460 IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. BME-23, NO.

6, NOVEMBER 1976

Scattering of Ultrasound by Blood

KoPING K. SHUNG, MEMBER, IEEE, RUBENS A. SIGELMANN, SENIOR MEMBER, IEEE, AND JOHN M. REID, MEMBER, IEEE

Abstract-The ultrasonic volumetric scattering cross section of the wave from blood with that of a perfect flat reflector. In their
erythrocyte has been measured over a range of frequencies by com- work a general scattering coefficient was defined. No specific
paring the nms value of the gated backscattered signal from the blood scattering mechanism was introduced.
with that of a wave reflected from a flat reflector with known reflec-
tion coefficient. It is found to be proportional to the fourth power
In this paper we will present the experimental result for blood
of the frequency predicted by the wave scattering theory for small par- systematically collected based on the substitution method.
ticles in the frequency range from 5 MHz to 15 MHz. The relation be- Twersky's wave scattering theories will be applied to account
tween the scattering coefficient and the hematocrit is also examined up for these results. It is our hope that this paper can furnish
to a hematocrit of 45 percent. This coefficient is found to increase some preliminary quantitative information about the scatter-
along with the hematocrit until it reaches a maximum around hemato-
crit = 26 percent and then decreases as the hematocrit increases.
ing mechanism of ultrasound in blood and thus provide a
Twersky's wave scattering theory is applied to describe this result. better understanding of the interaction of ultrasound and
blood.
I. INTRODUCTION In Section II a brief description of the blood as a random
medium will be given. In Section III we will discuss the theo-
LTRASOUND has been used increasingly in medicine retical background of the substitution method. Section IV
both as a therapeutic agent and as a diagnostic tool. In contains the experimental arrangements, including the elec-
these applications the accurate measurement of acoustic tronic system and the acoustical apparatus, and the experi-
wave propagation parameters in biological tissues such as ve-
mental results. In Section V an analysis of the results is given.
locity, absorption, and scattering is required for the optimal
performance of the equipment. For instance, the absorption II. BLOOD AS A RANDOM MEDIUM
and scattering characteristics of a certain tissue determine the
depth of penetration of an ultrasonic beam in that tissue. The Human blood is composed of a liquid called plasma in which
explicit information on velocity is essential for the interpreta- are suspended erythrocytes (red blood cells), leukocytes
tion of an A-scan picture. Some data on the velocity and ab- (white blood cells), and platelets. The erythrocyte is an elas-
sorption are available in the literature [1], [2] . The proper- tic, nonnucleated biconcave disk with an average diameter of
ties of scattering in biological tissues, unfortunately, are 7 gm and an average thickness of 2 jim. The mean corpuscular
virtually unknown. This is true even in the case of blood de- volume is 87 um3. There are about 5 X 109/cm3 of erythro-
spite the fact that the operation of ultrasonic Doppler flow- cytes, 7.5 X 106/cm3 of leukocytes and 3.5 X 108/cm3 of
meters in cardiology and blood flow studies depends directly platelets in an adult [9]. The number of red blood cells
on the scattering properties of the red blood cells.
(RBC's) is much larger than that of the white cells and the
The scattering mechanisms for waves propagating in a volume of a red cell is much larger than that of the platelet
dense distribution of scatterers are very complex. A com- [9]. Thus the scattering of ultrasound by blood presumably is
plete solution of this problem requires the knowledge of the due to the RBC's [10].
pair distribution function P(r5, rt), i.e., the probability of find- The velocity of propagation and absorption of ultrasound
ing sth particle at rs and tth particle at rt. However, the expli- has been measured by Carstensen and Schwan [11], [12].
cit form for P(rs, rt) is not available. Twersky solved the scat- Their results indicate that the absorption is approximately
tering of waves by random distribution of scatterers in terms linearly proportional to frequency and hematocrit (for hemat-
of the one particle distribution function [3]. For more dense ocrit < 40 percent). For a 5-MHz sound wave, the absorption
medium he introduced the heuristic "hole" approach [4]-[7] . in blood (hematocrit = 45) is about 0.8 dB/cm. They also
A mathematical formulation taking into account the presence show that the velocity dispersion in blood is negligible and the
of pulse length, absorption, and gatewidth for an experimental velocity increases by 0.7 percent for an increase of 10 hemato-
technique (the substitution method) was derived by Sigelmann crit points.
and Reid [8]. In this method the scattering coefficient can be Some of the acoustically significant properties of the eryth-
measured by comparing the rms value of the gated scattered rocyte, plasma, 0.9 percent normal saline, and water are shown
inTableI [1], [2], [13].
Manuscript received June 11, 1975. This work was supported in part
by the National Institutes of Health under Grant GM-16436. III. THEORETICAL BACKGROUND
K. K. Shung and J. M. Reid were with the Department of Electrical
Engineering, University of Washington, Seattle, WA 98195. They are Sigelmann and Reid [8] showed that for an ultrasonic beam
now with the Institute of Applied Physiology and Medicine, Providence with an effective width D impinging on a cylindrical region
Hospital, Seattle, WA 98122. that contains the scatterers (Figs. 1 and 2) the received power
R. A. Sigelmann is with the Department of Electrical Engineering,
University of Washington, Seattle, WA 98195. backscattered from the scatterers can be expressed as
SHUNG et al.: SCATTERING OF ULTRASOUND BY BLOOD 461

