Note: Descriptions are shown in the official language in which they were submitted.
2~
This is a division of Patent Application 262,273
filed September 29, 1976.
Ventricular fibrillation (VF) is a lethal cardiac
arrhythmia for which the only known efficacious treatment
is electrical countershock. A victim of VF outside of the
hospital setting has little chance of survival since treatment
must take place within a few minutes after the onset of the
episode.
Fortunately, new techniques and devices are being
devised to help deal with this life threatening condition.
Among these are computer techniques which aid in the identifica-
tion of high risk VF patients, anti-arrhythmic drugs which
can be prophylactiaally administered to these patients, programs
for widespread cardio-pulmonary resuscitation training and
implantable devices which can automatically detect VF and
deliver cardioverting countersocks.
Many of the known techniques, such as defibrillation
in a hospital setting, or defibrillation by a paramedic as
part of a resuscitation program, rely upon the human detection
of VF. This detection has typicall~ been accomplished by
a trained operator interpreting an ECG from an oscilloscope
tracing. However, there are situations where such an approach
to reversing VF is impossible or impractical. There is ac-
cordingly a great need for an electronic device able to ac-
curately detect VF or other life threatening arrhythmias from
an input ECG where such a traditional approach is unfeasible.
For example, an external defibrillator could be built with
an interlock to its discharge switch so that a shock can be
delivered only after the presence of VF has been confirmed
by a detector receiving an ECG signal from the paddles. Such
2~
a defibrillator could safely be used by even an untrained
operator.
With regard to the automatic implantable defibril-
lator, techniques have been developed which are generally
acceptable for detecting VF and discriminating between life
threatening arrhythmias and other cardiac malfunctions. ~et
there is considerable room for improvement with regard to
detecting and discriminating VF from other non-fatal arrhythmias.
Accordingly, another use for such a detector as noted above
would be in the totally implantable automatic defibrillator.
Previous approaches to VF detection for implantable
devices have had certain drawbacks. Fundamental questions,
particularly important to an automatic implantable defibrillator,
relate to potential failure modes, the risks to a patient
should the device reach one of these failure modes, and speci-
fically to whether failures should occur in a passive or an
active manner. Obviously, failures must be minimized, but
they still must be considered. In this regard, it is believed
preferable that potential sensing failures lead to inherent
passivity of a defibrillating device.
In many known VF detectors and automatic implantable
defibrillators, the primary detection schemes would result
in active mode failures unless other lock-out circuitry is
provided. Examples are R-wave sensors, pressure sensors,
and elastomeric contraction sensors.
There is accordingly a great need for a VF detector
which is accurate in its detection of VF or other life threaten-
ing arrhythmias, so that failure modes may be passive.
There is disclosed herein an accurate simple, VF
detector which at least partly mitigates at least some of
Z~L
the drawbacks of known VF detectors.
The present detector may be embodied in a system
for measuring the electrical activity of the heart which can
reliably discriminate between hemodynamically efficient and
inefficient arrhythmias, being particularly sensitive to ven-
tricular fibrillation. Though presented as a part of an auto-
matic implantable defibrillator, it should be appreciated
that the present detector is not limited to this specific
application. For example, certain other arrhythmias, or tachy-
arrhythmias can easily be identified by utilizing the present
teachings.
Customarily, the term electrocardiogram (ECG) implies
the use of electrodes on the body surface to obtain electrical
signal indicative of heart activity. The term electrogram,
on the other hand, generally refers to measurements made at
the surface of the heart. As used herein, "ECG" is defined
broadly, and refers to any measurement of the electrical activity
of the heart, notwithstanding the source or technique of the
measurement.
With the present detector, VF may be detected with
a degree of accuracy never before possible, and hence inherent
passive failure modes can be afforded. The detector may operate
independently of the concepts of QRS detection and heart rate
calculations to maximize accuracy. As is known, these concepts
are particularly difficult to define during ventricular fibril~
lation. Furthermore, high-amplitude P and T-waves can in-
accurately be sensed as R-waves, leading to false VF diagnosis.
