FIELD OF THE INVENTION

[ 0001] The invention relates to a method of and apparatus for processing

photoplethysmography signals during cardiopulmonary resuscitation (CPR) to detect the return or presence of spontaneous pulse.

BACKGROUND OF THE INVENTION

[ 0002] CPR is the emergency procedure for people suffering from a cardiac arrest.

During CPR, chest compressions are delivered to artificially generate circulation of blood, and ventilations are given to supply blood with oxygen. The goal of CPR is to achieve return of spontaneous circulation (ROSC). When ROSC has been achieved, the heart of the patient has resumed beating and generates a spontaneous circulation which is life-sustaining. CPR can be stopped after achieving ROSC.

[ 0003] FIGURE 1 presents a simplified schematic of the 30:2 CPR protocol 100 that is widely used during cardiac rescues. CPR is delivered in 2-min blocks, in which series of thirty chest compressions are alternated by two ventilations. A compression rate of 100 to 120 min ^"1 is targeted. During the ventilations, compressions are stopped. At the end of each 2-min block, clinicians determine whether ROSC has been achieved. If so, CPR can be stopped. If ROSC has not been achieved, a new 2-min block of CPR is initiated. International guidelines state that interruptions of the compressions for assessment of ROSC should last no longer than 10 s.

[ 0004] ROSC assessment involves pulse checks which are typically performed by manual palpation. Manual palpation interrupts the chest compressions, is time-consuming and is unreliable. Consequently, manual palpation can lead to long interruptions in the chest compressions which degrades the quality of CPR.

[ 0005] The 30:2 CPR protocol is typically followed when applying mouth-to-mouth or bag- mask ventilation. Once a patient has been intubated for ventilations, compressions are not paused for ventilations any longer, but are uninterruptedly delivered during a 2-min block of CPR. Ventilations are then given during ongoing compressions targeting a ventilation rate of about 10 min ^"1 .

[ 0006] Determining whether a patient has achieved ROSC requires an assessment of the

electrical as well as the mechanical activity of the heart. The electrical activity of the heart is assessed via electrocardiography (ECG). If the ECG rhythm shows fibrillation or asystole, the heart cannot generate any output and it is clear that CPR should be continued. If the ECG signal looks organized with R-peaks, this is considered a potentially perfusing rhythm. An organized ECG signal does not imply that the heart has resumed contracting and generates output. Presence of an organized ECG signal while the heart has not resumed contracting with output is called pulseless electrical activity (PEA). PEA occurs frequently. PEA has been reported in about 49% of cases of in-hospital cardiac arrest, and in about 21% of cases of out-of -hospital cardiac arrest. Therefore, if an organized ECG signal is observed, a pulse check should still be performed to determine whether the heart is actually contracting and generating output. In current clinical practice, pulse checks are typically performed by manual palpation at the neck, groin or wrist.

[ 0007] Manual palpation is known to be unreliable and time-consuming, and requires

interruption of the chest compressions. Manual palpation can take significantly longer than the 10 s recommended for ROSC assessment, especially when a spontaneous pulse is absent. As a result, manual palpation can lead to long interruptions in the chest compressions, which reduce the compression-generated blood flow and can thereby negatively impact CPR outcome. [ 0008] Methods exist for more objective and more continuous assessment of ROSC. These methods are for instance monitoring of end-tidal CO2, invasive blood pressure, or central venous oxygen saturation, any of which can also be applied during compressions. However, measurement of end-tidal CO2 requires intubation, and measurement of invasive blood pressure and central venous oxygen saturation require placement of catheters with associated risks for the patient. These more objective methods are less practical in use during CPR. Therefore, what is needed is an objective, rapid, non-invasive and easy-to-use method to detect presence or absence of a spontaneous pulse to support ROSC detection.

[ 0009] The discussion above indicates that what is needed is an improved apparatus and method to automatically detect pulse during the application of CPR. This invention proposes such a method and apparatus.