TABLE I Ar effective receiving aperture of the transducer,

KNOWN ACOUSTICALLY SIGNIFICANT PROPERTIES OF BLOOD aLm attenuation constant of sound in medium I,
Density, p Adiabatic Compressibility ,xlO12 ax attenuation constant of sound in medium II,
gm/cm3 cm /dyne (at 200C) S cross section of the beam = rrD2/4,
Water 0.998 46.1 T period of the bursts,
0.9% Saline 1.005 44.3 X wavelength,
Plasma 1.021 40.9 c velocity of propagation in medium II,
Erythrocyte 1.092 34.1
time required for the wave to travel from the front
wall of the chamber to the front face of the gated
volume,
t2- tl gatewidth,
r the pulsewidth, and
17 backscattering coefficient.
The physical meaning of the backscattering coefficient is evi-
dent from [8, eq. (13)]. It is the mean-square pressure scat-
tered by an elementary volume of scatterers per solid angle in
the direction of the transducer with a plane wave of unit ampli-
tude incident on it. If the concentration of the scatterers is
low, this coefficient should be proportional to nO (the number
of scatterers per unit volume), i.e., single scattering will suffice
to describe the scattering mechanism. However, when the con-
centration increases, the scattering process becomes more com-
plicated. It could arise from either multiple scattering or the
gradual disappearance of the random nature of the movement
of the scatterers. We will not present the detailed discussion of
this subject here, but will give it in Section V, where we will
compare the experimental results with the scattering theories.
It is important to note that by measuring the scattering coeffi-
Fig. 1. Ultrasonic beam incident upon a cylindricalvolume of scatterers.
cient, r1, the scattering characteristics of the scattering medium
can be revealed.
y
The received power reflected from a flat reflector of reflec-
tion coefficient -1 located at a distance R + a/2 away from
the transducer is
prA2 e-4Cm(R+a/2)
T4 2 (R + a/2)2 (2)
Dividing (1) by (2) and neglecting the reflection loss through
mediums I and II, the attentuation of sound beam in medium I
and a/2 (a/2 << R), we obtain an expression from which 71 can
be calculated if all the parameters are known:
Ps 27Sc(t2 - tl) (eTc2 - eTca\fecc(t2t1) -
Pr R2 2-rca 2cat(t2 - t1)
MEDIUM I *e'0ac2 +tl) (3)
Taking 10 logarithm on both sides of (3), we have
10 log PS - 10 log Pr = 10 log 71 + 10 log 2SC02
MEDIUM 2
Fig. 2. Systems of coordinates for the backscattering formulation. ( eTCa - e--7C eC002-ti) _ca(t2-ti)
2wca 2cax(t2 - tl) I

