Scatter correction based on an artificial neural network for 99mTc and 123I dual-isotope SPECT in myocardial and brain imaging

ArticleinAnnals of Nuclear Medicine 21(1):25-32 · February 2007with16 Reads
Impact Factor: 1.68 · DOI: 10.1007/BF03033996 · Source: PubMed
Abstract

The aim of this study was to elucidate the clinical usefulness of scatter correction with an artificial neural network (ANN) in 99mTc and 123I dual-isotope SPECT. Two algorithms for ANN scatter correction were tested: ANN-10 and ANN-3 employing 10 and 3 energy windows for data acquisition, respectively. Three patients underwent myocardial or brain SPECT with one of the following combinations of radiopharmaceuticals administered: 99mTc-tetrofosmin and 123I-metaiodobenzylguanidine (MIBG), 99mTc-methoxyisobutylisonitrile (MIBI) and 123I-beta-methyl-paraiodophenyl-pentadecanoic acid (BMIPP), or 99mTc-ethyl-cistainate dimmer (ECD) and 123I-iomazenil. The patients were also referred for single-isotope imaging incorporating conventional triple-energy window (TEW) scatter correction. Crosstalk- and scatter-corrected 99mTc- and 123I-SPECT images in dual-isotope acquisition with ANN were compared with those in single-isotope acquisition. The ANN method well separated 123I and 99mTc primary photons. Although ANN-10 yielded images of poor quality, ANN-3 offered comparable image quality with the single-isotope scan without significant increase of acquisition time. The proposed method is clinically useful because it provides various combinations of information without anatomical misregistration with one acquisition.

Full-text

Available from: Koichi Ogawa
Original Article 25Vol. 21, No. 1, 2007
Annals of Nuclear Medicine Vol. 21, No. 1, 25–32, 2007
ORIGINAL ARTICLE
Received June 19, 2006, revision accepted October 11, 2006.
For reprint contact: Jingming Bai, M.D., Department of
Radiology, School of Medicine, Keio University, Tokyo 160–
8582, JAPAN.
E-mail: bjingming@goo.ne.jp
INTRODUCTION
AN IMPORTANT FEATURE of radionuclide imaging is the
capability of quantitative analysis of regional function of
organs. Quantification with SPECT images is degraded
by many factors such as photon scattering and attenua-
tion, partial volume effect, collimator aperture, detector
response and patient motion.
1
Regarding scatter correc-
tion, various methods have been reported, including the
dual photopeak window (DPW), transmission-dependent
scatter correction (TDSC) and triple energy window (TEW)
methods.
2–4
Some methods are also applicable to crosstalk
correction in simultaneous dual-isotope imaging.
5,6
How-
ever, it is very difficult to compensate for crosstalk be-
tween
99m
Tc and
123
I because the two radionuclides emit
photons whose energies are very close to each other.
We have proposed a scatter correction method with an
artificial neural network (ANN) in single-isotope imag-
ing.
7
This method is applicable to crosstalk correction
between
99m
Tc and
123
I.
8
In addition, other investigators
have also reported a scatter correction method with ANN,
9
and allied it to the same dual-isotope imaging.
10,11
Fur-
thermore, some papers discussed the clinical use of ANN
in dual-isotope imaging.
12,13
However, the above men-
tioned methods have a common disadvantage in practica-
bility because energy windows more than 9 are required
for data acquisition; this window setting is not achievable
in most commercially available SPECT systems. To over-
come this problem, we have developed a scatter correc-
tion method with an ANN employing three energy win-
dows for data acquisition.
14
This energy setting resulted
in reduced image noise without impairing quantitative
Scatter correction based on an artificial neural network for
99m
Tc and
123
I dual-isotope SPECT in myocardial and brain imaging
Jingming BAI,* Jun HASHIMOTO,** Koichi OGAWA,*** Tadaki NAKAHARA,**
Takayuki SUZUKI** and Atsushi KUBO**
*21
st
Century Center of Excellence Program, **Department of Radiology, School of Medicine, Keio University
***Department of Electrical Informatics, Faculty of Engineering, Hosei University
The aim of this study was to elucidate the clinical usefulness of scatter correction with an artificial
neural network (ANN) in
99m
Tc and
123
I dual-isotope SPECT. Methods: Two algorithms for ANN
scatter correction were tested: ANN-10 and ANN-3 employing 10 and 3 energy windows for data
acquisition, respectively. Three patients underwent myocardial or brain SPECT with one of the
following combinations of radiopharmaceuticals administered:
99m
Tc-tetrofosmin and
123
I-
metaiodobenzylguanidine (MIBG),
99m
Tc-methoxyisobutylisonitrile (MIBI) and
123
I-beta-methyl-
paraiodophenyl-pentadecanoic acid (BMIPP), or
99m
Tc-ethyl-cistainate dimmer (ECD) and
123
I-
iomazenil. The patients were also referred for single-isotope imaging incorporating conventional
triple-energy window (TEW) scatter correction. Crosstalk- and scatter-corrected
99m
Tc- and
123
I-
SPECT images in dual-isotope acquisition with ANN were compared with those in single-isotope
acquisition. Results: The ANN method well separated
123
I and
99m
Tc primary photons. Although
ANN-10 yielded images of poor quality, ANN-3 offered comparable image quality with the single-
isotope scan without significant increase of acquisition time. Conclusion: The proposed method is
clinically useful because it provides various combinations of information without anatomical
misregistration with one acquisition.
