LABORATORY INVESTIGATION
MYOCARDIAL INFARCTION
Quantification of myocardial infarct size by
thallium-201 single-photon emission computed
tomography: experimental validation in the dog
FLORENCE PRIGENT, M.D.,* JAMSHID MADDAHI, M.D., ERNEST V. GARCIA, PH.D. ,*
YUICHI SATOH, M.D., KENNETH VAN TRAIN, B.S., AND DANIEL S. BERMAN, M.D.
Downloaded from http://circ.ahajournals.org/ by guest on June 15, 2017
ABSTRACT To evaluate the potential advantages of thallium-201 (201T1) single-photon emission
computerized tomography (SPECT) to assess myocardial infarct size in the experimental animal, six
normal dogs and 14 dogs with 6 to 8 hr closed-chest coronary occlusion (eight left anterior descending
and six left circumflex) were studied. Ten minutes after intravenous administration of 2 mCi of 201T1,
30 projections were obtained over 180 degrees. The dogs were killed and their hearts sliced and stained
by triphenyl tetrazolium chloride (TTC). Pathologic infarct size was calculated for each slice and for
the entire left ventricular myocardium as percent weight. Tomograms were quantified by automatically
generating maximum-count circumferential profiles, which were compared with normal limit profiles
derived from the six normal dogs. Tomographic infarct size was defined as the percentage of circumferential points falling below normal for each tomogram. SPECT and TTC infarct size on 71 slices
correlated highly (mean SD 27.9 ± 23.4% and 26.7 25.3%, respectively; r = .93, p < .001,
SEE = 9.4%). To determine SPECT infarct size as percent total left ventricular myocardial weight,
infarct sizes from each slice were added to one another after each was multiplied by a coefficient that
reflected the contribution of that slice to the total left ventricular weight. SPECT and TTC infarct size
for the entire left ventricle correlated closely (mean ± SD 20.5 7.6% and 19.3 8.3%, respectively; r = .86, p < .001, SEE = 4.5%). It is concluded that 201T1 SPECT is a valid method for the
noninvasive assessment of experimental myocardial infarct size.
Circulation 74, No. 4, 852-861, 1986.
±
PROGNOSIS in acute myocardial infarction has been
shown to be related to the extent of the necrotic myocardium.1-t Several noninvasive, nonimaging,6-8 and
imaging5' 121 methods have been developed for measurement of infarct size. With planar imaging or limited-angle tomography,9-13 thallium-201 (201T1) myocardial perfusion and technetium-99m-pyrophosphate
infarct-avid scintigraphy are suboptimal for quantitaFrom the Department of Medicine, Division of Cardiology, and the
Department of Nuclear Medicine, Cedars-Sinai Medical Center, and the
Department of Medicine, University of California, Los Angeles, School
of Medicine, Los Angeles.
Supported in part by NIH SCOR grant 17651 and grants from the
American Heart Association, Greater Los Angeles Affiliate.
Address for correspondence: Jamshid Maddahi, M.D., Division of
Cardiology, Cedars-Sinai Medical Center, P.O. Box 48750, Los Angeles, CA 90048.
Received May 6, 1985; revision accepted July 3, 1986.
Presented in part at the 57th Scientific Sessions of the American Heart
Association, Miami Beach, November 1984.
*Present address: Miami University, Miami, FL.
**Present address: Emory University, Atlanta, GA.
852
±
tion of infarct size because of overlap of normal and
abnormal regions. Single-photon emission computerized rotational tomography (SPECT) has a theoretical
advantage over planar imaging and limited-angle tomography for the accurate assessment of infarct size
because of 180 or 360 degree angular sampling of the
myocardium, which results in reduced overlap and
improved contrast of the images 22, 23 Although quantification of infarct size with SPECT using 20'T1 has been
shown to be feasible,14-18 these previous studies used
subjective measurements or computerized planimetric
approaches that were dependent on detection of myocardial edges of the normal and nonvisualized regions
(with severe perfusion defects).
In this study we sought (1) to develop an objective,
computerized method for quantification of infarct size
using maximum counts circumferential profile analysis, independent of edge detection, and (2) to validate
the method in a living canine closed-chest preparation.
CIRCULATION
LABORATORY INVESTIGATION-MYOCARDIAL INFARCTION
Methods
Downloaded from http://circ.ahajournals.org/ by guest on June 15, 2017
Experimental protocol. Twenty adult mongrel dogs (15.5 to
32 kg) were studied, six controls and 14 with induced acute
myocardial infarction. In all dogs, anesthesia was induced with
intravenous administration of sodium pentobarbital (25 mg/kg)
and was maintained with supplemental intravenous doses thereafter. No intubation or artificial ventilation was needed except
in two dogs that had large infarctions. In the 14 dogs assigned to
the infarction study, the protocol was as follows: Femoral arterial pressure and a peripheral limb electrocardiogram were monitored continuously. Heparin (100 IU/kg) was administered
hourly to ensure anticoagulation during the study period. The
left common carotid artery was exposed and a modified No. 7F
Judkins angiographic catheter with an indwelling inflatable balloon-tipped No. 2F Fogarty catheter was introduced into the
carotid artery. The angiographic catheter was advanced under
fluoroscopic control into the ostium of the left anterior descending coronary artery (LAD) in eight dogs or the left circumflex
artery (LCX) in 'six dogs. The balloon-tipped catheter was then
passed through the angiographic catheter and positioned in the
coronary artery at different levels from the ostium. The coronary artery was then occluded by inflating the balloon with
contrast material. The occlusion was maintained until the dog
was killed. Imaging started 6 to 8 hr after the initiation of
occlusion. The imaging protocol was the same for the control
dogs and the dogs with infarction: The animals were put on the
imaging table in the right anterior oblique position. Two millicuries of 201T1 was injected intravenously and tomography was
performed in vivo 10 min after injection. A large-field-of-view
camera (Siemens Rota for 18 dogs and Siemens Orbiter for two
dogs) equipped with 75 photomultiplier tubes, a 1/4 inch thick
Nal (Tl) crystal, and an all-purpose parallel-hole collimator was
used. Twenty percent and 10% energy windows were used,
positioned on the 68 to 80 keV and 160 keV photopeaks, respectively. The collimator rotated 180 degrees around the dog's
chest from the left lateral to the right lateral position, obtaining
30 projections spaced by 6 degrees. Each projection took 30 sec
to acquire. Three-hour delayed images were also acquired in
five of six controls and in 10 of 14 dogs with infarction. The
data were stored in a 64 x 64 x 16 bit matrix. The extrinsic full
width at half maximum measured with a line source at 8 inches
was 18 mm for 201T1. The pixel size was 0.434 x 0.434 cm.
