Download Videodensitometric ejection fraction from intravenous digital

JACe Vol. 5, No.6
June 1985 :1377-8 1
1377
Videodensitometric Ejection Fraction From Intravenous Digital
Subtraction Right Ventriculograms: Correlation With First Pass
Radionuclide Ejection Fraction
ROBERT DETRANO , MD , PHD, WILLIAM MAcINTYRE, PHD, ERNESTO E. SALCEDO, MD , FACC ,
JAMES O'DONNELL, MD, DONALD A. UNDERWOOD, MD, FACC,
CONRAD SIMPFENDORFER, MD, FACC, RAYMUNDO T. GO, MD, KATHERINE BUTTERS, RT,
SUSAN WITHROW, RT
Cleveland. Ohio
Thirty-one consecutive patients undergoing intravenous
blurred mask digital subtraction right ventriculography
were submitted to first pass radionuclide angiography.
Second order mask resubtraction of end-diastolic and
end-systolic right ventricular digital image frames was
executed using preinjection end-diastolic and end-systolic frames to rid the digital subtraction images of mlsregistration artifact. End-diastolic and end-systolic perimeters were drawn manually by two independent
observers with a light pen. Ejection fractions calculated
from the integrated videodensitometric countswithin these
perimeters correlated well with those derived from the
First pass radionuclide ventriculography has become the
most widely used and accepted clinical tool for the determination of right ventricular ejection fraction (1-3). The
method is noninvasive and is based on the integration of
counts within the right ventricular perimeter at end-diastole
and end-systole. It is thus a densitometric method and free
of any dependence on assumption s regarding the shape of
the right ventricular cavity .
Though the method suffers from poor spatial resolution
and difficulty locating the plane of the pulmonic and tricuspid valves, these problems have been minimized by the
use of phase analysis, and validation of the method has been
achieved by demonstrating a fair correlation with direct right
ventricular cineangiographic ejection fraction (4). Further
validation of the first pass method has been accomplished
by comparing its results with those·obtained from multipleFrom the Department s of Cardiology and Nuclear Medicine , Cleveland
Clinic Foundation , Cleveland , Ohio. Manuscript received April II , 1984;
revised manuscript received December 26 , 1984, accepted January 16,
1985 .
Address for reprints: Emesto E. Salcedo, MD, Cleveland Clinic Foundation , 9500 Euclid Avenue, Cleveland , Ohio 44106.
© 1985 by the American College of Cardiol ogy
first pass radionuclide right ventriculogram (r = 0.84)
and the interobserver correlation was acceptable (r =
0.91). Interobserver differences occurred more frequently in patients with atrial fibrillation and in those
whose tricuspid valve planes were difficult to discern on
the digital subtraction right ventriculograms.
These results suggest that videodensitometric analysis
of digital subtraction right ventriculograms is an accurate method of determining right ventricular ejection
fraction and may find wide clinical applicability.
(J Am Coli CardioI1985;5:1377-81)
gated equilibrium imaging (5) and by comparing first pass
techniques using different radiotracers (6).
Intravenous digital subtraction ventriculography provides
high resolution right ventriculograms from a peripherally
injected bolus of contrast medium . Because light is digitally
represented as densitometric counts for each picture element
within the right ventricular contour of the subtracted images,
integration of these densitometric counts can be performed
and the resulting integral will be approximately proportional
to the volume of the chamber if the digital signal produced
by the opacified ventricle varies linearly with the volume
of iodine filling the chamber. Under these conditions , digital
subtraction ventriculography would have potential in the
estimation of videodensitometric right ventricular ejection
fraction , which is independent of geometric assumptions.
Videodensitometric left ventricular ejection fraction determined using a linearly calibrated imaging system has shown
good correlation with that derived by the Dodge area-length
method on the same digital subtraction left ventriculograms
(7) . We have shown an acceptable correlation between videodensitometrically determined left ventricular ejection
fraction and that derived from radionuclide left ventricu0735-1097/85/$3 .30
1378
DETRANO ET AL.
VIDEODENSITOMETRIC EJECTION FRACTION
lography (8). Though there has been some validation of the
videodensitometric method for the determination of right
ventricular ejection fraction (9), we know of no previous
investigation demonstrating a correlation between these derived ejection fractions and those derived from first pass
radionuclide right ventriculograms.
