuation and scatter cor-
in a simultaneous protocol would be to co-register paired rection. In the same chest phantom, a “blood pool” distri-
PET-PET, SPET-SPET or PET-SPET scans using the at- bution of emission activity demonstrated an absolute ac-
tenuation reconstructions. The transformation matrix de- curacy of within 5% of the true value in the heart using
fined by this can then be applied to the emission data, as the same methods. Narita et al. [65] found that in both
they are perfectly spatially co-registered with the attenu- simulated and experimental data the accuracy of the final
ation data. In neuroscience research, emission data are reconstructed image was within 3% of the true values, al-
often transformed to a standard stereotaxic space such as though this is perhaps optimistic for human studies. Nev-
that of Talaraich and Tourneaux [76] for the purposes of ertheless, it appears that the algorithms that exist now are
group or single subject versus group comparisons [77]. capable of producing data with a high degree of accuracy
This necessitates an elastic transformation of each sub- in SPET under well-supervised and controlled condi-
ject’s brain to the standard space [78]. The high defini- tions. These methods were recently applied to SPET
tion provided by structural imaging helps to improve the [123I]-IMP regional cerebral blood flow (rCBF) measure-
accuracy of this process [79]. ments in comparison with [15O]-H2O PET rCBF mea-
Finally, the use of iterative reconstruction methods al- surements and found to be within 10% of each other [83].
lows incorporation of prior information, if it is known,
about the expected distribution of radioactivity. In the
simplest case this may be to constrain the reconstruction
to positive values within the body outline. More sophis-
ticated approaches use anatomical information from X-
ray CT or MRI that has been segmented to classify tis-
sues with different expected concentrations of the radio-
tracer. This information is then used to “encourage” the
reconstruction with predefined probabilities or weigh-
tings in certain areas [80–82]. In one sense this can be
thought of as an intrinsic partial volume correction tech-
nique. In the brain, differences are usually ascribed to
grey matter, white matter, and ventricles for blood perfu-
sion and metabolic radiotracers. Clearly, the course reso-
lution of attenuation data from PET and SPET would not
allow this degree of classification, but body boundaries
and lung fields would provide sufficient contrast for use
in this type of scheme. Sufficient contrast exists in the
head with SPET and PET attenuation images to classify
pixels into bone, soft tissue and sinuses.

Results

Accuracy of attenuation and scatter correction in PET


and SPET

Given the vast amount of effort that has focussed on at-


tenuation and scatter correction, a reasonable question to
ask is what the current capability of SPET and PET is in
producing accurate reconstructions.
It is clear that both attenuation and scatter correction
are required for SPET and 3D PET, as they both contain
a large scatter component (30%–50%) in the photopeak
energy window. In SPET, the paper by Mukai et al. [55]
quoted absolute accuracy after scatter correction within
9% of the known values in a multi-component test object
Fig. 7a–d. Examples of attenuation correction factor maps for
over a range of radioactivity concentrations with both at-
SPET and PET. The maps correspond to (a) 201Tl in SPET, (b)
tenuation and scatter correction. Meikle et al. [64] 99mTc in SPET, (c) 18F in SPET and (d) 18F in PET mode. The y-
showed that without scatter correction the values in lung axis values in the plots are the magnitude of the corrections,
in a heterogeneous chest phantom were overestimated by which range from maxima of approximately 9 for 201Tl down to 3
35% (using measured transmission data for attenuation for 18F in SPET to nearly 20 in PET. The plots are for profiles at
correction), and improved to within 4% of the true value the location of the horizontal cross-hair&ig.c:/f

European Journal of Nuclear Medicine Vol. 25, No. 7, July 1998


782

Fig. 8. Two examples of cardiac scans in


coincidence mode with [18F]-DG are
shown. Both subjects were males. The
data were acquired on the EXACT 3D
large field of view PET system at Ham-
mersmith Hospital, with transmission re-
corded in single-photon mode using a
137Cs source and processed with local

threshold segmentation. The images are


transverse sections of 5 mm thickness. In
the top row, the lack of attenuation cor-
rection (left) overestimates the uptake in
the anterolateral wall. The uptake in the
lateral (free) wall and the anterior wall of
the right ventricle are also overestimated
compared with the corrected case (right).
The reconstructed counts have been re-
distributed preferentially to areas of low-
er attenuation, noticeable in the enhance-
ment in the lung fields and the absence
of counts in the region of spine and later-
al chest wall. In the lower case, there is
an apparent decrease in the posterior and
septal walls, consistent with an attenua-
tion artefact (left). After correction
(right), there is more equal distribution
of counts in these areas relative to the an-
terior wall&ig.c:/f