PS A r e4amR C15 (e-2eti a - e-2Ct2cI) (eT¢r - eTCr) (1) *e-"(2 -tl)J (4)
8TX2Rac
From (4) it can be seen that if the difference between the scat-
where tered power and reflected power in decibels is obtained, i7 can
Pt power transmitted by the transducer, be calculated readily since S, c, R, r, a, t2, tI are either known
462 IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, NOVEMBER 1976

Fig. 3. Block diagram of the electronic equipment.

or measurable. In this way, the cumbersome procedures re-

quired to measure the incident power at the site of the scatter-
ers are no longer necessary.
LEVEL
IV. EXPERIMENTAL ARRANGEMENTS AND RESULTS
The electronic system used in the experiment is shown in
Fig. 3. Sine wave bursts are generated by the Arenberg pulsed
oscillator. The bursts are then introduced into water by a
PZT-5 air-backed transducer of 1 cm in diameter operated in
thickness mode. The attenuator placed in front of the ampli-
Fig. 4. Acoustical apparatus.
fier maintains the input of the amplifier at a level compatible
with the dynamic range of the amplifier. Time gating allows
only the scattered wave from a specified volume inside the The hematocrit of the blood sample is measured before and
blood chamber to be measured by an rms indicating meter. after the experiment. No significant hemalysis is observed.
The scattered wave is received by the same transducer. The temperature of the water tank is always maintained at
The acoustical apparatus consisting of the PZT transducer, 370C.
the blood container, and the flat reflector is shown in Fig. 4. Blood samples were prepared from outdated bank RBC's by
The apparatus is rigidly and vertically supported. It can be diluting them with Ringer's injection solution. Outdated bank
moved both horizontally and vertically along the bench. The RBC's were obtained by centrifuging outdated citrated bank
blood chamber is a cylindrical container made of condom blood supplied by a local blood bank.
rubber which is almost transparent to the ultrasonic energy.' The following procedures were followed for all blood
The blood is circulated through the chamber by pumping. Cir- samples. One ml of sodium heparin (1000 units) was added to
culation is required to prevent the clumping and settling of the each 100 ml of blood sample to prevent clotting. The blood
cells in the chamber. was then passed through two layers of gauze sponges to filter
The flow rate of the blood in our experiment is approxi- out the clots or any other impurities. Before any measure-
mately 4 ml/s. As can be seen from Fig. 4, the shape of our ments were performed all samples were heated to 370C.
chamber and the positions of the inlet and outlet indicate that This experimental technique requires, firstly, the determina-
the flow of blood in the chamber certainly is not a uniform tion of the cross section of the ultrasonic beam and the rela-
one. Severe turbulence could be present. We noted that this tion between the echo amplitude and range for a flat reflector.
does not have very significant effect on the scattering for flow The diameter of the ultrasonic beam is measured by hanging a
rate between 3 ml/s and 8 ml/s, provided the hematocrit is stainless steel ball at the end of a nylon thread from the mov-
below 40 percent. For the samples of hematocrit greater than able carrier placed in the position chosen for the center axis of
50 percent, the dependence of scattering upon flow rate be- the blood chamber. This method could cause a small error be-
comes evident and it seems then that the turbulence effect can cause the nylon thread intersects a different cross section of
no longer be neglected. We are, therefore, not able to obtain the beam. However, we assume this difference is negligible.
a consistant result for hematocrit greater than 50 percent. The beam widths for the transducers used in our measure-
SHUNG et al.: SCATTERING OF ULTRASOUND BY BLOOD 463

-3
10

/1
0

m
HMTC-26/
w
-
I-
F - 5MHZ
l
<- 5 z
0

w , F-i5MHZ
LUz
0
n
CD
2z
r
-10 F Ir-

Cl,

I I I I
10 15 20 25 30
RANGE, cm
Fig. 5. Echo amp]litude for a stainless steel flat reflector is plotted
versus range.

