relationship of gated spect ventricular function parameters to angiographic measurements

Relationship of gated SPECT ventricular function parameters to angiographic measurements

Kenne th Nichols, PhD, Jacquel ine Tamis, MD, E. G o r d o n DePuey , MD, Jennifer Mieres , MD, Sanjay Malhotra , MD, and Alan Rozanski, MD

Objectives. Left ventricular volumes and ejection fractions constitute important information in the diagnosis of cardiac disease. This investigation examined the relations of functional parameters computed with a recently published scintigraphic gated tomographic method with those from angiography, analyzing discrepancies arising from differences involved in modeling the left ventricle.

Background. While left ventricular ejection fractions obtained from myocardial perfusion gated single-photon emission computed tomography (SPECT) have demonstrated accurate comparisons with other imaging modalities, validations of volumes have not been examined as exten- sively, and some recent studies have reported a wide range of angiographic correlation. It is important to know how volumes obtained by a new class of methods compare with those from older, well-established techniques in order to interpret individual patients' results, particularly when scintigraphic images are severely hypoperfused.

Methods and Results. Tc-99m sestamibi myocardial perfusion gated SPECT data were processed retrospectively for 58 patients studied by single-plane angiography. Endocardial borders were generated automatically on paired vertical and horizontal long-axis Tc-99m sestamibi gated tomograms for compating ventricular volume using a Simpson's rule summation of elliptical slices. Linear regression and paired t tests were used to compare SPECT with angiographic parameters for all patients and for groups identified on the basis of tomogram visual exami- nation as hypoperfused, ischemic or nonischemic, with the latter category fur ther subgrouped as to fixed defects or normal perfusion. Linear regression analysis demonstrated Pearson correlation coefficients of 0.87 for end-diastolic volumes, 0.91 for end-systolic volumes, and 0.86 for ejection fraction; paired t test analysis showed end-systolic volumes to be nearly identical (p > 0.99) to angiographic values. However, paired t tests also revealed gated SPECT end-diastolic volumes and ejection fractions were significantly lower (p < 10-4) than angiography. Correlations and trends were essentially the same for all subgroups except for the small sample (n = 10) of patients with normal perfusion.

Conclusions. Gated SPECT provides ventricular volumes and ejection fractions that correlate well with angiography, even in hypoperfused and ischemic populations. However, gated SPECT end-diastolic volumes and ejection fractions are significantly lower than angiographic measurements, partly because of inclusion of greater outflow tract amounts in standard angiographic models. Because myocyte concentration decreases rapidly at the ventricular base, it is likely that most gated SPECT methods will produce endocardial borders encompassing less of the outflow tract than do angiographic outlines. (J Nucl Cardiol 1998;5:295-303.)

Key Words: gated SPECT • angiography • ventricular volumes • ejection fraction

The importance of left ventricular volumes 1,2 and ejection fractions 3-6 for diagnosis and prognosis has been thor- oughly established. After 30 years of experience with angiography, 7 clinicians are most familiar with these mea-

From the Division of Cardiology and Department of Radiology, St. Luke's-Roosevelt Hospital, New York, N.Y.; Columbia University College of Physicians and Surgeons, New York, N.Y.; and Division of Cardiology, North Shore University Hospital, Manhasset, N.Y.

Received for publication Sept. 2, 1997; revision accepted Oct. 27, 1997. Supported in part by grants from General Electric Medical Systems,

Milwaukee, Wis.

surements and are able to make meaningful comparisons of individual patients' results against established normal limits. 8

Since 1990, use of Tc-99m sestamibi has extended diagnostic capability by combining single-photon emis-

Presented in part at the 46th Annual Scientific Session of the American College of Cardiology, Anaheim, Calif., March 1997.

Reprint requests: Kenneth Nichols, PhD, Division of Cardiology, St. Luke's-Roosevelt Hospital, Amsterdam Ave. at 114th St., New York, NY 10025.

Copyright © 1998 by American Society of Nuclear Cardiology. 1071-3581/98/$5.00 + 0 43/1/87226

295

296 Nichols et al. Journal of Nuclear Cardiology Gated SPECT and angiographic measurements May/June 1998

sion computed tomography (SPECT) myocardial perfusion assessment with ejection fraction (EF) computed from the first pass transit of the same injected radio- pharmaceutical. 9-t2 Several methods of computing resting EF from gated sestamibi tomograms of the stress perfusion distribution have since been published, t3-18 demonstrating high reproducibility, 1739-21 and have been validated successfully against established EF methods. 18-27 Gated SPECT (GSPECT) ventricular volumes, from which EFs are derived, have also been reported to agree well with established methods, 23-26 although reported linear correlation coefficients have ranged from r = 0.9225 to r = 0.6526 in comparing GSPECT with angiographic end-diastolic volume (EDV). This wide range of results raises questions as to how well GSPECT ventricular volumes predict measurements that would be expected from other imaging modalities. This is especially of concern when GSPECT algorithms are applied to severely hypoperfused myocardium, for which it may be difficult to verify appropriateness of endocardial borders because of markedly reduced tracer uptake (Figure 1). Concerns also arise as to GSPECT calculation accuracy for patients with myocardial ischemia because it has been observed that wall motion abnormalities can persist21, 28 beyond the 30 minutes usually allowed from the end of stress testing to the commencement o f GSPECT data acquisition. 29

