 |
Description  |
|
|
FIELD OF THE INVENTION
This invention relates in general to pulse oximetry. It relates
specifically to the reduction of noise and artifact in a pulse oximeter by
the use of a correlated ECG signal.
DESCRIPTION OF THE PRIOR ART
Oxygen saturation level is an important measure of a patient's well-being.
It is of value during many surgical and anesthetic procedures. Pulse
oximetry is a well-known means of measuring the oxygen saturation level of
blood, non-invasively, by measuring the pulsatile modulation of two beams
of light of different wavelengths as they pass through body tissue, a
finger, for example.
The beams of light are applied and detected by a sensor containing light
emitting diodes and a photodetector. An oximeter computes saturation
depending upon the relative magnitudes of the AC components of the two
absorption waveforms or plethysmographs. These AC components are, in
general, small compared to the magnitude of the total waveform (AC+DC
components) representative of total absorption present. Any motion of the
finger relative to the sensor results in perturbations of the
plethysmographs known as motion artifact, which may cause the computed
saturation value to be in error. In addition, in poorly perfused patients
the circulation may not adequately modulate the light beams to the point
where they can be distinguished from system noise.
Most commercial pulse oximeters determine the amplitude of the
plethysmographs utilizing only the information contained in each waveform,
itself. Algorithms currently in use must determine both the period and
amplitude of each plethysmograph by analyzing such waveform parameters as
slope transition, minima and maxima, and rise and fall time. Additionally,
such algorithms may employ decision rules in order to reject data which
falls outside of certain limits.
A major problem with these algorithms is that in searching for peak
waveform excursions they inherently maximize the influence of motion
artifact. While in many circumstances these algorithms perform adequately,
the presence of motion artifact may disrupt their function to the point
where either the majority of the data is rejected or an erroneous
saturation value is calculated.
To compensate for this inadequacy, the computed saturation value is
conventionally often averaged with several recent values in order to
reduce the disruptive effect of corrupted data. Such averaging is only
useful if the number of bad data values is small relative to the number of
good data values. Thus, the usefulness of most commercial pulse oximeters
is limited to situations in which the patient is well perfused and subject
to minimal (or isolated instances of) motion.
Recently, Nellcor Incorporated introduced a pulse oximeter which seeks to
overcome this problem. This oximeter's operation involves the
synchronization of the plethysmographs with an ECG signal. The ECG signal
is used to determine the period of the plethysmographs and to guide the
measurement of the peak and valley values from which the AC components
and, ultimately, oxygen saturation are computed. By fixing the temporal
location of the peak and valley sampling relative to the R-wave of the
ECG, this solution eliminates the error introduced by searching for peaks
in the presence of artifact. However, it fails to reduce the level of such
artifact present in the plethysmographs at the peak and valley locations.
SUMMARY OF THE INVENTION
It is an object of the present invention to provide a new and improved
pulse oximetry method which greatly minimizes the effects of motion
artifact and poor perfusion in computed saturation readings.
Another object is to provide an oximeter and method which use a correlated
signal to reduce the effect of motion artifact by increasing the
signal-to-noise ratio of the plethysmographs through the use of ensemble
averaging.
Yet another object is to provide an oximeter and method which utilize a
correlated ECG signal for this purpose.
A further object is to provide an oximeter and method which increase the
validity of readings obtained from poorly-perfused patients by increasing
the signal-to-noise ratio of the plethysmographs through the use of
ensemble averaging.
Still another object is to provide an oximeter and method which distinguish
those pulses which are contaminated with motion artifact from those which
are not, and distinguishes those pulses which are the result of low
perfusion from those which are not, and applies a varying degree of
averaging limited to the extent necessary to improve signal quality to a
usable level.
The foregoing and other objects of the invention are realized by providing
a pulse oximeter including a sensor for the emission and detection of two
beams of light of different wavelengths. The beams are passed through skin
tissue and modulated by the flow of blood therein. The preferred
embodiment includes an apparatus for the amplification and detection of an
ECG, R-wave signal. This signal is used as a reference to guide the
averaging of subsequent optical pulse waveforms. The weight given to the
newest pulse waveform during the averaging process is determined by the
amplitude of that pulse waveform and by the degree of similarity between
it and the preceding pulse waveform. The composite, averaged pulse
waveform is then used in computing the oxygen saturation of the blood.