The VF detector described hereinafter has simple circuitry
to minimize component count and therefore the possibility
of electronic component failure. And, the circuitry of the
-- 3 --
z~
hereinafter described VF detector is easily adaptable to low
power operation.
Reference is also made hereinafter to the principle
of probability density function. Briefly, the probability
density function defines the fraction of time on the average
that a given signal spends between two amplitude limits. It
has been noted that the probability densit~ function of an
ECG changes markedly between ventricular fibrillation and
normal cardiac rhythm. A probability density need not be
represented by the entire function, but rather, can be sampled
at discrete values of amplitude. As employed herein, the
entire function and the sampled form of the function a_e used
interchangeably. As will become apparent from the following
description, VF may be detected by monitoring a sampled pro-
bability density. The probability density function can be
monitored at any number of levels, but in a simple arrangement
monitoring is accomplished at one level, near zero, which
can be defined as the ECG baseline. In this instance, the
ECG is filtered, providing a first derivative of the ECG,
and in this manner moving any secondary probability density
function peaks toward the desired zero.
Howevex, a detector for accurately discriminating
between a normal cardiac state and a cardiac state requiring
cardioversion may require a relatively large input power to
perform its function.
It is accordingly an object of the present invention
to provide apparatus for cardioverting a malfunctioning heart
which enables the heart to be constantly monitored for the
occurance of a cardiac state requiring cardioversion but without
the continuous consumption of a relatively large amount of power.
z~
According to the present invention, there is provided
a two-stage apparatus for cardioverting a malfunctioning heart
in such a manner that power is conserved, the apparatus com-
prising first-stage detector means for discriminating between
a normal cardiac state and a state requiring cardioversion;
power/ing means for continually powering the first-stage de-
tector means; second-stage detector means for further dis-
criminating between a normal cardiac state and a state re-
quiring cardioversion, normally in a stand-by condition and
going to an active full power sensing condition upon initiation
by the first-stage detector means; means for enabling the
second-stage detector means and for thereby bringing the second-
stage detector means from its stand-by condition to its full
power active sensing condition; control means for powering
the second-stage detector only upon the first-stage detector
means sensing a cardiac state requiring cardioversion; delivery
means for delivering cardioverting energy to the heart; and
means for actuating the delivery means upon the second-stage
detector means sensing a cardiac state requiring cardioversion.
The second-stage detector may be in the form of
an impedance sensor which measures impedance between cardiac
electrodes. It has been found that the impedance due to cardiac
contractions is related to stroke volume. The impedance sensor,
however, requires a relatively large input power to perform
its sensing function. By means of the present invention,
the impedance sensor may remain idle for the greater majority
of time.
Embodiments of the invention will now be more par-
ticularly described with reference to the accompanying drawings,
given by way of example, in which:
Figure la is a tracing of a s~uare wave given for
exemplary purposes;
Figure lb is a plot of the probability density
function of the wave illustrated in Figure la;
Figure 2a is a typical catheter sensed ECG trace;
Figure 2b is a plot of the pro~ability density function
of the ECG trace illustrated in Figure 2a;
Figure 3a is an ECG trace representing ventricular
fibrillation;
Figure 3b is a curve representing the probability
density function of the ECG trace illustrated in Figure 3a;
Figure 4 is a block diagram of a probability density
function detector;
Figure 5 is a detailed circuit schematic of the
detector illustrated in Figure 4;
Figure 6a is a curve of an exemplary input ECG sig-
nal to the detector circuit of Figures 4 and 5, showing both
normal cardiac rhythm and fibrillation;
Figures 6b through 6e are curves representing sig-
nals at select locations in the circuit illustrated in Figures
4 and 5 based upon the ECG input illustrated in Figure 6a;
Figure 7 is a block diagram of a circuit for develop-
ing probability density function traces for the input of an
oscilloscope;
Figures 8a through 8d are curves illustrating an
ideal example of filtering an ECG trace to move the probability
density function to zero;
Figure 9a is a curve similar to that illustrated
in Figure 2a, but representing the ECG trace after filtering;
Figure 9b is a probability density function similar
~$36~32~
to that shown in Figure 2b, but illustrating the function
of the filtered ECG of Figure 9a;
Figure 10 is a block diagram of a phase lock loop
second-stage detector;
Figures lla through llf represent signals at select
locations in the circuit illustrated in Figure 10;
Figure 12 is a block diagram of a second-stage im-
pedance sensor for detecting ventricular fibrillation; and
Figures 13a through 13f are traces for explaining
the operation of the impedance detector illustrated in Figure
12.