Methods
Study group. Thirty-one consecutive adult patients
undergoing intravenous digital subtraction ventriculography
in the right anterior oblique or frontal projection consented
to undergo first pass radionuclide ventriculography. There
were 26 men and 5 women whose mean age was 57 years
(range 25 to 79). The indications for digital subtraction
ventriculography were 1) assessment of wall motion abnormalities and global left ventricular function in patients
with coronary artery disease (22 subjects), 2) assessment of
right and left ventricular function in patients with known or
suspected dilated cardiomyopathy (3 patients), 3) assessment of right ventricular function in patients with chronic
obstructive pulmonary disease (3 patients) ; and 4) assessment of cardiac structural abnormalities in patients with
congenital heart disease (3 patients, 2 with atrial septal
defect and 1 with repaired ventricular septal defect) .
Digital Subtraction Yentriculagraphy
DigitaJ imaging system. The digital imaging system used
(Philips Medical System Polydiagnostic C and DYI 1) acquires, digitizes, logarithmically amplifies and mask subtracts at 30 frames/so The resulting runs of digitized and
subtracted image frames are displayed in real time during
acquisition and stored in analogue form on videotape. For
patient examinations, X-ray exposure factors used are between 65 and 90 rnA and between 75 and 95 kY. A 100
cm focal length, 36 lines/em, 8: I ratio grid was used with
the 9 inch image intensifier.
Image acquisition. The investigators conducting the
digital subtraction ventriculographic examination first instructed the patient on the importance of breath holding and
the avoidance of the Yalsalva maneuver. The X-ray camera
was positioned in the 30° right anterior oblique projection
and the patient was instructed to inspire deeply and hold
his breath. The digital system acquired a 320 ms mask image
and a brief run (l to 2 s) of visually satisfactory subtracted
images, after which 0 .5 cc/kg of contrast medium was injected at a flow rate of 10 to 15 cc/s through a 16 gauge
intravenous catheter with its tip in an antecubital vein. Image
acquisition continued until after the contrast medium had
cleared the right ventricle.
Postprocessing and computation. The videotaped runs
of subtracted images were redigitized in 256 x 256 pixel
matrixes and placed on a computer disk for post-processing.
JACCVol. 5, No.6
June 1985:1377-81
An observer chose a region of interest, which included the
right ventricle , by drawing an elliptical outline with a light
pen. By integrating over all the pixel. count values within
this outline for each frame of the study, the computer generated a time-activity curve whose maxima and minima
represented end-diastolic and end-systolic frames. Two segments were chosen from this curve , an initial one heartbeat
segment before the arrival of any dye in the heart and a
later segment at near peak filling of the right ventricle consisting of two to four heartbeats.
Computation steps. Figure 1 illustrates the segments of
the time-activity curve used for one subject's study. DB and
S8 denote, respectively, the assumed end-diastolic and endsystolic frames before the arrival of contrast medium in the
heart . We shall call these frames second order background
mask images. Dc and Sc denote, respectively, the enddiastolic and end-systolic frames when the right ventricle
was optimally opacified. Second order background-resubtracted images (DM, SM) were obtained from the subtractions:
These subtractions are phase-matched and thus not analogues to background subtraction methods used in radionuclide angiography. Figure 2A illustrates a second order
resubtracted right ventricular image in end-diastole.
The observer then drew perimeters with a light pen around
the tight ventricular endocardial edge of the background
resubtracted end-diastolic and end-systolic frames . The pulmonary and tricuspid valve planes were chosen with the
help of a playback of the real time motion study. The com-
Figure 1. Time-activity curve demonstrating early precontrast
segment (left) and latersegment whenthe rightventricle wasladen
with contrast medium. DB and SB = assumed end-diastolic and
end-systolic frames before arrival of contrast medium in the heart;
Dc and Sc = end-diastolic and end-systolic frames when right
ventricle is optimally opacified.
rn
IZ
:;)
oo
o
a:
lUI
::E
o
I-
iii
o.
~
l
z
2.8
TIME (seconds)
JACC Vol. 5, No.6
June 1985:1377-81
DETRANO ET AL.