Within the past year, some of the GSPECT methods have become commercially available, 17,20 proliferating throughout the medical community. Consequently, there is a need for clinicians obtaining GSPECT ventricular function parameters to be aware of the relationships between results of these newer techniques with measurements from older, established methods. Therefore we addressed the question of how ventricular volumes and EFs from one such widely available GSPECT technique 20 compare with angiography, particularly in hypoperfused and ischemic patients. In the process, we sought to understand measurement accuracy and bias, not only of this particular GSPECT method but for all such techniques.

METHODS

Patients. GSPECT Tc-99m sestamibi myocardial perfusion data were analyzed retrospectively for 58 patients (age 60 -+ 11 years; 57% men) studied from January 1, 1992, to February 7, 1997, who also had angiography within less than 1 month of scintigraphy, with a median absolute time difference of 5 days. No patient experienced any cardiac event or change in medical therapy between studies. The patient population exhibited a wide range of angiographic EFs and EDVs, with the majority exhibiting hypoperfusion (see Results section). Reasons for performing perfusion studies included myocardial infarction (38%), coronary artery disease (35%), nonanginal

chest pain (9%), hypertension (9%), angina (6%), and conges- tive heart failure (3%).

Perfusion Tomography. Tc-99m sestamibi injections were performed during peak exercise of a Bruce treadmill protocol or intravenous pharmacological coronary vasodilatation with dipyridamole (0.142 mg/kg per minute infused over 4 minutes) using 1.11 GBq (30 mCi) for a 1-day protocol or 814 MBq (22 mCi) for a separate-day protocol29; 64 x 64 tomograms (pixel size of 6.4 ram) were acquired with high-resolu- tion collimation for 20 seconds/projection for 64 projections over 180 degrees with a biplane camera (General Electric Optima). Perfusion tomograms were acquired gated at eight frames per RR interval 30 minutes after stress. Data were included for incoming R waves between 50% and 150% of average heart rate.

All acquired data were reviewed visually and discarded if motion, gating, or other artifacts were deemed likely to com- promise further analysis. 30 Standard clinical data processing parameters were used31: Butterworth (cutoff = 0.40, power = 10.0) prefilters, followed by quantitative ramp x-filtering, 32 interslice spatial averaging, and time-filtering among gated frames. Stress gated projection data also were summed and reconstructed along with resting tomograms, using commercially available Cequal software. 31 The resulting short-axis images were visually read by an experienced physician as to whether stress tomograms exhibited extensive, severe hypoperfusion, and by comparison with rest tomograms, whether defects were reversible (indicating ischemia) or fixed (indicating scar). Polar perfusion maps were constructed from short- axis images, 31 aiding in this analysis (Figure 1).

The same ventricular angles and limits used for summed tomograms were applied to gated data to produce reoriented midventricnlar vertical long-axis (VLA) and horizontal long- axis (HLA) cinematic tomograms. End-diastole (ED) and end- systole (ES) were defined by the rapid, reliable procedure of finding the frame numbers corresponding to minimum and maximum myocardial counts resulting from partial volume effects.16, 20 Left ventricle (LV) centers were estimated from activity centroids of ES-ED "time-difference" images, because the only counts that should vary synchronously with heartbeat are myocardial counts, thus helping to avoid extraneous hepa- tobiliary activity. 20 Endocardial borders were generated by fit- ting maxima locations with fifth-order, two-dimensional har- monics to reduce noise effects, searching inward to predeter- mined percentage count thresholds, and reconciling endocardial with valve plane points 2° (Figure 1). To further refine endocardial edge detection, an observer had the option of altering automated LV centers, identified ED and ES frame numbers, and endocardial borders. Previous studies revealed that the frequency with which observers find it necessary to alter automatic centers, frames, and borders is 28%, 7%, and 14%, respectively, in analyzing gated SPECT data. 20 The technique used paired VLA and HLA endocardial outlines, corrected for the camera's line spread function, to model each LV slice going from apex to base as an elliptical solid of constant thickness, defined by imaging matrix pixel size of 6.4 ram. Outlines were combined to compute LV volumes, and from these, ejection fractions by Simpson's rule summation, 15 analogous to standard echocar- diographic biplane ventricular modeling. 33

Journal of Nuclear Cardiology Nichols et al. 297 Volume 5, Number 3;295-303 Gated SPECT and angiographic measurements

Figure L An example of severely hypoperfused Tc-99m sestamibi stress data. Paired vertical (top left) and horizontal (top right) long- axis mid-ventricular tom•graphic sections at end-diastole are shown along with automatically generated endocardial borders. The polar maps for this same patient are displayed for stress perfusion (lower left) and rest perfusion (lower right), demonstrating significant reversible ischemia.