SUMMARY OF THE INVENTION

[ 0010] The inventors have discovered a method by which a PPG signal can be used to detect a return (or continued presence) of a spontaneous pulse during the application of CPR. PPG is a non-invasive, easy-to-use, optical technology which is commonly applied to measure pulse rate (PR) and oxygen saturation (Sp02) in the clinic. To date, however, it has not been commonly applied for use during CPR. In the case of CPR chest compressions, the inventors have discovered that the presence and absence of pulse can be rapidly detected using PPG during short interruptions in the chest compressions.

[ 0011] FIGURE 2 illustrates porcine data 200 obtained during a laboratory test. The top trace 210 is the band-pass filtered PPG signal. The bottom trace 212 is the invasive aortic blood pressure measured as a reference. Before the defibrillation shock, pre-shock interval 214, the animal is in cardiac arrest. Here, during the few-second ventilation pauses 220 indicated by V, the PPG signal directly shows that there is no spontaneous pulse, confirming cardiac arrest. After the defibrillation shock at 230 and during post-shock interval 216, the heart of the animal resumes beating, as is confirmed by the rising aortic blood pressure. Here, during the few-second ventilation pauses 240, the PPG signal directly shows presence of a spontaneous pulse.

[ 0012] Data such as described above indicates that it may be possible to accurately assess

pulsatile function in a subject undergoing a CPR protocol using PPG. Presence and absence of a spontaneous pulse in the PPG signal are indicated on a continuous scale, which can support the clinician in decision making. Furthermore, in case of 30:2 CPR, pauses in chest compressions are naturally available in the protocol, which can be used for the PPG-based pulse detection support.

[ 0013] Thus what is proposed is an apparatus and method which incorporates an algorithm that performs as follows:

[ 0014] Automatic detection of pauses in the chest compressions, e.g., by making use of a

compression reference signal such as trans-thoracic impedance, accelerometry, chest compression force, a radar signal, or a camera signal, or alternatively derived from the PPG signal itself. This is followed by an interpretation of the PPG signal during the pauses in the chest compressions to indicate to the rescuer one or more of the following indications, as well as the relative or absolute value of the indication: presence of a spontaneous pulse in the PPG signal, absence of a spontaneous pulse in the PPG signal, or artifacts in the PPG signal.

[ 0015] The benefits of the inventive method and apparatus over the prior art are several-fold.

First, the method improves the functionality and efficiency of a signal processing computer and circuit in a medical device such as a defibrillator. Second, by rapidly detecting absence of a spontaneous pulse with PPG, long unnecessary interruptions of the compressions for pulse checks in case of PEA can be prevented. This increases the fraction of time that chest compressions are delivered during CPR, which may improve CPR outcome.

[ 0016] A third benefit is provided in that by rapidly detecting presence of a spontaneous pulse with PPG, the clinician can be supported to determine if further assessment of ROSC is adequate and compression should be stopped, which may reduce the risk of compression-induced refibrillation. A fourth benefit is provided in that by rapidly detecting presence of a spontaneous pulse with PPG, the clinician can be supported to determine withholding or postponing administration of vasopressors, which may also reduce the risk of re-arrest. A fifth benefit is that PPG is an easy-to-use and non-invasive technology, which is more convenient to use during CPR than, e.g., a continuous invasive blood pressure measurement. And by including artifact detection, the clinician or rescuer can be warned when the PPG signal should not be interpreted.