3D PET has mostly been employed in cerebral studies ever, the maximum correction factors average around 20.
to date. In phantom data emulating this situation, we have This is due to the longer path length for the two photons
found 3D PET to be as accurate as 2D measurements at each projection. In fact, individual correction factors
[84]. The generally accepted gold standard, 2D PET mea- in the projections (where the corrections are applied in
surements, are usually accurate to within 5% [85]. A PET before reconstruction) for some lines of response
number of authors have reported on the accuracy of 3D through the body may reach factors of 70–100 [85]. This
PET measurements modelling neurotransmitter and re- is independent of whether the device recording the coin-
ceptor uptake and density parameters, by simulation, ex- cidence events is a full-ring conventional PET system or
perimentation and in human subjects [86–88] using a va- a modified dual-headed gamma camera. As the initial
riety of scatter correction approaches. These studies have impetus to commercially develop routine attenuation
again found that 3D PET with scatter and attenuation correction in SPET was to remove the attenuation arte-
correction is as accurate as 2D PET, with reported agree- facts which confounded myocardial perfusion studies it
ment to within better than 5%. The chest and abdomen would seem a retrograde step to commence cardiac coin-
have seen little published for scatter correction in 3D cidence work with gamma cameras without attenuation
PET to date and remain a challenge for the future [89]. correction, as the effect of attenuation is now very much
greater. Examples of an [18F]-DG cardiac scan recon-
structed with and without scatter correction are shown in
Comparison of attenuation correction in PET and SPET Fig. 8. At least one manufacturer has realised this and
offers a transmission measurement system for their coin-
The results and profiles through the maps of attenuation cidence gamma camera.
correction factors for the four situations in SPET and
PET are shown in Fig. 7. All of the corrections peak to-
wards the centre of the body along an axis incorporating Discussion
heart and spine. In the SPET measurements, the greatest
correction factors (~9–10) are required for the low-ener- Transmission scanning was “rediscovered” in nuclear
gy photons from 201Tl (lead X-rays), and the lowest medicine in the mid-1980s for the purposes of improv-
maximum correction factors for 18F (~3). In PET, how- ing the quantitative accuracy of SPET emission studies.