Figure 2. An angiographic patient example. End-diastolic (left) and end-systolic (right) endocardial borders were drawn manually by an exPerienced cardiologist.

Contrast Angiography. All patients underwent contrast angiography at rest in concert with arteriographic evaluation of coronary artery disease. Patients were premedicated with diphenhydramine and diazepam. Single-plane images were obtained in the approximate RAO 30 degree projection during rapid injection of a nonionic agent (Iopamidol or Iohexol) through a 5F or 6F pigtail catheter positioned in the mid-LV cavity. Cinematic film was exposed at 30 frames per second for approximately 10 seconds of imaging from a 9-inch fluoro- scopic image intensifier. Immediately after patient data acquisition, a standard-sized 1.5-inch ball was placed in the field of view at the estimated location of the LV center, and fluoroscop- ic images were recorded for several seconds.

Angiograms were reviewed by a cardiologist blinded to results of gated SPECT parameters. Patients were only considered if image quality was judged adequate to allow confident

300-

E v

> Q I i i i - O tu 0.

o o

(A)

i001

,o

-100

/" , /

/ /

I:3 , / • r'[/ / /

• / " / /

/ / /

• ~ / / 7 " / D / 9 / / 0 o •

J/ /° o • " .4 & / ' / • • [ ]

• / . ~ " ~ a A &

/ '

isch

scar

nl

Cath E D V (ml)

t

300

• isch

[] scar

* nl

• o . . - ' " • o . ~ . " D & . . . . - ' ° " •

A A ~ a .o."°'"0

. . . . . . . . . . . . . . . . . . . . 3;0 ---a" "~ . , o . Average E D V (ml)

. - •~ ec3 • o o

(B)

Figure 3. A, Linear regression plots of end-diastolic volume (EDV) gated scintigraphic tomography (SPECT) plotted versus values from contrast angiography ( Cath ). Dashed lines are hyperbolic 95% confidence limits and solid diagonals are least-square fits to the data. Data points are distinguished as isch for reversible perfusion defects, scar for nonreversible defects, and nl for normal perfusion. B, Volume differences of SPECT EDV-Cath EDV are plotted versus means of volume values in a Bland-Altman graph.

discrimination of endocardial walls from other structures and/or background activity. Studies were excluded if frequent ectopic beats occurred during the LV phase of contrast dye injection, resulting in inability to identify a normal systolic contraction. Endocardial borders were drawn manually by an experienced cardiologist using either a Vanguard LV analyzer or General Electric Advantax cardiac analysis software that projected ven- triculograms and digitized the images as 512 x 512 matrixes (Figure 2). Outlines were then used to compute LV volumes and subsequent ejection fractions by use of the Dodge-Sandier area- length formula. 7

Stat is t ical /~aly$1s. All numerical results are reported as

298 Nichols et ai. Journal of Nuclear Cardiology Gated SPECT and angiographic measurements May/June 1998

300-

E

W I-- O UJ n.

/ ; , /

• / " o,"

• • 2 / / , / [ ] f ,.¢"

q y / / / °

D

I 0 300

( A ) C a t h E S V ( m l )

• i sch

[] s c a r

A nl

1 0 0 []

, , . . . . . . • i sch

............... _ [] s c a r . . . . . . . . , [] []

ii ,--° . , . . ~ - ~ - - - ' ~ - - " ; . . . . . .

[]

m = • • A v e r a g e E S V ( m l ) D.

-10

(B)

Figure 4. A, Linear regression plots of scintigraphic (SPECT) end- systolic volume (ESV) are plotted versus angiographic (Cath) values. Dashed lines are hyperbolic 95% confidence limits and solid diagonals are least-square fits to the data. Data points are distinguished as isch for reversible perfusion defects, scar for nonreversible defects, and nl for normal perfusion. B, End-systolic volume differences of SPECT ESV-Cath ESV are plotted versus volume means in a Bland-Altman graph.

mean values _+ 1 standard deviation. Linear regression analysis was used to compare calculations of LV volumes and ejection fractions between modalities and is the preferred test to compare measurements of independent tests performed on the same group of patients. 34 Regression results included computation of standard error of the estimate (SEE), as well as the 95% confidence limits of the linear slopes bounding hyperbole of the associated graphs (Figures 3 through 5), and the probability of no association between the two variables being compared. Linear regression analysis was also performed in conjunction with Bland- Altman graphs of residuals (i.e., differences) plotted versus means of paired values to search for trends and systematic errors (Figures 3 through 5). 35,36 An advantage of Bland-Altman analysis is that it does not involve declaring which of two variables is

100-

A

u. uJ k- O uJ D. ¢n

75-

50-

(A)

25-

? • y "

& • O / " o o % , t ~

J [] • o - •

# I • 7 / / •

[] ,o" / ,

/ " []

2'5 5'0 7'5 100 Cath EF (%)

• isch

o scar ~, nl

A

14. LU

o I- O LU 13. O0

50-

0

• isch

= scar & A nl

o

. . . . . . . . . . . 1;0 " " "° © . '= "~ A v e r a g e E F ( % }

-50

(B)

F i g u r e 5. A, Linear regression plot of scintigraphic (SPECT) ejection fraction (EF) are plotted versus values from contrast angiography (Cath) EF. Dashed lines are hyperbolic 95% confidence limits and solid diagonals are least-square fits to the data. Data points are distinguished as isch for reversible perfusion defects, scar for nonreversible defects, and nl for normal perfusion. B, Differences of SPECT-Cath ejection fractions are plotted versus EF means in a Bland-Altman graph.

the standard against which to test the other variable but instead treats both variables on an equal basis.