[ 0017] Thus in accordance with the principles and objectives of the invention, a method is

described for detecting pulse during CPR, comprising the steps of providing a signal processing apparatus having a photoplethysmography (PPG) detector which outputs a PPG signal stream and a cardiopulmonary resuscitation (CPR) compressions detector which outputs a CPR compressions reference signal (CMP). The method further describes the steps of automatically detecting a cessation of CPR compressions and a subsequent CPR compressions-free time period based on the CMP from the CPR compressions detector, acquiring a stream of normalized PPG signal values from the PPG detector over the time period, and deriving a normalized PPG signal autocorrelation-function-prominence and periodicity from the stream of normalized PPG signal values acquired over the time period. The method further includes steps of calculating a normalized time derivative with respect to each of the normalized PPG signal values referred to as the normalized phase-space plot (nPSP), identifying and clustering zero- values of the normalized PPG signals in the nPSP, identifying and clustering zero-values of the normalized derivatives in the nPSP, and classifying a pulse condition consisting of one selected from the set of pulse, no pulse, and artifact based upon the deriving step and both of the identifying and clustering steps. The method concludes by outputting a result of the classified pulse condition. The classified pulse condition can be output on a continuous scale from, e.g., zero to one, where zero indicates pulse absence with a high likelihood and one indicates pulse presence with a high likelihood.

[ 0018] Other embodiments of the inventive method include identifying and clustering steps which include a zero crossing (ZC) analysis wherein a proportion of ZCs of normalized PPG signals and normalized PPG signal time-derivatives in the nPSP greater than a predetermined value are identified, and used in the classifying step to determine the pulse condition. That is, the nPSP is a graphical representation of the PPG signal where one coordinate (say x) is the normalized PPG signal and the other coordinate (say y) is the normalized PPG signal time- derivative. When identifying ZCs of one of the two coordinates, i.e., (0,y) or (x,0), it is determined whether the other coordinate, i.e., y or x respectively, is greater than a predetermined value.

[ 0019] In another embodiment, the area traversed by the traces in the nPSP can be analyzed rather than only the zero-values and zero-crossings, and a fraction of the traces can be required to traverse within set bounds in order to consider pulse to be present.

[ 0020] In an embodiment, features derived from the normalized PPG signal and the time- derivate of the normalized PPG signal can be used in a logistical regression classifier to determine the pulse condition on a continuous scale from e.g., zero to one, where zero indicates a high likelihood of pulse absence and one indicates a high likelihood of pulse presence.

Furthermore, the range spanned by the continuous classifier output can be subdivided in various regions, where, e.g., a classifier output smaller than threshold Tl should be interpreted as pulse absence, a classifier output larger than T2 should be interpreted as pulse presence, and a classifier output larger than or equal to Tl or smaller than or equal to T2 should be interpreted as indeterminate.

[ 0021] Another embodiment of the invention is an apparatus for detecting spontaneous pulse during CPR, comprising an input port for a photoplethysmography (PPG) signal, and an input port for a CPR compressions sensor providing a CPR compressions reference signal (CMP). The apparatus further includes a signal processing circuit having a processor configured to execute an algorithm for determining a cessation of CPR compressions from the CMP, then processing the PPG signal during compressions-free time period by identifying and clustering values of the PPG signal and the PPG signal derivative, classifying a pulse condition consisting of one selected from the set of pulse, no pulse, and artifact based upon the processing, and providing an output of the pulse condition, possibly on a continuous scale. The apparatus further includes a user output including one of an aural and a visual indication of the pulse condition. The apparatus may be a defibrillator.

[ 0022] Another embodiment of the invention may be a computer program product including a computer-readable storage medium storing instructions in non-volatile memory that when executed by a computer causes the computer to perform a method for using a computer system for detecting a spontaneous pulse during the application of cardiopulmonary resuscitation (CPR). The method of the computer program product may be the method described above.

[ 0023] Other advantageous embodiments are defined by the dependent claims.

BRIEF DESCRIPTION OF THE FIGURES

[ 0024] FIGURE 1 is an illustration of a typical prior art CPR protocol. [ 0025] FIGURE 2 illustrates the detection of spontaneous pulse during a CPR event using a PPG detector.

[ 0026] FIGURE 3 illustrates a first embodiment of the main elements of the inventive apparatus.

