Pulse oximeter with alternate heart-rate determination

The present invention relates to an oxygen densitometer, and more particularly to determining the number of pulses by a plurality of mechanisms in a waveform detected by a pulse oximeter.

Pulsed oximeters typically include a variety of blood chemistries, including hemoglobin blood oxygen saturation in arterial blood, the volume of individual pulses supplied to tissues, and the pulse rate corresponding to each patient's heartbeat. Used to measure properties, but not limited to them. Measurements of these properties are achieved by using non-invasive sensors. The non-invasive sensor scatters light through a portion of the patient's tissue that is perfused with blood and optoelectronically senses the absorption of light in such tissue. The absorbed amount of light of various wavelengths is then used to calculate the amount of blood component to be measured.

Light scattered through tissue is selected as one or more wavelengths absorbed by blood in an amount that represents the amount of blood components present in the blood. The amount of transmitted light scattered through the tissue can vary as the amount of blood components in the tissue changes and the amount of light absorbed associated with it changes. To measure blood oxygen levels, such sensors have followed well-known techniques for measuring blood oxygen saturation and are typically adapted to produce at least two different wavelengths of light. A light source and a photodetector capable of sensing both wavelengths are provided.

Well-known non-invasive sensors include devices that are anchored to parts of the body, such as the fingers, ears, or scalp. In animals and humans, some tissues of these bodies are perfused with blood, and the surface of the tissues is easily accessible to the sensor.

Patent Document 1, Patent Document 2, and Patent Document 3 show a plurality of methods for calculating the number of pulses in a pulse oxygen meter having a "best rate" module. The "best rate" module intervenes between multiple pulse count calculations and selects the best rate based on the confidence associated with each pulse count calculation. The reliability is calculated using various metrics that determine the reliability of different pulse number calculations. Further, Patent Document 4 shows a fetal heart rate monitor, in which the fetal heart rate monitor identifies a fetal heart rate by using a plurality of parallel filter paths and evaluates different heart rates by using a performance index. Is weighted.

(N-100). Until around 1985, N-100 technology accepts or rejects pulses based on pulse size, pulse shape, expected time of (frequency) occurrence, and pulse history for ratio R / IR. I was doing it.

In particular, the N-100 found the pulse by looking for the maximum signal, which was followed by a negative maximum slope followed by the minimum value. The process was performed on a state machine called a "munch". Each maximum value is a noise gate It was not qualified until the signal fell below the noise threshold called gage). This served as a adaptable filter, as the noise gate level was set by feedback from subsequent processing steps to match the expected different signal amplitudes. The pulse was then accepted or rejected in the "Level 3" process. The "level 3" process is the ratio of the ratio of the new pulse to the average of the amplitude, period, and values in the history buffer (ratio-of-ratios; the ratio of Red to IR, where Red and IR are the ratio of AC to DC. It was a filter that fits the changing signal by comparing (expressed as a ratio) and then determining if there is a difference within the confidence level. When a new pulse was accepted, the history buffer was updated with the value for the new pulse. The level3 processing served as a adaptable passband filter in which the center frequency and bandwidth (reliability limit) matched the feedback from the filter output.

(N-200). The N-200 was an improvement over the N-100 because it was able to synchronize with the ECG as well as the built-in ECG filtering. The N-200 also added interpolation to compensate for the baseline shifting between the maximum pulse measurement time and the minimum pulse measurement time. The N-200 also included other filtering characteristics, such as a "boxcar" filter that calculates the average of a variable number of signal samples.

After various filtering and scaling steps, the N-200 applies a digitized signal to the "boxcar" filter, which calculates the average of N samples. Here N is set according to the filtered heart rate by feedback from subsequent processing steps. New samples are averaged in the Boxcar filter, but old samples are excluded. The boxcar length (N) is used to determine three parameters: pulse threshold, absolute minimum pulse, and small pulse. The set mean (ensemble-averaging) filter then calculates the weighted average of the new sample and the previously set averaged sample one pulse period ago. The sample then passes through a "munch" state machine and a noise gate, like the N-100. Interpolation characteristics have been added to the N-100 process to compensate for changes at baseline level. The minimum and maximum values occur at different times. The changing baseline increases or decreases the minimum value and does not increase or decrease the maximum value. Or vice versa, the changing baseline increases or decreases the maximum value and does not increase or decrease the minimum value.