Because regression results depend on the range of compared variables, paired t tests were also used to compute two- tailed probabilities (p) that measurements were substantially different between the two techniques, with values of p < 0.05 considered to be statistically significant. 37 Specifically, these are the probabilities that samples were drawn from populations in which the mean difference between matched values = 0. Paired t tests also produced estimates of mean differences, as well as the probable range of differences within 95% confidence limits.

The Fisher z test was used to assess statistical significance of two different regression results, producing the two-tailed probability that the two Pearson coefficients (r) were the same. 3s All tests were applied to the entire patient population


Table 1. Linear regress ion analysis: End-diastolic v o l u m e

All Hypo l schemic Nonisch Scar Normal

n 58 47 33 25 15 10 r 0.87 0.90 0.89 0.84 0.85 0.77 Int -27.0 ml -41.5 ml -37.8 ml -15.4 m~ -50.1 ml 26.6 ml Slope 1.07 1.17 1.17 0.98 1.17 0.63 SEE 32.3 ml 30.6 ml 30.0 ml 35.6 ml 40.7 ml 22.1 ml 95% cl 0.91-1.23 1.00-1.35 0.94-1.39 0.71-1.25 0.74-1.60 0.16-1.09 p <10 -6 <10 -6 <10 --6 <10 --6 <10 -4 0.01

Hypo, hypoperfused; Nonisch, nonischemic; n, number of subjects; r, Pearson correlation coefficient; Int, intercept; SEE, standard error of the estimate; 95% cl, 95% confidence limit bounds of slope; p, probability of no association.

Table 2. Linear regress ion analysis: End-systolic v o l u m e

All Hypo l schemic Nonlsch Scar Normal

n 58 47 33 25 15 10 r 0.91 0.91 0.89 0.93 0.93 0.84 lnt -6.1 ml -7.5 ml -1.5 ml -11.0 ml -22.4 ml 11.9 ml Slope 1.10 1.12 1.01 1.19 1.29 0.69 SEE 22.4 ml 24.1 ml 21.4 ml 23.8 ml 28.8 ml 8.8 ml 95% d 0.97-1.24 0.97-1.27 0.82-1.20 0.98-1.39 0.97-1.60 0.32-1.05 p <10-6 <10-6 <10 --6 <10 --6 <10 --6 0.003

Abbreviations as in Table 1.

Table 3. Linear regress ion analysis: Ejection fraction

All Hypo l schemlc Nonisch Scar Normal

n 58 47 33 25 15 10 r 0.86 0.89 0.89 0.82 0.87 0.59 Int 9.0% 5.5% 7.1% 11.2% 3.0% 48.0% Slope 0.75 0.80 0.78 0.72 0.84 0.20 SEE 7.5% 7.1% 6.5% 8.8% 8.7% 4.0% 95% d 0.63-0.87 0.68-0.93 0.63-0.92 0.50-0.94 0.55-1.13 -0.02-0.42 p <10-6 <10--6 <10 -6 <10 -6 <10 -4 0.08


and to all subgroups. Statistical significance of differences of means compared among subgroups was assessed by the unpaired t test, 37 with the usual two-tailed p < 0.05 limit.

RESULTS

Scan Interpretation. On the basis of visual and quantitative scintigraphic analysis, patients were found to include 47 (81%) with severe myocardial perfusion defects, of whom 33 (57%) demonstrated ischemia and 25 (43%) were nonischemic. The latter group contained

15 (26%) with fixed defects and 10 (17%) exhibiting normal myocardial perfusion.

Linear Regression. In comparing GSPECT with angiography for all patients, linear regression demonstrated Pearson correlation coefficients r = 0.87 for EDV (Table 1), 0.91 for ESV (Table 2), and 0.86 for EF (Table 3). These correlations were strong, with negligible probabilities of no association (p < 10-6), although SEE was large, on the order of 32.3 and 22.4 ml for EDV and ESV, and 7.5% for EF. Angiographic EDV ranged from 40 to 282 ml, ESV from 8 to 204 ml, and EF from 24% to 80%.