The probability density function cardiac arrhythmia
detector will first be described. However, before embarking
upon a detailed explanation of the circuit, there follows
a brief discussion of the theory of probability density.
The detector system described hereinafter, is based
upon a series of measurements on the ECG. The measurements
are known in the literature as the probability density function,
denoted as KX(X). If X(t) is a function of time, then KX(X)
can be interpreted as a function that defines the fraction
of time on the average that X(t) spends between two limits.
For example, the area under KX(X) between X=Xl and X=X2 is
the fraction of time that X(t) spends between the limits X
and X2. Looking at the simplified example illustrated in
Figure la, it can be seen that X(t) is always either at the
levels X=B or X=A, and that the waveform spends half of its
time at each one of these limits. The probability density
function for this example is illustrated in Figure lb, wherein
the continuous function of time X(t) has been mapped into
a function of the amplitude-time distribution of X(t).
The present inventors have recognized that the pro-
bability density function of an ECG changed markedly between
normal cardiac rhythm and ventricular fibrillation. In this
regard, the attention of the reader is directed to Figure
2a which illustrates a typical ECG trace, Figure 2b which
shows the probability density function of the ECG illustrated
in Figure 2a, Figure 3a which illustrate~s an exemplary ECG
trace representing ventricular fibrillation, and Figure 3b
which is the probability density function of the trace illus-
trated in Figure 3a. It will be noted that when comparingnormal cardiac rhythm with ventricular fibrillation, the great-
est changes occur in the respective ECG traces at X=0, or
at the baseline of the ECG signal. This is markedly reflected
in the probability density functions as can be seen when com-
paring Figures 2b and 3b.
In a most simplified arrangement the probability
density is sampled at one value of x, namely X=0 or at the
baseline of a filtered ECG. As will be later explained when
reference is made to Figures 8 and 9, the filter in its most
basic form provides the derivative of the ECG. Physiologically
then, sampling the probability density of the filtered ECG
at X=0 corresponds to detecting the presence of relative
isoelectric segments in the ECG. These isoelectric segments
disappear during severe tachyarrhythmias such as fibrillation.
It should be noted that many other sampling levels are available
for X other than zero, as will be explained below, and hence
the number or level of sampling points are not in any way
intended to be limited.
The present probability density function detector
shown in block form in Figure 4, and in detailed schematic
2~
form in Figure 5. A representative set of waveforms is illus-
trated in Figure 6.
The detector circuit is shown generally at 10, having
a first stage ECG section 12, followed by a window stage 14,
integrator stage 16 and threshold detector stage 18. The
input to the detector 10 is by way of terminal 20 which leads
directly to an ECG preamplifier 22. The~ output from pre-
amplifier 22 is fed to a gain control circuit 24, and then
in parallel to a first filter 26 and combined peak-to-peak
detector and second filter 28.
The ECG section 12 has a bandpass filter character-
istic. Most important in this bandpass characteristic is
the highpass section which is designed to reject low frequency
ECG components such as ST segments and to provide an approxi-
mation of the first derivative. The automatic gain control
circuit 24 is provided to normalize the probability density
function over a known and fixed range of amplitude. To facili-
tate understanding of the simplified block diagram of Figure
4, the respective transfer characteristics for the four discrete
sections are provided immediately beneath each section.
With particular reference now to Figure 5, it can
be seen that amplifier 42 serves as the main gain block, with
capacitors 44, 46 and 48, and resistors 50, 52 and 54 serving
as the bandpass elements. Gain control is provided by N-
type junction field effect transistor 56 which shunts part
of the ECG signal to ground through capacitor 58. This partial
shunting results in a voltage divider effect with resistor
50. A typical endocardial electrogram which w~uld appear
at terminal 20 and the corresponding output of the ECG section
12 which would appear at terminal 30 are illustrated in Figures
g _
z~
6a and 6b, respectively. It should be apparent that the filters
in the ECG section 12 concentrate the cardiac signal to a
significant degree along the time axis.