BRIEF DESCRIPTION OF THE DRAWINGS
The invention, including its construction and method of operation, is
illustrated more or less diagrammatically in the following drawings, in
which:
FIG. 1 is a diagrammatic illustration of an oximeter embodying features of
the present invention attached to a patient, with attachment to the
patient's finger shown in an enlarged inset;
FIG. 2 is a block diagram illustration of oximeter components and their
relationship;
FIG. 3 is a block diagram illustration of the ECG processing component, its
sub-components, and their relationship;
FIG. 4 is a block diagram illustration of the optical pulse processing
component, its subcomponents, components, and their relationship;
FIG. 5 is a diagrammatic illustration of red, infra-red, and composite
pulsatile waveforms as utilized by the oximeter of the present invention;
FIG. 6 is a block diagram of the R-wave detection algorithm routine in the
software of the present invention;
FIG. 7 is an illustration of two ECG waveform templates utilized by the
R-wave detection algorithm;
FIG. 8 is a block diagram of the R-wave peak excursion finding algorithm;
FIG. 9 is a block diagram of an R-wave artifact rejection timing
sub-routine utilized in the software of the present invention;
FIG. 10 is a block diagram of a portion of the R-wave peak discrimination
algorithm;
FIG. 11 is a block diagram of another portion of the R-wave peak
discrimination algorithm;
FIG. 12 is a block diagram of the R-wave acquisition and motion
determination algorithm;
FIG. is a diagram of three sets of pulsatile waveforms illustrating
calculation steps employed by the software of the present invention;
FIG. 14 is a block diagram illustration of the pulsatile waveform weight
determination algorithm; and
FIG. 15 is a block diagram of the ensemble averaging algorithm in the
oximeter embodying features of the invention.
DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT
Referring now to the drawings, and particularly to FIG. 1, a pulse oximeter
200 is shown connected to a patient 201. The oximeter 200 includes an
optical sensor 202 containing red and infra-red LEDs 206, and a
photodetector 207. The sensor 202 is placed in contact with the finger
201a of a patient and provides the pulse oximeter 200 with red and
infra-red plethysmographic waveforms.
The patient 201 also wears three standard ECG electrodes 203 which provide
the pulse oximeter 200 with an ECG signal. The ECG signal, if present, is
used to enhance the quality of the optical waveforms. The oximeter 200
computes oxygen saturation from the enhanced waveforms and displays it
digitally on a screen 204. As will hereinafter be discussed, the oximeter
200 is also capable of displaying other information on the screen 204.
Referring now to FIGS. 2 and 3, as well as FIG. 1, the major components of
the pulse oximeter 200 are illustrated in block diagram. Generally
speaking, according to the invention, an optical pulse processing circuit
205 of a type hereinafter discussed drives the red and infra-red LEDs 206,
and derives red and infra-red plethysmographic waveforms from the output
of photodetector 207. An ECG amplifier 211 and an R-wave detection
algorithm routine 208 process the ECG signal provided by electrodes 203,
and determine the timing for an ensemble averaging algorithm routine 209.
An oxygen saturation value is calculated by a microcomputer 215 in a
calculation algorithm routine 210 using the ensemble averaged waveform as
input, and is then displayed digitally on a screen 204.
If an ECG signal is not present, the lack thereof is recognized in the
R-wave detection algorithm routine 208, which causes an ensemble averaging
algorithm routine 209 to be bypassed and the unenhanced optical pulse to
be input into the calculation algorithm routine 210. The microcomputer 215
executes software comprising the R-wave detection, ensemble averaging,
calculation, and display algorithm routines.
FIG. 3 is a block diagram illustrating the ECG amplifier 211 in greater
detail. The three-lead ECG signal is amplified by differential amplifier
212. This amplifier 212 amplifies the differential component of the
signal, which is the desired ECG waveform, while rejecting a large portion
of the common-mode voltage.
The output of this amplifier 212 is AC-coupled, by capacitor 213, to
amplifier 214, which provides further gain. The gain provided by amplifier
214 is adjustable and can be set to 1/2, 1 or 2 by microcomputer 215.
Amplifier 214 can also accept an additional high level input 220, which is
intended to be connected to the output of an external ECG monitoring
device, thus obviating the need for an additional set of ECG electrodes on
the patient.
The output of amplifier 214 is processed by low-pass filter 216 to remove
unwanted artifact such as 60HZ and electrocautery induced noise, and is
converted to a serial, digital signal by A/D converter 217. The digitized
signal then passes through opto-isolator 218 to serial port 219 which
resides on the bus of the microcomputer 215. The opto-isolator 218 serves
to isolate the patient ECG leads from the external power supply, and is
incorporated for reasons of patient safety.
