WO2021103796A1 - 心电信号处理方法、心电信号处理装置和电子设备 - Google Patents

心电信号处理方法、心电信号处理装置和电子设备 Download PDF

Info

Publication number
WO2021103796A1
WO2021103796A1 PCT/CN2020/117707 CN2020117707W WO2021103796A1 WO 2021103796 A1 WO2021103796 A1 WO 2021103796A1 CN 2020117707 W CN2020117707 W CN 2020117707W WO 2021103796 A1 WO2021103796 A1 WO 2021103796A1
Authority
WO
WIPO (PCT)
Prior art keywords
sequence
symbol
interval
value
time interval
Prior art date
Application number
PCT/CN2020/117707
Other languages
English (en)
French (fr)
Inventor
孟桂芳
Original Assignee
京东方科技集团股份有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 京东方科技集团股份有限公司 filed Critical 京东方科技集团股份有限公司
Priority to US17/289,082 priority Critical patent/US20220304610A1/en
Publication of WO2021103796A1 publication Critical patent/WO2021103796A1/zh

Links

Images

Classifications

    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
    • A61B5/316—Modalities, i.e. specific diagnostic methods
    • A61B5/318—Heart-related electrical modalities, e.g. electrocardiography [ECG]
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
    • A61B5/316—Modalities, i.e. specific diagnostic methods
    • A61B5/318—Heart-related electrical modalities, e.g. electrocardiography [ECG]
    • A61B5/346—Analysis of electrocardiograms
    • A61B5/349—Detecting specific parameters of the electrocardiograph cycle
    • A61B5/352—Detecting R peaks, e.g. for synchronising diagnostic apparatus; Estimating R-R interval
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/02—Detecting, measuring or recording pulse, heart rate, blood pressure or blood flow; Combined pulse/heart-rate/blood pressure determination; Evaluating a cardiovascular condition not otherwise provided for, e.g. using combinations of techniques provided for in this group with electrocardiography or electroauscultation; Heart catheters for measuring blood pressure
    • A61B5/024—Detecting, measuring or recording pulse rate or heart rate
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/02—Detecting, measuring or recording pulse, heart rate, blood pressure or blood flow; Combined pulse/heart-rate/blood pressure determination; Evaluating a cardiovascular condition not otherwise provided for, e.g. using combinations of techniques provided for in this group with electrocardiography or electroauscultation; Heart catheters for measuring blood pressure
    • A61B5/024—Detecting, measuring or recording pulse rate or heart rate
    • A61B5/02405—Determining heart rate variability
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
    • A61B5/316—Modalities, i.e. specific diagnostic methods
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
    • A61B5/316—Modalities, i.e. specific diagnostic methods
    • A61B5/318—Heart-related electrical modalities, e.g. electrocardiography [ECG]
    • A61B5/346—Analysis of electrocardiograms
    • A61B5/349—Detecting specific parameters of the electrocardiograph cycle
    • A61B5/361—Detecting fibrillation
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
    • A61B5/316—Modalities, i.e. specific diagnostic methods
    • A61B5/318—Heart-related electrical modalities, e.g. electrocardiography [ECG]
    • A61B5/346—Analysis of electrocardiograms
    • A61B5/349—Detecting specific parameters of the electrocardiograph cycle
    • A61B5/364—Detecting abnormal ECG interval, e.g. extrasystoles, ectopic heartbeats
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7203—Signal processing specially adapted for physiological signals or for diagnostic purposes for noise prevention, reduction or removal
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
    • A61B5/7267—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems involving training the classification device
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/08—Feature extraction
    • G06F2218/10—Feature extraction by analysing the shape of a waveform, e.g. extracting parameters relating to peaks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/12—Classification; Matching