Immediately after imaging, all 20 dogs were killed by intravenous injection of potassium chloride and the hearts were
excised and cut perpendicular to the long axis of the left ventricle into either 6 mm (78 of the total of 142 slices) or 10 mm (64
slices) thick slices from apex to base. The slices were incubated
for 10 min in triphenyl tetrazolium chloride (TTC). TTC was
used for the pathologic measurement because several studies
have demonstrated the validity of TTC staining for determination of myocardial inefarct size in dogs with 6 to 8 hr of closedchest coronary occlusion or open-chest permanent ligation.24 25
Determination of infarct size from TTC staining. The
UTC-defined infarct area was planimetered on the basal side of
each myocardial slice by an experienced technologist unaware
of the scintigraphic findings. The percent infarcted areas of the
basal and apical side (the latter being measured on the basal side
of the preceding slice) were averaged and the resulting percent
infarcted area was multiplied by the mass of the slice to express
the slice fTC infarct size in grams. To obtain the percent TTC
infarct size for the entire left ventricular myocardium, the infarct masses of all slices were added together and the total
infarct weight was divided by the total left ventricular weight.
For proper comparison of pathologic slices and short-axis tomograms, the actual thickness of each pathologic slice was
measured.
Vol. 74, No. 4, October 1986
Tomographic data processing. Each of the 30 projections
was corrected for nonuniformity with a 30 million count image
of a cobalt-57 source and adjusted for center of rotation. Projections were filtered-back-projected by a low-resolution Hanning
filter to obtain transaxial tomograms. These were nine-point
smoothed (4-2-1 weighting) and reoriented into vertical longaxis and short-axis tomograms, which were respectively parallel to and perpendicular to the long-axis of the left ventricle.26
No attenuation or scatter correction was used. All tomograms
were one pixel (0.43 cm) thick.
The tomograms were analyzed by a method described previously.27 Briefly, the operator selected the short-axis and vertical long-axis cuts that encompassed the left ventricular cavity.
After determination of the center of the left ventricle, circumferential profiles were generated automatically by performing a
radial search for maximum value on 60 equidistant points over
the entire circumference of each tomographic cut (figure 1). The
maximal values were then plotted for each angle as a percentage
of the maximum value of the profile. For proper comparison of
the different profile to the normal limit profile, an anatomic
landmark was defined at the inferior junction of the two ventricles for the short-axis cuts and at the midapical point for the
vertical long-axis cuts. The myocardial activity was analyzed
from the short-axis circumferential profiles for all regions, except the apex, which was analyzed from the apical portions of
the vertical long-axis profiles.
Quantification of infarct size on individual tomograms
Generation of a normal data base. A reference normal data
base was established for initial distribution of 201T1 based on the
data from the six normal dogs. The short-axis normal profiles
A
B
c
D
"-
ICO)
z
0
C.)
0
180
360
ANGLE
FIGURE 1. Diagram showing various steps of quantitative analysis of
myocardial activity in a short-axis tomogram: The center of the left
ventricle is manually assigned (A) and 60 radii 6 degrees apart are
generated to a predetermined length. The computer searches the maximum count value along each radius (black dots) (B) until 360 degrees
are covered (C). The maximum count values are then plotted (D) against
angular location on the myocardial periphery as a percent maximum
count value in a given tomogram. An anatomic landmark is determined
at the inferior junction of the two ventricles (x in C) for proper matching
of each profile to the normal limit profile.
853
PRIGENT et al.
Downloaded from http://circ.ahajournals.org/ by guest on June 15, 2017
were grouped into five regions extending from apex to base and
the mean myocardial '01T1 distribution was determined for each
region. Similarly, five mean normal profiles were generated in
the vertical long-axis orientation.
Determination ofinfarct size. The profiles of the 14 dogs with
infarction were compared with the corresponding mean normal
profiles (figure 2). Since the boundary between the normally
and abnormally perfused myocardium was not sharply demarcated on the 20ITI tomograms, definition of defect edge as the
points at which the experimental profile fell below the mean
normal profile resulted in overestimation of defect size (figure
2). To circumvent this problem, different thresholds below the
mean normal profile were evaluated in 15 short-axis tomograms
("calibration tomograms") to determine the best definition for
the edge of the abnormal zone. These calibration tomograms
were selected so that among the slices there were five normal
slices (four from the normals dogs and one from the dogs with
infarction) and 10 slices with different-sized regions with transmural necrosis. The extent of perfusion defect on these tomograms using various thresholds below the mean normal profile
was compared with the TTC pathologic infarct size. Similarly,
different thresholds were applied to the vertical long-axis profiles of three dogs (two with different-sized infarcts in the apex
and one normal dog). Since the optimal normal limit thresholds
were determined in the calibration tomograms, the remaining 94
tomograms were prospectively quantified with the adopted
threshold. For each heart, all short-axis tomograms were
matched to the pathologic slices corresponding to the same
depth along the long axis of the left ventricle.