Referring now to FIG. 4, a block diagram describing the optical pulse
processing circuit 205 is illustrated in greater detail. The microcomputer
215 transmits desired red and infra-red LED drive values to a two-channel
D/A converter 221. The two output channels of the D/A converter 221 are
alternately selected by gating circuit 222, producing pulse trains 227 and
228 (illustrated in FIG. 5).
The pulse trains 227 and 228 drive voltage-to-current converters 224 which
provide current sources to drive the red and infra-red LEDs 206. The
action of the gating circuitry is controlled by a timing generator 223.
The beams of light produced by the LEDs 206 pass through skin tissue 207
in the patient's finger 201a and are modulated by the flow of blood
therein.
The modulated light excites photodetector 207 which produces an output
current varying in accordance with the incident light. The current is
converted to voltage and amplified by an I/V converter 225. A differential
realization of the I/V converter is utilized to maximize the supply
margins for a given gain, thereby decreasing the likelihood of saturating
the amplifier with ambient light, before the signal can be AC-coupled by
capacitor 226a.
The AC-coupled output of the I/V converter 225 is amplified by an amplifier
226 which incorporates gain selection circuitry controlled by
microprocessor 215. The output of amplifier 226 is the composite signal
229, shown in FIG. 5, and is a superposition of the modulated red and
infra-red pulse trains 227 and 228.
The separate red and infra-red pulsatile waveforms are synchronously
demodulated from the combined signal by the action of a gating circuit 230
and low-pass amplifiers 231. Synchronous demodulation requires that the
gating action of circuit 230 be synchronized with that of the gating
circuit 222; synchronization which is provided by the timing generator
223.
The demodulated red and infra-red signals comprise the DC signals 235 and
236 shown in FIG. 5. The AC signals 237 and 238 shown there are derived by
filtering and amplifying the DC signals using high-pass amplifiers 234.
Clamping circuit 239 allows the microcomputer 215 to force the input to
high-pass amplifiers 234 to zero, to allow quick recovery in the event the
amplifiers become saturated. Analog multiplexer 232 selects one of the
four (IRAC, IRDC, RAC, RDC) signals as input to the A/D converter 233,
under the control of the microcomputer 215.
The hardware of the pulse oximeter 200, including its organization and
operation, has now been described in a manner which will be readily
understood by those skilled in the art. The oximeter 200 is software
driven and the operation of the software as it relates to the present
invention will now be discussed. This includes the process of removing
motion artifact and enhancing the waveform quality in low perfusion
situations according to the invention.
ECG synchronization is used to provide a reliable time frame upon which to
base ensemble averaging, and a robust and accurate R-wave detection
algorithm routine 208 is an integral part of the invention. FIG. 6 is a
block diagram illustrating the components of the R-wave detection
algorithm routine 208 which is incorporated in the preferred embodiment.
The R-wave detection process involves three stages of processing: a
low-pass digital filter 240, a peak excursion finding algorithm routine
241, and a peak discrimination algorithm routine 256. These algorithm
routines will now be discussed in detail.
Referring again to FIG. 3, the ECG input signal from the A/D converter 217
is sampled via a serial port 219 at a rate of 240HZ. The resulting digital
waveform is low-pass filtered, with a cut-off frequency of 12HZ, to remove
artifact such as 60HZ and muscle noise.
The filtered ECG waveform then undergoes transformation by the peak
excursion finding algorithm routine 241 depicted in a flowchart in FIG. 8.
The purpose of this transformation is to amplify those characteristics of
the ECG waveform which are inherent in QRS complexes while inhibiting
those which are not, QRS being the medical acronym for electrical shock
instigation of the heart's contraction cycle. The algorithm routine 241
continually matches the ECG waveform to one of two templates 242 and 243
shown in FIG. 7.
In template 242, P3 is the maximum value within the interval N, and P2 is
the minimum. P4 is the minimum value to the left of P3, and P1 is the
maximum value to the right of P2. Note that the relations P3>=P2, P1>=P2,
and P4<=P3 apply to this template.
In template 243, P3 is the minimum value within the interval N, and P2 is
the maximum. P4 is the maximum value to the left of P3, and P1 is the
minimum value to the right of P2. Note that the relations P3<=P2, P1>=P2,
and P4>=P3 apply to this template.