Definitions

  • the present disclosure generally relates to the field of computer technology, and in particular, to an electrocardiographic signal processing method, an electrocardiographic signal processing device, and electronic equipment.
  • the electrocardiogram (ECG) signal is a comprehensive reflection of the electrical activity of the heart on the body surface.
  • ECG electrocardiogram
  • One way of processing is to classify the ECG signal by training a classifier. It is necessary to prepare a large amount of sample data, and perform feature construction on the sample data, and then train the classifier based on the features of the sample data until the optimization goal is reached. This processing method is complicated in process and low in efficiency.
  • the embodiments of the present disclosure propose an ECG signal processing method, an ECG signal processing device, and electronic equipment.
  • an ECG signal processing method including: acquiring an ECG signal, the ECG signal including a plurality of designated waves. Determine the interval sequence for the above-mentioned multiple specified waves. Based on symbol dynamics, the interval sequence is converted into a symbol value sequence, and the Shannon entropy of the symbol value sequence is determined. Then, based on the Shannon entropy of the symbol value sequence, the obtained ECG signals are classified.
  • the foregoing classification of the ECG signal based on the Shannon entropy value of the symbol value sequence includes: when the Shannon entropy value is greater than a first predetermined threshold, determining that the ECG signal belongs to the first predetermined category.
  • the foregoing conversion of the interval sequence into a symbol value sequence based on symbol dynamics includes: calculating the instantaneous heart rate for each time interval in the interval sequence to obtain the instantaneous heart rate sequence. Then, the multiple instantaneous heart rates in the instantaneous heart rate sequence are respectively coded first to obtain multiple coded symbols, and the multiple coded symbols form a symbol sequence. Then, the plurality of coded symbols in the symbol sequence are respectively subjected to second encoding to obtain a plurality of symbol values, and the symbol value sequence is formed by the plurality of symbol values.
  • the above-mentioned first encoding of multiple instantaneous heart rates in the instantaneous heart rate sequence to obtain multiple encoding symbols includes: for any instantaneous heart rate in the instantaneous heart rate sequence, first based on the value of any instantaneous heart rate, Any instantaneous heart rate is classified.
  • the predetermined symbol is used as the encoding symbol of the any instantaneous heart rate.
  • the ratio of the any instantaneous heart rate to the first value is calculated, and the obtained ratio is used as the coding symbol of the any instantaneous heart rate.
  • performing the second encoding on the multiple encoding symbols in the symbol sequence to obtain multiple symbol values includes: for any encoding symbol in the symbol sequence, calculating the previous encoding symbol and the any encoding symbol of the any encoding symbol in the symbol sequence. A coded symbol and the weighted sum of the subsequent coded symbols of any coded symbol, and the obtained weighted sum is used as the symbol value for the any coded symbol.
  • the above determining the Shannon entropy of the symbol value sequence includes: preset multiple symbol value intervals. For any symbol value interval in the above multiple symbol value intervals, calculate the ratio between the number of symbol values in the symbol value sequence and the number of symbol values in the symbol value sequence, and calculate The obtained ratio is used as the distribution probability of the symbol value sequence with respect to any symbol value interval. Then, based on the distribution probability of the symbol value sequence with respect to each of the plurality of symbol value intervals, the Shannon entropy of the symbol value sequence is determined.
  • determining the Shannon entropy of the symbol value sequence based on the distribution probability of the symbol value sequence with respect to each of the plurality of symbol value intervals includes: calculating the symbol value sequence with respect to the plurality of symbols based on the distribution probability of the symbol value sequence with respect to the plurality of symbol value intervals. The distribution expectation of the value interval. Based on the number of symbol values in the symbol value sequence and the number of multiple symbol value intervals, the gain coefficient is determined. Then, based on the aforementioned distribution expectation and gain coefficient, the Shannon entropy of the symbol value sequence is determined.
  • the above method further includes: setting a sliding window, the length of the sliding window is smaller than the length of the symbol value sequence. Slide the sliding window from front to back in the symbol value sequence to extract multiple symbol value subsequences from the symbol value sequence.
  • each sliding step of the sliding window is 1 symbol value.
  • the Shannon entropy of any symbol value subsequence is determined, and when the Shannon entropy of any symbol value subsequence is greater than the first predetermined threshold, the The score of each symbol value in any symbol value subsequence is increased by one.
  • the total score of each symbol value in the symbol value sequence is determined. Finally, it is determined that the symbol value whose total score is greater than the second predetermined threshold belongs to the first predetermined category.
  • the above determination of the interval sequence for the multiple specified waves in the ECG signal includes: identifying the respective peaks of the multiple specified waves in the ECG signal, and then based on the difference between every two adjacent peaks in the respective peaks of the multiple specified waves. The time interval between to determine the interval sequence for multiple specified waves.
  • the above-mentioned plurality of designated waves includes M+1 designated waves.
  • the foregoing determination of the interval sequence for the multiple specified waves based on the time interval between every two adjacent peaks of the respective peaks of the multiple specified waves includes: every two adjacent peaks of the M+1 specified waves are adjacent to each other.
  • the time interval between wave crests constitutes an initial interval sequence containing M time intervals.
  • the initial interval sequence is preprocessed to obtain an interval sequence containing N time intervals.
  • M and N are both positive integers, and M is greater than or equal to N.
  • the foregoing preprocessing of the initial interval sequence to obtain the interval sequence including N time intervals includes: determining an average value of the M time intervals. For any time interval in the M time intervals, it is determined whether the any time interval belongs to the second predetermined category based on the difference between the any time interval and the above-mentioned average value. If it is determined that the any time interval belongs to the second predetermined category, then the any time interval and the subsequent time intervals of the any time interval are filtered from the initial interval sequence.
  • the foregoing preprocessing of the initial interval sequence to obtain the interval sequence including N time intervals includes: determining the average value of the M time intervals. For any time interval in the M time intervals, sum the any time interval and the subsequent time interval of the any time interval to obtain the combined interval. Then, based on the difference between the merging interval and the above average value, it is determined whether the any time interval belongs to the second predetermined category. If the any time interval belongs to the second predetermined category, then the any time interval and the subsequent time intervals of the any time interval are combined into one time interval.
  • the foregoing preprocessing of the initial interval sequence to obtain the interval sequence including N time intervals includes: determining an average value of the M time intervals. For any time interval in the M time intervals, it is determined whether the any time interval belongs to the second predetermined category based on the difference between the any time interval and the above-mentioned average value. If it is determined that the any time interval belongs to the second predetermined category, then the any time interval and the subsequent time intervals of the any time interval are filtered from the initial interval sequence. On this basis, for any time interval remaining after the above-mentioned screening in the initial interval sequence, the any time interval and the subsequent time interval of the any time interval are summed to obtain the combined interval.
  • the any time interval belongs to the second predetermined category. If the any time interval belongs to the second predetermined category, then the any time interval and the subsequent time intervals of the any time interval are combined into one time interval.
  • the foregoing preprocessing of the initial interval sequence to obtain the interval sequence containing N time intervals includes: for any time interval of the M time intervals, if the any time interval is equal to the any time interval The ratio of the previous time interval of the interval is greater than the second value, and the ratio of the subsequent time interval of any time interval to the any time interval is greater than the third value, then the any time interval and the any time interval The following time intervals are screened out.
  • the ratio of the any time interval to the previous time interval of the any time interval is greater than the fourth value, and the subsequent time interval of the any time interval and the any time If the ratio of the interval is less than the fifth value, the any time interval and the subsequent time interval of the any time interval are filtered out.
  • the second value is less than the third value, and the fourth value is greater than the fifth value.
  • the above method further includes: removing the baseline wandering noise in the ECG signal after acquiring the ECG signal.
  • the acquired ECG signal is a single-lead ECG signal.
  • an ECG signal processing device including: an acquisition module, a determination module, a conversion module, an entropy calculation module, and a classification module.
  • the acquisition module is used to acquire the ECG signal, and the ECG signal includes a plurality of designated waves.
  • the determination module is used to determine the interval sequence for a plurality of specified waves.
  • the conversion module is used to convert the interval sequence into a symbol value sequence based on symbol dynamics.
  • the entropy calculation module is used to determine the Shannon entropy of the symbol value sequence.
  • the classification module is used to classify the ECG signal based on the value of Shannon entropy.
  • an electronic device Including: memory and at least one processor.
  • the memory is configured to store instructions.
  • At least one processor executes instructions stored in the memory to implement the method described above.
  • the interval sequence that characterizes the time interval between specified waves in the ECG signal is converted into a coarse-grained, higher-resolution sequence of symbol values, and Shannon entropy is calculated. . Then classify the ECG signal according to the value of Shannon entropy to obtain the classification result for the ECG signal.