European Journal of Nuclear Medicine Vol. 25, No. 7, July 1998


783

The first widespread application that it has found is in present when different scanners with different scanning
attenuation correction of myocardial perfusion SPET couches and patient positioning are used. There have
studies [90], where the heterogeneous density of the tho- been some recent developments in simultaneous multi-
rax and chest wall could cause attenuation artefacts in a modality PET-MRI also. The PET group at UCLA have
variety of locations. These artefacts reduced the overall used a small animal PET system based on the new inor-
specificity of the test by creating false-positive or equiv- ganic scintillator lutetium oxyorthosilicate (LSO) [102]
ocal scans. It has already been shown that correction for in a conventional MRI scanner [103, 104]. The system
attenuation reduces these artefacts [91] and, in particu- uses optical fibres to transport the light from the scintil-
lar, has a dramatic impact on specificity in single-vessel lator crystals to the photomultiplier tubes, thus avoiding
disease [92, 93]. The widespread introduction of attenu- problems from the magnetic field which would perturb
ation-corrected cardiac SPET has been slow due to is- the electrons in the tubes. Their collaborators in London
sues such as re-education of observers reporting the have developed a prototype PET-NMR spectroscopy
scans and the need for new data bases of normals for scanner for studying the physiology of the isolated rat
comparative purposes. However, most gamma camera heart using similar technology [105].
manufacturers now offer some form of simultaneous
emission-transmission scanning facility for attenuation
correction. For clinical use, simultaneity would appear Conclusion
to be mandatory.
Alternative methods for improving attenuation cor- It is still quite early days for the routine use of transmis-
rection without the need for transmission scanning exist. sion scanning in SPET, and for simultaneous emission-
The consistency criteria described by Natterer [94] have transmission measurements in PET. A number of obsta-
been used to derive attenuation correction maps consis- cles still need to be addressed before the use of these ac-
tent with the measured projection data [95]. Alternative- quisition and processing techniques are widely accepted
ly, appropriately scaled CT [96, 97] or segmented MRI and become routine. These include:
data can be used. These methods suffer from the usual 1. Simultaneous measurements. Clearly, for practical
problems of working with multi-modality images, name- reasons, simultaneous emission-transmission measure-
ly, accurate co-registration from the different modalities. ments are highly desirable. Simultaneous acquisition
Another suggested method used a small injection of a adds little or no time to the scan duration, and therefore
lung perfusion radiotracer to outline the lungs, from will have a high compliance. In SPET, the scanning line
which an approximate correction could be derived [98]. source and the focussed line/fan-beam collimator are
However, the advantge of a transmission scan recorded flexible approaches that are now being used routinely
simultaneously with the emission scan, and therefore with both 99mTc and 201Tl in cardiac scanning. While
spatially co-registered and easily accessible, has an over- similar methods have been developed for PET, purely
whelming attraction in both SPET and PET, and is likely practical considerations, such as using appropriately low
to become the standard practice. This has been achieved amounts of radioactivity in coincidence transmission
in SPET with a number of the different arrangements for sources [35] or adequately separating the emission and
transmission scanning, but has not yet become a practi- transmission data when using a high activity 137Cs point
cal reality in PET. source, have delayed its widespread introduction. This is
Transmission data are finding uses in further improv- particularly true in whole body PET surveys. The ques-
ing the accuracy of quantification in emission tomogra- tion of whether a simultaneous measurement can be
phy, particularly in scatter correction. This is equally made with dual-headed gamma cameras in coincidence
true for SPET and PET. There is potential in the future mode is still open for debate.
to use the transmission data to monitor subject move- 2. Higher sensitivity and more efficient acquisition
ment and derive corrections, as the attenuation distribu- schemes. Transmission scanning in SPET and PET can
tion is unlikely to change in the course of a study. often lead to high noise in the attenuation correction data
There is also interest in developing high-resolution which propagates through into the final reconstructed
structural imaging to be used in conjunction with func- emission images. The transmission data often require
tional emission tomographic data. Hasegawa and col- some processing to suppress this, and clearly, higher
leagues have developed a combined SPET-CT system sensitivity and more efficient transmission acquisitions
[99, 100], and Townsend and colleagues are working on schemes would help. While 153Gd has many desirable
a combined rotating PET-CT system [101]. The systems properties for SPET transmission acquisition and attenu-
can use the X-ray CT data for attenuation and scatter ation correction, it is a relatively low-energy photon
correction, but also have particular attraction for inter- which is highly attenuated, and this leads to very low
ventional procedures such as combined functional/struc- count rates in some projections through the average sub-
tural guided fine-needle biopsy. As the functional emis- ject. The profiled line source approach [106] addresses
sion and structural anatomical images are acquired se- this issue to some extent by placing higher concentration
quentially or interleaved on the same device the task of transmission sources towards the centre of the field of
accurately co-registering these should be simpler than at view, which is often the thicker section of the body.