[ 0027] FIGURE 4 illustrates a method step for determining an autocorrelation-function- prominence and periodicity of a pulse signal in a PPG signal stream.

[ 0028] FIGURE 5 illustrates a method step for determining the presence of a pulse signal in a PPG signal stream.

[ 0029] FIGURE 6 illustrates a method for classifying a pulse signal in a PPG signal stream.

[ 0030] FIGURE 7 is a method flow chart according to one embodiment of the invention.

DETAILED DESCRIPTION OF EMBODIMENTS

[ 0031] Now turning to FIGURE 3, a system 300 for detecting a spontaneous pulse during

CPR is shown, which includes apparatus 350. The apparatus 350 may be a patient monitoring device, an automatic CPR compressions device, a defibrillator, or another medical device for use in or out of the hospital. In general, apparatus 350 will be contained in a housing which contains signal processing circuitry that includes computer processors and controllers. Input ports into the housing receive data signals from various external sensors. An output port may contain audible annunciators, lights, and displays for providing information to the user.

[ 0032] In a preferred embodiment, apparatus 350 includes an input port 310 for a

photoplethysmography (PPG) signal. A PPG signal is an optically obtained plethysmograph signal, which provides a measure of the variations of blood volume in the illuminated tissue. It can be obtained by a pulse oximeter which illuminates the skin and measures changes in light absorption. A conventional pulse oximeter monitors the perfusion of blood to the dermis and subcutaneous tissue of the skin. Besides the ECG, the PPG signal is one of the most often acquired signals in clinics, especially in anesthesia or intensive care. The PPG input thus may be from a contact sensor measured from the finger, ear, nose or forehead. The PPG signal can also be obtained from a non-contact sensor, such as a camera to remotely detect variations in red, infrared, or green light emanating from a subject's skin. In an alternate embodiment, an ultrasound or radar transducer signal may be substituted for the PPG input signal, and further used as a substitute for the PPG input signal.

[ 0033] Apparatus 350 also includes a second input port 320 for a CPR compressions sensor which provides a CPR compressions reference signal (CMP). The CMP reference signal can be of several types of compression sensors, based on for example accelerometry, trans-thoracic impedance from ECG or defibrillator electrodes placed on the patient, force, camera, ultrasound transducer, or a radar signal sensor. Alternatively, the CMP reference signal can be the PPG signal or a derived signal relating to the movement of the chest during CPR compressions. In yet another embodiment, the CMP reference signal can be a control signal provided by an automatic chest compression robot.

[ 0034] The PPG signal and CMP reference signal inputs are provided to a signal processing circuit 330 which is configured to determine a pulse condition of "pulse", "no pulse", or "artifact" from the signal input data. The signal processing circuit 330 executes computer program instructions residing in non-volatile memory in accordance with the algorithm and method described in detail in the following paragraphs. In general the circuit executes the algorithmic functions of determining a cessation of CPR compressions by the CMP and then processing the PPG signal during compressions-free time period by identifying and clustering values of the PPG signal and the PPG signal derivative. The algorithm then classifies a pulse condition consisting of one selected from the set of pulse, no pulse, and artifact based upon the processing, and provides an output of the pulse condition. [ 0035] Signal processing circuit 330 conveys the output of the pulse condition to a user output 340 for information and control purposes. User output 340 is preferably one or more of an aural and a visual indication of the pulse condition. FIGURE 3 shows a simple light and text display clearly indicating one of "pulse" 360, "No pulse" 370, or "Artifact" 380. Pulse presence and absence can be indicated on a continuous scale as illustrated in FIGURE 3. The continuous scale can range from e.g., zero to one, where zero indicates a high likelihood of pulse absence and one indicates a high likelihood of pulse presence. Furthermore, the range spanned by the continuous classifier output can be subdivided in various regions, where, e.g., a classifier output smaller than threshold Tl should be interpreted as pulse absence, a classifier output larger than T2 should be interpreted as pulse presence, and a classifier output larger than or equal to Tl or smaller than or equal to T2 should be interpreted as indeterminate. In addition, user output 340 may be configured to provide guidance instructions that are appropriate to the determined pulse condition. For example, output 340 may provide an instruction to "further assess pulse presence" for a "pulse" condition, or to provide an instruction to "continue CPR" in the case of a "no pulse" or "indeterminate" condition.