  • the calculation speed can be improved, and on the other hand, by selecting an appropriate encoding method, the essential characteristics of the ECG signal can be discarded while the influence of irrelevant noise can be discarded, so that the symbol
  • the Shannon entropy of the value sequence can more accurately measure the uncertainty of the ECG characteristics, so as to obtain more accurate classification results.
  • the algorithm complexity of this process is low, no need to train a classifier, and it is simple and easy to use.
  • Fig. 1 schematically shows a flowchart of an ECG signal processing method according to an embodiment of the present disclosure
  • FIG. 2A schematically shows a flowchart of an ECG signal processing method according to another embodiment of the present disclosure
  • Fig. 2B schematically shows a flowchart of an ECG signal processing method according to another embodiment of the present disclosure
  • FIG. 3 schematically shows a flowchart of an ECG signal processing method according to another embodiment of the present disclosure
  • FIG. 4 schematically shows an example diagram of a process of converting an interval sequence into a symbol value sequence according to an embodiment of the present disclosure
  • Fig. 5 schematically shows an example diagram of a voting decision process according to an embodiment of the present disclosure
  • Fig. 6A schematically shows an exemplary diagram of an ECG signal processing process according to an embodiment of the present disclosure
  • Fig. 6B schematically shows an exemplary diagram of an ECG signal processing process according to another embodiment of the present disclosure
  • Fig. 7 schematically shows a block diagram of an electrocardiographic signal processing device according to an embodiment of the present disclosure.
  • Fig. 8 schematically shows a block diagram of an electronic device according to an embodiment of the present disclosure.
  • At least one of the “systems” shall include, but is not limited to, systems having A alone, B alone, C alone, A and B, A and C, B and C, and/or systems having A, B, C, etc. ).
  • At least one of the “systems” shall include, but is not limited to, systems having A alone, B alone, C alone, A and B, A and C, B and C, and/or systems having A, B, C, etc. ).
  • the ECG signal is a comprehensive reflection of the electrical activity of the heart on the body surface. For accurate classification of ECG signals, it can provide effective assistance for clinical monitoring and telemedicine scenarios.
  • One way of processing the ECG signal is to classify the ECG signal by training a classifier. This method requires the preparation of a large amount of sample data, the feature construction of the sample data, and then the training of the classifier based on the features of the sample data until the optimization goal is reached. This processing method is complicated in process and low in efficiency.
  • an ECG signal processing method is provided, and the method will be described below. It should be noted that the sequence number of each step in the following method is only used to indicate the step for description, and should not be regarded as indicating the execution order of the various steps. Unless explicitly indicated, the method does not need to be executed exactly in the order shown.
  • Fig. 1 schematically shows a flowchart of an ECG signal processing method according to an embodiment of the present disclosure.
  • the method may include the following steps S110 to S150.
  • step S110 an ECG signal is acquired.
  • the acquired ECG signal includes multiple designated waves.
  • the ECG signal can include a variety of ECG waveforms, such as P wave, Q wave, R wave, S wave, T wave, and so on.
  • the "designated wave” as used herein refers to at least one of the above-mentioned various electrocardiographic waveforms.
  • step S120 an interval sequence for a plurality of specified waves in the electrocardiogram signal is determined.
  • the interval sequence for the multiple specified waves includes multiple time intervals to characterize the time interval between the multiple specified waves in the ECG signal.
  • step S130 based on symbolic dynamics, the interval sequence is converted into a symbol value sequence.
  • the process of converting an interval sequence into a symbol value sequence can be a process of symbolizing and encoding the interval sequence in the amplitude domain based on symbol dynamics, thereby simplifying the interval sequence to a finite number of symbol values.
  • Symbol value sequence a process of symbolizing and encoding the interval sequence in the amplitude domain based on symbol dynamics, thereby simplifying the interval sequence to a finite number of symbol values.
  • step S140 the Shannon entropy of the symbol value sequence is determined.
  • Shannon entropy is a measure of uncertainty, and is usually used to solve the problem of quantification of information.
  • step S150 the ECG signal is classified based on the value of Shannon's entropy.
  • the ECG signal processing method converts the interval sequence that characterizes the time interval between specified waves in the ECG signal into a coarse-grained, higher-separated symbol based on symbol dynamics. Value sequence and calculate Shannon entropy on it. Then classify the ECG signal according to the value of Shannon entropy to obtain the classification result for the ECG signal.
  • the calculation speed can be improved, and on the other hand, by selecting an appropriate coding method, the essential characteristics of the ECG signal are preserved while the influence of irrelevant noise is discarded, so that the symbol value sequence
  • the Shannon entropy can more accurately measure the uncertainty of ECG characteristics, so as to obtain more accurate classification results.
  • the algorithm complexity of this process is low, no need to train a classifier, and it is simple and easy to use.
  • FIG. 2A schematically shows a flowchart of an ECG signal processing method according to another embodiment of the present disclosure, which is used to exemplarily describe the implementation process of step S120 shown in FIG. 1.
  • the process of determining the interval sequence for a plurality of specified waves in the ECG signal in step S120 may include the following sub-steps S121 to S122.
  • sub-step S121 the respective peaks of a plurality of designated waves in the electrocardiogram signal are identified.
  • various identification algorithms may be used to identify the respective crests of multiple designated waves based on the characteristics of the designated waves. Since the electrocardiogram signal can be characterized as the distribution of the amplitude of the electrocardiogram signal with respect to time, by identifying the respective crests of multiple specified waves, it is possible to determine the time points for the multiple crests.
  • sub-step S122 based on the time interval between every two adjacent peaks of the respective peaks of the plurality of designated waves, an interval sequence for the plurality of designated waves is determined.
  • Fig. 2B schematically shows a flowchart of an ECG signal processing method according to another embodiment of the present disclosure, which is used to exemplarily describe the implementation process of the sub-step S122 shown in Fig. 2A.
  • the sub-step S122 determines the time interval between every two adjacent peaks of the respective peaks of the plurality of designated waves.
  • the process of specifying the interval sequence of the wave may include the following sub-steps S1221 to S1222.
  • the time interval between every two adjacent peaks of the respective peaks of the M+1 designated waves forms an initial interval sequence including M time intervals.
  • sub-step S1222 the initial interval sequence is preprocessed to obtain an interval sequence containing N time intervals.
  • M and N are both positive integers, and M is greater than or equal to N.
  • the QRS complex composed of Q wave, R wave and S wave is the most prominent ECG waveform in the ECG signal, reflecting the electrical behavior of the heart when the ventricles contract.
  • the designated wave is set to R wave.
  • the time interval between every two adjacent R wave crests is regarded as an RR interval, and the initial interval sequence is composed of M RR intervals. It can be expressed as: ⁇ RR 1 , RR 2 ,..., RR M ⁇ .
  • the initial RR interval sequence is preprocessed to remove the abnormal time intervals in the initial RR interval sequence to obtain the interval sequence, which can be expressed as: ⁇ RR 1 , RR 2 ,..., RR N ⁇ .
  • the PT (Pan-Tompkins) algorithm is used to locate the R wave crest in the ECG signal.
  • the foregoing process of preprocessing the initial interval sequence containing M time intervals to obtain the interval sequence containing N time intervals may be performed in the following manner.
  • the sequence average determines the average value of M time intervals in the initial interval sequence, which can be called the sequence average.
  • the any time interval belongs to the second predetermined category based on the difference between the any time interval and the sequence mean, and if so, from the initial interval sequence In this, any time interval and the subsequent time interval of the any time interval are filtered out.
  • any time interval in the M time intervals sum the any time interval and the subsequent time interval of the any time interval to obtain the merge interval, and based on the merge interval and the sequence mean The difference between determines whether the any time interval belongs to the second predetermined category, and if so, the any time interval and the subsequent time interval of the any time interval are combined into one time interval.
  • the second predetermined category may, for example, characterize the time interval for ectopic heartbeats.
  • the average mean (RR) of M time intervals is determined according to formula (1).
  • i is an integer greater than 1 and less than or equal to M.
  • a new interval sequence is recombined, which is represented by ⁇ RR 1 , RR 2 ,..., RR N ⁇ , the unit is seconds (s), including N time intervals, which can also be called Including N heartbeat lengths, N is less than or equal to M.
  • the foregoing process of preprocessing the initial interval sequence containing M time intervals to obtain the interval sequence containing N time intervals may be performed in the following manner.
  • any time interval in M time intervals if the ratio of the any time interval to the previous time interval of the any time interval is greater than the second value, and the subsequent time interval of the any time interval If the ratio to the any time interval is greater than the third value, then the any time interval and the subsequent time intervals of the any time interval are filtered from the initial interval sequence.
  • the ratio of the any time interval to the previous time interval of the any time interval is greater than the fourth value, and the subsequent time of the any time interval The ratio of the interval to the any time interval is less than the fifth value, then the any time interval and the subsequent time intervals of the any time interval are filtered from the initial interval sequence.
  • the second value is less than the third value
  • the fourth value is greater than the fifth value.
  • RR i corresponds to a premature beat and is ectopic
  • i is an integer greater than 1 and less than M.
  • i is an integer greater than 1 and less than M.
  • d is the fourth value
  • e is the fifth value