European Journal of Nuclear Medicine Vol. 25, No. 7, July 1998


784

3. Flexible emission-transmission radionuclide com- 15. Chang LT. A method for attenuation correction in radionuclide
binations. Attenuation correction will find application computed tomography. IEEE Trans Nucl Sci 1978; NS-25:
for all forms of SPET and PET studies, and therefore re- 638–643.
strictions such as having the transmission radionuclide 16. Maeda H, Itoh H, Ishii Y, Mukai T, Todo G, Fujita T, Torizuka
K. Determination of the pleural edge by gamma-ray transmis-
energy lower than the emission radionuclide in SPET sion computed tomography. J Nucl Med 1981; 22: 815–817.
will prove a limitation. The scanning line source ap- 17. Morozumi T, Nakajima M, Ogawa K, Yuta S. Attenuation cor-
proach provides this flexibility already. rection methods using the information of attenuation distribu-
Transmission scanning in emission tomography pro- tion for single photon emission CT. Med Imag Tech 1984; 2:
vides more than just artefact reduction in myocardial 20–28.
perfusion scanning, and will see increasing utility in pro- 18. Malko JA, Van Heertum RL, Gullberg GT, Kowalsky WP.
viding quantitative measurements in SPET and PET in SPECT liver imaging using an iterative attenuation correction
the realm of functional imaging. algorithm and an external flood source. J Nucl Med 1986; 27:
701–705.
&p.2:Acknowledgements. I am grateful to Steve Meikle, PhD, of the 19. Greer KL, Harris CC, Jaszczak RJ, Coleman RE, Hedland
Department of Nuclear Medicine, Royal Prince Alfred Hospital, LW, Floyd CE, Manglos SH. Transmission computed tomog-
Sydney, for the valuable comments and suggestions he made re- raphy data acquisition with a SPECT system. J Nucl Med Tech
garding this paper. 1987; 15: 53–56.
20. Bailey DL, Hutton BF, Walker PJ. Improved SPECT using si-
multaneous emission and transmission tomography. J Nucl
Med 1987; 28: 844–851.
References 21. Kaufman L, Gamsu G, Savoca C, Swann S, Murphey L,
Hruska B, Palmer D, Ullman J. Measurement of absolute lung
1. Mayneord WV. The radiography of the human body with ra- density by Compton-scatter densitometry. IEEE Trans Nucl
dioactive isotopes. Br J Radiol 1952; 25: 517–525. Sci 1976; NS-23: 599–605.
2. Kuhl DE, Hale J, Eaton WL. Transmission scanning: a useful 22. Tung C-H, Gullberg GT, Zeng GL, Christian PE, Datz FL,
adjunct to conventional emission scanning for accurately key- Morgan HT. Non-uniform attenuation correction using simul-
ing isotope deposition to radiographic anatomy. Radiology taneous transmission and emission converging tomography.
1966; 87: 278. IEEE Trans Nucl Sci 1992; 39: 1134–1143.
3. Anger HO. Whole-body scanner, Mark II [abstract]. J Nucl 23. Manglos SH. Truncation artifact suppression in cone-beam ra-
Med 1966; 7: 331. dionuclide transmission CT using maximum likelihood tech-
4. Anger HO, McRae J. Transmission scintiphotography. J Nucl niques: evaluation with human subjects. Phys Med Biol 1992;
Med 1968; 9: 267–269. 37: 549–562.
5. McRae J, Anger HO. Transmission scintiphotography and its 24. Hawman EG, Ficaro EP, Hamill JJ, Schwaiger M. Fan beam
applications. In: Medical Radioisotope Scintigraphy, Vienna: collimation with off center focus for simultaneous emis-
IAEA. 1968; 1: 57–69. sion/transmission SPECT in multi-camera systems [abstract].
6. Sorenson JA, Briggs RC, Cameron JR. 99mTc point source for J Nucl Med 1994; 35 (Suppl): 92P.
transmission scanning. J Nucl Med 1969; 10: 252–253. 25. Chang W, Loncaric S, Huang G, Sanpitak P. Asymmetrical fan
7. Tothill P, Galt JM. Quantitative profile scanning for the mea- transmission CT on SPECT systems. Phys Med Biol 1995; 40:
surement of organ radioactivity. Phys Med Biol 1971; 16: 913–928.
625–634. 26. King MA, Luo D, Dahlberg ST, Villegas BJ. Transmission im-
8. Phelps ME, Hoffman EJ, Mullani NA, Ter-Pogossian MM. aging of large attenuators using a slant hole collimator on
Application of annihilation coincidence detection to transaxial a three-headed SPECT system. Med Phys 1996; 23: 263–
reconstruction tomography. J Nucl Med 1975; 16: 210–224. 272.
9. Derenzo SE, Zaklad H, Budinger TF. Analytical study of 27. Tan P, Bailey DL, Meikle SR, Eberl S, Fulton RR, Hutton BF.
a high-resolution positron ring detector system for transaxial A scanning line source for simultaneous emission and trans-
reconstruction tomography. J Nucl Med 1975; 16: 1166– mission measurements in SPECT. J Nucl Med 1993; 34:
1173. 1752–1760.
10. Graham LS, Neil R. In vivo quantitation of radioactivity using 28. Waters SL, Forse GR, Horlock PL, Woods MJ. The half-life of
the anger camera. Radiology 1974; 112: 441–442. 68Ge. Int J Appl Radiat Isot 1981; 32: 757.

11. Myers MJ, Lavender JP, deOliviera JB, Maseri A. A simpli- 29. Casey ME, Nutt R. A multicrystal two dimensional BGO de-
fied method of quantitating organ uptake using a gamma cam- tector system for PET. IEEE Trans Nucl Sci 1986; 33:
era. Br J Radiol 1981; 54: 1062–1067. 460–463.
12. Macey D, Marshall R. Absolute quantitation of radiotracer up- 30.Carroll LR, Kretz P, Orcutt G. The orbiting rod source: improv-
take in lungs using a gamma camera. J Nucl Med 1984; 23: ing performance in PET transmission correction scans. In: Es-
731–735. ser PD, ed. Emission computed tomography – current trends.
13. Budinger TF, Gullberg GT. Transverse section reconstruction New York: Society of Nuclear Medicine; 1983: 235–247.
of gamma-ray emitting radionuclides in patients. In: Ter-Po- 31. Carson RE, Daube-Witherspoon ME, Green MV. A method
gossian MM, Phelps ME, Brownell GL, eds. Reconstruction for postinjection PET transmission measurements with a rotat-
tomography in diagnostic radiology and nuclear medicine. ing source. J Nucl Med 1988; 29: 1558–1567.
Baltimore: University Park Press; 1977: 315–342. 32. Daube-Witherspoon ME, Carson RE, Green MV. Post-injec-
14. Jaszczak RJ, Chang LT, Stein NA. Whole body single photon tion transmission attenuation measurements for PET. IEEE
emission computed tomography using large field of view scin- Trans Nucl Sci 1988; 35: 757–761.
tillation cameras. Phys Med Biol 1979; 24: 1123–1143. 33. Huesman RH, Derenzo SE, Cahoon JL, Geyer AB, Moses