Key words: artificial neural network, dual-isotope SPECT, scatter correction, brain SPECT,
myocardial SPECT
Page 1
Annals of Nuclear Medicine26 Jingming Bai, Jun Hashimoto, Koichi Ogawa, et al
accuracy in phantom experiments. In the current study,
this three-window ANN method was applied to clinical
myocardial and brain SPECT to elucidate its practicabil-
ity and usefulness.
MATERIALS AND METHODS
Energy window setting and structure of artificial neural
network
We used the following two methods to estimate the counts
of primary photons (C
Tc-prim
for
99m
Tc and C
I-prim
for
123
I)
after
99m
Tc and
123
I dual-isotope acquisition: an ANN
with 10 inputs (ANN-10), and an ANN with 3 inputs
(ANN-3).
The ANN-10 has three layers: one input layer with ten
units, one hidden layer with twenty units and one output
layer with two units (Fig. 1 (A)). For input values, we used
ten ratios of counts determined by the two energy win-
dows each: one was count C
k
acquired with the k-th
narrow window and the other was count C
s
acquired with
the broad window ranging from 120 to 180 keV. The
narrow windows with a width of 6 keV were sequentially
positioned in the energy range from 120 to 180 keV. The
value input into the k-th window was the ratio R
k
(= C
k
/
C
s
). The outputs were two count ratios, R
I/2
and R
I+Tc
.
R
I/2
was the count ratio of half of the primary photons for
123
I to the photons measured at the upper half of the
123
I
photo-peak window (159–180 keV) (C
IH
). R
I+Tc
was the
count ratio of the sum of the primary photons for
99m
Tc
and
123
I to the total photons C
s
. With the calculated ratio
R
I/2
on the basis of the neural network and the measured
count C
IH
, we estimated the primary count C
I-prim
for
123
I
at each pixel in a planar image.
C
I-prim
= 2 * C
IH
* R
I/2
With the calculated ratio R
I+Tc
and the measured count C
s
,
we calculated the primary count C
Tc-prim
for
99m
Tc at each
pixel in the planar image.
C
Tc-prim
= C
s
* R
I+Tc
C
I-prim
In ANN-3, three energy windows (from 90 to 119 keV,
from 120 to 150 keV, and from 151 to 183 keV, respec-
tively) were set for scatter count (C
scat
),
99m
Tc count (C
Tc
)
and
123
I count (C
I
), respectively (Fig. 1 (B)). This ANN
Fig. 1 Structures of ANN-10 (A) and ANN-3 (B) algorithms designed for separating
99m
Tc and
123
I
primary photons in dual-isotope SPECT imaging.
AB
Page 2
Original Article 27Vol. 21, No. 1, 2007
comprises also three layers: one input layer with three
units, one hidden layer with six units and one output layer
with two units (Fig. 1). The inputs in the ANN-3 are three
count ratios (R
scat
, R
Tc
and R
I
) which were calculated from
the above counts acquired with three energy windows and
the total count (C
s
) in a broad window (90–183 keV), i.e.,
R
scat
= C
scat
/C
s
, R
Tc
= C
Tc
/C
s
and R
I
= C
I
/C
s
. The output
layer provides the ratios of estimated primary to total
photons for
99m
Tc (R
Tc-prim
) and
123
I (R
I-prim
) in each pixel,
respectively. The primary counts for
99m
Tc (C
Tc-prim
) and
123
I (C
I-prim
) at each pixel in the planar image were
obtained as follows:
C
Tc-prim
= C
s
* R
Tc-prim
C
I-prim
= C
s
* R
I-prim
Both ANN-10 and ANN-3 were designed and trained
with a back-propagation algorithm.