  • a new interval sequence is reconstituted, which is represented by ⁇ RR 1 , RR 2 ,..., RR N ⁇ , the unit is seconds, including N time intervals, which can also be called N hearts Stroke length, N ⁇ M.
  • the obtained interval sequence eliminates the interference of the ectopic heartbeat on the ECG signal classification process, and can effectively improve the accuracy of subsequent classification.
  • RR i and RR i+1 are deleted together in the above two embodiments is that after determining that RR i corresponds to an ectopic heartbeat, RR i+1 is usually also affected by RR i . However, it is different from the normal heartbeat length. For example, RR i+1 is a compensatory pause, so both can be deleted together.
  • FIG. 3 schematically shows a flowchart of an ECG signal processing method according to another embodiment of the present disclosure, which is used to exemplarily describe the implementation process of step S130 shown in FIG. 1.
  • step S130 is based on symbol dynamics, and the process of converting an interval sequence into a symbol value sequence may include the following sub-steps S131 to S133.
  • the instantaneous heart rate for each time interval in the interval sequence is calculated to obtain the instantaneous heart rate sequence.
  • sub-step S132 the first encoding is performed on the multiple instantaneous heart rates in the instantaneous heart rate sequence to obtain multiple encoding symbols, and the multiple encoding symbols form a symbol sequence.
  • sub-step S133 a second encoding is performed on the plurality of coded symbols in the symbol sequence to obtain a plurality of symbol values, and the symbol value sequence is formed by the plurality of symbol values.
  • HR i 60/RR i
  • bpm beats per minute
  • the number of strokes, i is an integer greater than or equal to 1 and less than or equal to N.
  • the foregoing process of separately encoding multiple instantaneous heart rates in the instantaneous heart rate sequence to obtain multiple encoding symbols may be performed in the following manner.
  • any instantaneous heart rate in the instantaneous heart rate sequence classify any instantaneous heart rate based on the value of the instantaneous heart rate.
  • the predetermined symbol is used as the encoding symbol of the any instantaneous heart rate.
  • the ratio of the any instantaneous heart rate to the first value is calculated, and the obtained ratio is used as the coding symbol of the any instantaneous heart rate.
  • the foregoing process of classifying any instantaneous heart rate based on the value of any instantaneous heart rate may be: when the value of the instantaneous heart rate is greater than or equal to a predetermined value, determining that the instantaneous heart rate belongs to the first classification, and when the instantaneous heart rate When the value of is less than the predetermined value, it is determined that the instantaneous heart rate belongs to the second category.
  • f is the predetermined value for classifying each instantaneous heart rate
  • g is the first value
  • the ratio between the predetermined value and the first value is the predetermined symbol in this example.
  • the above parameters can be set according to actual needs.
  • [] means rounding operation, 0 ⁇ SY i ⁇ 63.
  • the foregoing process of performing second encoding on multiple code symbols in the symbol sequence to obtain multiple symbol values can be performed in the following manner: For any code symbol in the symbol sequence, calculate the any code symbol The weighted sum of the preceding coded symbol, the any coded symbol, and the subsequent coded symbol of the any coded symbol, and the obtained weighted sum is used as the symbol value for the any coded symbol.
  • the encoding symbols SY i in the symbol sequence ⁇ SY 1 , SY 2 ,..., SY N ⁇ can be second encoded according to formula (7) to obtain the symbol value SYV i for which the encoding symbol SY i is directed.
  • This second encoding process can also be referred to as a symbol value templating process.
  • i is an integer greater than 0 and less than N.
  • h 1 , h 2 and h 3 respectively represent the first weight, the second weight and the third weight, and h 1 , h 2 and h 3 can be set according to actual needs.
  • the symbol sequence ⁇ SY 1 , SY 2 ,..., SY N ⁇ is converted into a symbol value sequence ⁇ SYV 1 , SYV 2 ,..., SYV N ⁇ .
  • FIG. 4 schematically shows an example diagram of a process of converting an interval sequence into a symbol value sequence according to an embodiment of the present disclosure.
  • 4-1 is an example schematic diagram of the interval sequence, showing the amplitude change from the first time interval to the 61st time interval.
  • 4-2 is an example schematic diagram of an instantaneous heart rate sequence, showing the amplitude change from the first instantaneous heart rate to the 61st instantaneous heart rate.
  • 4-3 is an example schematic diagram of the symbol sequence, showing the amplitude change from the first code symbol to the 61st code symbol.
  • 4-4 is an example schematic diagram of the symbol value sequence, showing the amplitude change from the second symbol value to the 60th symbol value.
  • the above-mentioned process of determining the Shannon entropy of the symbol value sequence may be performed in the following manner.
  • Preset multiple symbol value intervals Preset multiple symbol value intervals. Then, for any symbol value interval in the plurality of symbol value intervals, calculate the ratio between the number of symbol values in the symbol value sequence and the number of symbol values in the symbol value sequence, and The ratio is used as the distribution probability of the symbol value sequence with respect to any symbol value interval. Next, the Shannon entropy of the symbol value sequence is determined based on the distribution probability of the symbol value sequence with respect to each of the plurality of preset symbol value intervals.
  • the minimum value of the symbol value is 0, and the maximum value is 262143.
  • the value space of [0, 262143] is called the symbol value space (SYV space), and the SYV space can be divided into 128 symbol value intervals at 2048 intervals, as the preset 128 symbol value intervals.
  • the first symbol value interval is [0, 2047]
  • the second symbol value interval is [2048, 4095], and so on, so I won’t repeat it.
  • the SYV space may be divided into 64 symbol value intervals at an interval of 4096, as the preset 64 symbol value intervals. With different symbol value space ranges and different division intervals, different numbers of symbol value intervals can be preset, which is not limited here.
  • the number of symbol values falling in the j-th symbol value interval in the symbol value sequence ⁇ SYV 1 , SYV 2 ,..., SYV N ⁇ is n j
  • j is an integer greater than or equal to 1 and less than or equal to Q, and n j ⁇ N.
  • the symbol value sequence ⁇ SYV 1 , SYV 2 ,..., SYV N ⁇ can be determined Shannon Entropy.
  • the definition of Shannon entropy is revised.
  • the above process of determining the Shannon entropy of the symbol value sequence based on the distribution probability of the symbol value sequence with respect to each of the plurality of symbol value intervals may be performed in the following manner.
  • the expected distribution of the symbol value sequence with respect to the plurality of symbol value intervals is calculated.
  • the gain coefficient is determined based on the number of symbol values in the symbol value sequence and the number of multiple symbol value intervals. Then, based on the aforementioned distribution expectation and gain coefficient, the Shannon entropy of the symbol value sequence is determined.
  • the number of symbol value intervals is Q
  • the symbol value sequence ⁇ SYV 1 , SYV 2 ,..., SYV N ⁇ has the distribution probability of the jth symbol value interval as p j , which can be calculated according to the formula (8 ) Calculate the Shannon entropy SE of the symbol value sequence ⁇ SYV 1 , SYV 2 ,..., SYV N ⁇ .
  • the type of the ECG signal can be determined by comparing the magnitude relationship between the value of the Shannon entropy and the first predetermined threshold ThSe.
  • the foregoing process of classifying the electrocardiogram signal based on the value of Shannon entropy may include: when the value of Shannon entropy is greater than a first predetermined threshold, determining that the electrocardiogram signal belongs to the first predetermined category.
  • the ECG signal when the Shannon entropy of the symbol value sequence of the ECG signal is small, the fluctuation of the ECG signal is more stable. The larger the Shannon entropy of the symbol value sequence of the ECG signal is, it means the The fluctuation of the ECG signal is more unstable. Under normal circumstances, the unstable ECG fluctuation may be caused by some diseases.
  • the first predetermined threshold (also referred to as “Shannon entropy threshold”) is used as a boundary to determine whether the ECG signal belongs to the first predetermined category.
  • the first predetermined category may be related to a certain type of disease, for example, the first predetermined category is related to atrial fibrillation (AF).
  • AF atrial fibrillation
  • the setting of the length N of the symbol value sequence and the Shannon entropy threshold ThSe will directly affect the accuracy of the classification result. If the Shannon entropy threshold is too low, some normal ECG signals will be classified into the first predetermined category. If the Shannon entropy is too high, it will lead to insensitivity to some abnormal ECG signals. Improper selection of the length of the symbol value sequence will not only cause inaccurate classification results, but the too long symbol value sequence will also affect the real-time and complexity of the algorithm.
  • the Shannon entropy threshold ThSe can be set in the range of 0.2 to 0.5, and the length N of the symbol value sequence can be set in the range of 15 to 60. At this time, both sensitivity and accuracy can be considered.
  • the ECG signal data can be used for about one minute to realize the classification of the ECG signal, so as to know the correlation between the ECG signal and a certain type of disease (such as atrial fibrillation), which can be used for clinical monitoring and telemedicine scenarios. provide assistance.
  • a certain type of disease such as atrial fibrillation
  • the following voting decision mechanism can be used to more accurately locate the position of the heartbeat in the ECG signal belonging to the first predetermined category.
  • the ECG signal processing method according to the embodiment of the present disclosure may further include the following manners.
  • the length of the sliding window needs to be less than the length of the symbol value sequence.
  • each sliding step of the sliding window is 1 symbol value.
  • the Shannon entropy of any symbol value subsequence is determined, when the Shannon entropy of any symbol value subsequence is greater than the first predetermined threshold ,
  • the score of each symbol value in the any symbol value subsequence is increased by a predetermined score value, and the predetermined score value can be set as required, and there is no limitation here.
  • the total score of each symbol value in the symbol value sequence is determined, and it is determined that the symbol value with the total score greater than the second predetermined threshold belongs to the first predetermined category.
  • Fig. 5 schematically shows an exemplary diagram of a voting decision process according to an embodiment of the present disclosure.
  • the length of the sliding window in this example is 20, which corresponds to the length of 20 symbol values.