[ 0036] If apparatus 350 is embodied as a defibrillator, externally applied ECG electrodes may provide a concurrent stream of ECG signals via a third input port (not shown) to signal processing circuit 330. In prior art defibrillators, an analysis of the ECG signal data completely determines whether the defibrillator is to be armed for providing electrotherapy or not. In these defibrillators, PEA and normal ECG rhythm cannot be distinguished. But in an

apparatus/defibrillator 350 that also captures spontaneous pulse activity, a PEA condition can be identified. Signal processing circuitry 330 may control the arming of the defibrillator for electrotherapy based on both of the analysis of the ECG signal and of the pulse condition. For example, if the ECG signal is analyzed as shockable, the circuit 330 may initiate arming of the apparatus/defibrillator 350 for delivery of electrotherapy only if the pulse condition is also classified as "no pulse" or "indeterminate". But if the ECG signal is analyzed as shockable and the pulse condition is classified as "pulse", circuit 330 may prevent arming, or dis-arm high voltage circuitry to prevent an inappropriate delivery of electrotherapy.

[ 0037] As used herein for purposes of the present disclosure, the term "processor" is used

generally to describe various apparatus relating to the operation of a medical apparatus, system, or method. A processor can be implemented in numerous ways (e.g., such as with dedicated hardware) to perform various functions discussed herein. A processor is also one example of a controller which employs one or more microprocessors that may be programmed using software (e.g., microcode) to perform various functions discussed herein. A controller may be

implemented with or without employing a processor, and also may be implemented as a combination of dedicated hardware to perform some functions and a processor (e.g., one or more programmed microprocessors and associated circuitry) to perform other functions. Examples of controller components that may be employed in various embodiments of the present disclosure include, but are not limited to, conventional microprocessors, application specific integrated circuits (ASICs), and field-programmable gate arrays (FPGAs). "Outputs" and "signals" may be understood to be electrical or optical energy impulses which represent a particular detection or processing result.

[ 0038] In various implementations, a processor or controller may be associated with one or more computer storage media (generically referred to herein as "memory," e.g., volatile and nonvolatile computer memory such as RAM, PROM, EPROM, and EEPROM, floppy disks, compact disks, optical disks, magnetic tape, etc.). In some implementations, the storage media may be encoded with one or more programs that, when executed on one or more processors and/or controllers, perform at least some of the functions discussed herein. Various storage media may be fixed within a processor or controller or may be transportable, such that the one or more programs stored thereon can be loaded into a processor or controller so as to implement various aspects of the present invention discussed herein. The terms "program" or "computer program" are used herein in a generic sense to refer to any type of computer code (e.g., software or microcode) that can be employed to program one or more processors or controllers.

[ 0039] In various implementations, the processor or controller may be a computer that is

configured to execute instructions stored in a computer-readable storage medium, the instructions causing the computer to perform the general steps of automatically detecting (730) a cessation of CPR compressions from an input of CPR compressions reference signal (CMP) and a subsequent CPR compressions-free time period based on the CMP, acquiring (740) a stream of normalized PPG signal values from a photoplethysmography (PPG) detector which outputs a PPG signal stream over the time period, deriving (750) a normalized PPG signal autocorrelation- function-prominence and periodicity from the stream of normalized PPG signal values acquired over the time period, calculating a normalized derivative (760) with respect to each of the normalized PPG signal values, identifying and clustering zero-values of the normalized PPG signals (770), identifying and clustering zero-values of the normalized derivatives (770), and classifying (780) a pulse condition consisting of one selected from the set of "pulse", "no pulse", and "artifact" based upon the deriving step and both of the identifying and clustering steps. The condition of "pulse" and "no pulse" can be provided on a discrete or continuous scale. The computer instructions may further control the outputting (790) of a result of the classified pulse condition and/or additional instructional guidance.