European Journal of Nuclear Medicine Vol. 25, No. 7, July 1998


785

WW, Uber DC, Vuletich T, Budinger TF. Orbiting transmis- 52. Lange K, Carson R. EM reconstruction algorithms for emis-
sion source for positron emission tomography. IEEE Trans sion and transmission tomography. J Comput Assist Tomogr
Nucl Sci 1988; NS-35: 735–739. 1984; 8: 306–316.
34. Ranger NT, Thompson CJ, Evans AC. The application of a 53. Hudson HM, Larkin RS. Accelerated image reconstruction us-
masked orbiting transmission source for attenuation correction ing ordered subsets of projection data. IEEE Trans Med Imag
in PET. J Nucl Med 1989; 30: 1056–1068. 1994; MI-13: 601–609.
35. Meikle SR, Bailey DL, Hooper PK, Eberl S, Hutton BF, Jones 54. Lewellen TK, Miyaoka RS, Kaplan MS, Kohlmyer SK, Costa
WF, Fulham MJ. Simultaneous emission and transmission W, Jansen F. Preliminary investigation of coincidence imaging
measurements for attenuation correction in whole-body PET. J with a standard dual-headed SPECT system [abstract]. J Nucl
Nucl Med 1995; 36: 1680–1688. Med 1995; 36: 175P.
36. Rhodes CG, Wollmer P, Fazio F, Jones T. Quantitative mea- 55. Mukai T, Links JM, Douglass KH, Wagner HN Jr. Scatter cor-
surement of regional extravascular lung density using positron rection in SPECT using non-uniform attenuation data. Phys
emission and transmission tomography. J Comput Assist Tom- Med Biol 1988; 33: 1129–1140.
ogr 1981; 5: 783–791. 56. Ljungberg M, Strand S-E. Attenuation correction in SPECT
37. Townsend DW, Spinks TJ, Jones T, Geissbühler A, Defrise M, based on transmission studies and Monte Carlo simulations of
Gilardi M-C, Heather JD. Three dimensional reconstruction of build-up functions. J Nucl Med 1990; 31: 493–500.
PET data from a multi-ring camera. IEEE Trans Nucl Sci 57. Ljungberg M, Strand S-E. Scatter and attenuation correction
1989; 36: 1056–1065. in SPECT using density maps and Monte Carlo simulated
38. deKemp RA, Nahmias C. Attenuation correction in PET using scatter functions. J Nucl Med 1990; 31: 1560–1567.
single photon transmission measurement. Med Phys 1994; 21: 58. Bailey DL, Hutton BF, Meikle SR, Fulton RR, Jackson CB. It-
771–778. erative scatter correction incorporating attenuation data [ab-
39. Karp JS, Muehllehner G, Qu H, Yan X-H. Singles transmis- stract]. Eur J Nucl Med 1989; 15: 452.
sion in volume-imaging PET with a 137Cs source. Phys Med 59. Bergström M, Eriksson L, Bohm C, Blomqvist G, Litton J-E.
Biol 1995; 40: 929–944. Correction for scattered radiation in a ring detector positron
40. Yu SK, Nahmias C. Single-photon transmission measurements camera by integral transformation of the projections. J Com-
in positron emission tomography using 137Cs. Phys Med Biol put Assist Tomogr 1983; 7: 42–50.
1995; 40: 1255–1266. 60. Axelsson B, Msaki P, Israelsson A. Subtraction of Compton-
41. Jones T, Bailey DL, Bloomfield PM, et al. Performance char- scattered photons in single-photon emission computerized to-
acteristics and novel design aspects of the most sensitive PET mography. J Nucl Med 1984; 25: 490–494.
camera built for high temporal and spatial resolution [ab- 61. Siegel JA, Maurer AH, Wu RK, Blasius KM, Denenberg BS,
stract]. J Nucl Med 1996; 37: 85P. Gash AK, Carabello BA, Spann JF, Malmud LS. Absolute left
42. Bailey DL, Miller MP, Spinks TJ, Bloomfield PM, Livieratos ventricular volume by an iterative build-up factor analysis of