Determination of infarct size for the entire left ventricular
myocardium. Two methods were used to determine total left
ventricular myocardial infarct size from infarct size on individual tomograms. ln method 1, it was assumed that each slice
contributed equally to the total left ventricular mass, and thus
the number of abnormal profile points of all slices were added
up and then were divided by the total number of profile points,
according to the formula:
APapex + API + AP2 +
TPapex + 60 x nsa
where IS = infarct size, AP - number of abnormal points for
the apical profile (APapex) and short-axis profiles from apex to
base (AP1, AP2 ...), TPapex = total number of apical profile
points, and nsa = number of short-axis slices, and 60 total
number of profile points in the short-axis slices. This method
will be referred to as the uncorrected method.
In method 2, the difference in mass of the myocardial slices
from apex to base was taken into consideration. As shown in
figure 3, when the left ventricle is sliced at equidistant intervals
perpendicular to its long axis, slice mass differs from apex to
base because of varying slice radius and myocardial wall thickness. Because the extent of infarction is given as a percentage
with the maximum counts circumferential profile analysis, it is
necessary to account for myocardial slice mass before summation of individual slice infarct size. Thus it was postulated that
each myocardial slice would represent a certain fraction of the
left ventricular mass according to its fractional distance along
the long axis of the left ventricle. Therefore the number of
abnormal profile points on each tomogram (APaPeX5 API,
AP2, ...) was multiplied by a coefficient, K, reflecting the
contribution of each myocardial slice to the total left ventricular
mass and then summed. This sum was then divided by the total
number of left ventricular profile points corrected for K according to the formula:
=
%IS =
APapex(Kapex) + APl(K1) + AP2(K2)
TPapex(Kapex) + 60(K1 + K2 +
+
where AP = number of abnormal points in the apex (APapex)
and short-axis cuts numbered from the most apical from the
:TOMOGRAPHEY
PATHOLOGY
E.
A
i
...E...
TOOGAPY_
1 50
Et400_
M: OFILE
Bi
D
.3
0i;180
IOFILE
E: ::::;:
36..0. An ...le
Z: ;E
10
AAL .:00:
..;|: ....;: ::;:
1180
InfarCit :size
18%
:i
36(f Angle
Infarct :s.ize 22%
FIGURE 2. Comparison of pathologic (A and B) and tomographic (C and D) infarct size in a myocardial slice with anterior wall
infarction. The pale area on the TTC-stained slice (A) represents the region with myocardial necrosis. By planimetry, the infarct
region was shown to occupy 18% of the slice area (B). On the tomographic slice (C), the border of the infarcted region was not
discrete (arrows). When the circumferential profile generated from the tomogram (D) was compared with a proper threshold, the
tomographic infarct size was measured to be 22%.
854
CIRCULATION
LABORATORY INVESTIGATION-MYoCARDIAL INFARCTION
The comparison of the infarct size linear regressions for the
uncorrected and corrected methods was made with F tests as
described by Smith and Choi.30 All values were expressed as the
mean ± SD. The alpha level of significance was .05. All p
values were two sided.
Downloaded from http://circ.ahajournals.org/ by guest on June 15, 2017
FIGURE 3. Varying masses of myocardial slices from apex to base of
the left ventricle. Each slice mass is dependent on the mean radius (R)
and thickness (t) of the slice.
most basal (API, AP2), K = correction factor for the apex
(Kapex) and short-axis cuts (K,, K2), Tpapex = total number of
points in the apex, 60 = number of points per short-axis slice,
and K1, K2 ... = K factor for the short-axis slices 1, 2 ...
This method for determination of infarct size of the entire left
ventricle will be referred to as the corrected method.
To determine K, in the five control dogs for which the pathologic data were available, the dependence of myocardial slice
mass on fractional distance from the left ventricular apex was
assessed. The mass of each slice was expressed as a percentage
of the heaviest slice in a given heart and plotted as a function of
the fractional distance of the slice along the left ventricular long
axis. Polynomial regression analysis was applied to define the
relationship between the percent mass (K) and fractional distance for each dog and for all dogs.
Statistical analysis. Statistical analysis was carried out with
BMDP and SAS statistical software packages.28' 29 Pathologic
and tomographic infarct sizes for individual slices and for the
entire myocardium were compared by least-squares linear regression analysis and by simultaneously testing whether the
slope differed from 1 and whether the y intercept differed from 0
(F test). The standard error of the regression estimate (SEE)
expressed the random variability in tomographic infarct size
given the pathologic infarct size. In the slice-by-slice comparison, the ratio of the SEE over the mean slice infarct size was
used as an index of measurement variability.
Polynomial regression analysis was applied to define a relationship between slice mass and its fractional distance along the
left ventricular long-axis and to determine the polynomial degree that best fit the data. Comparison of individual dog polynomial regressions was carried out with analysis of variance.28 The
absolute error in the measurement of slice infarct size was defined as the difference between the pathologic and tomographic
infarct size (%), multiplied by the infarct weight defined by the
pathologic measurement.
Vol. 74, No. 4, October 1986
Results
Pathologic infarct size. The mean heart weight was 82
± 9 g (range 68 to 95) in the normal dogs and 93 ± 27
g in the dogs with infarction. The infarct weights
ranged from 7 to 40 g, representing 8% to 31% of the
left ventricle. The mean infarct weight was 21 ± 10 g
in dogs with LAD occlusion and 16 ± 7 g in those with
LCX occlusions. There were 107 slices in the dogs
with infarction, 83 of which were found to contain
infarcted tissue by TTC staining. Necrosis was transmural in 51 of 83 slices (61%), involved more than
50% of the myocardial wall thickness in 24 of 83 slices
(29%), and involved less than 50% of the wall thickness in the remaining eight of 83 slices (10%).