[ 0040] FIGURE 7 illustrates the preferred method 700 for detecting the pulse condition in the presence of CPR. FIGURE 4 refers in detail to a method step 750 for deriving a PPG signal prominence and periodicity. FIGURE 5 refers in detail to a method step 770 for identifying and clustering zero-values of normalized PPG signals and their time derivatives. FIGURE 6 refers in detail to a method step 780 of classifying a pulse condition based on the derived prominence and periodicity and the identifying and clustering. Thus together these FIGURES illustrate in detail a preferred algorithm for detecting a pulse condition in the presence of CPR.

[ 0041] Method 700 for detecting a pulse condition in the presence of CPR begins with a step of providing 710 of a PPG signal stream and a CMP as inputs. The inputs may be provided in accordance with a defibrillator as described with regards to FIGURE 3, although different types of apparatus' may be used. The PPG and CMP signal inputs to the defibrillator may optionally be disposed as a camera-based sensor. Alternatively, the CMP signal input can be provided by a compression robot, and can be derived from a compression control signal used in the compression robot.

[ 0042] The CMP input signal stream is then monitored at monitoring step 720 for the particular purpose of identifying periods of signal that represent ongoing chest compressions, and periods of relative low signal or noise that represent a lack of chest compression activity. One method of detecting compression signal is by analysing the power in an accelerometer or a trans-thoracic impedance signal. Another method of detection compression signal in an accelerometer or a trans-thoracic impedance signal is by analysing the shape of the signal and determining the periodicity of the signal. The PPG signal stream may be captured at all times, with values stored in computer memory. Non-usable PPG signals, i.e. during chest compressions periods, may optionally be discarded, while usable PPG signals, i.e. during pauses in chest compressions, may be retrieved for subsequent analysis.

[ 0043] Method 700 continues at step 730 by automatically detecting pauses in, or cessations of, ongoing chest compressions as indicated by the CMP. As previously described, a pause may be initiated in order to ventilate a non-intubated patient and per protocol may last up to ten seconds when assessing ROSC. Step 730 further determines whether the pause duration exceeds a minimum analysis time period. A preferable minimum time period is four seconds of pause, which correlates to the size of sequential analysis time windows. If the time period is too short, the method returns and continues to monitor for a subsequent cessation of CPR at step 720.

[ 0044] If a sufficiently long analysis time window of cessation is detected by CMP, acquisition of a PPG signal stream values begins at acquisition step 740. PPG values from start of cessation may be retrieved from memory. The acquisition step 740 may also provide a PPG signal artifact detection to ascertain whether the PPG signal is usable. PPG signal artifact may for example be detected by comparing predetermined thresholds to one or more of the PPG signal transmission ratios, e.g. the detected photocurrent versus the applied LED current [nA/mA], that is higher than would be expected with tissue between the LEDs and photodiode. The PPG signal can also be considered to be in artifact when the PPG signal saturates or clips, or too large a signal range is spanned by the raw PPG signal in too short a period of time. PPG signals with too much artifact may be non-indicative of a pulse condition. When detected at decision step 742 then, the pulse condition is classified as "artifact" at step 744 and is provided to the output display at step 790.