  • the length N of the symbol value sequence ⁇ SYV 1 , SYV 2 ,..., SYV N ⁇ is greater than 20. Slide the sliding window from front to back in the sequence of symbol values to extract multiple subsequences of symbol values.
  • the first symbol value subsequence is ⁇ SYV 1 ,SYV 2 ,...,SYV 20 ⁇
  • the second symbol value subsequence is ⁇ SYV 2 ,SYV 2 ,...,SYV 21 ⁇
  • the third symbol value subsequence It is ⁇ SYV 3 , SYV 2 ,..., SYV 22 ⁇ , and so on, so I won’t repeat it.
  • the Shannon entropy of each extracted symbol value sub-sequence is calculated separately, and the calculation method is the same as the above-mentioned method of calculating the Shannon entropy of the symbol value sequence.
  • the predetermined score is 1.
  • the Shannon entropy of the first symbol value sub-sequence ⁇ SYV 1 , SYV 2 ,..., SYV 20 ⁇ is calculated as SE 1 , and if SE 1 is greater than the first predetermined threshold ThSe, then the symbol value sub-sequence ⁇ SYV 1 ,
  • the score of each symbol value in SYV 2 ,..., SYV 20 ⁇ is increased by 1.
  • the score of the symbol value SYV i is represented by AF i.
  • the score of each symbol value in the symbol value sub-sequence ⁇ SYV 1 , SYV 2 ,..., SYV 20 ⁇ remains unchanged.
  • the total score of each symbol value in the symbol value sequence ⁇ SYV 1 , SYV 2 ,..., SYV N ⁇ can be obtained.
  • the second predetermined threshold is set to 12. In other examples, the second predetermined threshold can be set as needed, and in principle, it is sufficient to ensure that the second predetermined threshold is less than or equal to the length of the sliding window.
  • the ECG signal processing method can not only classify the ECG signal, but also more accurately locate the heartbeat position in the ECG signal belonging to the first predetermined category through the voting decision mechanism. The location is of great significance to the fields of medical diagnosis and monitoring.
  • the ECG signal will be interfered by noise, which will bring difficulties to the ECG signal processing.
  • the baseline drift caused by low-frequency interference such as the breathing and electrode movement of the measured object causes the actual measurement center electrical signal to deviate from the normal baseline position, and the phenomenon of slow fluctuations up and down appears.
  • the ECG signal itself contains rich low-frequency components, and the baseline drift will conceal useful information and affect the accuracy of the analysis, recognition, classification, and positioning of the ECG signal.
  • the baseline wandering noise in the ECG signal can be removed.
  • Various methods can be used to remove the baseline drift noise, such as adaptive filtering, Kalman filtering, wavelet transformation, etc., which are not limited here.
  • median filtering is used to quickly and effectively suppress baseline drift.
  • the single-lead ECG signal can be processed to obtain the classification result about whether the heartbeat is abnormal, and it is not necessary to process based on the complex signal in the multi-lead ECG signal, which is very simple reliable.
  • FIG. 6A schematically shows an exemplary diagram of an ECG signal processing process according to an embodiment of the present disclosure.
  • Fig. 6B schematically shows an exemplary diagram of an ECG signal processing process according to another embodiment of the present disclosure. It should be noted that the ECG signal processing procedures shown in FIGS. 6A and 6B are only used as examples to facilitate understanding of the present disclosure, and the present disclosure does not limit this.
  • the ECG signal processing process may include steps S601 to S610.
  • step S601 the original ECG signal is acquired.
  • step S602 the original ECG signal is preprocessed to remove baseline wandering noise, so as to obtain the preprocessed ECG signal.
  • step S603 the R wave in the ECG signal is identified.
  • step S604 the initial RR interval sequence is extracted.
  • step S605 the ectopic heartbeat in the initial RR interval sequence is processed to obtain the RR interval sequence.
  • step S606 the instantaneous heart rate sequence is calculated.
  • step S607 symbol conversion is performed to obtain a symbol sequence.
  • step S608 the symbol value templating is performed to obtain a symbol value sequence.
  • step S609 the Shannon entropy of the sequence of symbol values is calculated.
  • step S610 based on the first predetermined threshold, the Shannon entropy is thresholded, so that the category to which the ECG signal belongs is determined according to the determination result, and the ECG signal is labeled.
  • the types of ECG signals can be marked as "atrial fibrillation related”. Then step S601 is repeated until all the ECG signals are marked.
  • the electrocardiogram processing process further includes step S611 to make a voting decision to locate abnormal heartbeats within the electrocardiogram signals belonging to the first predetermined category, so as to make more specific annotations.
  • one or more heartbeat positions within the ECG signal that are determined to belong to the first predetermined category may be marked as "atrial fibrillation related". The voting decision has been explained in detail above, so I won’t repeat it here. Then step S601 is repeated until all the ECG signals are marked.
  • the embodiment of the present disclosure combines symbol dynamics and custom Shannon entropy in the process of processing the ECG signal, which is more sensitive to the ECG characteristics related to certain types of diseases, and the algorithm complexity is not high, and there is no need
  • the training classifier effectively improves the accuracy and real-time performance of the ECG signal classification and the positioning of the abnormal heartbeat within the ECG signal during the ECG signal monitoring process.
  • Fig. 7 schematically shows a block diagram of an electrocardiographic signal processing device according to an embodiment of the present disclosure.
  • the ECG signal processing device 700 may include: an acquisition module 710, a determination module 720, a conversion module 730, an entropy calculation module 740, and a classification module 750.
  • the obtaining module 710 is used to obtain an electrocardiographic signal, and the electrocardiographic signal includes a plurality of designated waves.
  • the determining module 720 is used to determine an interval sequence for a plurality of specified waves in the ECG signal.
  • the conversion module 730 is used to convert the interval sequence into a symbol value sequence based on symbol dynamics.
  • the entropy calculation module 740 is used to determine the Shannon entropy of the symbol value sequence.
  • the classification module 750 is used to classify the ECG signal based on the value of Shannon's entropy.
  • any number of the modules, sub-modules, units, and sub-units, or at least part of the functions of any number of them may be implemented in one module. Any one or more of the modules, sub-modules, units, and sub-units according to the embodiments of the present disclosure may be split into multiple modules for implementation.
  • any one or more of the modules, sub-modules, units, and sub-units according to the embodiments of the present disclosure may be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), System-on-chip, system-on-substrate, system-on-package, application-specific integrated circuit (ASIC), or hardware or firmware in any other reasonable way that integrates or encapsulates the circuit, or can be implemented by software, hardware, and firmware. Any one of these implementations or an appropriate combination of any of them can be implemented.
  • FPGA field programmable gate array
  • PLA programmable logic array
  • ASIC application-specific integrated circuit
  • any one of these implementations or an appropriate combination of any of them can be implemented.
  • one or more of the modules, sub-modules, units, and sub-units according to the embodiments of the present disclosure may be at least partially implemented as a computer program module, and the computer program module may perform corresponding functions when it is executed.
  • Fig. 8 schematically shows a block diagram of an electronic device suitable for implementing the above-described method according to an embodiment of the present disclosure.
  • the electronic device shown in FIG. 8 is only an example, and should not bring any limitation to the function and scope of use of the embodiments of the present disclosure.
  • the electronic device 800 includes one or more processors 810 and a computer-readable storage medium 820.
  • the electronic device 800 can execute a method according to an embodiment of the present disclosure.
  • the processor 810 may include, for example, a general-purpose microprocessor, an instruction set processor, and/or a related chipset, and/or a special-purpose microprocessor (for example, an application specific integrated circuit (ASIC)), and so on.
  • the processor 810 may also include on-board memory for caching purposes.
  • the processor 810 may be a single processing unit or multiple processing units for executing different actions of a method flow according to an embodiment of the present disclosure.
  • the computer-readable storage medium 820 may be a non-volatile computer-readable storage medium. Specific examples include but are not limited to: magnetic storage devices, such as magnetic tapes or hard disks (HDD); optical storage devices, such as optical disks (CD-ROM) ; Memory, such as random access memory (RAM) or flash memory; etc.
  • magnetic storage devices such as magnetic tapes or hard disks (HDD)
  • optical storage devices such as optical disks (CD-ROM)
  • Memory such as random access memory (RAM) or flash memory; etc.
  • the computer-readable storage medium 820 may include a computer program 821, which may include code/computer-executable instructions, which, when executed by the processor 810, causes the processor 810 to perform a method according to an embodiment of the present disclosure or any modification thereof.
  • the computer program 821 may be configured to have, for example, computer program code including computer program modules.
  • the code in the computer program 821 may include one or more program modules, for example, including 821A, module 821B,... It should be noted that the division and number of modules are not fixed. Those skilled in the art can use appropriate program modules or program module combinations according to the actual situation. When these program module combinations are executed by the processor 810, the processor 810 can Perform the method according to the embodiment of the present disclosure or any modification thereof.
  • the present disclosure also provides a computer-readable storage medium.
  • the computer-readable storage medium may be included in the device/device/system described in the above embodiment; or it may exist alone without being assembled into the device/ In the device/system.
  • the aforementioned computer-readable storage medium carries one or more programs, and when the aforementioned one or more programs are executed, the method according to the embodiments of the present disclosure is implemented.
  • the computer-readable storage medium may be a non-volatile computer-readable storage medium, for example, may include but not limited to: portable computer disk, hard disk, random access memory (RAM), read-only memory (ROM) , Erasable programmable read-only memory (EPROM or flash memory), portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination of the above.
  • a computer-readable storage medium may be any tangible medium that contains or stores a program, and the program may be used by or in combination with an instruction execution system, apparatus, or device.