7,8,14
Several energy
spectra generated by the Monte Carlo method were used
as a training data set.
7,8,14
The data were acquired simultaneously with spectrum
SPECT acquisition employing 125 energy channels from
70 to 210 keV. After acquisition, the spectrum data were
rearranged to 3 and 10 windows for ANN-3 and ANN-10,
respectively. In ANN-10, to eliminate the scattered pho-
tons from the high energy peak (529 keV) of
123
I, the mean
count ranging from 180 to 210 keV was subtracted from
the measured total counts in each pixel. In ANN-3, as the
count in the energy window from 90 to 119 keV contains
the scatter fraction from the above high energy peak, no
additional subtraction of photon scattering was performed.
Patients
Three patients were enrolled in the present study. In-
formed consent was obtained from each patient after a
detailed explanation of the study. The form to obtain
consent was approved by the institutional committee of
Keio University Hospital.
The first patient had Parkinson’s disease without a
history of myocardial infarction. The patient underwent
both
123
I-metaiodobenzylguanidine (MIBG, Daiichi Ra-
dioisotope Labs., Tokyo, Japan) and
99m
Tc-tetrofosmin
(Nihon Medi-Physics plc. Tokyo, Japan) SPECT at rest.
The second was a patient with myocardial infarction
referred for
123
I-beta-methyl-paraiodophenylpenta-
decanoic acid (BMIPP, Nihon Medi-Physics plc. Tokyo,
Japan) and
99m
Tc-methoxyisobutylisonitrile (MIBI,
Daiichi Radioisotope Labs., Tokyo, Japan) SPECT. Coro-
nary angiography revealed stenosis in the circumflex
artery.
The third was a patient with epilepsy referred for
123
I-
iomazenil (Nihon Medi-Physics plc. Tokyo, Japan) and
99m
Tc-ethyl-cistainate dimmer (ECD, Daiichi Radioiso-
tope Labs., Tokyo, Japan) SPECT.
Protocol
Imaging sequences
Figure 2 shows the examination protocols of the three
patients. The first patient was injected with 111 MBq of
123
I-MIBG at rest and SPECT imaging was performed 3
hours after injection. Soon after the single-isotope imag-
ing, 555 MBq of
99m
Tc-tetrofosmin was administered at
rest followed by simultaneous
123
I-MIBG and
99m
Tc-
tetrofosmin SPECT.
The second patient was injected with
123
I-BMIPP at
rest and SPECT imaging was performed 15 minutes after
injection. After completing BMIPP SPECT,
99m
Tc-MIBI
was administered to obtain simultaneous BMIPP and
MIBI SPECT. The patient also underwent MIBI SPECT
Fig. 2 Imaging protocols of the three patients.
Page 3
Annals of Nuclear Medicine28 Jingming Bai, Jun Hashimoto, Koichi Ogawa, et al
two days after.
The last patient was injected with 222 MBq of
123
I-
iomazenil at rest and SPECT imaging was performed 3
hours after injection. After completing iomazenil SPECT,
555 MBq of
99m
Tc-ECD was administered to obtain
simultaneous iomazenil and ECD SPECT. The patient
also underwent ECD SPECT two days after.
Instrumentation, data acquisition and image processing
A three-headed rotating gamma camera Toshiba GCA-
9300A/DI with low-energy high-resolution parallel-hole
collimators was used for data acquisition, and image
processors DELL DIMENSION 8300 and Toshiba
GMS5500U/DI were employed for image processing.
Single-isotope acquisition in each case was carried out
with the TEW scatter correction method employing a 20%
main photopeak window and two 7% subwindows adja-
cent to the main window.
3
In the projection data, the
scatter fraction was estimated by the area approximation
based on the counts of the subwindows, and it was
removed from the counts in the main window to obtain the
counts of primary photons.
3
The gamma camera rotated
continuously for 16 minutes in each acquisition. SPECT
data were arranged into 60 projections over 360 degrees.
In dual-isotope SPECT (ANN), 60 projections over
360 degrees (stepped by 6 degrees) were obtained. Data
acquisition time for each projection angle was 50 seconds
and total acquisition time was 19 minutes.
The counts of primary photons separated by the ANN-
3 and ANN-10 from the acquired dual-isotope projection
data were reconstructed into images of a 64 × 64 matrix
using an OS-EM method with 5 iterations and 10 subsets.
Attenuation correction was not conducted to either the
dual-isotope or single-isotope imaging data.