ments range from 2 mm to 6 mm. The relations between the lo-5

5 10 20
echo amplitude and range for a flat reflection for transducers FREQUENCY, F(MHZ)
of 5 and 15 MHz are plotted in Fig. 5. The far field region for Fig. 6. Backscattering coefficient is plotted versus frequency.
these two transducer starts at 10 cm and 24.5 cm, respectively.
The backscattering coefficient for blood samples with he-
matocrit of 8 was measured using transducers with resonant tween two red cells is about 10 percent of its diameter for a
frequency at 5, 7, 8.5, 10, and 15 MHz. 15-MHz bursts were blood sample with a hematocrit of 45 percent. Their motion
obtained by operating a 5-MHz transducer at its third har- will be hindered by the presence of other particles. Thus, it
monic. All the measurements were done with a pulsewidth of seems that the particles can no longer be considered indepen-
4 ,us. In this length, the burst has reached steady-state, there- dent of one another.
fore, it can be assumed monochromatic. It is short enough
that the scattering from the blood can be separated from the V. ANALYSIS OF THE RESULTS
reflections of the chamber walls. The backscattering coeffi- In this section we attempt to express the scattering coeffi-
cients also were measured at three different frequencies; 5, cient in terms of the concentration of the scatterers, the fre-
8.5, and 15 MHz for other hematocrits. These results are quency of the incident wave, and the physical characteristics
plotted in Fig. 6. The fourth power of frequency is also of the scatterer.
plotted. It shows that the scattering from blood obeys Ray- In Section III we mentioned that the backscattering coeffi-
leigh's scattering theory for small particles and its fourth cient defined in [8] is the mean-square pressure scattered by
power frequency dependence will not be affected by the in- an elementary volume of scatterers per solid angle in the back-
creasing hematocrit. ward direction with a wave of unit amplitude incident on it.
The backscattering coefficients for blood samples with dif- Twersky has shown that as a plane wave propagates through
ferent hematocrits were measured at frequencies of 5, 8.5, and a slab of randomly distributed scatterers the coherent energy
15 MHz. The results are shown in Figs. 7-9. They all indicate for particles with dimension <K the wavelength ()) can be
that the scattering increases along with the hematocrit until it written as [3]
reaches a maximum in the region where hematocrit ranges
from 24 to 30. The scattering is approximately linearly de- (P(Z) P*(Z) > (P(O) P*(O) > e-n(aa+ars)Z
= (5)
pendent upon the number of particles present in the medium where ( ) means ensemble average.
for hematocrits below 8 percent. After this point, compli- n function of volume concentration of the scatterers
cated scattering process appears. It could arise from multiple (Wo) = n0 for WO 0, = n0/(l - WO), for otherwise,
-

scattering or from the gradual disappearance of random nature Ua absorption cross section of the scatterer, and
of the motions of these scatterers, i.e., we can no longer con- as scattering cross section of the scatterer.
sider that the scatterers are randomly distributed. Especially
as the concentration gets very high, such as in normal whole Equation (5) shows that the energy loss per unit volume due
blood, the particles are very close. The average distance be- to scattering is given by
464 IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, NOVEMBER 1976

14 0 as scattering coefficient = n as (6)