Landscapes

  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Engineering & Computer Science (AREA)
  • Cardiology (AREA)
  • Physics & Mathematics (AREA)
  • Biophysics (AREA)
  • Surgery (AREA)
  • Veterinary Medicine (AREA)
  • Public Health (AREA)
  • General Health & Medical Sciences (AREA)
  • Animal Behavior & Ethology (AREA)
  • Molecular Biology (AREA)
  • Medical Informatics (AREA)
  • Pathology (AREA)
  • Biomedical Technology (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Artificial Intelligence (AREA)
  • Signal Processing (AREA)
  • Physiology (AREA)
  • Psychiatry (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Computation (AREA)
  • Fuzzy Systems (AREA)
  • Mathematical Physics (AREA)
  • Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)

Abstract

本公开提供了一种心电信号处理方法、装置以及电子设备。该心电信号处理方法包括：获取心电信号，该心电信号包括多个指定波。然后确定针对上述多个指定波的间期序列。基于符号动力学，将间期序列转换为符号值序列，并确定符号值序列的香农熵。接着基于符号值序列的香农熵的数值，对所获取的心电信号进行分类。

Description

心电信号处理方法、心电信号处理装置和电子设备
本申请要求申请日为2019年11月29日、申请号为CN201911211864.7的中国专利申请的优先权，其内容一并在此作为参考。
技术领域
本公开总体上涉及计算机技术领域，具体地，涉及一种心电信号处理方法、心电信号处理装置和电子设备。
背景技术
心电(electrocardiogram，ECG)信号是心脏电活动在体表的综合反映。对于心电信号的准确分类，可为临床监护和远程医疗等场景提供有效协助。一种处理方式在于通过训练分类器对心电信号进行分类。需要准备大量样本数据，并对样本数据进行特征构造，再基于样本数据的特征进行分类器的训练，直至达到优化目标。该处理方式过程复杂，效率较低。
发明内容
本公开实施例提出了一种心电信号处理方法、心电信号处理装置和电子设备。
根据本发明的一个方面，提出了一种心电信号处理方法，包括：获取心电信号，该心电信号包括多个指定波。确定针对上述多个指定波的间期序列。基于符号动力学，将间期序列转换为符号值序列，并确定符号值序列的香农熵。接着基于符号值序列的香农熵的数值，对所获取的心电信号进行分类。
例如，上述基于符号值序列的香农熵的数值，对心电信号进行分类包括：当香农熵的数值大于第一预定阈值时，确定心电信号属于第一预定类别。
例如，上述基于符号动力学，将间期序列转换为符号值序列包括：计算间期序列中每个时间间隔所针对的瞬时心率，以得到瞬时心率序列。然后对瞬时心率序列中的多个瞬时心率分别进行第一编码，以得到多个编码符号，并由该多个编码符号组成符号序列。接着对符号序列中的多个编码符号分别进行第二编码，以得到多个符号值，并由该多个符号值组成符号值序列。
例如，上述对瞬时心率序列中的多个瞬时心率分别进行第一编码，以得到多个编码符号包括：对于瞬时心率序列中的任一瞬时心率，先基于该任一瞬时心率的数值，对该任一瞬时心率进行分类。当该任一瞬时心率属于第一分类时，将预定符号作为该任一瞬时心率的编码符号。当该任一瞬时心率属于第二分类时，计算该任一瞬时心率与第一数值的比值，并将所得到的比值作为该任一瞬时心率的编码符号。
例如，上述对符号序列中的多个编码符号分别进行第二编码，以得到多个符号值包括：对于符号序列中的任一编码符号，计算该任一编码符号的在前编码符号、该任一编码符号、以及该任一编码符号的在后编码符号的加权和，并将所得到的加权和作为针对该任一编码符号的符号值。
例如，上述确定符号值序列的香农熵包括：预设多个符号值区间。对于上述多个符号值区间中的任一符号值区间，计算符号值序列中落入该任一符号值区间的符号值的数量与符号值序列中符号值的数量之间的比例，并将所得到的比例作为符号值序列关于该任一符号值区间的分布概率。然后，基于符号值序列关于多个符号值区间各自的分布概率，确定符号值序列的香农熵。
例如，上述基于符号值序列关于多个符号值区间各自的分布概率，确定符号值序列的香农熵包括：基于符号值序列关于多个符号值区间各自的分布概率，计算符号值序列关于多个符号值区间的分布期望。基于符号值序列中符号值的数量和多个符号值区间的数量，确定增益系数。然后，基于上述分布期望和增益系数，确定符号值序列的香农熵。
例如，上述方法还包括：设置滑动窗口，该滑动窗口的长度小于所述符号值序列的长度。将滑动窗口在符号值序列中从前至后进行滑动，以从符号值序列中提取出多个符号值子序列。其中，滑动窗口的每次滑动步长为1个符号值。对于所提取的多个符号值子序列中的任一符号值子序列，确定该任一符号值子序列的香农熵，当该任一符号值子序列的香农熵大于第一预定阈值时，将该任一符号值子序列中的每个符号值的评分加1。在滑动窗口滑动结束后，确定符号值序列中的每个符号值的总评分。最后确定总评分大于第二预定阈值的符号值属于第一预定类别。
例如，上述确定针对心电信号中多个指定波的间期序列包括：识别心电信号中多个指定波各自的波峰，然后基于该多个指定波各自的波峰中每两个相邻波峰之间的时间间隔，确定针对多个指定波的间期序列。
例如，上述多个指定波包括M+1个指定波。上述基于多个指定波各自的波峰中每 两个相邻波峰之间的时间间隔，确定针对多个指定波的间期序列包括：由M+1个指定波各自的波峰中每两个相邻波峰之间的时间间隔，组成包含M个时间间隔的初始间期序列。然后对初始间期序列进行预处理，以得到包含N个时间间隔的间期序列。其中，M和N均为正整数，M大于等于N。
例如，上述对初始间期序列进行预处理，以得到包含N个时间间隔的所述间期序列包括：确定所述M个时间间隔的平均值。对于M个时间间隔中的任一时间间隔，基于该任一时间间隔与上述平均值之间的差异确定该任一时间间隔是否属于第二预定类别。如果确定该任一时间间隔属于第二预定类别，则从初始间期序列中将该任一时间间隔以及该任一时间间隔的在后时间间隔筛除。
例如，上述对初始间期序列进行预处理，以得到包含N个时间间隔的所述间期序列包括：确定M个时间间隔的平均值。对于M个时间间隔中的任一时间间隔，对该任一时间间隔与该任一时间间隔的在后时间间隔进行求和以得到合并间隔。然后，基于该合并间隔与上述平均值之间的差异确定该任一时间间隔是否属于第二预定类别。如果该任一时间间隔属于第二预定类别，则将该任一时间间隔以及该任一时间间隔的在后时间间隔合并为一个时间间隔。