Table 4. Linear regression analysis: Bland-Altman graphs

EDV ESV EF

n 58 58 58 r 0.39 0.41 0.25 lnt -44.5 ml -11.7 2.6% Slope 0.22 0.20 -0.14 SEE 29.9 ml 20.7 ml 8.2°/0 95% d 0.08-0.36 0.08-0.31 -0.29-0.00 p 0.003 0.001 0.06


For volumes, slopes > 1.0 and intercepts < 0, and for EF slope < 1.0 and intercept > 0.0 (Tables 1 through 3 and Figures 3 through 5).

B l a n d - A l t m a n Analys i s . Bland-Altman trend graphs for EDV differences versus EDV means (Figure 3, B) demonstrated that the positive regression slope = 0.22 and negative intercept = -44.5 ml (Table 4) were largely due to GSPECT EDVs being lower than angiographic values for small volumes (< 100 ml) but more nearly comparable for larger volumes. The Bland-Altman graph for ESV was similar (Figure 4, B), with positive regression slope = 0.20 and a less pronounced negative intercept = -11.7 ml (Table 4). Consequently, the EF trend was for a negative Bland-Airman slope = -0.14 (Table 4 and Figure 5, B). Ideal results would have been for volume and EF Bland-Altman graphs to have slopes = 0 and associated Pearson coefficients r = 0 and for probabilities of no association = 1.0, but instead these were p = 0.003 for EDV, 0.001 for ESV, and 0.06 for EF (Table 4).

Paired tTests . By paired t test analysis, mean ESVs were nearly identical (p > 0.99), a useful finding given the prognostic power of end-systolic volumes. 2 However, EDV was significantly lower by -18.2 ml (p = 10 -4) and EF significantly lower by -5.3% (p = 10 -5) than angiographic results (Tables 5 through 7). Given the Bland- Altman volume graphs (Figures 3, B and 4, B), these paired t test findings were partly due to the preponder- ance of subjects studied with angiographic volumes < 100 ml. EFs were underestimated throughout the measurement range, as shown by Bland-Altman EF graph of Figure 5, B.

Intergroup C o m p a r i s o n s . Linear regression results were essentially the same for all subgroups except for patients with normal perfusion, for which the small sample size (n = 10) is inadequate for reliable regression analysis, 34 compounded by most EF values clustering around a narrow angiographic EF range from 60% to 80% (Figure 5, B). Application of the Fisher z test to all other subgroups indicated no statistical difference (p <

10 -6 ) between Pearson coefficients for any parameters (Tables 1 through 3). That GSPECT correlated strongly with angiography in this predominantly hypoperfused patient population is significant because of concerns about the accuracy of measurements derived from images in which large myocardial segments evince greatly reduced tracer uptake (Figure 1). Thus by the Fisher z test, correlations were equally strong for ischemic versus nonischemic subgroups, which is significant because exercise-induced regional wall motion abnormalities can persist beyond 30 minutes after stress.21, 28 The preva- lence of this phenomenon may vary between 12% 21 and 19% 28 of all patients, depending on population charac- teristics. The most likely reason that such effects were not seen in this investigation is that too few patients experienced long-lasting effects of sufficient magnitude to be detected.

Paired t test results were essentially similar within each subgroup, with mean differences approximately 1 ml for ESV, -20 ml for EDV, and approximately -5% for EF (Tables 5 through 7), all of which were statistically significant except within the normal subgroup for EDV (p = 0.18) (Table 5) and EF (p = 0.32) (Table 7).

In comparing mean scintigraphic results among subgroups, the unpaired t test revealed that the only statistically significant difference (p = 0.01) found was for EF between "scar" and "normal" subgroups (Table 7). Of note, by this same test there were no differences between ischemic and nonischemic subgroups for any scintigraphic parameters. In comparing mean angiographic results among subgroups, by the unpaired t test the only significant differences (p = 0.01) were for EDV, ESV, and EF between "scar" and "normal" subgroups (Tables 5 through 7).

DISCUSSION

C o m p a r i s o n With Other Inves t iga t ions . The Fisher z test revealed no significant difference between GSPECT and angiographic EF correlations found in this study (r = 0.86; n = 58) and all other reports,1422,2526 for which Pearson correlation coefficients ranged from 0.8426 to 0.93. 22 The Fisher z test also demonstrated no statistical difference between GSPECT versus angiographic EDV correlation reported here (r = 0.87; n = 58) and the only other two published reports,25, 26 with results of the present investigation agreeing more closely with the study of the larger patient size (r = 0.92; n = 25) 25 than with the study demonstrating weaker correlation of r = 0.65, based on only 11 patients 26 (Fisher p = 0.31 and 0.14, respectively). Hence, based on z test results, the present investigation is consistent with all other published reports for both volumes and ejection fractions. Assumptions underlying different GSPECT methods


Table 5. Paired t test: End-diastolic v o l u m e

All Hypo Ischemic Nonisch Scar Normal

n 58 47 33 25 15 10 SPECT mn 109 _+ 65 ml 114 _+ 53 ml 102 +_ 64 ml 120 + 65 ml 139 + 75 ml 90 + 31 ml Cathmn 128_+52ml 131 + 7 0 m l 1 2 0 + 4 9 m l 1 3 8 + 5 6 m l 162+55m1" 102_+36m1" Mn dif -18.2 ml -17.9 ml -18.1 ml -18.5 ml -23.1 ml -11.6 ml 95% cl -27-10 ml -27-9 ml -29-7 ml -33-4 ml -45-1 ml -29-6 ml p 10- 4 10 -4 10- 3 10- 2 0.04 0.18

Abbreviations as in Table 1. *Statistically significant difference (p = 0.01) be tween subgroups by the unpaired t test.