O DATA POINT and
12 ---- THEORETICAL CURVE
- ~~~0 FITTED
/I Os = J I f(¢, o) dQ
M where Q represents solid angle.
D
10 _

ft
/
0 In f(l, 8) is the scattering ampli-
tude in direction o with a plane
I
I' wave of unit amplitude incident
0
0 a N in Pdirection.
I
x
UI
,,w 6 I; The backscattering coefficient is, therefore,
0
z
71 = n I f(+Il )12. (7)
4
1 , ~~~~~f*s5 MHZ
It should be noted that Twersky has derived two expressions
for f [3], [4]. One representation (the free space formalism)
2 is limited only to very sparse concentration of scatterers. The
other representation (the two space formalism) is more general
I I f I and considers the exciting wave propagating in K space but the
0 10 20 30 40 50 60 70 scattered field is analytically constrained to radiate into free
IHMTC %
K-space where K is the bulk propagation coefficient for the
Fig. 7. Scattering coefficient is plotted versus hematocrit at 5 MHz.
slab medium and can be expressed as K = k + 27rnf(il i)/k, and
k = free space propagation coefficient. We use the two-space
representation for f in (7) and obtain
o DATA POINT I? = °W o 9 k4 a ( | of3 +| |P
P e
(8)
10
- THEORETICAL
CURVE FITTED
where
,,_
3

8 a radius of the scatterer,

IE /
3 adiabatic compressibility of the scatterer,
0 /s o:
0!)
QO \
,B' bulk adiabatic compressibility of the scattering medium,
6 _ o'Q
'IOx0
U
L / \c.
~~~~0 Pe density of the scatterer, and
80! 4 -
,
0 p' bulk density of the scattering medium.
z /
Twersky [5] also proved that the mixture relation holds for
42
U
_ f * 8.5MHZ compressibility and density if PeIP - 1, where p is the density
C/ of the imbedding medium, i.e.,

0
I
10 20
I 1
30
1
40 50 60
fI'=(1- Wo)il3+Wo1Pe
HMTC % p'=(l- Wo)p+Wope. (9)
Fig. 8. Scattering coefficient is plotted versus hematocrit at 8.5 MHz.
This condition is satisfied by blood.as can be seem from Table
I, Pe/P 1. Therefore,
-

10 H- 0 DAATA POINT 1 4 3 WO - (3elWo)3-

e Wofe
E
0
__- T}HEORETI CAL
C URVE FITTED =127rk - WO a - Wo)3 Jr WO Pe
8
IL
0
0 %
| Pe - WO)p -( Wope 2 W
x _1, 0000 o ,
10 6
£$ 0° |
+ a ~~Pe /
Li oe
'0 " =-k4 a3 WO (I WO) (| W + Wof3e
W e
-lWo)3 e
\
I.
0 \ --ka3 0(lW0
U-
El 4 0
c,
4 0
Pe P
0 +
(10)
2 15 MHZ Pe
cit
,/ I l l -
V/ I I
Equation (10) indicates that for WO -
0
0
10 20 30 40 50 60
HMTC %
In 12 k4a3 WO(( ep + PeP (11)
Fig. 9. Scattering coefficient is plotted versus hematocrit at 15 MHz.
SHUNG etal.: SCATTERING OF ULTRASOUND BY BLOOD 465

The single scattering process will suffice to describe the scat- |Pe~ 2 Pe-P 2

tering mechanism. For WO I -

(1- Wo)t3 + Wo le Pe
XZ 1 P
k4 a3 (( | Pe (s = P3e 43 + Pe
- P
221
1 27jy
W)
0 Pe
(12) \|(1 - Wo)3 + Woie Pe /

where 1 - WO is the volume of the cavity. Thus, scattering =Y1y7 (16)

may be considered occurring at these cavity sites or due to where
these "holes." Also note that (10) shows the maximum scat-
tering happens as WO 0-5. Pe Pf3eU3
~2 iPe P 2

In deriving these equations, independent particles are as-

sumed. One particle distribution function is assumed to be 'y='y(WO)=4 r (IWO) 3+Wo Pe Pe
e Pe -P
no/N = 1/V where N = total number of scatterers in a volume (1- Wo) P +WoPe Pe
V and nO = N/V for sparse concentration and N/(l - Wo) V for
more dense concentration. Two-particle distribution function The only unknown quantity in (16) is the fitting constant b.
is assumed to be the product of two one-particle distribution This can be calculated from our experimental data. For fre-
functions. Obviously, this is no longer valid for very dense quency = 5.2 MHz, our result indicates a. 12.25 X 10-5 cm'- -

medium. The particles have finite volume as concentration as Wo 0.25. Substituting these data into (16), b is found to
-

gets very high. They can no longer be considered to be point be 1.72 or /lb = 0.58. Substituting this value into (16), we
scatterers. They will be affected by the presence of other par- obtain
ticles, thus, more or less correlated. However, the explicit ex-
pression for the two-particle distribution function is not avail- °s 0.21 k4 a3 WO (1 - WO)(1 - 1.72 WO)
able. In order to explain their experimental result, Beard et al. ( fTae _ Pe P 2

introduced the heuristic "hole" approach [6] . As can be seen (17)