The "geometric mean" is an integral part of C-Lock, a registered trademark of NELLCOR, that averages samples from multiple pulses to form a synthetic pulse. This process is also known as "cardiac-gated averaging". The above process requests a "trigger" event to mark the start of each pulse.<patcit num="1"><text>U.S. Pat. No. 6,083,172</text></patcit><patcit num="2"><text>U.S. Pat. No. 5,853,364</text></patcit><patcit num="3"><text>U.S. Pat. No. 6,411,833</text></patcit><patcit num="4"><text>U.S. Pat. No. 5,524,631</text></patcit>

<p> (A Brief Summary of the Invention) The present invention determines a plurality of heart rates and makes a selection among a plurality of heart rates based on a metric for only one of the plurality of heart rate calculations. Regarding. The first heart rate calculation method is selected and the first heart rate calculation method is used unless the accuracy of the metric is doubtful. If the accuracy of the above metrics is questionable, an alternative heart rate calculation is available and will be used instead.</p><p> In one embodiment, the first heart rate calculation method does not use the geometric mean waveform, while the alternative heart rate calculation uses the geometric mean waveform. An alternative heart rate calculation is used when the first calculation does not qualify the pulse for the first calculation that was detected immediately before.</p>

(Detailed description of preferred embodiments) FIG. 1 illustrates an embodiment of an oxygen measurement system incorporating the present invention. The sensor 10 includes a red LED and an infrared LED as well as a photodetector. These are connected to the substrate 14 by a cable 12. The LED drive current is provided by the LED drive interface 16. The photocurrent received from the sensor is provided to IV interface 18. The IR voltage and the red voltage are then provided to the sigma delta interface 20 incorporating the present invention. The output of the sigma-delta interface 20 is provided to the microcontroller 22. Microcontroller 22 includes a 10-bit A / D converter. The controller 22 includes a flash memory for programs and an EEPROM for data. The processor also includes a controller chip 24 connected to flash memory 26. Finally, a clock 28 is used to provide an interface 30 to digital calibration on the sensor 10. The isolated host 32 receives the processed information and at the same time receives an analog signal on line 34 to provide an analog display.

(Summary of Design) The design of the present invention is intended to deal with unwanted noise. A metric of the signal is measured and used to determine the weighting of the filter. The metric of the signal is that the pulse is plethysmograph or noise, for example frequency (frequency that is in the range of human heart rate), shape (shape similar to the pulse of the heart), rise time, etc. It suggests whether or not it is likely to be. A similar technique was used in the Nellcor N200 in the context of this specification. The new design adds many different features and modifications, such as the two set averages claimed in the present invention.

The details of the architecture are shown in the figure in Figure 2. This design calculates oxygen saturation and pulse count. These are described separately below.

(I. Calculation of Oxygen Saturation) (A. Signal Adjustment) Digital red and IR signals are received and adjusted in this block by: Keep the ratio by (1) taking the first derivative and removing the baseline shift, (2) filtering through low frequencies using a constant coefficient, and (3) dividing by the DC value. The function of the signal conditioning subsystem emphasizes the frequencies that occur in human plethysmographs and attenuates the low frequencies where motion artifacts are often concentrated. The signal tuning subsystem selects its filter coefficients based on the hardweer characteristics identified during initialization. Input-Digital Red and IR Signal Output-Preprocessed Red and IR Signals.

(B. Pulse Identification and Qualification) Low-pass filtered digital red and IR signals identify pulses and qualify them to resemble arterial pulsations. Provided to this block. This is done using a pre-trained neural network and is primarily done over IR signals. Pulses are identified by examining their amplitude, shape and frequency, as was done with the Nellcor N-100. The input to this block is the average pulse period from block D. This function was similar to the N-100, which used the number of pulses to change the pre-qualification. The output suggests the degree of arrhythmia and the quality of the individual pulses. Inputs-(1) Preprocessed red and IR signals, (2) Average pulse period, (3) Lowpass waveform from lowpass filter. Output-(1) Degree of arrhythmia, (2) Fluctuations in pulse amplitude, (3) Quality of individual pulses, (4) Pulse beep notification, (5) Eligible pulse period and age ).

(C. Calculation of signal quality metrics) This block is related to pulse shape (derivative skew), period variability, pulse amplitude and variability, ratio ratio variability, and pulse count. Determine the frequency content to be used. Inputs-(1) raw digital red and IR signals, (2) degree of arrhythmia, individual pulse quality, pulse amplitude variation, (3) preprocessed red and IR signals, (4) ) Average pulse period. Output-(1) Filter weights for low pass and geometric mean, (2) Metrics for sensor-off detectors, (3) Normalized pre-processed waveforms, (4) Percent modulation.

(D. Average pulse period) This block calculates the average pulse period from the received pulse. Input-Eligible pulse period and age. Output-Average pulse period.