Lesion-to-normal count ratios (mean count in the re-
gion-of-interest placed on the lateral wall divided by that
in the septum) were obtained from the single- and dual-
isotope images of the second patient undergoing
123
I-
BMIPP and
99m
Tc-MIBI SPECT.
RESULTS
Figure 3 shows
123
I-MIBG and
99m
Tc-tetrofosmin
transaxial SPECT images obtained with single- and dual-
isotope acquisition. Dual-isotope imaging showed no
cardiac uptake of
123
I-MIBG and well preserved uptake of
99m
Tc-tetrofosmin. Single-isotope imaging using
123
I-
MIBG also manifested no cardiac uptake. Compared with
ANN-10, ANN-3 offered low-noise image both in MIBG
and tetrofosmin SPECT. ANN-3 provided images of
comparable quality with single-isotope imaging with TEW.
Figure 4 shows
99m
Tc-MIBI and
123
I-BMIPP transaxial
SPECT images. A mismatch in the lateral wall can be
easily recognized between myocardial perfusion and fatty
acid metabolism (preserved perfusion and impaired me-
tabolism) in ANN-3 and single-isotope (arrow). In ANN-
3, findings in single- and dual-isotope imaging are almost
the same. On the other hand, ANN-10 failed to visualize
Fig. 3 Transaxial SPECT images of
99m
Tc-tetrofosmin and
123
I-MIBG in single- and dual-isotope
acquisition. ANN: dual-isotope imaging using artificial neural network with 10 inputs (ANN-10) and
3 inputs (ANN-3); Tc:
99m
Tc-tetrofosmin; I:
123
I-MIBG; Single-I: single-isotope imaging with
123
I-
MIBG.
Page 4
Original Article 29Vol. 21, No. 1, 2007
Fig. 5 Transaxial SPECT images of
99m
Tc-ECD and
123
I-iomazenil in single- and dual-isotope
acquisition. ANN: dual-isotope imaging using artificial neural network with 10 inputs (ANN-10) and
3 inputs (ANN-3); Tc:
99m
Tc-ECD; I:
123
I-iomazenil; Single-Tc: single
99m
Tc-ECD imaging; Single-I:
single
123
I-iomazenil imaging.
Fig. 4 Transaxial SPECT images of
99m
Tc-MIBI and
123
I-BMIPP in single- and dual-isotope
acquisition. ANN: dual-isotope imaging using artificial neural network with 10 inputs (ANN-10) and
3 inputs (ANN-3); Tc:
99m
Tc-MIBI; I:
123
I-BMIPP; Single-Tc: single
99m
Tc-MIBI SPECT; Single-I:
single
123
I-BMIPP SPECT; arrow: lesion site; arrowhead: biliary discharge of MIBI.
Page 5
Annals of Nuclear Medicine30 Jingming Bai, Jun Hashimoto, Koichi Ogawa, et al
the mismatch clearly because of inferior image quality of
both SPECT images. The lateral-to-septal count ratios
were 0.56 and 0.46 in the single
99m
Tc-MIBI and single
123
I-BMIPP SPECT images, respectively. The count ra-
tios were 0.63 (
99m
Tc) and 0.46 (
123
I) in the ANN-3
images, and 0.74 (
99m
Tc) and 0.60 (
123
I) in the ANN-10
images.
Figure 5 includes
123
I-iomazenil and
99m
Tc-ECD
transaxial SPECT images. In dual-isotope imaging with
ANN-3, both
123
I-iomazenil and
99m
Tc-ECD SPECT
showed normal uptake in the cerebral cortex. On the other
hand, marked differences in the uptake were observed in
the basal ganglia and thalamus: preserved uptake in
99m
Tc-
ECD SPECT and almost no uptake in
123
I-iomazenil
SPECT. Compared with ANN-3, ANN-10 yielded infe-
rior image quality especially in
99m
Tc-ECD transaxial
SPECT. Although image quality in the basal ganglia and
thalamus in ECD SPECT of ANN-3 is inferior to that of
single-I-123 imaging, visualization of the cerebral cortex
is of acceptable quality in ANN-3.