L, Young HE, Jones T. Experience with fully 3D PET and im- gated radionuclide images. Radiology 1984; 151: 477–481.
plications for future high resolution 3D tomographs. Phys 62. Wu RK, Siegal JA. Absolute quantitation of radioactivity us-
Med Biol 1998; 43: 777–786. ing the buildup factor. Med Phys 1984; 11: 189–192.
43. Xu EZ, Mullani NA, Gould KL, Anderson WL. A segmented 63. Siegel JA, Wu RK, Maurer AH. The buildup factor: effect of
attenuation correction for PET. J Nucl Med 1991; 32: scatter on absolute volume determination. J Nucl Med 1985;
161–165. 26: 390–394.
44. Meikle SR, Dahlbom M, Cherry SR. Attenuation correction 64. Meikle SR, Hutton BF, Bailey DL. A transmission dependent
using count-limited transmission data in positron emission to- method for scatter correction in SPECT. J Nucl Med 1994; 35:
mography. J Nucl Med 1993; 34: 143–150. 360–367.
45. Xu M, Luk WK, Cutler PD, Digby WM. Local threshold for 65. Narita Y, Eberl S, Iida H, Hutton BF, Braun M, Nakamura T,
segmented attenuation correction of PET imaging of the tho- Bautovich GJ. Monte Carlo and experimental evaluation of
rax. IEEE Trans Nucl Sci 1994; NS-41: 1532–1537. accuracy and noise properties of two scatter correction meth-
46. Meikle SR, Eberl S, Hooper PK, Fulham MJ. Simultaneous ods for SPECT. Phys Med Biol 1996; 41: 2481–2496.
emission and transmission (SET) scanning in neurological 66. Welch A, Gullberg GT, Christian PE, Datz FL. A transmis-
PET studies. J Comput Assist Tomogr 1997; 21: 487–497. sion-map-based scatter correction technique for SPECT in in-
47. Bailey DL, Young HE, Bloomfield PM, et al. ECAT ART – a homogeneous media. Med Phys 1995; 22: 1627–1635.
continuously rotating PET damera: performance characteris- 67. Ollinger JM, Johns GC. Model-based scatter correction for
tics, comparison with a full ring system, initial clinical stud- fully 3D PET. In: Klaisner L, ed. Proceedings of the 1993
ies, and installation considerations in a nuclear medicine de- IEEE Nuclear Science Symposium and Medical Imaging Con-
partment. Eur J Nucl Med 1997; 24: 6–15. ference, San Francisco, USA. 1993; 3: 1264–1268.
48. Bailey DL, Jones WF, Brun T, Young J, Nutt R. A spiral CT 68. Watson CC, Newport D, Casey ME. A single scatter simula-
approach to recording accurate single photon transmission da- tion technique for scatter correction in 3D PET. In: Grangeat
ta in PET [abstract]. J Nucl Med 1997; 38: 113P. P, Amans J-L, eds. Three-dimensional image reconstruction in
49. Yu SK, Nahmias C. Segmented attenuation correction using radiology and nuclear medicine. Dordrecht: Kluwer Academ-
artificial neural networks in positron tomography. Phys Med ic (Viergever MA, ed. Computational Imaging and Vision;
Biol 1996; 41: 2189–2206. 1996; 4: 255–268).
50. Moore SC. Attenuation compensation. In: Ell PJ, Holman BL, 69. Bailey DL, Robinson M, Meikle SR, Bye PTP. Simultaneous
Computed emission tomography. London: Oxford University emission and transmission measurements as an adjunct to dy-
Press, 1982: 339–360. namic planar gamma camera studies. Eur J Nucl Med 1996;
51. Shepp LA, Vardi Y. Maximum likelihood reconstruction for 23: 326–331.
emission tomography. IEEE Trans Med Imag 1982; MI-1: 70. Smith AM, Bruckbauer T, Wienhard K, Pietrzyk U, Byars
113–122. LG. Spatial transformation during 3D reconstruction in posi-