In all 14 dogs, the infarction involved at least 50%
of the myocardial wall thickness in at least two-thirds
of the infarcted slices and at least one slice per heart
had a totally transmural infarct.
Mean normal profile. Figure 4 illustrates the shortaxis mean normal profiles for the five different anatomic regions along the long axis of the left ventricle.
The pattern of initial myocardial distribution of 201TI
was similar for the first three regions; the region extending from 90 to 270 degrees, corresponding to the
interventricular septum, showed relatively lower activity than the remaining portions of the left ventricular
circumference, presumably because of photon attenuation by the right ventricle. In region IV, which was
closer to the left ventricular base, the arc between 45
and 180 degrees had relatively lower counts, presumably because of the increased distance of this region
from the detector in frontal views and increased attenuation from the left ventricle. In the basal region (region
V), the reduction of 201T1 activity in the 45 to 180
degree arc became even more pronounced.
Comparison of different thresholds for infarct sizing in
the pilot calibration sample. Different thresholds below
the normal profile (40%, 45%, 55%, 60%, and 65%)
were tested to calculate infarct size in the short-axis
and vertical long-axis tomograms. The best thresholds
were 60% of the mean for the short-axis tomograms
with LAD necrosis, 45% of the mean for the short-axis
tomograms with LCX necrosis, and 40% of the mean
normal profile for the apical portion of the vertical
long-axis profiles for both LAD and LCX infarction.
The LAD territory was considered to include the septum and anterior wall on all the short-axis slices. How855
PRIGENT et al.
100
100
50
50
0
0
Cc
z
D
0 0
z 0
REGION
I
1IIIIIVV
1
0
180
360
ANGLE
z
0
.
.
.
180
*
360
11
0
180
360
Downloaded from http://circ.ahajournals.org/ by guest on June 15, 2017
Ill
0
180
180
360
Iv
360
V
FIGURE 4. Mean normal profiles for different regions of the left ventricle. Portions of profiles with decreased activity (90 to 270
degrees in regions I through III representing septum and anterior walls and 45 to 225 degrees in regions IV and V representing
septum and inferior walls) are more likely the result of photo attenuation.
the proportion of the anterior wall vascularized
by the LAD progressively narrowed from apex to base,
representing 33% of the short-axis periphery in region
I, 25% in region II, 23% in region III, and 22% in
regions IV and V. The circumflex territory was considered to include the rest of the left ventricular myocardium (inferolateral region) on the short-axis slices. With
these thresholds, the correlation coefficient was .95
(p < .001) for the 18 calibration slices and .91 (p <
.001) when only the 12 calibration slices with infarction were considered.
Figure 2 illustrates an example of correlation between pathologic and tomographic infarct size in an
individual slice with LAD necrosis. On the pathologic
slice (A), the infarct size (pale area) was measured by
planimetry to be 18% (B). On the circumferential profile (D) generated from the tomogram (C), the infarct
size was measured to be 22% when the adopted threshold of 60% was used.
ever,
Relationship between TTC and SPECT estimates of infarct size in individual slices. The optimum thresholds
derived from the 18 calibration tomograms were applied prospectively to the remaining 94 tomograms.
When all 94 prospective slices (whether or not they
contained infarction) of the dogs with infarction were
considered together, the correlation coefficient was
.94 (p < .001) and linear regression yielded an SEE of
8.2%. When only the 71 prospective slices with infarc856
tion were considered, the correlation coefficient was
.93 (p < .001) and the SEE was 9.4% (figure 5). The
regression line (slice tomographic infarct size vs pathologic infarct size) did not differ significantly from the
line of identity. All 24 normal slices (no infarct by
TTC) of the dogs with infarction as well as all slices of
the control dogs were also normal by quantitative
analysis.
When the slices were divided into three groups with
regard to the type of necrosis transmural (group l),
subendocardial involving more than 50% of the myocardial wall thickness (group 2), and subendocardial
involving less than 50% of the myocardial wall thickness (group 3) (table 1)
the correlation between
tomographic and pathologic infarct size was excellent
for group 1 (mean SD 38.1 + 27.5% and 35.2
28.3% for pathology [P] and tomography [T], respectively; T = -1.6 + .97P, r = .94, p < .001, SEE =
-
±
9.6%, ratio SEE/mean slice infarct size = .27); the
correlation was good in group 2 (mean SD 20.7 +
11.0% for P and 22.4
15.8% for T; T = -3.2 +
=
1.24 P, r .86, p < .001, SEE = 8.3%, ratio SEE/
mean infarct size
.37), and there was no correlation
ingroup3(mean SD5.7 3.1%forPand3.4
5.7% for T; r = .44, p = .14, SEE = 5.4%. ratio
SEE/mean infarct size = 1.6).
In 65 of 71 slices (92%), the absolute error between
pathologic and tomographic infarct size was 0.5 g or
±
CIRCULATION
LABORATORY INVESTIGATION-MYOCARDIAL INFARCTION
100
TABLE 1
.
Pathologic (P) and tomographic (T) infarct size in 107 slices divided
into four groups according to the type of myocardial infarctionA
80
P(%)
Ca
b.o
60
0
0
t
40
.