例如，上述对初始间期序列进行预处理，以得到包含N个时间间隔的所述间期序列包括：确定所述M个时间间隔的平均值。对于M个时间间隔中的任一时间间隔，基于该任一时间间隔与上述平均值之间的差异确定该任一时间间隔是否属于第二预定类别。如果确定该任一时间间隔属于第二预定类别，则从初始间期序列中将该任一时间间隔以及该任一时间间隔的在后时间间隔筛除。在此基础上，对于初始间期序列中经上述筛除后剩余的任一时间间隔，对该任一时间间隔与该任一时间间隔的在后时间间隔进行求和以得到合并间隔。再基于该合并间隔与上述平均值之间的差异确定该任一时间间隔是否属于第二预定类别。如果该任一时间间隔属于第二预定类别，则将该任一时间间隔以及该任一时间间隔的在后时间间隔合并为一个时间间隔。
例如，上述对初始间期序列进行预处理，以得到包含N个时间间隔的所述间期序列包括：对于M个时间间隔中的任一时间间隔，如果该任一时间间隔与该任一时间间隔的在前时间间隔的比值大于第二数值、且该任一时间间隔的在后时间间隔与该任一时间间隔的比值大于第三数值，则将该任一时间间隔以及该任一时间间隔的在后时间间隔筛除。并且/或者，对于任一时间间隔，如果该任一时间间隔与该任一时间间隔的在前时 间间隔的比值大于第四数值、且该任一时间间隔的在后时间间隔与该任一时间间隔的比值小于第五数值，则将该任一时间间隔以及该任一时间间隔的在后时间间隔筛除。其中，所述第二数值小于所述第三数值，所述第四数值大于所述第五数值。
例如，上述方法还包括：在获取心电信号之后，去除心电信号中的基线漂移噪声。
例如，所获取的心电信号为单导联心电信号。
根据本公开实施例的另一方面，提供了一种心电信号处理装置，包括：获取模块、确定模块、转换模块、熵计算模块、以及分类模块。获取模块用于获取心电信号，心电信号包括多个指定波。确定模块用于确定针对多个指定波的间期序列。转换模块用于基于符号动力学，将间期序列转换为符号值序列。熵计算模块用于确定符号值序列的香农熵。分类模块用于基于香农熵的数值，对心电信号进行分类。
根据本公开实施例的另一方面，提供了一种电子设备。包括：存储器和至少一个处理器。存储器配置为存储指令。至少一个处理器执行存储在存储器中的指令，以实现上文所述的方法。
根据本公开实施例的技术方案，基于符号动力学将表征心电信号中指定波之间时间间隔的间期序列转换为粗粒度的、分离度更高的符号值序列，并对其计算香农熵。然后依据香农熵的数值对心电信号进行分类，以得到针对心电信号的分类结果。其中，通过间期序列在幅度域的符号化处理，一方面能够提高计算速度，另一方面通过选取恰当的编码方式，在保存心电信号的本质特征的同时能够丢弃无关噪声的影响，使得符号值序列的香农熵能够较为准确地度量心电特征的不确定性，从而得到更加准确的分类结果。该过程的算法复杂度较低，不必训练分类器，简单易用。
附图说明
为了更清楚地说明本公开实施例或传统的技术方案，下面将对实施例描述中所需要使用的附图作简单地介绍。显而易见地，下面描述中的附图仅仅是本公开的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据这些附图获得其他的附图。下图中：
图1示意性示出了根据本公开实施例的心电信号处理方法的流程图；
图2A示意性示出了根据本公开另一实施例的心电信号处理方法的流程图；
图2B示意性示出了根据本公开另一实施例的心电信号处理方法的流程图；
图3示意性示出了根据本公开另一实施例的心电信号处理方法的流程图；
图4示意性示出了根据本公开实施例的间期序列转换为符号值序列的过程的示例示意图；
图5示意性示出了根据本公开实施例的投票决策的过程的示例示意图；
图6A示意性示出了根据本公开实施例的心电信号处理过程的示例示意图；
图6B示意性示出了根据本公开另一实施例的心电信号处理过程的示例示意图；
图7示意性示出了根据本公开实施例的心电信号处理装置的框图；以及
图8示意性示出了根据本公开实施例的电子设备的框图。
具体实施方式
为使本公开实施例的目的、技术方案和优点更加清楚，下面将结合本公开实施例中的附图，对本公开实施例中的技术方案进行清楚、完整的描述。显然，所描述的实施例是本公开的一部分实施例，而不是全部。基于所描述的本公开实施例，本领域普通技术人员在无需创造性劳动的前提下获得的所有其他实施例都属于本公开保护的范围。应注意，贯穿附图，相同的元素由相同或相近的附图标记来表示。在以下描述中，一些具体实施例仅用于描述目的，而不应该理解为对本公开有任何限制，而只是本公开实施例的示例。此外，在以下说明中，省略了对公知结构和技术的描述，以避免不必要地混淆本公开的概念。
在此使用的术语仅仅是为了描述具体实施例，而并非意在限制本公开。在此使用的术语“包括”、“包含”等表明了所述特征、步骤、操作和/或部件的存在，但是并不排除存在或添加一个或多个其他特征、步骤、操作或部件。
在此使用的所有术语(包括技术和科学术语)具有本领域技术人员通常所理解的含义，除非另外定义。应注意，这里使用的术语应解释为具有与本说明书的上下文相一致的含义，而不应以理想化或过于刻板的方式来解释。
在使用类似于“A、B和C等中至少一个”这样的表述的情况下，一般来说应该按照本领域技术人员通常理解该表述的含义来予以解释(例如，“具有A、B和C中至少一个的系统”应包括但不限于单独具有A、单独具有B、单独具有C、具有A和B、具有A和C、具有B和C、和/或具有A、B、C的系统等)。在使用类似于“A、B或C等中至少一个”这样的表述的情况下，一般来说应该按照本领域技术人员通常理解该表述 的含义来予以解释(例如，“具有A、B或C中至少一个的系统”应包括但不限于单独具有A、单独具有B、单独具有C、具有A和B、具有A和C、具有B和C、和/或具有A、B、C的系统等)。
心电信号是心脏电活动在体表的综合反映。对于心电信号的准确分类，可为临床监护和远程医疗等场景提供有效协助。一种心电信号的处理方式在于通过训练分类器对心电信号进行分类。这种方式需要准备大量样本数据，并对样本数据进行特征构造，再基于样本数据的特征进行分类器的训练，直至达到优化目标。该处理方式过程复杂，效率较低。
根据本公开实施例，提供了一种心电信号处理方法，下面对该方法进行说明。应注意，以下方法中各个步骤的序号仅作为该步骤的表示以便描述，而不应被看作表示该各个步骤的执行顺序。除非明确指出，否则该方法不需要完全按照所示顺序来执行。
图1示意性示出了根据本公开实施例的心电信号处理方法的流程图。
如图1所示，该方法可以包括以下步骤S110～S150。
在步骤S110，获取心电信号。
其中，所获取的心电信号包括多个指定波。通常情况下，心电信号中可以包括多种心电波形，例如P波、Q波、R波、S波、T波等。本文所述的“指定波”是指上述多种心电波形中的至少一种。
然后，在步骤S120，确定针对心电信号中多个指定波的间期(interval)序列。
其中，针对多个指定波的间期序列包括多个时间间隔，以表征心电信号中多个指定波之间的时间间隔。
接着，在步骤S130，基于符号动力学(symbolic dynamics)，将间期序列转换为符号值序列。
例如，将间期序列转换为符号值序列的过程可以是一种基于符号动力学在幅度域上对间期序列进行符号化编码的过程，从而将间期序列简化为由有限个符号值组成的符号值序列。
接着，在步骤S140，确定符号值序列的香农熵(Shannon entropy)。
在信息论中，香农熵是对不确定性的度量，通常用于解决信息的量化问题。一个事件的香农熵的数值越大，表征该事件携带越多的信息；一个事件的香农熵的数值越小， 表征该事件携带越少的信息。
接着，在步骤S150，基于香农熵的数值，对心电信号进行分类。