European Journal of Nuclear Medicine Vol. 25, No. 7, July 1998


786

tron emission tomography. Eur J Nucl Med 1997; 24: 1413– Jones T, eds. Quantification of brain function using PET. San
1417. Diego: Academic Press; 1996: 82–87.
71. Phipps PR, Gonda I, Bailey DL, Borham PW, Bautovich GJ, 87. Townsend DW, Price JC, Mintun MA, Kinahan PE, Jadali F,
Anderson SD. Comparison of planar and tomographic scintig- Sashin D, Simpson N, Mathis CA. Scatter correction for
raphy to measure the penetration index of inhaled aerosols. brain receptor quantitation in 3D PET. In: Myers R, Cun-
Am Rev Respir Dis 1989; 139: 1516–1523. ningham VJ, Bailey DL, Jones T, eds. Quantification of
72. Regnis JA, Robinson M, Bailey DL, Cook P, Hooper P, Chan brain function using PET. San Diego: Academic Press; 1996:
H-K, Gonda I, Bautovich GJ, Bye PTP. Mucociliary clearance 76–81.
in patients with cystic fibrosis and in normal subjects. Am J 88. Trébossen R, Bendriem B, Fontaine A, Frouin V, Remy P.
Respir Crit Care Med 1994; 150: 66–71. Quantitation of the [18F]fluorodopa uptake in the human stri-
73. IIda H, Kanno I, Takahashi A, et al. Measurement of absolute atum in 3D PET with the ETM scatter correction. In: Myers
myocardial blood flow with H215O and dynamic positron- R, Cunningham VJ, Bailey DL, Jones T, eds. Quantification
emission tomography. Strategy for quantification in relation to of brain function using PET. San Diego: Academic Press;
the partial-volume effect. Circulation 1988; 78: 104–115. 1996: 88–92.
74. Iida H, Rhodes CG, de Silva R, Yamamoto Y, Araujo LI, Ma- 89. Badawi RD. 3D-Mode acquisition in clinical PET. Nucl Med
seri A, Jones T. Myocardial tissue fraction-correction for par- Commun 1997; 18: 801–804.
tial volume effects and measure of tissue viability. J Nucl Med 90. Tsui BM, Gullberg GT, Edgerton ER, Ballard JG, Perry JR,
1991; 32: 2169–2175. McCartney WH, Berg J. Correction of non-uniform attenua-
75. Slomka PJ, Hurwitz GA, Stephenson J, Cradduck T. Automat- tion in cardiac SPECT imaging. J Nucl Med 1989; 30:
ed alignment and sizing of myocardial stress and rest scans to 497–507.
three-dimensional normal templates using an image registra- 91. Roach PJ, Hutton BF, Meikle SR, et al. Improved accuracy
tion algorithm. J Nucl Med 1995; 36: 1115–1122. of Tl-201 myocardial scintigraphy using transmission based
76. Talairach J, Tournoux P. Co-planar stereotactic atlas of the quantitative SPECT [abstract]. Aust NZ J Med 1994; 24: 509.
human brain: 3-dimensional proportional system: an ap- 92. Ficaro EP, Fessler JA, Shreve PD, Kritzman JN, Rose PA,
proach to cerebral imaging. Stuttgart: Georg Thieme, 1988. Corbett JA. Diagnostic accuracy of attenuation-corrected
77. Friston KJ, Frith CD, Liddle PF, Frackowiak RSJ. Comparing 99mTc-sestamibi single-photon emission computed tomogra-