(l-Wo)1 + WoI3e Pe
from (1 1)
This equation is plotted along with the data points in Figs. 7-9
r71 V'Wo for WO -0 (13) for frequency = 5.2, 8.5, and 15.2 MHz, respectively. The
where V' = volume of the scatterer. volumetric scattering cross section is obtained by dividing (17)
Thus if we visualize starting from a uniform medium com- with n,. This result is plotted in Figs. 10-12 for frequency =
posed of the scatterers (WO 1) and then introduce the holes
-
5.2, 8.5, and 15.2 MHz, respectively, along with the experi-
of volume Vh as the scatterer, then (13) suggests mental data. Data points in Figs. 10-12 are obtained by divid-
ing the data points in Figs. 7-9 by the corresponding number
n -(1- WO)vh. (14) of blood cells. A hematocrit point is equivalent to 1.07 X 108
Note here we have assumed that the WO 1 could be realized. red cells/cm3. Thus the volumetric scattering cross section for
Vh can be represented by frequency = 5.2 MHz, hematocrit = 25 percent is

V - NV' V( - WO) 12.25 X10 5 = 4.6 X 10-14 cm2.

Vh- N W)
-WO if WO I.
fl!0
However, WO is usually not realizable. Therefore, Beard et Note here we have assumed that the fitting constant b is inde-
-
1

al. assumed pendent of frequency. These figures indicate that the scatter-
ing contribution due to a scatterer decreases as the hematocrit
Vh V'(1 - bWO) for W -Wm increases. The experimental points follow the theoretical
curves approximately.
where b is a fitting constant and W1,m is the maximum volume The considerable spreading of the experimental data arises
concentration realizable for certain particles. For rigid spheres from several factors. 1) The inaccuracy in hematocrit deter-
it was found Wm 0.64 [6]. -
minations. 2) The age of the blood cells-the morphology of
Using these relations, (1) can be written as the erythrocytes changes considerably as they grow older.
3) The agglutination of the blood cells. Because of the surface
7=1 214 _re
k4a3W(l-W)(l-bW
3
l2
- 4)( charge possessed by these cells, they tend to aggregate, clump,
or form rouleaux. 4) Error due to misalignment of the trans-
ducer. 5) The spreading of the beam. In deriving (3), constant
+ PeP ) (15) beam cross section was assumed. Actually, in the far-field, the
beam diverges at an angle 0 about the central axis given by
where b is a fitting constant. As can be seen from (15), if sin 0 = 0.61 X/r, where ro is the radius of the transducer [15],
WO l/b, -+0. [6]. Sometimes there are bubbles generated on the surfaces
Since the scattering for small scatterers (k a << 1) is angular of the transducer, the chamber wall, and the reflector. These
dependent and the backward scattering is larger than the for- bubbles will cause a mismatch of the acoustic impedance be-
ward direction [5], [14], the scattering coefficient is given by tween the transducer and water, thus reducing the power
466 IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, NOVEMBER 1976

TABLE II
COMPARISON OF THEEXPERIMENTAL AND THEORETICAL RESULTS OF
E
SCATTERING CROSS SECTION
to 10 a\ 0
---
dota point
theory
0 \
x
z
8 0

U)
w
z
_
0
0 6 \< O f - 5MHZ transmitted into the water. Similarly, bubbles on the reflector
z

and chamber wall will change the reflection and transmission

w4 \8\ coefficient. Despite these possible errors, the scattering cross
C.)