After initial amplification and filtering in ECG
section 12, the signal at terminal 30 passes through window
section 14 which comprises a window comparator 32. Window
comparator 32 is designed to provide a digital signal at its
output terminal 34, the sense of which depends upon whether
the input to the comparator 32 lies inside or outside a band
centered about a given window level introduced at terminal
60. In the simplified embodiment of the present invention,
the window level at terminal 60 is chosen at the ECG base-
line. The band can be seen between "+a" and "-a" in Figure
6b, and the resultant digital signal developed by window com-
parator 32 and appearing at terminal 34 can be seen in Figure
6c. It will be noted that the digital output of comparator
32 goes to a fixed level whenever the filtered signal leaves
the designated band. The sizes of resistors 62 and 63 set
the band width "a". It can also be seen in Figure 6c that
upon the onset of fibrillation, very little time is spent
by the filtered ECG signal inside the designated band, corres-
ponding to the lower value of the probability density function
at X=0 as shown in Figures lb and 3b.
The digital signal appearing at terminal 34 is then
integrated by integrator 36 with respect to a bias level,
and produces a signal at output terminal 38 such as that illus-
trated in Figure 6d. As can be seen, this output signal takes
the form of a ramp when fibrillation begins. This output
signal at termianl 38 in turn becomes an input to the threshold
detector, or comparator 40. Detector 40 then switches when
-- 10 --
~-f~ Z~
the ramp signal at terminal 38 reaches a given threshold level.
Hysteresis is provided in the threshold detector stage 40
for a latching function so that the ramp must fall past level
Vt to VL (shown on the transfer characteristic beneath de-
tector 40) for fibrillation detection to cease. This is in-
dicated in Figure 6 during the period of inactive fibriIla-
tion shown in Figure 6a wherein the trace of Figure 6d falls
beneath the upper switching threshold of detector 40. Still,
the output of detector 40 is high, resulting from the noted
hysteresis characteristic.
It should be noted that the above-described simpli-
fied detector configuration provides inherent passive failure
mode behaviour and remains inactive if no ECG is applied.
Also, the inventive detector is independent of heart rate
definition and its inherent ambiguity during VF. Accordingly,
the probability density function VF detector overcomes major
disadvantages common in known VF detectors.
From the previous discussion, it should be apparent
that the probability density function provides another tool
for viewing the original time-amplitude function. All of
the discrete characteristics of the original signal are re-
tained, but are displayed in a different format. Thus informa-
tion of general diagnostic significanceis inherent in the
presentation and in some instances can be more readily seen
or measured automatically. The attention of the reader is
therefore directed to Figure 7, which illustrates a circuit
in greatly amplified block form which can be used to provide
complete displays of probability density functions. These
traces of probabilitydensity provide a great deal of informa-
tion in the detection and study of tachyarrythmias.
-- 11 --
In Figure 7, an input signal is introduced at inputterminal 62, to be then passed through an automatic gain control
circuit 64. In this manner, input signals of different ampli-
tude can be handled by the overall circuit. On the probability
density display, signal amplitude appears on the abscissa,
and therefore, AGC will normalize the width of the display.
A digital storage element 66 follows AGC 64, and
serves to provide a repetitive sou~ce of the input signal.
The storage element 66 stores approximately two-seconds of
ECG data in a digital memory, and continually repeats this
data. In this manner, the same data is repeatedly provided
to a window comparator 68.
The window comparator 68 provides a logical "1"
whenever its input signal lies within a narrow band centered
around a band center "X". By then passing the output of the
window comparator 68 through a simple low-pass filter 70,
a voltage is developed, proportional to the average time that
the input stays within the designated band. This signal is
fed to the "vertical" input of oscilloscope 72. This is pre-
cisely analogous to the definition of the distribution asdefined above. Sweeping the band center "X" slowly through
the range of the input signal by means of a wave generator
74, provides a continuous display on the oscilloscope screen.