functional (PET) images: the assessment of significant phy. Circulation 1996; 93: 463–473.
change. J Cereb Blood Flow Metab 1991; 11: 690–699. 93. Kluge R, Sattler B, Seese A, Knapp WH. Attenuation correc-
78. Friston KJ, Ashburner J, Frith CD, Poline J-B, Heather JD, tion by simultaneous emission-transmission myocardial sin-
Frackowiak RSJ. Spatial registration and normalisation of im- gle-photon emission tomography using a technetium-99m-la-
ages. Human Brain Mapping 1995; 2: 165–189. belled radiotracer: impact on diagnostic accuracy. Eur J Nucl
79. Allain P, Travère JM, Baron JC, Bloyet D. Accurate PET posi- Med 1997; 24: 1107–1114.
tioning with reference to MRI and neuroanatonical data bases. 94. Natterer F. Determination of tissue attenuation in emission
In: Uemura K, Lassen NA, Jones T, Kanno I, eds. Quantifica- tomography of optically dense media. Inverse Problems
tion of brain function: tracer kinetics and image analysis in 1993; 9: 731–736.
brain PET. Amsterdam: Elsevier; 1993: 401–408. 95. Welch A, Clack R, Natterer F, Gullberg GT. Towards accu-
80. Ardekani BA, Braun M, Hutton BF. Improved quantification rate attenuation correction in SPECT without transmission
with the use of anatomical information in PET image recon- measurements. IEEE Trans Med Imag 1997; MI-15: 532–
struction. In: Uemura K, Lassen NA, Jones T, Kanno I, eds. 541.
Quantification of brain function: tracer kinetics and image 96. Fleming JS. A technique for using CT images in attenuation
analysis in brain PET. Amsterdam: Elsevier; 1993: 351– correction and quantification in SPECT. Nucl Med Commun
362. 1989; 10: 83–97.
81. Ouyang X, Wong WH, Johnson VE, Hu X, Chen C-T. Incor- 97. Perring S, Summers Q, Fleming JS, Nassim MA, Holgate
poration of correlated structural images in PET image recon- ST. A new method of quantification of the pulmonary region-
struction. IEEE Trans Med Imag 1994; MI-14: 627–640. al distribution of aerosols using combined CT and SPECT
82. Ardekani BA, Braun M, Hutton BF, Kanno I. Minimum cross- and its application to nedocromil sodium administered by
entropy reconstruction of PET images using prior anatomical metered dose inhaler Br J Radiol 1994; 67: 46–53.
information obtained from MR. In: Myers R, Cunningham VJ, 98. Madsen MT, Kirchner PT, Edlin JP, Nathan MA, Kahn D. An
Bailey D, Jones T, eds. Quantification of brain function using emission-based technique for obtaining attenuation correc-
PET. San Diego: AcademicPress; 1996: 113–117. tion data for myocardial SPECT studies. Nucl Med Commun
83. Iida H, Narita Y, Kado H, Kashikura A, Sugawara S, Shoji Y, 1993; 14: 689–195.
Kinoshita T, Ogawa T, Eberl S. Effects of scatter and attenua- 99. Lang TF, Hasegawa BH, Liew SC, Brown JK, Blankespoor
tion correction on quantitative assessment of regional cerebral SC, Reilly SM, Gingold EL, Cann CE. Description of a pro-
blood flow with SPECT. J Nucl Med 1998; 39: 181–189. totype emission-transmission computed tomography imaging
84. Bailey DL, Jones T. A method for calibrating three-dimen- system J Nucl Med 1992; 30: 1881–1887.
sional positron emission tomography without scatter correc- 100. Kalki K, Blankespoor SC, Brown JK, Hasegawa BH, Dae
tion. Eur J Nucl Med 1997; 24: 660–664. MW, Chin M, Stillson C. Myocardial perfusion imaging with
85. Huang SC, Hoffman EJ, Phelps ME, Kuhl DE. Quantitation a combined x-ray CT and SPECT system. J Nucl Med 1997;
-in positron emission tomography. 2. Effect of inaccurate at- 38: 1535–1540.
tenuation correction. J Comput Assist Tomogr 1979; 3: 804– 101. Beyer T, Kinahan PE, Townsend DW, Sashin D. The use of
814. X-ray CT for attenuation correction of PET data. In: Trendler
86. Rakshi J, Bailey DL, Morrish PK, Brooks DJ. Implementation RC, ed. Proceedings of the 1994 IEEE Nuclear Science Sym-
of 3D acquisition, reconstruction and analysis of dynamic flu- posium and Medical Imaging Conference, Norfolk, VA,
orodopa studies. In: Myers R, Cunningham VJ, Bailey DL, USA. 1994; 4: 1573–1577.

European Journal of Nuclear Medicine Vol. 25, No. 7, July 1998


787

102. Melcher CL, Schweitzer JS. Cerium-doped lutetium oxy- 105. Marsden PK, Shoa Y, Cherry SR, et al. Simultaneous acqui-
orthorthsilicate: a fast, efficient new scintillator. IEEE Trans sition of PET images and NMR spectra in high field magnet.
Nucl Sci 1992; 39: 502–505. J Nucl Med 1997; 38 [5(Suppl)]: 45P.
103. Shao Y, Cherry SR, Farahani K, Meadors K, Siegel S, Silver- 106. Celler A, Sitek A, Stoub E, Lyster D, Dykstra C. Develop-
man RW, Marsden PK. Simultaneous PET and MR imaging. ment of multiple line-source array for SPECT transmission
Phys Med Biol 1997; 42: 1965–1970. scans. J Nucl Med 1997; 38 [5(Suppl)]: 215P–216P.
104. Shao Y, Cherry SR, Farahani K, et al. Development of a PET
detector system compatible with MRI/NMR systems. IEEE
Trans Nucl Sci 1997; NS-44: 1167–1171.

European Journal of Nuclear Medicine Vol. 25, No. 7, July 1998

You might also like