[ 0045] If the PPG signal is clean (no artifact detected at 742), the PPG signal is analysed for prominence and periodicity at step 750. The normalized PPG signal data is grouped into sequential time windows for analysis, preferably 4 seconds long. One method 400 for determining prominence and periodicity is illustrated in FIGURE 4 in a normalized

autocorrelation function (nACF) 410 as a function of time 420. In this exemplary method, the PPG signal prominence 430 is determined as the amplitude of the signal relative to the largest of the two adjacent minima in the surrounding intervals, here shown as minima 432 and 434. The location of the prominence peak at 440 indicates periodicity of the signal, which further may be used to determine rate, i.e. pulse rate in beats per minute (BPM). Other features of the nACF analysis may be determined, such as the number or sum of the prominent peaks in the ACF window, given a threshold for the prominence. Alternatively, pulse rate can be determined from other methods than the nACF, such as via Fourier Analysis or other types of spectral analysis.

[ 0046] A pulsatility-derived measure of the PPG signal may also be determined at step 750 by root mean square (RMS) analysis. The RMS pulsatility can for instance be obtained as

PP9trp [n]

PPdrx [n]

PPSMM

[ 0048] with ppgbp[n] the band-pass filtered PPG signal, ppgbl[n] the baseline of the PPG signal extracted via a low-pass filter, ppgn[n] the normalized PPG signal, and Nwdw the window duration in samples. The unit of the pulsatility, pit, is mNP which stands for milli-Normalized- Pulsatility. By using the RMS value, there is no requirement of detecting individual pulses first to arrive at an amplitude measure, which makes the RMS value convenient to apply during pauses without spontaneous pulse as well.

[ 0049] The PPG pulse rate is further assessed at rate decision step 752 to ascertain whether the calculated pulse is outside of an expected range. One expected range of rates is from about 30 BPM to about 300 BPM. Artifact-free rate values outside of this range may indicate that the PPG signal is not associated with pulse, and therefore a "no pulse" or "indeterminate" classification is set at step 754, which is further provided to the output/display step 790.

[ 0050] If within an expected pulse range, outputs of step 750 of a normalized PPG signal

prominence and a PPG pulse rate are provided to calculating step 760 and identify /clustering step 770. Calculating step 760 calculates a normalized time-derivative with respect to each of the normalized PPG signal values. The sets of normalized PPG signal values and their derivatives are then processed at identifying step 770 in a normalized phase-space plot (nPSP) analysis, which analyses "zero-crossings", i.e. zero-values, of PPG signals versus their derivatives and vice-versa.

[ 0051 ] PPG-based pulse analysis methods using normalized phase-space plots are described in detail in co-assigned and co-pending International Patent Application No. PCT/IB2010/055397, entitled "METHOD OF AND APPARATUS FOR PROCESSING PHOTO- PLETHYMOGRAPH SIGNALS" and filed on November 24 2010, and herein incorporated by reference. A similar method is applied in step 770, which is illustrated graphically by the nPSP plot 500 shown in FIGURE 6.

[ 0052] The PPG signals and their derivatives over each time period, e.g. four second windows, are plotted on the signal axis 510 and the derivative axis 520 respectively. Signals and derivatives are normalized by dividing by their absolute maxima in the window. Then the values of the derivatives for the zero-values of the normalized PPG signals, shown as zero-crossings (ZCs) 530 are identified. The ZCs 530 are then clustered according to the quadrant and analysed for their distance from the origin, i.e. magnitude of derivative values, and for the distance of each ZC from its cluster center. Similarly, the values of the normalized PPG signals for the zero- values of the normalized derivatives, shown as zero-crossings (ZCs) 540 are identified. The ZCs 540 are then clustered according to the quadrant and analysed for their distance from the origin, i.e. magnitude of PPG values, and for the distance of each ZC from its cluster center.