Note further that any continuous function on an interval N must map into
one of these two templates. The largest closed excursion on the interval N
is defined as the quantity P3-P2-(P4-P1) for template 242, and
P2-P3-(P1-P4) for template 243. Since a QRS complex is characterized by a
large positive or negative excursion followed immediately by a return to
the baseline, the largest closed excursion in an interval on the order of
a typical QRS duration is a good indication of the presence of an R-wave
spike.
The algorithm routine 241 calls for N=8, 12, 16, 20, and 24, with
individual excursion values being summed so as to give a total
transformation value. More weight is placed on lower values of N in order
to emphasize narrower spikes over wider ones.
The algorithm routine 241 maintains a queue, of a length N, which is
searched in order to determine the parameters P1, P2, P3, and P4. Starting
at the top of the flowchart diagram, block 244 describes the manipulation
of the queue. The newest sample value is inserted at the head of the queue
and the oldest is removed from the tail.
Next, at block 245, the queue is searched and the maximum value and its
position in the queue are found. Similarly, at block 246, the queue is
searched again and the minimum value and its position in the queue are
found. Decision diamond 247 then determines which of the two templates 242
or 243 are matched by the current interval contained in the queue by
asking the question, "Does the maximum occur before the minimum?" If it
does, the waveform matches template 242; otherwise it matches template
243.
Following this, the parameters P2 and P3 are assigned the appropriate
maximum or minimum values in blocks 248 and 249. Next, the value of P4 is
found in block 250 or block 251. If template 242 was matched, it is given
by the maximum value after P2, as in block 253. Otherwise, it is given by
the minimum value before P3 as in block 252. Finally, the peak closed
excursion on the interval N is computed in blocks 254 and 255 using the
formula appropriate to the currently matched template.
After transformation of the ECG waveform in the manner described, the peak
discrimination algorithm 256 routine (see FIG. 6) classifies the spikes
found in the transformed waveform as either QRS complexes or artifact. The
peak discrimination algorithm routine 256 is a state machine with three
states: peak, valley, and noise peak. If the algorithm routine 256 is in
the valley or noise peak state, it enters the peak state when the waveform
exceeds a threshold which is based upon the past history of the ECG
waveform. If the algorithm routine 256 is in the peak state, it exits this
state and enters the valley state when the waveform drops below one fourth
of the maximum value attained in the peak state.
If the algorithm routine 256 is in the valley state, it enters the noise
peak state whenever the waveform climbs above four times the minimum value
attained during the valley state. If the algorithm routine 256 is in the
noise peak state, it enters the valley state whenever the waveform drops
half the distance between the maximum value attained during the noise peak
state and the minimum value attained during the previous valley state. The
detection of a QRS spike is signalled upon the transition into the peak
state.
The algorithm routine 256 maintains an average of the last eight QRS peaks
in order to set the threshold for detecting the next peak in the waveform.
In addition, the algorithm routine 256 maintains an average of the noise
peak levels found between the last four QRS peaks, in order to aid in the
rejection of artifact while accepting valid QRS spikes. These averages are
updated upon the transition between the peak and valley states.
Additional rejection of artifact is gained by examining the length of time
289 which has elapsed between a new peak 290 and the last accepted peak
291, as illustrated in FIG. 9. If it is less than 5/8 of the previous R--R
interval 292, the spike 290 is assumed to be noise and is not counted as a
QRS spike. If it is greater than 5/8, but less than 7/8, of the previous
R--R interval 292, the spike 290 is accepted "on probation" as long as it
exceeds a second threshold which takes into account the level of noise
encountered during the last four beats. It is counted as a valid QRS spike
but the previous state information is saved in order to "undo" acceptance
of spike 290 if a better candidate is found.
A probation interval 293 is calculated in this case during which any spike
294 meeting the threshold requirement overrides the acceptance of the
spike 290 in question. This probation interval is equal to 9/8 of the
previous R--R interval 393, minus the length of time 289 which has elapsed
since the last accepted peak 291. If the new spike 290 arrives after 7/8
of the previous R--R interval 292, it is accepted unconditionally, and no
chance is allowed for a subsequent spike to override it.
FIGS. 10 and 11 illustrate flowcharts detailing the algorithm routine 256.
At the top of FIG. 10 we have a decision diamond 257, which determines
whether the algorithm routine 256 is in the peak state. If it is not, the
algorithm routine 256 must check whether conditions are satisfied for
entering this state.