Table 6. Paired t test: End-systolic v o l u m e

All Hypo Ischemic Nonisch Scar Normal

n 58 47 33 25 15 11 SPECT mn 59 e 54 ml 64 + 58 ml 54 + 47 ml 66 + 62 ml 85 + 76 ml 37 + 14 ml Cath nun 59 e 44 ml 64 _+ 47 ml 55 + 41 ml 65 + 49 ml 84 + 53 ml* 36 + 19 ml* Mn dif 0.0 ml 0.4 ml -0.9 ml 1.1 ml 1.4 ml 0.6 ml 95% cl -8-6 ml -8-7 ml -7-8 ml -9-11 ml -16-19 ml -7-8 ml /9 >0.99 0.92 0.82 0.83 0.97 0.86

Abbreviations as in Table 1. 7Statistically significant difference (p = 0.01) be tween subgroups by the unpaired t test.

Table 7. Paired t test: Ejection fraction

All Hypo l schemic Nonisch Scar Normal

n 58 47 33 25 15 10 SPECT mn 52 + 14% 50 + 15% 53 + 14% 52 _+ 15% 46 e 16% ~r 61 -+ 5%4 Cath mn 58 + 8% 56 + 17% 58 + 16% 57 + 17% 51 + 17%* 65 + 13%* Mn dif -5.3% -5.5% -5.7% -4.6% -5.1% -3.9% 95% cl -8%-3% -8%-3% -8%-3% -9%-1% -10%-0% -12%-5% p 10 -5 1 0-5 1 0-5 1 0-z 0.04 0.32

Abbreviations as in Table 1. *tStatistically significant difference (p = 0.001 ) be tween subgroups by the unpaired t test.

may, however, lead to different regression trends, insofar as the current investigation found volume intercepts <0.0 and slopes >1.0, whereas the only other reported linear regression analysis of GSPECT to angiographic volumes found intercepts >0.0 and slopes <1.0. 26

Comparison of Ventricular Models. It is likely that scintigraphic EDV is significantly lower due to our observation that angiographic endocardial drawings sometimes include more outflow tract than is visible rou- tinely on sestamibi gated tomograms, as illustrated by comparing Figure 1 with Figure 2. Whereas GSPECT outlines are typified by simple ovoid shapes, angiograph-

ic outlines can include a portion of the outflow tract at the base of the heart, angled at 45 degrees relative to the LV apex-to-base axis. The literature comparing GSPECT with angiography reinforces this impression in that angiographic endocardial borders are seen to extend into the outflow tract substantially further than in corresponding scintigraphic ventricular borders (e.g., Figure 1 in reference 22). Addit ional angiographic volume con- tributed by inclusion of a greater portion of the outflow tract would give rise to a larger percent volume difference between the two methods in the lower volume range than for larger volumes, consistent with the Bland-Altman


graph of Figure 3, B. This may also account for ESV from scintigraphy being more comparable with angiography, as systolic ventricular shapes are generally simpler and more symmetrical than at end-diastole. Valve leaflets themselves are not visible in sestamibi tomograms. Rather, radioactive counts end abruptly at the point at which myocyte concentration falls below detectable limits. For angiographic drawings it is more reliable to define the valve plane over valve leaflets when these are visible, but for sestamibi images it is natural to define valve plane points at locations where myocardial counts drop off substantially, 15,17,2° and consequently it is likely that endocardial outlines generated by most GSPECT methods will include less of the outflow tract than do angiographic endocardial borders.

Also contributing to LV volume underestimation for the specific GSPECT method studied here is that for the smallest ventricles, systolic counts from opposite myocardial walls appear to spill into one another and into the LV cavity. Although algorithms attempted to correct for this by using count subtractions varying with LV cavity background activity, 20 this approach may be inadequate for the smallest ventricles, especially when these are also hypercontractile. Fortunately, the clinical implications for those patients are minor, as their relatively smaller volumes and higher ejection fractions are associated with reduced probability of coronary artery disease. 1-4

Study limitations. Single-plane angiographic values were used for this study, and these would not be expected to be as accurate as those using biplane contrast angiography. Single-plane values are known to overestimate both biplane measurements and true ventricular volumes as measured from LV autopsy casts. 8,39 Given this limitation, it is likely that GSPECT comparisons with biplane angiography of Volumes would correlate even more closely than the already strong correlations found by this investigation. It is unfortunate that relatively few patients exhibited normal perfusion in this study, thereby limiting the extent to which conclusions can be drawn for that group, but this is an inevitable consequence of referral bias.