20
N=71
Y=-. 1+X
r=.93
p<.OO 1
SEE=9.4%
0
20
40
60
80
100
100
100
94
91
89
78
77
77
76
67
58
56
56
Group I: Transmural myocardial infarction (n = 51)
T(%) P(%) T(%) P(%) T(%) P(%) T(%)
100
89
100
79
84
60
75
63
100
55
60
57
42
56
50
49
49
44
42
42
38
35
31
28
28
27
43
57
52
50
33
42
43
35
38
27
23
13
23
26
25
23
23
21
19
19
18
18
18
18
17
17
7
25
8
0
35
17
0
17
22
23
27
23
22
17
15
15
14
14
13
12
10
10
9
8
7
10
20
11
32
14
13
0
13
0
5
7
3
Downloaded from http://circ.ahajournals.org/ by guest on June 15, 2017
PATHOLOGIC INFARCT SIZE (%)
FIGURE 5. Relationship between pathologic and tomographic infarct
size in the 71 prospective slices with infarction.
less, regardless of the slice infarct weight. In the remaining six (8%), this absolute error was greater than
0.5 g and ranged from 0.6 to 1.2 g.
Dependence of myocardial slice mass on its distance from
the apex. A relationship was observed between slice
fractional distance and slice mass in all five normal
dogs in that the mass of each slice increased with
increasing distance from the apex with a plateau or
reduction at the region of the base (figure 6). A seconddegree equation best fit the data for all five dogs. The
100
Group 2: Infarction involving >50% of myocardial wall thickness
(n = 24)
T(%)
P(%)
T(%)
P(%)
45
42
35
34
30
30
25
25
24
23
22
19
17
38
47
37
32
50
42
42
25
27
30
42
17
15
17
15
13
13
13
11
11
10
8
20
16
14
8
10
9
7
6
0
5
0
6
7
Group 3: Infarction involving <50% of myocardial wall thickness
T(%) P(%) T(%) P(%)
*
90
10
8
8
7
80 _*/
70
X60
17
0
0
4
6
3
3
1
0
3
0
3
AAll 24 slices containing no infarction by pathology (group 4) were
also normal by tomography.
50 _,
to 40
*
30
y=-.023x2 +
o
1.2
20
10
.
°
10
20
30 40 50 60 70
80
DISTANCE OF SLICE FROM LV APEX
(% TOTAL LONG AXIS LUNGTIO
90
100
FIGURE 6. Dependence of myocardial slice weight on its distance
from the left ventricular apex in the five control dogs. A second-degree
equation best described this relationship.
Vol. 74, No. 4, October 1986
individual equations for each dog were then compared
to determine whether one equation would suffice to
describe the relationship between slice fractional distance and slice mass for all five dogs. The five equations did not differ significantly so a pooled seconddegree polynomial equation was fit to all of the data as
K = -1.2 + 2.9x - .023x2 (figure 6).
Total myocardial infarct size in vivo: tomography vs pathology. There was a high correlation between the pathologic and tomographic infarct size assessed by the
uncorrected method (r = .82, p < .001, SEE = 6.3%)
857
PRIGENT et al.
(CORRECTION FOR MYOCARDIAL SLICE WEIGHT)
(NO CORRECTION FOR MYOCARDIAL SLICE WEIGHT)
40
40
LAD *
LCX M.
//
N
I-
AN
* ,~~~~~~~~//X
<: E
_
/
E
-.m
0o0~~~~0
0
0
zC>20 _
a
30
0
30 _
0
LADLCX E
W
//
N
T
20
C: >
10
t/
O>
10
0~~~~~~~
>
10
-
U
Z
N=14
yAY-.8+1.16X
r=.82
y=. 1+.94X
r=.86
P<.OO1
SEE=4.5%
a
z
//Up<.OOl
SEEE=6.3%
Downloaded from http://circ.ahajournals.org/ by guest on June 15, 2017
0
40
20
10
30
PATHOLOGIC INFARCT SIZE
(% Entire Myocardial Weight)
FIGURE 7. Relationship between the tomographic and pathologic in0
farct size expressed as a percentage of the entire left ventricular weight
by the uncorrected method.
(figure 7). With this method, however, all tomographic infarct sizes in the eight dogs with LAD occlusion
were above the line of identity, signifying overestimation by tomography, and four of six with LCX necrosis
were underestimated by tomography. The regression
line had a slope greater (1.16) but insignificantly different from the line of identity (p = .29).
When correction for differing myocardial slice mass
was done, the correlation between the pathologic infarct size and tomographic infarct size was improved
(r = .86, p < .001, SEE - 4.5%) and the slope was
closer to unity (.94) (figure 8). Figure 9 shows the
effect of slice mass correction on the relationship between pathologic and tomographic infarct size. In all
eight dogs with LAD necrosis, the correction reduced
the tomographic infarct size and in six of eight this
correction resulted in a closer one-to-one relationship
between TTC and SPECT infarct size. The LCX infarct size increased slightly in five of six dogs with
LCX infarction and in four of these five, this correction partially compensated for the underestimation by
SPECT. Overall, with correction, 11 of 14 data points
came closer to the line of identity.
Table 2 summarizes the variables of the regression
line for the two methods. The mean percent infarct size
was (mean + SD) 20.5 + 7.6% for pathology, 23.1
± 10.7% for the uncorrected method, and 19.3 +
8.3% for the corrected method. The correlation coeffi858
U
I1
0
1
10
30
20
PATHOLOGIC INFARCT SIZE
(% Entire LV)
J
440
FIGURE 8. Relationship between the tomographic and pathologic infarct size expressed as a percentage of the entire left ventricular weight
by the corrected method.
cient was better for the corrected than for the uncorrected method (.86 vs .82). Even though none of the
regression lines significantly differed from the line of
identity, the slope was slightly closer to the line of
identity with correction (slope = .94, intercept = . 1,
and p = .56 for comparison of the regression line with
the line of identity) than without correction (slope =
1. 16, intercept = .8, p = .29). The SEE, the relative
bias, and the relative precision were smaller with correction (4.5%, - 1.2%, and 4.3%, respectively) than
without correction (6.3%, 2.6%, and 6.2%).