DISCUSSION
Dual-isotope imaging has two main advantages. The first
is that it provides two different kinds of information with
one scan and accordingly reduces examination time and
burden to patients and medical staff. The second is that
there is no problem of anatomical misregistration be-
tween the two images. In contrast, the crosstalk hampers
the use of two radionuclides whose photon energies are
close to each other. Thus, most clinical SPECT protocols
employ the combination of
201
Tl and
99m
Tc, or
201
Tl and
123
I. Since
201
Tl cannot be used to label compounds,
information obtained from it is limited to myocardial
perfusion or tumor cell viability. On the other hand,
99m
Tc
and
123
I are available for labeling various compounds, and
dual-isotope imaging with
99m
Tc and
123
I engenders a vast
array of information. Therefore, we contrived a scatter
correction method (ANN) enabling
99m
Tc and
123
I dual-
isotope acquisition and applied it to myocardial and brain
SPECT.
When we first developed a scatter correction method
with ANN (ANN-10 in the present study), it was used to
compensate for scattered photons in
99m
Tc single-isotope
acquisition.
7
However, other easier methods such as TEW
can eliminate scattered photons in single-isotope acquisi-
tion or dual-isotope acquisition with
201
Tl and
99m
Tc, or
201
Tl and
123
I.
2–4
On the other hand, a distinguishing
feature of the method with ANN is its applicability in
99m
Tc and
123
I dual-isotope imaging.
8,14
Since a series of
simulation and phantom experiments showed its accept-
able image quality and quantitative accuracy,
7,8,14
clinical
myocardial and brain SPECT imaging was conducted in
the present study. We compared the performance of the
original method (ANN-10) and its revised version (ANN-
3).
Regarding the difference between ANN-10 and ANN-
3, ANN-3 has some advantages compared with its coun-
terparts. First, the reduced number of energy windows in
ANN-3 results in a reduced amount of acquired data and
shortened time for data processing. Second, increasing
the widths of energy windows offers easy uniformity
maintenance of the camera system and increased signal-
to-noise ratios leading to improved image quality. The
images obtained in the three cases clearly reveal the
difference in image quality between ANN-3 and ANN-
10. Third, improved accuracy in quantifying myocardial
uptake can be expected. Quantitative analysis of the
severity of the infarcted myocardium in the second case
shows the superiority of ANN-3 over ANN-10. Finally,
routine clinical application of ANN-10 cannot be ex-
pected yet because commercially available SPECT cam-
eras do not fit the setting of so many energy windows.
Therefore, ANN-3 has considerable superiority over ANN-
10 in clinical practice.
In comparing ANN-3 and single-isotope acquisition,
acquisition time of dual-isotope imaging is comparable
(19 minutes for ANN-3 and 16 minutes for single acqui-
sition) and image quality obtained by ANN-3 is also
acceptable. Therefore, ANN-3 has clinical feasibility for
dual-isotope imaging.
The first patient underwent simultaneous
123
I-MIBG
and
99m
Tc-tetrofosmin SPECT. The patient contracted
Parkinson’s disease without a history of cardiac disease,
and it is well known that patients with Parkinson’s disease
often manifest no cardiac uptake or severely reduced
uptake of MIBG despite the absence of cardiac disease
like this case.
15,16
In such cases, concurrent myocardial
perfusion imaging is useful to judge whether this reduced
uptake is caused by Parkinson’s disease itself or cardiac
disease. Simultaneous acquisition reduces the burden to
patients especially when the patient suffers from parkin-
sonism making it difficult to keep still for a long time
during data acquisition.
The second patient was referred for simultaneous
123
I-
BMIPP and
99m
Tc-MIBI SPECT. The discrepancy be-
tween BMIPP and perfusion images (so-called perfusion-
metabolism mismatch) in patients with unstable angina or
myocardial infarction suggests the metabolic switch from
fatty acid metabolism to glycolysis in the myocardium.
17,18
It is reported that assessing this mismatch is of clinical
value because it is related to reversibility of impaired
cardiac function and the occurrence of cardiac events.
19
By using
99m
Tc instead of
201
Tl, it would be easier to
estimate the metabolism-perfusion mismatch, because
99m
Tc is less susceptible to photon attenuation and allows
images of excellent quality.
Dual-isotope brain SPECT with
123
I-iomazenil and
99m
Tc-ECD was performed on the third patient suspected
of having epilepsy. Iodine-123-iomazenil is a benzodiaz-
epine receptor imaging agent and the benzodiazepine
receptor density decreases in epileptic foci.
20,21
It has
Page 6
Original Article 31Vol. 21, No. 1, 2007
been confirmed that the distribution of benzodiazepine
receptors is ubiquitous in cerebral cortex and that the
receptor density is very low in basal ganglia and thala-
mus.