This is determined in part by decision diamond 268 in FIG. 11, which asks
the question, "Does the waveform exceed threshold 2?" If the waveform does
not exceed this threshold, it does not enter the peak state. If it does,
the algorithm routine 256 then determines the length of time 289 which has
elapsed since the last accepted spike and its relationship to the previous
R--R interval 292 (see FIG. 9). Decision diamond 269 first checks to see
if the last spike was accepted "on probation". If it was, decision diamond
272 determines whether the current spike falls within the probation
interval. If it does, block 275 corrects state variables such as the
average R--R interval for the longer pulse length. Block 277 then starts
the ensemble averaging with the current overriding spike 294, causing any
plethysmographic data collected since the last spike 290 to be discarded.
This does not substantially affect the quality of the ensemble averaged
waveform since the discarded portion generally falls during the decay
portion of the waveform 296 which is well after the valley 297 and peak
298 used to determine the saturation reading.
The overriding spike 294 (see FIG. 9) is then accepted unconditionally in
block 279, i.e., without any chance to be itself overridden. If the last
spike was not accepted on probation or the current probation interval has
expired, the length of time 289 which has elapsed since the last accepted
pulse must exceed 5/8 of the previous R--R interval 292. Decision diamond
270 determines whether or not this is the case. If the current pulse 290
is too soon, it is classified as noise by box 273. If the current pulse
290 is acceptable, it must then be determined whether or not it should be
accepted "on probation". This is determined by decision diamond 271, which
checks whether the length of time 289 elapsed since the last accepted
spike exceeds 7/8 of the previous R--R interval 292.
If this is the case, block 279 accepts the pulse unconditionally. If not,
decision diamond 274 checks to determine whether the waveform exceeds
threshold 1, which takes into account the noise level encountered during
the last four beats. If the waveform does not exceed this threshold, it
does not enter the peak state. If this threshold is exceeded, the spike is
accepted "on probation". Block 276 calculates the probation interval and
accepts the spike. In all cases in which the spike was accepted, the peak
state is entered by block 281.
If the peak state was not entered as a result of tests at decision diamonds
268 or 274, decision diamond 278 determines whether the algorithm routine
256 is in the valley or the noise peak state. If it is in the valley
state, decision diamond 282 and block 286 keep track of the minimum value
encountered during the valley state. If the waveform is not at a new local
minimum, decision diamond 283 determines whether the noise peak state can
be entered. If the waveform has climbed above four times the last local
minimum set by block 286, the noise peak state is entered by block 285.
If the algorithm routine 256 is found to be in the noise peak state as a
result of test 278, decision diamond 280 checks to see whether the
waveform has fallen to halfway between the last local maximum saved by
block 288 and the last local minimum saved by block 286. If it has, the
valley state is entered by block 284. Otherwise, decision diamond 287 and
block 288 keep track of the maximum value encountered during the noise
peak state.
Referring again to FIG. 10, if the algorithm routine 256 was initially
found to be in the peak state as a result of test 257, decision diamond
258 and block 259 keep track of the maximum value encountered during the
peak state. If the waveform is not at a local maximum, decision diamond
260 then checks to see whether the waveform has fallen to one quarter of
the last local maximum set by block 259. If it has, decision diamond 261
then determines whether the current peak was a noise peak or a QRS peak.
If it was a noise peak, block 263 updates the average of the noise levels
over the last four beats. If it was a QRS peak, block 262 then updates the
average of the last eight QRS peaks using the local maximum found by block
259. It also updates the average of the noise levels over the last four
beats using the largest local maximum found by block 288 since the last
accepted beat.
Blocks 264 and 265 compute the threshold values needed to detect the next
QRS peak. Threshold 1 is halfway between the current eight beat peak
average and the current four beat noise average. Threshold 2 is one-half
of the current eight beat peak average.
In all cases before exiting algorithm routine 256, parameters reflecting
the quality of the ECG waveform are tested. Decision diamond 266 checks to
see if the time elapsed since the spike was accepted exceeds four times
the last R--R interval. Next, decision diamond 299 checks to determine
whether the baseline of the transformed signal has exceeded one-half the
peak value. If any of these test levels are exceeded, the ECG waveform is
assumed to be lost and block 267 disengages the synchronization.
The operation of the R-wave detection algorithm routine 208 of the oximeter
software has now been discussed in detail. Referring now to FIGS. 12-15,
the ensemble averaging algorithm routine 209 utilizes the output of R-wave
peak discrimination algorithm routine 256 to enhance that component of the
red and infra-red plethysmographic waveforms which is correlated with the
ECG, while diminishing all which is unrelated, to yield a signal with an
improved signal-to-noise ratio.