Discussion
In this study, we attempted to quantify myocardial
infarct size by 201T1 tomography in a living canine
preparation with 180 degree acquisition without scatter
or attenuation correction. Quantification of myocardial
infarct size with SPECT and 99mTc-pyrophosphate has
been shown to be feasible.5' 14, 19 Although this approach is useful in the setting of acute myocardial
infarction, there is a delay of several hours to days
before 99mTc-pyrophosphate is optimally taken up by
the acutely necrotic myocardium and the study usually
becomes negative several days to weeks after the
event. On the contrary, 20MT1 scintigraphy can be applied to estimate the infarct size at the onset of acute
myocardial infarction, and the study usually remains
positive thereafter. Thus the two techniques are comCIRCULATION
LABORATORY INVESTIGATION-MYOCARDIAL INFARCTION
(EFFECT OF CORRECTION FOR SLICE MASS)
40rr
Before After Correction
LAD o
LCX o
W
30
N
I-
_
U>
_i
Z
L.-
z
20
g W
10
Downloaded from http://circ.ahajournals.org/ by guest on June 15, 2017
0
With Correction:
y=. 1+.94X
r=.86
p<.00 1
SEE=4.5%
30
20
10
PATHOLOGIC INFARCT SIZE
(% LV Myocardium)
FIGURE 9. Effect of tomographic data correction on the relationship
between the pathologic and tomographic infarct size. Correction consistently reduced the infarct size in the eight dogs with LAD occlusion and
slightly increased the infarct size in five of six dogs with LCX infarction. With correction, the infarct size was better estimated in 11 of 14
dogs (values closer to the line of identity).
plementary and useful for assessing infarct size and
prognosis.
Keyes et al. 15 demonstrated that viable and infarcted
myocardium could be measured accurately in excised
canine hearts by rotational 20'Tl SPECT with 360 degree acquisition and no attenuation correction or scatter correction. Tamaki et al. ,16 using 360 degree acquisition and attenuation correction, applied the technique
TABLE 2
Variables of the regression line of the two tomographic methods for
determination of infarct size vs TTC
Mean IS ± SD (%)
r
P, (<)
Intercept
Slope
Comparison with line of
identity (P2)
SEE (%)
Relative bias (%)
Relative precision (%)
Uncorrected
Corrected
23.1 ± 10.7
.82
.001
- .8
1.16
19.3 ± 8.3
.86
.001
.1
.94
.29
6.3
2.6
6.2
.56
4.5
Pathology
(TTC)
20.5 ± 7.6
-1.2
4.3
r = correlation coefficient, P, - value in testing that the true correlation coefficient is equal to 0; P2 - p value for the comparison of the
regression line with the line of identity; IS = infarct size.
Vol. 74, No. 4, October 1986
to human subjects and showed a good correlation between the tomographic infarct size and the enzymederived estimate. Caldwell et al.'7 measured the relative myocardial perfusion defect size in dogs, using
180 degree acquisition without scatter or attenuation
correction, and found a good correlation with the measurement in vitro. In all of these studies, endocardial
and epicardial myocardial borders as well as the edge
of infarcted area were assigned manually or by a computerized edge-detection algorithm.
In this study, a previously described quantitative
technique using maximum counts circumferential profile analysis was applied.28 The technique requires the
intervention of the operator only for the selection of
tomograms and for the assignment of the center of the
left ventricle, apex, and the interventricular junction
and has been shown to be reproducible. We chose to
acquire 30 projections over 180 degrees rather than
360 degrees to obtain higher spatial and contrast resolution.3'-33 The tomograms were reconstructed without
correction for scatter or attenuation because methods
for correction of the variable attenuation of the 20 11
activity in the thorax have not been optimized. Although scatter and attenuation may be partly responsible for the lack of definition of the edges of the perfusion defect on tomographic images and on the profiles,
establishment of normal 20'T1 distribution profiles in
various regions of the left ventricle and comparison of
the experimental profiles to these normal profiles partly compensated for this problem. This compensation is
evidenced by the fact that despite the appearance in
normal tomograms of reduced counts in the more basal
portions of the septal and inferior walls, none of the
tomograms showing noninfarcted tissue was considered abnormal after their circumferential profiles were
compared with the respective lower limits of normal.
To avoid partial volume effect, only the short-axis
slices containing the left ventricular cavity were analyzed, and the apex was analyzed from the vertical
long-axis cuts. This aspect of our approach to assess
the left ventricular apex is particularly important in
cases with necrosis in the LAD distribution, which
frequently involves the apex, and may go undetected if
apical partial volume correction is not done.
A good correlation was noted between the tomographic and pathologic infarct size in individual slices.
Although the SEE was 8.2%, the absolute error of the
measurement was small (<0.5 g) regardless of the size
of the infarction. We observed that subendocardial
infarction that involved less than 50% of the myocardial wall thickness and less than 10% of the myocardial
periphery were either missed or underestimated by to859
PRIGENT et al.
Downloaded from http://circ.ahajournals.org/ by guest on June 15, 2017
mography (table 1, group 3). This is believed to be
mainly caused by the system spatial resolution, 345
individual tomograms. Assessment of total left ventricular infarct size requires further correction account-
since there is a significant reduction in the reconstructed voxel value for objects smaller than twice the full
width at half maximum. It is likely that this dependence on the object size would tend to mask nontransmural perfusion or small perfusion defects.
We observed that when the uncorrected method was
used to determine left ventricular infarct size, the LAD
infarcts were generally overestimated while LCX infarcts were underestimated. This was due to the fact
that this method assumes that all slices contribute to
the same degree to the total left ventricular mass and
overestimates the relative contribution of the apical
part of the left ventricle. Thus infarction involving the
apex, such as LAD infarction, will be overestimated.