22,23
The images obtained in this patient are compat-
ible with the above fact. Although SPECT did not reveal
the epileptic focus in this patient, the dual-isotope imag-
ing with
123
I-iomazenil and
99m
Tc-ECD may be useful in
diagnosing and treating patients with epilepsy because it
provides information about the cause of receptor loss
(associated cerebral infarction and so on). In addition,
when using this protocol in patients with cerebrovascular
disease, it would reveal the accurate sites of ischemic
penumbra because of the lack of no anatomical misregis-
tration.
24,25
The crosstalk between two radionuclides in dual-iso-
tope imaging depends on the ratio of the administered
doses. In the present study, we injected 111 MBq or 222
MBq of
123
I and 555 MBq of
99m
Tc, which are standard
doses used in the clinical setting. However, further study
is needed to elucidate the feasibility of ANN under ex-
traordinary combinations of injected doses for employing
newly developed radiopharmaceuticals or imaging proto-
cols.
CONCLUSION
We applied scatter correction with an artificial neural
network to
123
I and
99m
Tc dual-isotope SPECT in clinical
myocardial and brain imaging. The correction method
(ANN-3) well separated
123
I and
99m
Tc primary photons
and yielded images of acceptable quality without a sig-
nificant increase of the acquisition time. The proposed
method provides various combinations of information
through simultaneous acquisition of
123
I and
99m
Tc, both
of which can label various compounds.
ACKNOWLEDGMENTS
This research is partially supported by the Ministry of Educa-
tion, Culture, Sports, Science and Technology, Grant-in-Aid for
the 21
st
Century Center of Excellence (COE) Program entitled
“Basic study and clinical application of human stem cell biology
and immunology: Approaches based on the development of
experimental animal models” (Keio University). This study is
also partially supported by Grant-in-Aid for Scientific Research
(15591302 and 18790912) from the Ministry of Education,
Culture, Sports, Science and Technology.
REFERENCES
1. Ogawa K. Image distortion and correction in single photon
emission CT. Ann Nucl Med 2004; 18: 171–185.
2. King MA, Hademenos GJ, Glick SJ. A dual-photonpeak
window method for scatter correction. J Nucl Med 1992; 33:
605–612.
3. Ogawa K, Harata Y, Ichihara T, Kubo A, Hashimoto S. A
practical method for position-dependent Compton-scatter
correction in single photon emission CT. IEEE Trans Med
Imag 1993; 10: 408–412.
4. Meikle SR, Hutton BF, Bailey DL. A transmission-depen-
dent method for scatter correction in SPECT. J Nucl Med
1994; 35: 360–367.
5. Ichihara T, Ogawa K, Motomura N, Kubo A, Hashimoto S.
Compton scatter compensation using the triple-energy win-
dow method for single- and dual-isotope SPECT. J Nucl
Med 1993; 34: 2216–2221.
6. Tsuji A, Kojima A, Matsumoto M, Oyama Y, Tomiguchi S,
Kira T, et al. A new method for crosstalk correction in
simultaneous dual-isotope myocardial imaging with Tl-201
and I-123. Ann Nucl Med 1999; 13: 317–323.
7. Ogawa K, Nishizaki N. Accurate scatter compensation
using neural networks in radionuclide imaging. IEEE Trans
Nucl Sci 1993; 40: 1020–1025.
8. Ishii M, Ogawa K, Nakahara T, Hashimoto J, Kubo A.
Quantification of
123
I and
99m
Tc in dual-isotope SPECT
with an artificial neural network. Med Imag Tech 2004; 22:
155–163.
9. Maksud P, Fertil B, Rica C, Fakhri GE, Aurengo A. Artificial
neural network as a tool to compensate for scatter and
attenuation in radionuclide imaging. J Nucl Med 1998; 39:
735–745.
10. Fakhri GE, Maksud P, Kijewski MF, Haber MO, Todd-
Pokropek A, Aurengo A, et al. Scatter and cross-talk correc-
tions in simultaneous
99m
Tc/
123
I brain SPECT using con-
strained factor analysis and artificial neural networks. IEEE
Trans Nucl Sci 2000; 47: 1573–1580.
11. Fakhri GE, Moore SC, Maksud P, Aurengo A, Fijewski MF.
Absolute activity quantitation in simultaneous
123
I/
99m
Tc
Brain SPECT. J Nucl Med 2001; 42: 300–308.
12. Zheng XM, Zubal IG, Seibyl JP, King MA. Correction for
scatter and cross-talk contamination in dual radionuclide
Tc-99m and I-123 images using an artificial neural network.
IEEE Trans Nucl Sci 2004; 51: 2649–2653.