(E1. Lowpass Filter and Geometric Mean) Block E1 filters the lowpass filter and geometric mean of the signal adjusted by block A and normalized by block C to identify the number of pulses. The weight for the low pass filter is determined by the signal measurement reference block C. The signal is also aggregate averaged (which attenuates frequencies other than the frequency of interest near the number of pulses and their harmonics), and the weight of the aggregate mean filter is also determined by the signal measurement reference block C. .. When flagged as signal down, it is assigned a lower weight. When the signal is flagged as arrhythmic, the geometric mean is not appropriate during the arrhythmia and is therefore assigned a higher weight. Red and IR are treated separately, but with the same filter weight. Filtering is delayed by about 1 second so that the signal metrics are calculated first.

The filter uses continuously fluctuating weights. If the samples are not aggregate averaged, the weights for the already filtered samples are set to zero in the weighted average, and new samples are processed through the code. This block tracks the age of the signal, the cumulative filtering (sum of response time and delay in processing). Results that are too old can be flagged (if good pulses have not been detected for some time). Inputs-(1) Normalized and preprocessed red and IR signals, (2) Average pulse period, (3) Lowpass filter weights and Geometric mean filter weights, (4) If available ECG trigger, (5) IR fundamental wave for zero crossing trigger. Output-(1) Filtered red and IR signals, (2) Age.

(F. Evaluation of Filtered Waveform Correlation and Calculation of Average Weights) This uses the same noise metrics used for the N100 and N200 described above, but without feedback. The variable weighting for the filter is controlled by the difference in the ratio of ratios. The effect of this variable weight filtering is that the ratio of ratios changes slowly as the artifacts increase, and the ratio of ratios changes rapidly as the artifacts decrease. The subsystem has two response modes. Filtering in fast mode targets metrics with an age of 3 seconds. In normal mode, the target age is 5 seconds. In fast mode, the minimum weighting of the current value is reduced to a higher level. In other words, a low weight is assigned to the new ratio ratio calculation in the presence of noise, and a high weight is assigned in the absence of noise. Input-(1) Filtered red and IR signals and age, (2) Calibration factor, (3) Response mode (user speed setting). Output-Ratio weight averaging for ratio calculation.

(H. Calculation of Saturation) Saturation is calculated using an algorithm that has a ratio of the calibration factor to the averaged ratio. Input-(1) Averaged ratio ratio, (2) Calibration factor. Output-Saturation.

(II. Calculation of the number of pulses) (E2. Low-pass filter and geometric mean) Block E2 filters the signal adjusted by block A by low-pass filter and geometric mean to identify the number of pulses. The weight for the low pass filter is determined by the signal measurement reference block C. The signal is also aggregate averaged (which attenuates frequencies and harmonics other than the frequency of interest near the number of pulses), and the weight of the aggregate mean filter is also determined by the signal measurement reference block C. When flagged as signal down, it is assigned a lower weight. If the signal is flagged as arrhythmic, filtering is not appropriate during the arrhythmia and therefore a higher weight is assigned. Red and IR are treated separately. The processing of this block is delayed by about 1 second so that the metric of the signal is calculated first.

The filter uses continuously fluctuating weights. If the samples are not aggregate averaged, the weights for the already filtered samples are set to zero in the weighted average, and new samples are processed through the code. This block tracks the age of the signal, the cumulative filtering (sum of response time and delay in processing). Results that are too old can be flagged (if good pulses have not been detected for some time). Inputs-(1) preprocessed red and IR signals, (2) average pulse period, (3) lowpass filter weights and geometric mean filter weights, (4) ECG triggers, if available, (5) IR fundamental wave for the trigger of zero intersection. Output-(1) Filtered red and IR signals, (2) Age.

(I. Identifying and qualifying filtered pulses) This block identifies and qualifies the pulse period from the filtered waveform. Also, this result is only used if the pulse was not qualified by block B. Inputs-(1) filtered red and IR signals and age, (2) average pulse period, (3) hardware ID or noise floor. Output-Eligible pulse period and age.

(J. Calculation of average pulse period and number of pulses) This block calculates the number of pulses and the average pulse period. Input-Eligible pulse period and age. Output-(1) average pulse period, (2) number of pulses.

(III. Vein pulsation) (K. Detection of venous pulsation) Block K receives the preprocessed red signal and IR signal from block A as well as the age and the number of pulses as input and as output. Provides indicators of venous pulsation. This subsystem uses a single-tooth comb filter to generate the IR fundamental waveform, which is the output to the set average filter, in the time domain. Inputs-(1) filtered red and IR signals and age, (2) number of pulses. Output-Indicator of venous pulsation, IR fundamental wave.