The band center is coupled into the "horizontal" input of
the oscilloscope 72.
The respective traces ofFigures 2, 3 and 9 were
developed from the circuit illustrated in Figure 7. The trace
of Figure 2a, as noted above, represents an electrogram recorded
from an intracardiac catheter. The corresponding density
function is shown in Figure 2b. Several regions have been
~p~z~
identified on the respective traces, representing the same
cardiac events but on the two different display patterns.
For example, t~e region "B" of Eigure 2a is the most negative
peak of the R-wave. The signal spends very little time at
this value, thus the corresponding peak of the probability
density curve of Figure 2b is small. The peak at "A" is repre-
sentative of the ST segment, and as is readily seen, the ECG
signal spends more time on the average at this level than
at region "B". Accordingly, the peak is higher at "A" in
Figure 2b. The ECG dwells longest at the baseline identified
at "C" in Figure 2a, and the zero peak is the largest in Figure
2b.
It should by now be evident that the absence of
a peak at zero in the probability density function may be
utilized as being characteristic of abnormal cardiac rhythym.
By taking the derivative of the original ECG input signal,
the function of the filter in the arrangement described above,
the zero peak is considerably emphasized. Following up the
example of Figure 1, reference should be made to Figure 8.
In Figure 8a, there is illustrated a square wave alternating
between "+A" and "-A". The probability density function of
this square wave is given in Figure 8b and is similar to that
shown in Eigure lb. Since the square wave spends no time
at X=0, the probability density function has no peak at X=0.
Figure 8c represents an impulse train which is developed by
taking the derivative of the square wave illustrated in Figure
8a. The distribution function (probability density function)
of the impulse train, unlike that illustrated in Figure 8b,
is a unit impulse at zero, as shown in Figure 8d. Thus, the
effect of taking the derivative of the original square wave
- 13 -
2:~
input and then evaluating the probability density function
of the derivative is to shift peaks to X=0.
The same principle as explained in the ideal case
of Figure 8 is applicable to the filtered ECG shown in Figure
9a. It will be recalled that the trace of Figure 9a represents
the curve of Figure 2a after filtering. As can be seen, the
peak "A" corresponding to the ST segment which appears in
Figure 2b has been eliminated from the probability density
function wave of Figure 9b. Furthermore, the Figure 9b zero
peak is considerbly larger than that of Figure 2b. Thus,
the filter improves the detection accuracy by enhancing the
zero peak of the probability density function and thereby
emphasizing the measure of the differences between VF and
normal cardiac rhythm.
As mentioned previously, the distribution need not
only be sampled at zero as in the embodiment of the VF detector
described above. If two sampling points, say Xl and X0 are
defined as shown in Figures 3b and 9b, more discrimination
resolution becomes available by taking a ratio. As illustrated,
approximate measurement would show the value of the pro-
bability density function at these two points on the waveform
for the two examples to be:
For Normal Rhythm
Kx(Xl) = .012 x 0)
Kx(Xl) .012
m Kx(X ~ 2.5
C = .0048
m
- 14 -
For Ventricular Fibrillation
Kx(Xl) = 08 KX(Xo) = .11
KX(Xl)
m KX(Xo) .72
It can readily be seen that this measure yields over two orders
of magnitude difference between normal cardiac rhythm and
fibrillation. Sensing a value of C near 1.0 corresponds
to the detection of a severe arrhythmia.
In view of the high degree of reliability necessary
for the successful application of an implantable automatic
defibrillator, it may become desirable to improve the accuracy
of the detection system even relative to that described im-
mediately above. This can be done by adding stages of sensing
devices responsive to other parameters. One such parameter
which can aid in the discrimination of very severe tachy-
arrhythmias and fibrillation, is the variability in the R-
to-R wave interval. As noted above, even during extremely
high rate tachyarrhythmias, R-waves can be identified and
generally occur at a stable rate. During fibrillation, on
the other hand, all regularity in the autput of an R-wave
detector is lost. It is therefore possible to discriminate
between fibrillation and tachyarrhythmias by measuring the
variability of the R-wave intervals by means of an R-wave
detector. By combining the probability density function de-
tector and an R-wave interval detector, it becomes a practi-
cality to discriminate between fibrillation and even severe
tachyarrhthmias with an accuracy never before attained.