2K
\IQ p sections measured experimentally for low hematocrits com-
pare favorably with those using the formula derived by Morse
and Ingard [14] for the scattering cross section of a viscous
l
0 10 20 30 40 50 60 particle suspended in a frictionless medium. Both the experi-
HMTC % mental results for hematocrit = 3 and the values calculated
Fig. 10. Scattering cross section is plotted versus hematocrit at 5 MHz. from Morse and Ingard's formula at frequencies of 5.2, 8.5,
and 15 MHz are listed in Table II.
The experimental results for hematocrits > 45 percent were
not obtained for the reasons stated in Section III. In order to
eliminate these effects, it is very desirable to redesign the flow
E chamber so that the flow inside the chamber is not turbulent,
I010
cn
o 0 data point but laminar. The relation between the scattering and flow rate
theory
x
z
---
should be established.
o
P-_.)
s -o
0 O
VI. CONCLUSION
cn 0
ua
0 6
0R The scattering of ultrasound by blood was found to be pro-
0
0% portional to the fourth power of the frequency predicted by
z
_ Io ~~~~~f8,3.5 MHZ
N
%
the wave scattering theory .
for small scatterers and dependent
4
upon the hematocrit of the blood. The relation between the
cn
O6oc3,..
"
00
o'-, scattering
O and the number of RBC's was also examined for
2 _ 0 blood samples with hematocrits up to 45 percent. The results
o

for hematocrit greater than 45 percent were not obtainable

.

I I I I 1 due to experimental difficulties. The scattering was found to

0 10 20 30 40 50 60 increase along with hematocrit until it reaches a maximum in
HMTC %
the hematocrit range from 24 to 30, then to decrease as he-
Fig. 11. Scattering cross section is plotted versus hematocrit at 8.5 MHz. matocrit increases. Twersky's wave scattering theory and a
heuristic "hole" approach were applied to describe these
results.
Comparing the attenuation constant and the scattering coef-
ficient for blood, it is concluded that the attenuation of ultra-
sound in blood is primarily due to absorption. The scattering
is so weak that it may be neglected for all practical purposes
E 10
C4

to 0
0 data point
in the frequency range less than 15 MHz. Since the scattering
x 8 theory
increases
---
much faster than absorption as frequency increases,
z
0
-0
-- D o the scattering will become comparable to absorption as fre-
C.)
"o
quency goes beyond 15 MHz.
Cl
6 )\No The experimental technique discussed in this paper has the
C') _p-
CA)0
o potential of being applied to other biological tissues such as
f 15MHZ
cardiac muscle, lung, etc.
-
-
0 "

Z4
L&J
I- - c JA.
en
4 (IoPQ5 REFERENCES
[1] D. E. Goldman and T. F. Heuter, "Tabular data of the velocity
absorption of high frequency sound in mammalian tissues,"
I ' J. Acoust. Soc. Amer., vol. 28, p. 35, 1956; also vol. 29, p. 655,
0
10 20 30 40 50 60 1957.
HMTC % [21 H. P. Schwan, Biological Engineering. New York: McGraw-Hill,
Fig. 12. Scattering cross section is plotted versus hematocrit at 15 MHz. 1969.
IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. BME-23, NO. 6, NOVEMBER 1976 467

[3] V. Twersky, "On scattering of waves by random distributions-I: ultrasound backscattering from an ensemble of scatterers excited
Freespace scatterer formalism," J. Math. Phys., vol. 3, p. 700, by sine-wave bursts," J. Acoust. Soc. Amer., vol. 53, p. 1351,
1962. 1973.
[4] -, "On scattering of waves by random distributions-Il: Two- [9] W.R.Platt,ColorAtlasandTextbookofHematology. Philadel-
space scatterer formalism," J. Math. Phys., vol. 3, p. 724, 1962. phia, PA: Lippincott, 1969.
[5] -, "Acoustic bulk parameters of random volume distributions [10] J. M. Reid and R. A. Sigelmann, private communication.
of small scatterers," J. Acoust. Soc. Amer., vol. 36, p. 1314, [11] E. L. Carstensen, K. Li, and H. P. Schwan, "Determination of the
1964. acoustic properties of blood and its components," J. Acoust. Soc.
Amer., vol. 25, p. 286,1953.
[6] C. I. Beard, T. H. Kays, and V. Twersky, "Scattering by random [121 E. L. Carstensen and H. P. Schwan, "Absorption of sound arising
distribution of spheres vs. concentration," IEEE Trans. Antennas from the presence of intact cells in blood," J. Acoust. Soc.Amer.,
Propagat. (Special Issue on Partial Coherence), vol. AP-15, pp. vol. 31, p. 185, 1959.
99-118, Jan. 1967. [131 R. J. Urick, "A sound velocity method for determining the com-
[71 S. W. Hawley, T. H. Kay, and V. Twersky, "Comparison of dis- pressibility of fmely divided substances," J. Appl. Phys., vol. 18,
tribution functions from scattering data on different sets of p. 983, 1947.
spheres," IEEE Thans. Antennas Propagat. (Special Issue on [14] P. M. Morse and K. U. Ingard, TheoreticalAcoustics. New York:
Partial Coherence), vol. AP-15, pp. 118-135, Jan. 1967. McGraw-Hill, 1968.
[81 R. A. Sigelmann and J. M. Reid, "Analysis and measurement of [15] R. H. Bolt and T. F. Heuter, Sonics. New York: Wiley, 1955.