CONCLUSIONS

That strong correlations between GSPECT and angiography were found in the severely hypoperfused patient population studied in this investigation is encour- aging because this is the most clinically relevant group, in which it can be difficult to verify endocardial borders. Recent work in enhancing myocardial appearance in severely hypoperfused sestamibi studies may further extend the ability of GSPECT methods to provide accurate, reproducible functional parameters for this most challenging patient group. 40 Given any new diagnostic

test, once normal limits have been established, the question becomes whether test results significantly contribute to improving diagnostic accuracy. For the particular gated SPECT method used in this investigation, 20 normal limits have been documented, 41 on the basis of which it has been proven useful to diagnose functional abnormalities in hypertensive patients with no perfusion abnormalities. 42 EDV values from this technique in conjunction with normal limits have also helped to identify subgroups of coronary artery disease patients with LV dysfunction not suspected on clinical grounds. 43 Therefore, our preliminary experience with incorporating volume information suggests that gated SPECT volume measurements will indeed have a significant role in enhancing prognostic capability.~,2

We thank Helene Salensky for valuable assistance with scintigraphic processing.

References

1. Fuster V, Gersh B J, Giuliani ER, et al. The natural history of idiopathic dilated cardiomyopathy. Am J Cardiol 1981 ;47:525-31.

2. White HD, Norris RM, Brown M, et al. Left ventricular end-systolic volume as the major determinant of survival after recovery from myocardial infarction. Circulation 1987;76:44-51.

3. Pryor DB, Harrel FE Jr., Lee KL, et al. Prognostic indicators from radionuclide angiography in medically treated patients with coronary artery disease. Am J Cardiol 1984;53:18 -22.

4. Iskandrian AS, Hakki AH, Goel IP, et al. The use of rest and exercise radionuclide ventriculography in risk stratification in patients with suspected coronary artery disease. Am Heart J 1985;110:864-72.

5. Jones RH. Use of radionuclide measurements of left ventricular function for prognosis in patients with coronary artery disease. Semin Nucl Med 1987;12:95-103.

6. Shah PK, Maddahi J, Staniloff HM, et al. Variable spectrum and prognostic implications of left and right ventricular ejection fractions in patients with and without clinical heart failure after acute myocardial infarction. Am J Cardiol 1986;58:387-93.

7. Sandler H, Dodge HT. The use of single plane angiocardiograms for the calculation of left ventricular volume in man. Am Heart J 1968;75:325-34.

8. Fifer MA, Grossman W. Measurement of ventricular volumes, ejection fraction, mass and wall stress. In: Grossman W, ed. Cardiac catheterization and angiography. 3rd ed. Philadelphia: Lea and Febiger, 1986:282-300.

9. Borges-Neto S, Coleman RE, Jones RH. Perfusion and function at rest and treadmill exercise using Tc-99m sestamibi: comparison of one and two day protocols in normal volunteers. J Nucl Med 1990;31:1128-32.

10. Berrnan DS, Kiat H, Maddahi J. The new 99mTc myocardial perfusion imaging agents: 99mTc-sestamibi and 99mTc-teboroxime. Circulation 1991;84(suppl I):I-7-21.

11. Friedman J, Berman DS, Kiat H, et al. Rest and treadmill exercise first pass radionuclide ventriculography: validation of left ventricular ejection fraction measurements. J Nucl Cardiol 1994;1:382-8.

12. Parameshwar J, Keegan J, Sparrow J, Sutton GC, Poole-Wilson PA. Predictors of prognosis in severe chronic heart failure. Am Heart J 1992;123:421-6.

13. Faber TL, Akers MS, Peshock RM, Corbett JR. Three-dimensional motion and perfusion quantification in gated single-photon emission computed tomograms. J Nucl Med 1991;32:2311-7.


14. Piriz JM, Kiernan FJ, Eldin A, et al. Correlation of left ventricular ejection fraction by gated SPECT Tc-99rn sestamibi imaging with contrast ventriculography at subsequent cardiac catheterization [abstract]. J Nucl Meal 1996;37:105E

15. DePuey EG, Nichols K, Dobrinsky C. Left ventricular ejection fraction assessed from gated technetium-99m-sestamibi SPECT. J Nucl Med 1993;34:t871-6.

16. Cooke CD, Garcia EV, Cullom SJ, Faber TL, Pettigrew RI. Deter- mining the accuracy of calculating systolic wall thickening using a fast Fourier transform approximation: a simulation study based on canine and patient data. J Nucl Med 1994;35:1185-92.

17. Germano G, Kiat H, Kavanaugh PB, et al. Automatic quantification of ejection fraction from gated myocardial perfusion SPECT. J Nucl Med 1995;36:2138-47.

18. Everaert H, Franken PR, Flamen P, Momen A, Bossuyt A, Goris ML. Radial distribution of myocardial count rate density to measure left ventricular ejection fraction from gated perfusion SPECT studies [abstract]. J Nucl Meal 1996;37:211E

19. Nichols K, Rozanski A, Salensky H, DePuey EG. Accuracy and reproducibility of automated tomographic ventricular function measurements [abstract]. J Am Coil Cardiol 1996;27:215A

20. Nichols K, DePuey EG, Rozanski A. Automation of gated tomographic left ventricular ejection fraction. J Nucl Cardiol 1996;3:475-82.