A technique of ascertaining R-to-R wave interval
;
1 5
2~
variability by way of a phase lock loop will now be described.
The phase lock loop circuit has the capability of "locking"
onto periodic input signals and providing an AC output voltage
which is at a constant phase and an integral multiple fre-
quency with respect to the input. If the input is not periodic,
however, the loop cannot "lock", and this condition is easily
detected. By utilizing the probability density function de-
tector as a first detector stage and a phase lock loop de-
tector as a second detector stage, the absence o a locked
state in the phase lock loop detector, coupled with the condi-
tion of the first detector stage having issued a fibrillation
output, verifies the presence of VF with an exceedingly high
degree of accuracy. Phase lock loop circuits are well described
in the literature, and an example of a low power version with
lock indication, directly applicable to fibrillation detection,
can be found in Application Note ICAN-6101, RAC COS/MOS Inte-
grated Circuits, 1975 Databook Series, pp. 471-478. Accordingly,
the phase lock loop circuitry is shown only in block form
in Figures 11 and 12. Its application to a fibrillation de-
tector is, however, a novel concept.
With reference now to Figures 10 and 11, the useof a phase lock loop in a fibrillation detection circuit will
be described. The previously discussed probability density
function fibrillation detector is an integral part of a first
stage detector shown at 76. The input to the first stage
detector 76 reaches an ECG amplifier 78, and is processed
by the probability density function detector 80. If fibrilla-
tion is sensed by the detector 80, then a signal is issued
at line 82, and is fed to one terminal of an AND gate 84.
The second input terminal 86 of AND gate 84 is associated
- 16 -
2:~
with the second stage of the detector combination, and in
particular, the phase lock loop circuit shown generally at
92.
The signal issued by the ECG amplifier 78 also serves
as an input to a filter 88 which feeds filtered signals to
an R-wave detector 90, each being of conventional design.
The ECG signal to filter 88 is illustrated in Figure lla,
while the filtered signal serving as the input to the R-wave
detector 90 is shown in Figure llb.
The R-wave detector 90 senses the presence of R-
waves, and for each R-wave, issues a pulse of finite period.
If the R-waves are regular in interval, the output of the
R-wave detector 90 is a periodic train of pulses. Figure
llc illustrates the output of the R-wave detector 90 based
upon the input of the filtered ECG shown in Figure llb. It
will be noted that the first three pulses of R-wave detector
90 are periodic.
The phase lock loop 92 includes a phase detector
94, the output of which is filtered by a low-pass filter 96,
in turn feeding signals to the control side of a voltage con-
trolled oscillator 98. The oscillator 98 issues a regular
train of square wave pulses and feeds the same to the phase
detector 94, which then compares the phase of these regular
pulses with the input from the R-wave detector 90.
The pase lock loop 92 may be of numerous designs,
as these circuits are well known. In any event, it is contem-
plated that the phase detector 94 provide output information
for a lock detector 100 which is indicative of the phase rela-
tionship between the R-wave detector pulses and the oscillator
pulses, and in turn, which indicates whether the phase lock
Z'~
loop 92 is able to lock upon the input from the R-wave de-
tector 90.
Upon the lock detector 100 receiving an indication
from the phase detector that the loop is locked, the detector
100 will, for example, issue a logical "0". Under this set
of condition, the AND gate 84 will remain idle, even if the
probability density detector 80 has indicated on lead 82 that
fibrillation is present. On the other hand, if the lock de-
tector 100 receives a signal indicating that the loop 92 is
not locked, then a logical "l" will be issued to AND gate
84, and if this logical "l" occurs simultaneously with a similar
signal from the probability density detector 80, then the
AND gate 84 will issue a signal on line 102 which will trigger
the de~ibrillating electronics. Still, however, if the phase
lock loop 92 cannot lock, and yet the probability density
detector 80 has sensed no irregularity, then the AND gate
84 will remain idle. This is illustrated in Figures lld through
llf. Figure lld shows the output of the probability density
function detector 80, Figure lle represents the output of
the lock detector 100, and Figure llf represents the output
of the AND gate 84 at lead 102.