Conversely, infarction involving the mid and basal
portion of the left ventricle, such as LCX infarction,
will be underestimated. As expected, correction for
slice mass consistently decreased the LAD infarct size
by decreasing the relative contribution of the apical
slices to the total left ventricular mass and slightly
increased the LCX infarct size in five of six dogs by
increasing the relative contribution of the mid and basal slices to the total left ventricular mass. In the remaining dog with LCX infarction, the infarct size decreased by 1% with correction; in that experiment, the
infarct was larger in the apical part of the left ventricle
than in its basal part. Considering all dogs, correction
for slice mass was beneficial in 11 of 14 cases by
bringing their data points closer to the line of identity.
When comparing the two methods, the regression line
was closer to the line of identity with the corrected
method; the correlation coefficient was greater and the
SEE, the relative bias, and the relative precision were
smaller with correction.
In a separate study,36 we have developed and validated a scintigraphic-based algorithm in normal dogs,
which expresses the relationship between slice mass
and fractional distance from the apex. This approach
was then applied to eight normal subjects and a relationship between slice mass and fractional distance
from the apex in human beings, close to the relationship in dogs, was found. The utility of the approach
outlined in the manuscript for infarct sizing in human
subjects by 201T1 tomography requires further validation, which is currently underway in our institution.
In conclusion, a method for quantifying myocardial
infarct size by 201T1 SPECT and circumferential profiles analysis has been described. Analysis of maximum-count circumferential profiles appears to be a
suitable method for quantification of infarct size in
ing for differing myocardial slice mass. Extension of
this method for quantifying the extent of myocardial
infarction in human beings requires further validation.
860
We are indebted to Dr. William Ganz for making his animal
laboratory available to us and to Dr. Michael Fishbein for his
valuable suggestions. We also acknowledge Christopher Wong,
C.N.M.T., for acquisition of the studies, Gretchen Edwards for
pathologic measurements, Kenneth Resser, B.A., for statistical
analysis, and Gregory Kuhl,M.. A., for assistance in preparation
of the manuscript.
References
1. Page DL, Caulfield JB, Kastor JA, Desanctis W, Sanders C: Myocardial changes associated with cardiogenic shock. N Engl J Med
285: 133, 1971
2. Sobel BE, Bresnahan GF, Shell WE, et al: Estimation of infarct
size in man and its relation to prognosis. Circulation 46: 640, 1972
3. Holman BL, Chisholm RJ, Braunwald E: The prognostic implication of acute myocardial infarct scintigraphy with 99mTc-pyrophosphate. Circulation 57: 320, 1978
4. Sobel BE: Infarct size, prognosis, and causal contiguity. Circulation 53 (suppl I): 1- 146, 1976
5. Holman BL, Goldhaber SZ, Kirsh C-M, Polak JF, Friedman BJ,
English RJ, Wynne J: Measurement of infarct size using single
photon emission computed tomography and technetium-99m pyrophosphate: a description of the method and comparison with patient
prognosis. Am J Cardiol 50: 503, 1982
of the
6. Shell WE, KjekshusJK, Sobel BE: Quantitative assessment
extent of myocardial infarction in the conscious dog by means of
analysis of serial changes in serum creatine phosphokinase activity.
J Clin Invest 52: 2579, 1973
7. Norris RM, Whitlock RM, Barratt-Boyes C, Small C: Clinical
measurement of myocardial infarct size: modification of a method
for the estimation of total creatine phosphokinase release after
myocardial infarction. Circulation 51: 614, 1975
8. Roberts AJ, Cipriano PR, Alonso DR, Jacobstein JG, Combes JR,
Gay WA: Evaluation of methods for the quantification of experimental myocardial infarction. Circulation 57: 35, 1978
9. Poliner LR, Buja LM, Parkey RW, Stokely EM, Stone MJ, Harris
R, Saffer SW, Templeton GH, Bonte FJ, Willerson JT: Comparison of different noninvasive methods of infarct sizing during experimental myocardial infarction. J Nucl Med 18: 517, 1977
10. Botvinick EH, Shames D, Lappin H, Tyberg JV, Townsend R,
Parmley WW: Noninvasive quantitation of myocardial infarction
with technetium-99m pyrophosphate. Circulation 52: 909, 1975
11. Nelson AD, Khullar S, Leighton RF, Budd GC, Gohara A, Ross
JN Jr, Andrews LT, Windham J: Quantification of thallium-201
scintigrams in acute myocardial infarction. Am J Cardiol 44: 664,
1979
12. Okada RD, Lim YL, Chesler DA, Kaul S, Pohost GM: Quantitation of myocardial infarct size from thallium-201 images: validation of a new approach in an experimental model. J Am Coll
Cardiol 3: 948, 1983
13. Lewis SE, Stokely EM, Devous MD Sr, Bonte FJ, Buja LM,
Parkey RW, Willerson JT: Quantitation of experimental canine
infarct size with multipinhole and rotating-slanthole tomography. J
Nucl Med 22: 1000, 1981
14. Keyes JW Jr, Leonard PF, Brody SL, Svetkoff DJ, Rogers WL,
Lucchesi BR: Myocardial infarct quantification in the dog by single
photon emission computed tomography. Circulation 58: 227, 1978
15. Keyes JW Jr, Brady TJ, Leonard PF, Svetkoff DB, Winter SM,
Rogers WL, Rose EA: Calculation of viable and infarcted myocardial mass from thallium-201 tomograms. J Nucl Med 22: 339, 1981
16. Tamaki S, Nakajima H, Murakami T, Yui Y, Kambara H, Kadota
K, Yoshida A, Kawai C, Tamaki N, Mukai T, Ishii Y, Torizuka K:
Estimation of infarct size by myocardial emission computed tomography with thallium-201 and its relation to creatine kinase-MB
CIRCULATION
LABORATORY INVESTIGATION-MYOCARDIAL INFARCTION
17.
18.
19.
20.
21.
22.