(IV. Sensor Off) (L. Sensor Off Detection and Pulse Amplitude Loss) The pulse loss and sensor off detection subsystem uses a pre-trained neural network and the sensor is detached from the patient (L. sensor off detection and pulse amplitude loss). off) Determines whether or not. Input to the neural network is a metric that quantifies several aspects of the behavior of IR and red values over a few seconds. The sample is ignored by a number of oximeter algorithm subsystems while the signal state is neither Pulse Present nor Sensor Maybe Off. The values of the signal state variables are "pulse present state, discontinuous state, pulse loss state, sensor can be off state, and sensor off state". Input-(1) Metrics, (2) Front-end servo settings and ID. Output-Signal state including sensor off indicator.

(Pulse Counting Subsystem) The subsystem averages the qualifying pulse periods from the pulse identification and qualification subsystems. The subsystem outputs the average period and the corresponding number of pulses.

The oxygen measurement algorithm includes two examples of this subsystem. The first example receives input from an example relating to pulse identification and qualification. The input waveform is processed by the signal conditioning subsystem and then filtered through the low end, but not geometrically averaged by the geometric mean subsystem. The second example of the pulse count subsystem receives inputs from two examples of pulse count identification and qualification subsystems. The first example is described above and the second example of receiving input is set averaged.

(Selection of Sources for Pulse Period) An example of a subsystem receives qualified pulses from two sources. The subsystem chooses which of these two sources to use in the pulse count calculation based solely on the analysis of one source, the "first" source. The oxygen measurement algorithm specifies a pulse identification and qualification example that does not receive the geometric mean waveform as the first source, and another example of pulse identification and qualification for the qualifying pulse period. Specify as an "alternative" source. Qualified pulse periods from alternative sources are used only for whether or not the previous pulse from the first source was rejected. When a qualified pulse period is received by the first source, the pulse period is always used to update the pulse count calculation and is an alternative source until the first source rejects the pulse period again. Qualified from the pulse period can be prevented from being used.

(Calculation of average pulse period and evaluation of number of pulses) When using Qualified_Pulse_Period, the subsystem updates the average pulse period Avg_Period by using a pulse-based variable weight IIR filter, and then calculates the Rate output from Avg_Period. To do. The steps for this filtering operation are

<maths num="1"><img file="JP2007527770A_D0001.tif" /></maths>Is. Here, in BPM, r_tIs the number of pulses corresponding to Qualified_Pulse_Period, and the subscript t-1 indicates the previously qualified pulse. Δt is the sample interval in seconds of the oxygen measurement algorithm. 60 / Δt is the number of samples per minute. x is the filter weight for a response time of 7 seconds for a typical adult pulse count.

k is the weight of the final filter, x and a continuous value r_tBased on both with the difference between. During the first few pulses, k increases to at least 1 / Total_Qualified so that the first qualified pulse can be weighted equally.

Consecutive_Qualified is the number of continuously qualified pulses, and Total_Qualified is the total number of qualified pulses since the subsystem was reinitialized. Before k is calculated, both Consecutive_Qualified and Total_Qualified are incremented each time Qualified_Pulse_Period is used. Consecutive_Qualified is set to zero when the pulse is rejected by the source of the pulse period currently in use.

Avg_Period in step 6 above_tThe official update for Avg_Period_tAnd Qualified_Pulse_Period are geometric mean. The geometric mean allows the subsystem to respond to large fluctuations over the duration of pulses-to-pulse, as well as large changes in persistence in the number of pulses.

Once the Rate is initialized to a non-zero value, Rate_Age increases for all samples, whether or not the Rate is updated.

(Context diagram) FIG. 3 is a context diagram of the pulse number calculation subsystem. The above subsystem updates Avg_Period and Rate output from Qualified_Pulse_Periods. The above subsystem uses Qualified_Pulse_Periods from Alternative_Period_Source only when Notify_Pulse_Rejected is last received from the first source. The above subsystem updates the Rate_Age output based on Qualified_Period_Age. When the Rate is updated, the subsystem sets the Pulse_Rate_Updated flag. Reinitialize commands the subsystem to reinitialize itself. Once the Rate is initialized, Increment_Rate_Age tells the subsystem to increase the Rate_Age for all samples.

<figref num="1">FIG. 1 is a block diagram of an oxygen densitometer incorporating an embodiment of the present invention.</figref><figref num="2">FIG. 2 is a diagram of a software processing block of an oxygen densitometer including an embodiment of the present invention.</figref><figref num="3">FIG. 3 is a contextual diagram of the pulse number calculation subsystem.</figref>