VIDEODENSITOMETRIC EJECfION FRAcnON
1379
right heart frames of the first pass image series . The approximate region of the right ventricular chamber was outlined by manual generation of a fixed region of interest.
The computer then calculated a counts/frame profile to identify the end-diastolic and end-systolic images . Refined regions of interest were then generated manually to coincide
with the right ventricular chamber outlines to obtain enddiastolic and end-systolic ventricular counts. Ejection fraction (EF) was calculated from the differences between enddiastolic and end-systolic counts according to the standard
formula :
EF == Cd - Cs/Cd,
Figure 2. A, Second order resubtracted end-diastolic image used
in the calculation of the relative end-diastolic volume from the
right ventriculogram. B, Background-corrected end-diastolic right
ventricular image from first pass radionuclide ventriculogram.
puter then summed these pixel values to obtain counts/image.
The formula used to obtain the ejection fraction (EF) was:
EF == Counts (DM ) - counts (SM).
Counts (DM )
where Cd = end-diastolic counts and Cs = end-systolic
counts .
Statistical calculation. Paired t tests were used to compare the mean ejection fractions by the two methods. Digital
subtraction ventriculographic eject ion fractions were compared with the radionuclide results using linear regression
and Pearson correlation coefficients. Interobserver variability was measured by Pearson correlation and intraclas s
correlation coefficients ( 10).
These results were averaged over the two to four chosen
beats . All post-processing and computation steps were repeated by a second independent and blinded observer.
Radionuclide Ventriculograms
First pass radionuclide angiograph y was performed using
a commercial computerized multicrystal scintillation camera
system (Baird System-77). With the patient sitting upright
and in the anterior position, a rapid intravenous injection
was made by placing a 20 gauge indwelling catheter into
the right antecubital vein using 15 /LCi of technetium-99m
diethylenetriaminepentaacetic acid (DTPA) dissolved in less
than I ml of normal saline solution. An additional 20 ml
of normal saline solution was used to flush the tracer bolus
into the central circulation . Data were then recorded at 20
frames/s for 30 s and stored on magnet ic disc .
Data analysis. After appropriate corrections for crystal
uniformity and background activity, distinct right ventricular images (Fig. 2B) could be identified by displaying the
Results
Table I shows the mean ejection fractions by either method
for the entire group and for those patients with coronary
artery disease. dilated cardiomyopathy . chronic obstructive
pulmonary disease and congenital heart disease. There was
no significant difference in the mean ejection fractions calculated by the two methods. Figure 3 shows the scatter gram
for the right ventricular ejection fractions determined by the
calculations of the two independent observers. A satisfactory correlation (r = 0.91) resulted from both the Pearson
and intraclass correlation methods . The standard error was
5.5. Examination of the graph , however , reveals that for
eight patients there were significant differences (>6%) between the two observers' calculations. Review of the right
ventricular images of these particular subjects with the superimposed ventricular outline s showed that in every case
there were either discrepancies in the determination of the
Table 1. Mean Ejection Fractions (%) Determined by Both Methods in 31 Cases
No. of
Subjects
Entire group
Coronary arter y disease
Dilated cardiomyopathy
Chronic obstructive
pulmonary disease
Congenital heart disease
31
22
3
3
3
Mean Ejection
Fraction-DSV
45.1
49 .7
25.3
36 .3
±
±
±
±
12.6
14.6
0 .5
5.7
40 .0 ± 5.7
Mean Ejection
Fraction-RNV
43 .0 :!:
50 .0:!:
16 .0 ±
34 .3 ±
15.0
17.3
4 .0
7.8
38.3 :!: 7.6
Data are represented as mean values ± standard dev iation . DSV = digital subtraction ventriculography;
RNV = radionuclide ventriculography.
DETRANO ET AL.
VIDEODENSITOMETRIC EJECTION FRACTION
1380
100
90
JACC Vol. 5, No.6
June 1985:1377-81
"(0182 +0 .95 X
r o O.9 1
80
70
N
60
:
50
°40
30
20
10
o
10
20
30
40
50
60
70
80
90
100
OBS I
Figure 3. Scattergram and regression line for digital subtraction
right ventricular ejection fractions calculated by the two independent observers (08S) in 31 cases.
tricuspid valve plane or an irregular cardiac rhythm (atrial
fibrillation) .