13. El Fakhri G, Habert MO, Maksud P, Kas A, Malek Z,
Kijewski MF, et al. Quantitative simultaneous (
99m
)Tc-
ECD/
123
I-FP-CIT SPECT in Parkinson’s disease and mul-
tiple system atrophy. Eur J Nucl Med Imaging 2006; 33: 87–
92.
14. Ogawa K, Suzawa K. Quantification of two radionuclides in
simultaneous
123
I/
99m
Tc SPECT with artificial neural net-
works. IEEE Nuclear Science Symposium and Medical
Imaging Conference 2004; 6: 3689–3693.
15. Satoh A, Serita T, Seto M, Tomita I, Satoh H, Iwanaga K,
et al. Loss of
123
I-MIBG uptake by the heart in Parkinson’s
Disease: Assessment of cardiac sympathetic denervation
and diagnostic value. J Nucl Med 1999; 40: 371–375.
16. Orimo S, Amino T, Ozawa E, Kojo T, Uchihara T, Takahashi
A, et al. A useful marker for differential diagnosis of
Parkinson’s disease—MIBG myocardial scintigraphy.
Rinsho Shinkeigaku 2004; 44: 827–829.
17. Opie LH, Owen P, Riemersma RA. Relative rates of glucose
and free fatty acids by ischemic and nonischemic myocar-
dium after coronary artery ligation in the dog. Eur J Clin
Invest 1973; 3: 419–435.
18. Tadamura E, Tamaki N, Kudoh T, Hattori N, Konishi J.
BMIPP compared with PET metabolism. Int J Card Imag-
ing 1999; 15: 61–69.
19. Taki J, Nakajima K, Matsunari I, Bunko H, Takata S,
Page 7
Annals of Nuclear Medicine32 Jingming Bai, Jun Hashimoto, Koichi Ogawa, et al
Kawasuji M, et al. Assessment of improvement of myocar-
dial fatty acid uptake and function after revascularization
using iodine-123-BMIPP. J Nucl Med 1997; 38: 1503–
1510.
20. Savic I, Widen L, Thorell JO, Blomqvist G, Ericson K,
Roland P. Cortical benzodiazepine receptor binding in
patients with generalized and partial epilepsy. Epilepsia
1990; 31: 724–730.
21. Savic I, Pauli S, Thorell JO, Blomqvist G. In vivo demon-
stration of altered benzodiazepine receptor density in pa-
tients with generalized epilepsy. J Neurol Neurosurg Psy-
chiatry 1994; 57: 797–804.
22. Shinotoh H, Yamasaki T, Inoue O, Itoh T, Suzuki K,
Hashimoto K, et al. Visualization of specific binding sites of
benzodiazepam in human brain. J Nucl Med 1986; 27:
1593–1599.
23. Woods SW, Seibyl JB, Goddard AW, Dey HM, Zoghbi SS,
Germine M, et al. Dynamic SPECT imaging after injection
of the benzodiazepine receptor ligand [
123
I]iomazenil in
healthy human subjects. Psychiatry Res 1992; 45: 67–77.
24. Hatazawa J, Satoh T, Shimosegawa E, Okudera T, Inugami
A, Ogawa T, et al. Evaluation of cerebral infarction with
iodine 123-iomazenil SPECT. J Nucl Med 1995; 36: 2154–
2161.
25. Hashimoto J, Sasaki T, Itoh Y, Nakamura K, Kubo A,
Amano T, et al. Brain SPECT imaging using three different
tracers in subacute cerebral infarction. Clin Nucl Med 1998;
23: 275–277.