[ 0053] Although four total clusters are indicated in FIGURE 6, more or less clusters may be identified. For example, two separate clusters in the top quadrant of FIGURE 6 might be identified by the particular ZCs. [ 0054] Various ZC statistical parameters may be developed from the data in each window. For example a proportion of ZCs farther than a particular fraction from the origin, e.g. 0.65 from the origin may be used as a factor to assess absence or presence of pulse. Other parameters that may be useful for classifying pulse/no-pulse are:

[ 0055] The bounding box of the phase-space plot, i.e., the smallest rectangle encompassing a defined number of trajectories in the phase-space plot;

[ 0056] The aspect ratio of both sides of the bounding box;

[ 0057] The number of zero-crossings close to the origin divided by total number of zero

crossings in the non-normalized phase-space plot;

[ 0058] K-means clustering of zero-crossings in the non-normalized phase-space plot; and

[ 0059] Pulse rate determined from the number of zero-crossings per unit of time in one quadrant of the phase-space plot.

[ 0060] Determination of the fraction of the traces in the nPSP that traverse within set bounds.

[ 0061] The desired parameters from deriving step 750 and identifying/clustering step 770 are provided to a classifying step 780 for final determination of a pulse condition of "pulse", "no pulse", or "artifact." Preferably, classifying step 770 conducts a logistical regression of the parameters to arrive at a "no-pulse" or "pulse" decision on a continuous scale from, e.g., zero (high likelihood of pulse absence) to one (high likelihood of pulse presence), and for conflicting or confounded parameters to arrive at an "artifact" decision. In one preferred example, the parameters of ZC zero-value spread (nPSP_kmeansclusterdistance), proportion of ZCs clustered outside a numerical threshold from the origin (nPSP_zerocrossings), and the PPG signal prominence (max_prominence) combine to determine a classifier output (classifier_output) by the following calculation: %% The classifier output (interpreted as posterior likelihood)

h = theta_0 + theta_l * max_prominence + theta_2 * nPSP_zerocrossings + theta_3 * nPSP_kmeansclusterdistance

classifier_output = 1 / (1 + exp(-h) );

%% Coefficients for logistic regression model

theta_0 = -6.9166;

theta_l = 8.8664;

theta_2 = 2.0269;

theta_3 = -2.843;

[ 0062] The classifier output from step 780 is selected accordingly to indicate absence of spontaneous pulse, or to indicate presence of pulse on a continuous scale from, e.g., zero (high likelihood of pulse absence) to one (high likelihood of pulse presence). The output is provided to the user at display step 790.

[ 0063] FIGURE 6 illustrates the performance and output of the method 700 described above. The parameters of prominence 630, distance from cluster center 640, and proportion of ZC outside a distance from the origin 650 are plotted for each time window segment tested. The respective classifier output on a continuous scale 610 is shown in the lower part of FIGURE 6. When a true spontaneous pulse appears in the PPG signal, as indicated by the reference label "X"s appearing at the right side of output line 610, prominence 630 increases, distance to cluster centers 640 decreases, and the fraction of distant ZCs 650 increases. As can be seen, the resulting classifier output 610 correctly follows the transition from a "no-pulse" condition to a "pulse" condition. Furthermore, it may be noted that the range spanned by the continuous classifier output can be subdivided in various regions, where, e.g., a classifier output smaller than threshold Tl should be interpreted as pulse absence, a classifier output larger than T2 should be interpreted as pulse presence, and a classifier output larger than or equal to Tl or smaller than or equal to T2 should be interpreted as indeterminate.

[ 0064] It may be noted that the classifier is able to detect presence or absence of a spontaneous pulse in a single four-second window, a quicker time than manual palpation. Thus, the rescuer or clinician can rapidly decide whether or not a further assessment of ROSC is adequate, and whether or not CPR chest compressions should be resumed.

[ 0065] While the invention has been illustrated and described in detail in the drawings and

foregoing description, such illustration and description are to be considered illustrative or exemplary and not restrictive; the invention is not limited to the disclosed embodiments. Other variations to the disclosed embodiments can be understood and effected by those skilled in the art in practicing the claimed invention, from a study of the drawings, the disclosure, and the appended claims. In the claims, the word "comprising" does not exclude other elements or steps, and the indefinite article "a" or "an" does not exclude a plurality. The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage. Any reference signs in the claims should not be construed as limiting the scope.