IEEE TRANSACTIONS ON MEDICAL IMAGING, VOL. MI-1, NO. 2, OCTOBER 1982
teristics of a quad BGO detector for positron emission computed
tomographs," in Proc. 3rd Symp. Physical and TechnicalAspects
of Transmission and Emission Computed Tomography, Tokyo,
Japan, Sept. 1980, pp. 94-95.
[271
H. Murayama,
E. Tanaka, N. Nohara, and T. Hayashi, ''Timing
and positioning circuit for a quad BGO detector for a positron
computed tomograph," in Proc. 3rd Symp. Physical and Tech-
nical Aspects of Transmission and Emission Computed To-
mography, Tokyo, Japan, Sept. 1980, pp. 96-97.
(28]
N. Nohara,
E. Tanaka, H. Murayama, K. Ishimatsu, andT. Hayashi,
"Performances of quad BGO detectors for positron emission
computed tomography," in Proc. 4th Symp. Physical and Tech-
nical Aspects of Transmission and Emission Computed To-
mography, Tokyo, Japan, 1981, pp. 75-76.
(29] H. Murayama,
N. Nohara, E. Tanaka, and T. Hayashi, "A quad
BGO detector and its timing and positioning discriminator for
positron computed tomography," Nucl. Instrum. Meth., vol. 192,
pp. 501-5 11, 1982.
[30] T. F. Budinger, S. E. Derenzo, G. T. Gullberg, W. L. Greenberg,
and R. H. Huesman, "Emission computer assisted tomography
with single-photon and positron annihilation photon emitters,"
J. Comput. Assist. Tomogr., vol. 1, no. 1, pp. 13 1-145, 1977.
[311
K. Takami, K. Ishimatsu, T. Hayashi, K. Ueda, F. Kawaguchi,
K. Okajima, A. Ohgushi, S. Inoue, Y. Takakusa, S. Nakase, and
E. Tanaka, "Design consideration for a continuously rotating
positron computed tomograph," IEEE Trans. Nucl. Sci., vol.
NS-29, to be published.
Estimation of the Local Statistical Noise in
Emission Computed Tomography
N. M. ALPERT, D. A. CHESLER, J. A. CORREIA, R. H. ACKERMAN, J. Y. CHANG, S. FINKLESTEIN,
S. M. DAVIS, G. L. BROWNELL, AND J. M. TAVERAS
Abstract-A simple modification of the filtered backprojection algo-
rithm is presented for the computation of the local statistical noise in
emission computed tomography. The technique is general in that any
distribution of radioactivity may be accommodated. When applied to
positron emission tomography, it is shown that the effects of photon
absorption, random coincidences, radioactive decay, and detector non-
uniformity may be included. Calculations have shown the effects of
resolution, object size, and photon absorption on the statistical noise of
disk-shaped emitters. Comparison of calculation and experiment show
close agreement both in magnitude and spatial variation. Measurements
of the noise level in tomograms of the brain obtained during continuous
inhalation of 150-C02 demonstrate that estimates of radioactivity
concentration with a precision of a few percent are readily attainable.
INTRODUCTION
W
rITHIN the limits imposed by finite spatial resolution,
emission computed tomography provides an estimate
of the radioactivity concentration at any point in the tomo-
graphic plane. In order to take full advantage of the quantita-
tive potential of emission computed tomography, it is essential
to have reliable estimates of the precision of measurement.
Without estimates of the precision, one cannot determine the
significance of differences in radioactivity concentrations within
a slice, compute the error propagated in the calculation of
quantitative values, or compare results obtained with different
techniques. By precision of measurement, we mean the varia-
Manuscript received March 26, 1982; revised August 12, 1982. This
work was supported by the National Institutes of Health, U.S. Public
Health Services, under Research Grant NS 10828.
The authors are with the Department of Radiology, Massachusetts
General Hospital, Boston, MA 02114.
tion in the computed value of local activity concentration due
to the Poisson fluctuations in the projection data.
Several investigators have studied the effects of statistical
noise in emission computed tomography. Theoretical analyses
have established the relationship among the rms noise, the
number of detected events, and the spatial resolution for the
case of a disk with uniform activity distribution [1], [2].
Somewhat more complicated geometries were systematically
studied by Budinger et al. [3] using computer simulation, and
algorithms for arbitrary activity distributions have been pre-
viously derived [4]. The formulas and nomograms developed
to date provide information useful in designing experiments,
but cannot be directly applied to practical measurements.
In the experimental situation, it is necessary to consider
several additional contributions to the statistical noise. These
include the effect of photon attenuation, corrections for varia-
tion in detector uniformity and/or linearity, corrections for
random coincidences in positron emission tomography (PET),
corrections for scattered radiation, and the changes in statisti-
cal quality due to radioactive decay during measurements with
short-lived radionucides.
In PET, the measured coincidence rate (prompt coincidences)
can be considered to be the sum of the true coincidence rate,
the random coincidence rate (time uncorrelated), and the
prompt scatter rate. Prompt scatter is not easily measured,
and is usually minimized by appropriate shielding of the
tomograph, whereas random coincidence rates can be routinely
estimated during each experiment. Neglecting the prompt
scatter contribution, the true coincidence rate, which is directly
and linearly proportional to the underlying quantity of radio-
0278-0062/82/1000-0142$00.75
©
1982 IEEE
142