Downloaded from http://circ.ahajournals.org/ by guest on June 15, 2017
23.
24.
25.
release after myocardial infarction in man. Circulation 66: 994,
1982
Caldwell JH, Williams DL, Harp GD, Stratton JR, Ritchie JL:
Quantitation of size of relative myocardial perfusion defect by
single-photon emission computed tomography. Circulation 70:
1048, 1984
Thompson CJ, Mena I, Maublant JC, Narahara KA: Thallium-201
tomographic estimation of left ventricular mass in normal and infarcted canine hearts using a new automated edge-detection program. J Am Coll Cardiol 5: 533, 1985 (abst)
Lewis SE, Devous MD Sr, Corbett JR, Izquierdo C, Nicod P,
Wolfe C, Parkey RW, Buja LM, Willerson JT: Measurement of
infarct size in acute myocardial infarction by single photon emission computed tomography with technetium-99m pyrophosphate.
Am J Cardiol 54: 193, 1984
Doherty PW, Lipton MJ, Beminger WH, Skioldebrand CG, Carlsson E, Redington RW: Detection and quantitation of myocardial
infarction in vivo using transmission computed tomography. Circulation 63: 597, 1981
Weiss ES, Ahmed SA, Welch MJ, Williamson JR, Ter-Pogossian
MM, Sobel BE: Quantification of infarction in cross sectional
canine myocardium in vivo with positron emission transaxial tomography and 11C-palmitate. Circulation 55: 66, 1977
Jasczak RZ, Murphy PH, Huard D, Burdine J: Radionuclide emission computed tomography of the head with Tc-99m and a scintillation camera. J Nucl Med 18: 373, 1977
Keyes JW Jr, Orlandea N, Heetderks JW, Leonard P, Rogers W:
The humogontron-A scintillation camera transaxial tomograph. J
Nucl Med 18: 381, 1977
Fishbein MC, Meerbaum S, Rit J, Lando U, Kanmatsuse K, Mercier JC, Corday E, Ganz W: Early phase acute myocardial infarct
size quantification: validation of the triphenyl tetrazolium chloride
tissue enzyme staining technique. Am Heart J 101: 593, 1981
Vivaldi MT, Kloner RA, Schoen FJ: Simulated autopsy detection
Vol. 74, No. 4, October 1986
26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
of early myocardial infarcts: effects of duration of ischemia and
autolysis. J Am Coll Cardiol 5: 400, 1985 (abst)
Borello JA, Clinthorne NH, Rogers WL, Thrall HJU, Keyes JW Jr:
Oblique angle tomography: a restructuring algorithm for transaxial
tomographic data. J Nucl Med 22: 471, 1981
Garcia EV, Van Train KV, Maddahi J, Prigent F, Friedman J,
Areeda J, Waxman A, Berman DS: Quantification of rotational
thallium-201 myocardial tomography. J Nucl Med 26:17, 1985
Dixon WJ, editor: BMDP statistical software. Los Angeles, 1983,
University of California Press
SAS Institute Inc.: SAS User's Guide: Statistics. Cary, NC, 1982
Smith P, Choi S: Simple tests to compare two dependent regression
lines. Technometrics 24: 123, 1982
Tamaki N, Mukai T, Ishii Y, Fujita T, Yamamoto K, Minato K,
Yonekura Y, Tamaki S, Kambara H, Kawai C, Torizuka K: Comparative study of thallium emission myocardial tomography with
1800 and 3600 data collection. J Nucl Med 23: 661, 1982
Coleman RE, Jaszczak RJ, Cobb FR: Comparison of 1800 and 3600
data collection in thallium-201 imaging using single-photon emission computerized tomography (SPECT): concise communication.
J Nucl Med 23: 655, 1982
Hoffman EJ: 1800 compared with 3600 sampling in SPECT. J Nucl
Med 23: 745, 1982
Caldwell JH, Williams DL, Hamilton GW, Ritchie JL, Harp GD,
Eisner RL, Gullberg GT, Nowak DJ: Regional distribution of myocardial blood flow measured by single photon emission tomography: comparison with in vitro counting. J Nucl Med 23: 490, 1982
Hoffman EJ, Huang SC, Phelps ME: Quantification in positron
emission tomography. 1. Effect of object size. J Comput Assist
Tomogr 3: 299, 1979
Prigent F, Resser K, Rozanski A, Garcia E: Quantification of the
percent hypoperfused myocardium by thallium-201 tomography
requires correction for myocardial slice mass. Circulation 72 (suppl
III): 111-136, 1985 (abst)
861
Quantification of myocardial infarct size by thallium-201 single-photon emission
computed tomography: experimental validation in the dog.
F Prigent, J Maddahi, E V Garcia, Y Satoh, K Van Train and D S Berman
Downloaded from http://circ.ahajournals.org/ by guest on June 15, 2017
Circulation. 1986;74:852-861
doi: 10.1161/01.CIR.74.4.852
Circulation is published by the American Heart Association, 7272 Greenville Avenue, Dallas, TX 75231
Copyright © 1986 American Heart Association, Inc. All rights reserved.
Print ISSN: 0009-7322. Online ISSN: 1524-4539
The online version of this article, along with updated information and services, is located on
the World Wide Web at:
http://circ.ahajournals.org/content/74/4/852
Permissions: Requests for permissions to reproduce figures, tables, or portions of articles originally
published in Circulation can be obtained via RightsLink, a service of the Copyright Clearance Center, not the
Editorial Office. Once the online version of the published article for which permission is being requested is
located, click Request Permissions in the middle column of the Web page under Services. Further
information about this process is available in the Permissions and Rights Question and Answer document.
Reprints: Information about reprints can be found online at:
http://www.lww.com/reprints
Subscriptions: Information about subscribing to Circulation is online at:
http://circ.ahajournals.org//subscriptions/