Page 8
    • "In published 123I-MIBG studies, we could divide into two groups of values of H/M, although the precise collimator information was not available from all studies. In one group, the delayed H/M ratio ranged from 2.1 to 2.4, using LEHR or LEGP collimators [7,14,24,25]. Another group showed a comparatively higher delayed H/M ratio, from 2.8 to 3.0. "
    [Show abstract] [Hide abstract] ABSTRACT: Although the heart-to-mediastinum (H/M) ratio in a planar image has been used for practical quantification in (123)I-metaiodobenzylguanidine (MIBG) imaging, standardization of the parameter is not yet established. We hypothesized that the value of the H/M ratio could be standardized to the various camera-collimator combinations. Standard phantoms consisting of the heart and mediastinum were made. A low-energy high-resolution (LEHR) collimator and a medium-energy (ME) collimator were used. We examined multi-window correction methods with (123)I- dual-window (IDW) acquisition, and planar images were obtained with IDW correction and the LEHR collimator. The images were obtained using the following gamma camera systems: GCA 9300A (Toshiba, Tokyo), E.CAM Signature (Toshiba/Siemens, Tokyo) and Varicam (GE, Tokyo). Cardiac phantom studies demonstrated that contamination of the H/M count ratio was greater with the LEHR collimator and least with the ME collimator. The corrected H/M ratio with the LEHR collimator was similar to that with ME collimators. The uncorrected H/M ratio with the ME collimator was linearly related to the H/M ratio with IDW correction with the LEHR collimator. The relationship between the uncorrected H/M ratios determined with the LEHR (E.CAM) and the ME collimators was y = 0.56x + 0.49, where y = H/M ratio with the E.CAM and x = H/M ratio with the ME collimator. The average normal values for the low-energy collimator (n=18) were 2.2+/-0.2 (initial H/M ratio) and 2.42+/-0.2 (delayed H/M ratio), and for the low/medium-energy (LME) collimator (n=14) were 2.63+/-0.25 (initial H/M ratio) and 2.87+/-0.19 (delayed H/M ratio). H/M ratios in previous clinical studies using LEHR collimators are comparable to those with ME collimators. The IDW-corrected H/M ratios determined with the LEHR collimator were similar to those determined with the ME collimator. This finding could make it possible to standardize the H/M ratio in planar imaging among various collimators in the clinical setting.
    Full-text · Article · Dec 2008 · European Journal of Nuclear Medicine
  • [Show abstract] [Hide abstract] ABSTRACT: Radionuclide imaging has the potential to be used in quantitative analysis of the regional function of organs. However, quantification of SPECT images is degraded by many factors such as Compton photon scattering. This could have a destructive effect on clinical reports so it is important to do scatter correction to get better quality SPECT images. We intended to determine how scatter correction with the TEW method can help physicians who look at heart SPECT images, get better reports. This study used the TEW method for scatter correction, which was proposed by Ogawa et al.,(9) using the two narrow windows on either side of the photopeak (20% down and 20% up of the photopeak respectively). Injection of radiopharmaceutical 99mTc was used for medical imaging. In the Shariati Hospital, Tehran, we studied a total of 80 patients with heart disease indications (43 men and 37 women) over the ages of 30-80 years. Contrast and sharpness were considerably improved after scatter correction so physicians could look at defects better. In a few cases scatter correction changed heart defect reports to normal. Using TEW, sensitivity and specificity increased from 86% to 94% and from 61% to 84% respectively. This method was simple to use in clinics.
    Preview · Article · Feb 2008 · Journal of Applied Clinical Medical Physics
  • [Show abstract] [Hide abstract] ABSTRACT: Simultaneous multi-isotope SPECT imaging has a number of applications, for example, cardiac, brain and cancer imaging. The major concern in simultaneous multi-isotope is the significant crosstalk contamination between the different isotopes used. The current study focuses on a method of crosstalk compensation between two isotopes in simultaneous dual isotope SPECT acquisition applied to cancer imaging using 99mTc/111In. Monte Carlo (MC), which is thought to offer the most realistic crosstalk and scatter compensation modeling, in typical implementations, has inherent long calculation times (often several hours or days) associated with it. This makes MC unsuitable for clinical applications. We have previously incorporated convolution based forced detection into SIMIND Monte Carlo program which have made MC feasible to use in clinical time frames. In order to evaluate the accuracy of our accelerated MC program a number of point source simulation results were compared to experimentally acquired data in terms of spatial resolution and detector sensitivity. We have developed an iterative MC-based image reconstruction technique that simulates the photon downscatter from one isotope into the acquisition window of a second isotope. The MC based estimation of scatter contamination contained in projection views is then used to compensate for the photon contamination during iterative reconstruction. We use a modified ordered subset-expectation maximization (OS-EM), named as simultaneous ordered subset-expectation maximization (Sim-OSEM), to perform this step. We have undertaken a number of simulation tests and phantom studies to verify this approach. The proposed reconstruction technique also evaluated by reconstruction of experimentally acquired projection phantom data. Reconstruction using Sim-OSEM showed very promising results in terms of crosstalk and scatter compensation and uniformity of background compared to analytical attenuation based reconstruc- ion after triple energy window (TEW) based scatter correction of projection data. In our case images obtained using Sim-OSEM showed better scatter compensation and more uniform background when compared to the images reconstructed for separately acquired projection data using analytical attenuation based reconstruction.
    No preview · Conference Paper · Oct 2011
Show more