21. Nichols K, DePuey EG, Salensky H, Rozanski A. Reproducibility of ejection fractions from stress versus rest gated perfusion SPECT [abstract]. J Nucl Med 1996;37:115E

22. Williams KA, Taillon LA. Left ventricular function in patients with coronary artery disease assessed by gated tomographic myocardial perfusion images: comparison with assessment by contrast ventriculography and first-pass radionuclide angiography. J Am ColI Cardiol

1996;27:173-81. 23. Vansant JE Faber TL, Folks RD, Rao JM, Garcia EV. Resting left van-

tricular volumes and ejection fraction from gated SPECT: correlation to first pass [abstract]. Circulation 1995;92(suppl I):I-11.

24. Faber TL, Cooke CD, Pettigrew RI, Garcia EV. Left ventricular volumes and mass from gated perfusion tomograms using a standard processing package [abstract]. J Nucl Med 1995;36:12E

25. Mochizuki T, Murase K, Tanake H, et al. Assessment of left ventricular volume using ECG-gated SPECT with technetium-99m-MIBI and technetium-99m-tetrofosmin [abstract]. J Nucl Med 1997;38:53-7.

26. Ttnajillo NP, Quaife RA, Adiseshan P, Groves B, Corbett JR. A new automated method for assessment of left ventricular function and myocardial perfusion using gated Tc-99m sestamibi imaging: comparison with cardiac catheterization [abstract]. J Nucl Med 1996;37: 179E

27. Klodas E, Rogers PJ, Sinak LJ, et al. Quantitation of regional ejection fractions using gated tomographic imaging with Tc-99m-sestamibi [abstract]. J Am Coll Cardiol 1996;27:215A.

28. Johnson LL, Verdesca SA, Xavier RC, Chang KK, Nott LT, Noto RB. Postischemic stunning affects walt motion scores on post stress gated sestamibi tomograms [abstract]. J Am Coll Cardiol 1996;27(suppl):

241A. 29. DePuey EG, Nichols KI, Slowikowski JS, et al. Fast stress and rest

acquisitions for technetium-99m-sestamibi separate-day SPECT. J Nucl Med 1995;36:569-74.

30. DePuey, EG. Artifacts in SPECT myocardial imaging. In: DePuey EG, Berman DS, Garcia EV, eds. Cardiac SPECT imaging. New York: Raven Press, 1995:169-200.

31. Van Train KF, Areeda J, Garcia EV, et al. Quantitative same-day rest- stress technetium-99m sestamibi SPECT: definition and validation of stress normal limits and criteria for abnormality. J Nucl Med 1993; 34:1494-502.

32. Crawford CR. CT filtration aliasing artifacts. IEEE Trans Meal Imaging 1991;10:99-102.

33. Tortoledo FA, Quinones MA, Fernandez GC, Waggoner AD, Winters WL. Quantification of left ventricular volumes by two-dimensional echocardiography: a simplified and accurate approach. Circulation 1983;67:579-84.

34. Motulsky H. Intuitive Biostatistics. New York: Oxford University Press, 1995:170-8.

35. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986; 1:307-10.

36. Bland JM, Altman DG. A note on the use of the interclass correlation coefficient in the evaluation of agreement between two methods of measurement. Comput Biol Med 1990;20:337-40.

37. Glanz SA. Primer of Biostatistics, New York: McGraw-Hill, Inc, 1992:67-91,278-87.

38. Elston RC, Johnson WD. Essentials of Biostatistics. Philadelphia: EA. Davis Company, 1994:205-10.

39. Wynne J, Green LH, Mann T, Levin D, Grossman W. Estimation of left ventricular volumes in man from biplane cineangiograms filmed in oblique projections. Am J Cardiol 1978;41:726-34.

40. Nichols K, DePuey EG, Salensky H, Rozanski A. Image enhancement of severely hypoperfused myocardia for computation of tomographie ejection fraction. 3 Nuct Med 1997;38:1411-7.

41. Nichols K, Rozanski A, Malhotra S, Cohen R, DePuey EG. Normal limits of left ventricular function parameters from myocardial perfusion gated SPECT [abstract]. J Nucl Med 1996;37:10517.

42. Rozansld A, Nichols K, Malhotra S, Cohen R, DePuey EG. Functional abnormalities in hypertensive patients with normal myocardial perfusion demonstrated by gated SPECT [abstract]. J Nucl Med 1996;37: 115P.

43. Malhotra S, Nichols K, DePuey EG, Cohen R, Rozanski A. Determination of occult left-ventricular dysfunction by gated myocardial perfusion SPECT [abstract]. Circulation 1996;94:I-240.

relationship of gated spect ventricular function parameters to angiographic measurements

Documents