Now, the impedance sensor VF detection circuit will
be described. It has been found that the impedance between
cardiac electrodes varies in accordance with the volume of
blood in the heart. When in normal rhythm, the heart regularly
contracts and fills, and hence the impedance change is periodic.
During fibrillation, however, stroke volume essentially goes
to zero, and a severe drop in pulsatile impedance change can
be seen. The circuit illustrated in block form in Figure
12 is able to detect the absence of pulsatile impedance changes,
- 18 -
analogous to a drop in stroke volume and hence ventricular
pressure. The traces of Figures 13a through 13f relate to
12
the circuit illustrated in Figure ~
With reference then to Figures 12 and 13, the impedance
VF detector is shown generally at 104. The detector 104 is
powered by a power supply on line 106, actuated by a gate
108 which is, in turn, controlled by the probability density
function detector shown at 110. As noted previously, the
impedance VF detector requires a substantial amount of power
from the implanted battery source. Therefore, so as not to
drain the battery, the detector 104 is designed to remain
idle until the probability density function detector 110 senses
an abnormality, and triggers the impedance detector 104 by
actuating its power supply. In this way, the circuit of Figure
12 provides an implied "AND" function. That is, the second-
stage circuit 104, which triggers the defibrillating electronics,
is only actuated upon command from the first-stage probability
density function detector 110. Therefore both circuits must
agree that fibrillation is present before a fibrillation output
is generated.
The basic element in the impedance VF detector 104
is illustrated schematically as impedance 112. The impedance
112 is, for example, related to the impedance of the blood
and tissue measured across intracardiac electrodes spaced
apart on a catheter. A current source 114 associates with
the impedance 112 and provides a current input of constant
value. An oscillator 116 feeds the current source 114 so
that source 114 generates an AC current to the impedance 112.
In this manner, the voltage across the impedance 112 will
be proportional to the current multiplied by the impedance
-- 19 --
2~
value. As typical values, the oscillator 116 is set to lOOKHz,
with the current source 114 supplying lOO~a. The impedance
112 is typically on the order of 50 ohms, and therefore ap-
proximately 5 mV appears across impedance 112. The voltage
across the impedance 112 is then amplified by means of a voltage
amplifier 118, and the amplified voltage from amplifier 118
is then demodulated by means of a synchronous demodulator
120.
The amplified and demodulated output of demodulator
120 is fed to a bandpass amplifier 122, and then to a trigger
network 124, a ramp generator 126, and a threshold detector
128. The output of the threshold detector 128, if present,
appears at terminal 120, and serves to trigger the defibrillation
circuitry into operation.
Figure 13a represents an ECG which is at first normal,
and then indicates fibrillation. Figure 13b shows, in an
exaggerated form so as to appear on the same time scale, the
output of oscillator 116, and Figure 13c represents a trace
of the voltage across impedance 112 after amplification by
-~ 20 amplifier 118 and corresponding to the ECG in Figure 13a.
,,:
r~ It can be seen in Figure 13c that the voltage across impedance
112 increases for each normal beat of the heart as blood is
ejected from the heart.
The output of demodulator 120, after amplification
by amplifier 122, is illustrated in Figure 13d where a negative-
going signal is indicated for each reduction in voltage, or
pulse, across impedance 112. Ramp generator 126 develops
a ramp which is shown in Figure 13e. It will be noted that
the ramp returns to its baseline each time the demodulated
and amplified output of amplifier 122 represented in Figure 13c
- 20 -
2:~ :
crosses a set threshold level. Accordingly, during normal
cardiac rhythm, the threshold detector 128 remains inactive.
However, once fibrillation commences, where indicated in Figure
13a, the curve of Figure 13d smoothes out, without the threshold
being reached, and therefore the ramp of Figure 13e continues
to elevate until it exceeds the threshold of detector 128.
At this occurrence, detector 128 is triggered, and a fibrilla-
tion output is issued on line 130.
.- ~
.:~
'
"~ :
: .
i.
,....................................................................... .
. 20
.- :
. , .
-
- 21 -