Figure 4 shows the correlation between the digital subtraction right ventricular ejection fractions calculated videodensitometrically and the radionuclide right ventricular
ejection fraction s. The correlation coefficient is 0.84, the
standard error 8.3 .
Discussion
We have previously reported (8) a satisfactory correlation
between videodensitometrically calculated left ventricular
ejection fractions derived from intravenous digital subtraction ventriculograms and those derived from radionuclide
ventriculograms. The data presented in this report similarly
support an acceptable correlation between videodensitometric right ventricular ejection fractions derived from digital subtraction angiograms and first pass radionuclide right
Figure 4. Scattergram and regression line for radionuclide right
ventricular (RNV) ejection fraction (vertical axis) versus digital
subtraction right ventricular (DSV) ejection fraction (horizontal
axis) in 31 cases.
100
90
,,(""i.44+0.99X
r=0 .84
80
70
•
60
::>
z
a:
50
40
30
20
10
o
10
20
30
40
50
DSV
60
70
80
90
tOO
ventricular ejection fractions . Videodensitometry , like
radionuclide ventriculography, offers the advantage of the
lack of dependence on the shape of the right ventricular
cavity . The validity of the method holds even when this
cavity does not conform to a convenient geometric form.
This may explain the rather good correlation between the
two densitometric techniques .
Limitation of the method. Physical factors related to
scatter of X-ray and imperfections in the linear transfer of
energy from the input surface of the image intensifier to the
television camera have been implicated as causes of error.
Our observations (II) and those of others (9,12) indicate
that these nonlinearities of signal response contribute a relatively small amount to the errors of videodensitometrically
calculated ventricular ejection fraction. However, they cannot be ignored because their contribution to the error is
unpredictable and might conceivably be large in individual
cases. Fortunately, technical solutions (13,14) to these difficulties of nonlinearity have been proposed and will soon
be available for use.
A relatively large error in calculation of right ventricular
ejection fraction might be introduced by overlap of the right
ventricle by part of the right atrium making accurate location
of the tricuspid valve plane difficult or impossible. Review
of our right ventriculographic studies revealed that in six
cases in which significant interobserver variability occurred
(> 6%), there were obvious differences in the tricuspid valve
determination . We believe that this is the largest source of
error in the videoden sitometric determination of right ventricular ejection fraction.
Clinical applications. Digital subtraction right ventriculography can be performed using relatively small amounts
of angiographic contrast medium. We have achieved excellent qualitative and quantitative results with as little as
15 to 20 ml of Renografin-76 injected in a peripheral arm
vein. In fact, we have found peripheral intravenous injections to be as good as if not superior to injections in the
superior vena cava (15) . The ease of performance makes
this type of study applicable in many clinical situations. For
example , there are some data to suggest that deteriorating
right ventricular function may be an important determinant
of outcome in patients with ventricular septal infarction and
rupture (16). Patients with chronic obstructive pulmonary
disease could be followed up with these studies to determine
the beneficial effect of bronchodilator therapy (17) . Involve ment of the right ventricle by an inferior myocardial infarction could also be easily assessed . The effect of essential
hyperten sion on right ventricular function , previously studied using radionuclide ventriculography (18) , could be assessed with digital subtraction right ventriculography. Assessing improvement in right ventricular systolic function
after mitral valve repair or replacement and closure of ventricular septal defect is another possible application (19,20) .
The addition of videodensitometric analysis to the inter-
DETRANO ET AL.
VIDEODENSITOMETRIC EJECTION FRACTION
JACC Vol. 5, No.6
June 1985:1377-81
pretation of intravenous digital subtraction right ventriculograms adds the advantage of quantitative estimate of right
ventricular function to the high spatial resolution of radiographic images. Because the ejection fractions so calculated
are independent of geometric assumptions, their accuracy
is reliable even when these assumptions are not valid. These
factors promise to make digital subtraction right ventriculography a highly useful clinical tool in the noninvasive
laboratory.