The algorithm routine 209 relies on the assumption that instances of
moderate to severe motion, and of low perfusion, can be detected as the
plethysmographic waveforms are being sampled. To do this it has been found
to be advantageous to buffer these waveforms while they are being sampled,
and to delay the actual averaging until the next R-wave peak is detected.
The averaging weight of the current waveform cycle can then be adjusted,
depending upon whether the plethysmographic waveform just acquired is weak
or exhibits the influence of excessive motion artifact. An additional
benefit of this buffering stage is that the oximeter 200 is able to
discard waveform pulses during which optical pulse processing circuitry
205 has saturated and distorted the waveform. Yet another benefit of this
buffering stage is that it allows a level of error tolerance in the R-wave
detection process whereby algorithm routine 256 can accept certain
marginal QRS spikes "on probation", as described above, while maintaining
the flexibility to correct the error if a better candidate is subsequently
detected.
In order to give less weight to waveform pulses which are distorted by
motion artifact, a criteria by which motion can be measured is
established. FIG. 13 illustrates examples of waveforms which will be
discussed in relation to these criteria.
Referring to FIG. 13, the algorithm routine 209 assumes that a
plethysmograph 301 unaffected by motion varies only slightly between one
pulse and the next. In addition, a change in the amplitude, not shape, of
the pulse comprises the majority of the observed difference between one
pulse and the next. A plethysmograph 302 containing artifact, however,
differs greatly from the previous signal.
Signals 303 and 304 are obtained from plethysmographs 301 and 302,
respectively, by subtracting the latest pulse from that immediately
preceding it, in a point-by-point fashion. Note that the average amplitude
of signal 303 is less than that of signal 304. Note further that the value
of signal 303 is principally constant; while that of 304 changes rapidly,
and often.
The integrations of the change in these two signals are shown by signals
305 and 306. Note that the final value of integrated signal 306 is much
higher than that of integrated signal 305. Thus, a good indication of the
amount of motion present in a pulse is given by the integration of the
absolute value of the derivative of the point-by-point difference between
the current pulse and the preceding pulse, taken over the length of the
pulse.
FIG. 12 illustrates a flowchart of the algorithm routine 307 which acquires
a plethysmographic waveform while, simultaneously, determining whether or
not the waveform is unduly affected by motion. This algorithm routine 307
sums the change in succeeding, point-by-point differences, taken between
the pulse being currently acquired and the previous pulse.
Referring to the top of this flowchart, block 307a samples the
plethysmograph. Block 308 computes the difference between the current
point and, the corresponding point in the last pulse. Block 309 then
inserts the current point into the buffer. Next, block 310 sums the
absolute value of the difference between the last difference and the
current difference. Decision diamond 311 then checks for the occurrence of
an R-wave spike which would be detected by the R-wave detection algorithm
routine 208. If a spike was detected, block 312 saves the accumulated sum
as an indication of the level of motion present in the pulse, and
initializes the variables to prepare for the next pulse.
The ensemble averaging algorithm routine 209 utilizes a variable-weight
average in order to provide flexibility over a broad spectrum of pulsatile
waveforms. It attempts to give large weight to waves which are largely
motion-free, while diminishing the weight given to those which have
motion. Additionally, if a low perfusion situation is detected, less
weight is given to all pulses until several strong pulses are found.
Furthermore, the algorithm routine 209 takes into account the pulse rate
when determining the averaging weight. Since the averaging occurs each
time a beat is detected, more averaging can be used on a patient with a
fast pulse rate than one with a slow pulse rate while maintaining a
constant response time. More averaging is needed in cases of motion
artifact and low perfusion because the signal-to-noise ratio of these
pulses is less than with normal pulses.
It is known generally that ensemble averaging with a set of N waveforms
increases signal-to-noise level by a factor of the square root of N for
uncorrelated, random noise. Thus, ensemble averaging will decrease the
influence of uncorrelated motion artifact and will enhance a low perfusion
signal which may be buried in noise at the expense of response time. A
weight determination algorithm routine 313, for which a flowchart is shown
in FIG. 14, intelligently trades response time for signal-to-noise ratio.
At the same time, a maximum limit on response time is set in order to
ensure that the displayed saturation value is reasonably current.
The weight de | | |