Techniques for Improving the Selectivity of

Electromyographic Recordings
ISAK GATH AND ERIK V. STALBERG

Abstract-Several methods aimed at obtaining highly selective elec- A needle electrode inserted into the muscle will record ac-
tromyographic recordings are studied in human muscle with a multi- tion potentials from many units, depending upon the type
electrode technique. It was found that a small electrode size had a of electrode used, its pick-up area, the geometrical arrange-
major effect in obtaining selective recording. High-pass filtering of the
signal resulted in a further increase in the amplitude ratio of action ment of the muscle fibers in relation to the recording elec-
potentials generated by close and remote fibers due to different fre- trode, etc. In recent years, it has been shown [2], [3] that
quency content of the action potentials. Additional improvement in muscle fibers belonging to different motor units are inter-
selectivity could be achieved by bipolar recording. mingled. It has been demonstrated [4] with single fiber
INTRODUCTION EMG electrode, that an average of 1.4-1.6 fibers from the
THE term motor unit [11] applies to the cell body of the same motor unit are within the electrode pick up radius of
alpha motor neutron located in the anterior horn of the 250 pm.
spinal cord (or in the motor nuclei of the brain stem), its There has been an increasing interest in the firing pattern
axon, and all the muscle fibers innervated by this axon. A of motor units [5] - [9] since the introduction of the con-
pulse generated in the motor neuron will excite all the muscle centric needle electrode by Adrian and Bronk in 1929 [101.
fibers belonging to the same motor unit, resulting in an almost The analysis of the firing pattern of the motor units requires
synchronous activity of all these fibers. The summated ac- 1) a highly selective recording from only one or a small num-
tivity of the action potentials of these single fibers belonging ber of units and 2) recording conditions that will minimize
to the same motor unit within the uptake area of the EMG the recorded action potential duration, and therefore, will
electrodes is termed the "motor unit action potential." Dur- reduce the chance of superposition of action potentials from
ing the increase of muscle tension, more motor units are different motor units. Achieving this the recording will be
activated ("recruitment") and the firing rate of each of the comprised of a series of distinct pulses, action potentials
units increases; however, motor units fire asynchronously. [Fig. 1(a)]. When the time of occurrence of each pulse is
measured, the record is transferred into a stochastic array
of Dirac pulses [Fig. 1 (b)] . Different methods of statistical
Manuscript received July 24, 1974; revised June 3, 1975 and Sep-
tember 24, 1975. This work was supported by the Swedish Medical analysis may be applied in order to study the firing pattern
Research Council under Grant 135. of the motor units, e.g., point process analysis [11] .
I. Gath is with the Department of Electrical Engineering, Technion, The aim of the present study is to establish techniques for
Israel Institute of Technology, Haifa, Israel.
E. V. St°alberg is with the Department of Clinical Neurophysiology, obtaining recordings suitable for the analysis of the firing
University Hospital, Uppsala, Sweden. pattern of one or a few motor units.