The secretarial assistance of Paula LaManna and the statistical calculations
of Judy Leatherman are deeply appreciated.
1381
analysis of digital subtraction angiograms. J Am Coli Cardiol
1983;52:871-5.
8. Detrano R, MacIntyre W, Salcedo EE, et al. Videodensitometric ejection fraction from intravenous digital subtraction left ventriculograms:
correlation with conventional direct contrast and radionuclide ventriculography. Radiology 1985;155:19-23.
9. Nissen SE, Friedman B, Waters J, Booth D, DeMaria A. Right ventricular ejection fraction by videodensitometry of intravenous digital
subtraction angiograms: experimental validation and initial clinical
results (abstr). J Am Coli Cardiol 1984;3:589.
10. Bartko JJ. The intraclass correlation coefficient as a measure of reliability. Psychol Rep 1966;19:3-11.
II. Detrano R, MacIntyre W, MacIntyre R, et al. How do videodensitometric ejection fractions from intravenous digital subtraction ventriculograms compare with first pass radionuclide ejection fractions?
(technical aspects). Proc SPIE 1985 (in press).
12. Nalcioglu 0, Roeck WW, Pearce JG, Gillan GD, Milne XK. Quantitative fluoroscopy. IEEE Trans Nucl Sci 1981;28:219-30.
References
I. Berger HJ, Zaret BL. Noninvasive radionuclide assessment of right
ventricular performance in man. Cardiovasc Clin 1979;10:91-104.
13. Seibert JA. Characterization and Correction of Veiling Glare in Xray Image Intensified Fluoroscopy. PhD Thesis, 1983. University of
California.
14. Naimudden N, Hasegawa BH, Dobbins JT, et al. Chest radiography
using a digital beam attenuation (abstr). Radiology 1984;53:38.
2. Berger HJ, Matthay RA, Loke J, Marshall RC, Gottschalk A, Zaret
BL. Assessment of cardiac performance with quantitative radionuclide
angiocardiography: right ventricular ejection fraction with reference
to findings in chronic obstructive pulmonary disease. Am J Cardiol
1978;41:897-905.
15. Detrano R, Simpfendorfer C, MacIntyre R, Salcedo EE. Cardiac digital subtraction angiography: peripheral versus central intravenous dye
injection. Cardiovasc Intervent Radiol 1985 (in press).
3. Johnson LL, McCarthy DM, Sciacca RR, Cannon PJ. Right ventricular
ejection fraction during exercise in patients with coronary artery disease. Circulation 1979;60:1284-91.
16. Radford MJ, Johnson RA. Ventricular septal rupture. A review of
clinical and physiologic feature and an analysis of survival. Circulation
1981;64:545-53.
4. Steele P, Kirch D, LeFree M, Battock D. Measurement of right and
left ventricular ejection fractions by radionuclide angiocardiography
in coronary artery disease. Chest 1976;70:51-6.
17. Matthay RA, Berger HJ, Loke J, Gottschalk A, Zaret BL. Effects of
aminophylline upon right and left ventricular performance in chronic
obstructive pulmonary disease. Am J Med 1978;65:903-10.
5. Maddahi J, Berman DS, Matsuoka DT, et al. A new technique for
assessing right ventricular ejection fraction using rapid multiple-gated
equilibriumcardiac blood pool scintigraphy. Circulation 1979;60:581-9.
18. Ferlinz J. Right ventricular performance in essential hypertension.
Circulation 1980;61:156-62.
6. Goldberg MJ, Mantel J, Friedon M. Intravenous xenon!" for the
determination of radionuclide first pass right ventricular ejection fraction. Am J Cardiol 1981;47:626-32.
7. Tobis J, Nalcioglu 0, Seibert A, Johnston WD, Henry WL. Measurement of left ventricular ejection fraction by videodensitometric
19. Goldman ME, Horowitz SF, Meller J, Mindbach B, Teichholtz LB.
Recovery of right ventricular function following repair of acute ventricular septal defect. Chest 1982;82:59-63.
20. Johnston DL, Lesoway R, Kostuk WJ. Changes in ventricular function
after mitral valve surgery: assessment by radionuclide ventriculography (abstr). Circulation 1983;68(suppl 3):344.