Detection Of Spontaneous Pulse Using The Acceleration Signals During Cardiopulmonary Resuscitation

Cardiac arrest（CA）is a medical emergency with sudden stop of cardiac mechanical activity and loss of consciousness,accompanied with disappearance of apparent breathing and pulse.Cardiopulmonary resuscitation（CPR）is a major first aid technique provided to the patients with cardiac arrest in order to restore the patient’s spontaneous circulation through rapid identification and effective treatment.Pulse detection is not only one of the main methods to identify cardiac arrest,but also an important means to judge whether patients resume spontaneous circulation during CPR.During CPR,rapid and accurate pulse detection is vital important that the inaccurate pulse detection will seriously affect the survival of patients.False negative judgment will lead to unnecessary sustained compression or defibrillation,on the contrary,detection of pulse with long interruptions of chest compressions and false positive judgment will make patients lose the best opportunity for treatment.For out-of-hospital cardiac arrest（OHCA）patients,pulse detection is mainly achieved by palpation,but previous studies showed that palpation is time-consuming and inaccurate even for experienced rescuers.To solve this problem,lots of automatic pulse detection methods have been investigated including the use of thoracic impedance（TI）signals,photoplethysmography（PPG）and near-infrared spectroscopy（NIRS）and other physiological signals that may be related to the pulse.However no practical algorithm or device has been applied to the actual emergency treatment of OHCA patients yet.The blood pressure and end-tidal carbon dioxide（ETCO₂）monitoring recommended by the resuscitation guidelines are obviously not suitable for patients in out-of-hospital settings.Automated external defibrillator（AED）is a key medical instrument for rescuing patients with cardiac arrest.During CPR,AED is mainly used for rhythm analysis and performing defibrillation.Now some AEDs also integrate a feedback unit for monitoring the depth and frequency of chest compression in real time during CPR.With the feedback unit,AED can guide rescuers to carry out effective chest compression,and greatly improve the quality of CPR.The feedback unit has two major forms,one is based on the change of the patient’s thoracic impedance during chest compression,and the other is based on the accelerometer sensor placed on the surface of the patient’s chest over the heart.The shock advice algorithms in AEDs are able to identify shockable rhythms and non-shockable rhythms with high accuracy,but can’t distinguish a perfusing rhythm（PR）from a non-perfusing rhythm when an organized rhythm（OR）is present.The accelerometer-based CPR feedback sensor monitors the depth and frequency of chest compression by sensing thoracic movement.When the patient’s spontaneous circulation is restored,the heart beat will cause the chest wall movement which might be captured by the accelerometer-based CPR sensor that placed on the surface of the chest.It is possible to detect the pulse automatically by extending the algorithm with the powerful computing ability of AED.If this hypothesis can be confirmed,the pulse can be further detected by the automatic pulse detection algorithm while OR is identified by AED.To confirm this hypothesis,the following work has been done:1.Animal experiment and data extractionThe experimental data were retrospectively collected from 49 male adult domestic pigs experienced cardiac arrest and CPR in previous study,of which 44 were ventricular fibrillation models and 5 were asphyxia models.A total of 1025 segments with 3 seconds were extracted.Each segment contained electrocardiogram（ECG）,blood pressure（ABP）and acceleration（ACC）signals.123 segments were excluded because of changes of state or abnormal data,and the remaining segments were classified by emergency doctors in Southwest Hospital.Finally,216 OR,included 63 PR and 153 PEA,and 686 VF were included in the database.The database was randomly divided into training set and testing set for the study of ECG rhythm analysis algorithm.The training set consisted of 108 OR and 346 VF and the testing set inlcuded 108 OR and 340 VF.For classification of PR and PEA by pulse detection algorithm,all OR segments,including 63 PR and 153 PEA,were randomly divided into training set（32 PR and 74 PEA）and testing set（31 PR and 79 PEA）.2.Rhythm analysis algorithmRhythmanalysis algorithm is mainly used to classify VF and OR.The algorithm extracts five characteristic parameters of ECG signal,including peak fold-line-rate（F_p）,peak angle fold-line-rate（Fα）,heart rate（HR）,amplitude distribution rate（F_a）and autocorrelation rate（F_r）.BP neural network model was used to classify VF and OR,the extracted characteristic parameters were used as the inputs level of neural network,and the output was rhythm category.The parameters of neural network model are optimized by training set.The test set is used to evaluate the rhythm detection algorithm.Finally,the detection accuracy of QRS 97.5%was achieved.3.Pulse detection classifierPulse detection algorithmis provided after the ECG rhythm is classified as OR.The peak correlation coefficient（CCp）denoting the maximum of the normalized cross-correlation sequence,was used to quantify the degree of correlation between ECG and ACC and to discriminate PR from PEA.If CCp is greater than the threshold set b y the classifier,it is considered that pulse exists at this time,otherwise there is no pulse.The training set was used to optimize the filtering parameters of acceleration signal and the testing set was used for the assessment of pulse classifier.The accuracy,sensitivity and specificity of the testing set were 96.7%,93.6%and 97.5%,respectively.4.The scheme of automatic pulse detectionThe ECG rhythm analysis algorithm and pulse detection algorithm presented in the current study are integrated to form a complete automatic pulse detection scheme.The data segments during compression pause or after defibrillation are extracted first,thenthe ECG rhythm is detected by the rhythm analysis algorithm to discriminate OR and VF.While OR is comfirmed,the pulse detection algorithm is provided to detect the spontaneous pulse.To verify the performance of the scheme,the testing dataset of rhythm analysis is used.Correct classification of pulse followed by OR rhythm that detected by rhythm analysis algorithm for the segmentsin the testing set is considered to be successful pulse detections.The accuracy of pulse detection is 98.9%.The experimental results demonstrated that the acceleration signal used for resuscitation quality monitoring carries pulse information.The pulse detection algorithm based on acceleration signal and ECG signal proposed in this study can be used to detect the presence of spontaneous pulse during CPR.