本领域技术人员可以理解，根据本公开实施例的心电信号处理方法基于符号动力学将表征心电信号中指定波之间时间间隔的间期序列转换为粗粒度的、分离度更高的符号值序列，并对其计算香农熵。然后依据香农熵的数值对心电信号进行分类，以得到针对心电信号的分类结果。其中，通过间期序列在幅度域的符号化处理，一方面能够提高计算速度，另一方面通过选取恰当的编码方式在保存心电信号的本质特征的同时丢弃无关噪声的影响，使得符号值序列的香农熵能够较为准确地度量心电特征的不确定性，从而得到更加准确的分类结果。该过程的算法复杂度较低，不必训练分类器，简单易用。
图2A示意性示出了根据本公开另一实施例的心电信号处理方法的流程图，用于示例性地对图1所示的步骤S120的实施过程进行说明。
如图2A所示，步骤S120确定针对心电信号中多个指定波的间期序列的过程可以包括以下子步骤S121～S122。
在子步骤S121，识别心电信号中多个指定波各自的波峰。
例如，本步骤S121可以基于指定波的特征，采用各种识别算法来识别多个指定波各自的波峰。由于心电信号可以表征为心电信号的幅值关于时间的分布，因此通过识别多个指定波各自的波峰可以确定多个波峰所针对的时间点。
在子步骤S122，基于该多个指定波各自的波峰中每两个相邻波峰之间的时间间隔，确定针对多个指定波的间期序列。
图2B示意性示出了根据本公开另一实施例的心电信号处理方法的流程图，用于示例性地对图2A所示的子步骤S122的实施过程进行说明。
如图2B所示，以包含M+1个指定波的心电信号为例，子步骤S122基于多个指定波各自的波峰中每两个相邻波峰之间的时间间隔，确定针对该多个指定波的间期序列的过程可以包括以下子步骤S1221～S1222。
在子步骤S1221，由M+1个指定波各自的波峰中每两个相邻波峰之间的时间间隔，组成包含M个时间间隔的初始间期序列。
在子步骤S1222，对初始间期序列进行预处理，以得到包含N个时间间隔的间期序列。
其中，M和N均为正整数，M大于等于N。
例如，通常情况下，由Q波、R波和S波构成的QRS波群是心电信号中最显著的心电波形，反映了心室收缩时心脏的电行为。本例中设置指定波为R波。识别心电信号中的M+1个R波的波峰，每两个相邻R波波峰之间的时间间隔作为一个RR间期(RR interval)，由M个RR间期组成初始间期序列，可表示为：{RR 1，RR 2，…，RR M}。对该初始RR间期序列进行预处理，把初始RR间期序列中异常的时间间隔去除，从而得到间期序列，可表示为：{RR 1，RR 2，…，RR N}。识别心电信号中R波波峰的方法包括多种，包括幅度法、斜率法和面积法等等，在此不做限制。本例中采用P-T(Pan-Tompkins)算法在心电信号中定位R波波峰。
下面分别通过以下实施例来说明上述对初始间期序列进行预处理，以得到包含N个时间间隔的间期序列的过程。
示例性地，上述对包含M个时间间隔的初始间期序列进行预处理，以得到包含N个时间间隔的间期序列的过程可以按照如下方式进行。
首先，确定初始间期序列中M个时间间隔的平均值，可称为序列均值。一方面，对于M个时间间隔中的任一时间间隔，基于该任一时间间隔与序列均值之间的差异确定该任一时间间隔是否属于第二预定类别，如果是，则从初始间期序列中将该任一时间间隔以及该任一时间间隔的在后时间间隔筛除。
附加地或替换地，对于M个时间间隔中的任一时间间隔，对该任一时间间隔与该任一时间间隔的在后时间间隔进行求和以得到合并间隔，并基于合并间隔与序列均值之间的差异确定该任一时间间隔是否属于第二预定类别，如果是，则将该任一时间间隔以及该任一时间间隔的在后时间间隔合并为一个时间间隔。第二预定类别例如可以表征针对异位心搏的时间间隔。
例如，对于初始间期序列{RR 1，RR 2，…，RR M}，根据公式(1)确定M个时间间隔的平均值mean(RR)。
Figure PCTCN2020117707-appb-000001
RR i表示第i个时间间隔，i为大于等于1小于等于M的整数。若RR i与mean(RR)之间的差异满足公式(2)，则认为RR i过大，对应于异位心搏，需要将RR i以及RR i+1筛除。在i＝M的情况下，则仅将RR i筛除。
Figure PCTCN2020117707-appb-000002
公式(2)中，a为预置参数，可以根据需要进行调整，原则上需要保证该预置参数大于1，本例中a＝1.8。
此外，若合并间隔(RR i-1+RR i)与mean(RR)之间的差异满足公式(3)，即(RR i-1+RR i)比RR i-1更接近mean(RR)且(RR i-1+RR i)比RR i更接近mean(RR)，则认为RR i-1和RR i均过小，在两个正常心搏之间可能插入了一个误识别的心搏，需要将这两个相邻RR值(即RR i-1和RR i)合并为一个RR间期。
Figure PCTCN2020117707-appb-000003
公式(3)中，i为大于1且小于等于M的整数。
经过上述筛除和合并两步操作，重新组成新的间期序列，用{RR 1，RR 2，…，RR N}表示，单位为秒(s)，包括N个时间间隔，也可称为包括N个心搏长度，N小于等于M。
此外，示例性地，上述对包含M个时间间隔的初始间期序列进行预处理，以得到包含N个时间间隔的间期序列的过程可以按照如下方式进行。
一方面，对于M个时间间隔中的任一时间间隔，如果该任一时间间隔与该任一时间间隔的在前时间间隔的比值大于第二数值、且该任一时间间隔的在后时间间隔与该任一时间间隔的比值大于第三数值，则从初始间期序列中将该任一时间间隔以及该任一时间间隔的在后时间间隔筛除。另一方面，对于M个时间间隔中的任一时间间隔，如果该任一时间间隔与该任一时间间隔的在前时间间隔的比值大于第四数值、且该任一时间间隔的在后时间间隔与该任一时间间隔的比值小于第五数值，则从初始间期序列中将该任一时间间隔以及该任一时间间隔的在后时间间隔筛除。其中，第二数值小于第三数值，第四数值大于第五数值。
例如，对于初始间期序列{RR 1，RR 2，…，RR M}中的时间间隔RR i，若RR i满足公式(4)，则认为RR i对应于早搏(premature beat)，属于异位心搏，需要将RR i以及RR i+1筛除。如果i＝M，则仅将RR i筛除。
Figure PCTCN2020117707-appb-000004
公式(4)中，i为大于1且小于M的整数。b为第二数值，c为第三数值，b小于c，可以根据需要进行设置。本例中b＝0.9，c＝1.6。
此外，若RR i满足公式(5)，则认为RR i对应于逸搏(escape beat)，也属于异位心搏，需要将RR i以及RR i+1筛除。如果i＝M，则仅将RR i筛除。
Figure PCTCN2020117707-appb-000005
公式(5)中，i为大于1且小于M的整数。d为第四数值，e为第五数值，d大于e，可以根据需要进行设置。本例中d＝1.3，e＝0.6。
经过上述筛除异位心搏后，重新组成新的间期序列，用{RR 1，RR 2，…，RR N}表示，单位为秒，包括N个时间间隔，也可称为N个心搏长度，N≤M。
可以理解，经过上述各实施例的预处理过程，所得到的间期序列排除了异位心搏对于心电信号分类过程的干扰，能够有效提升后续分类精确度。需要说明的是，上述两个实施例中将RR i与RR i+1一同删除的原因是：在确定RR i对应于异位心搏后，通常RR i+1也会因受到RR i的影响而有异于正常的心搏长度，例如RR i+1为代偿间歇(compensatory pause)，故可以将二者一同删除。
图3示意性示出了根据本公开另一实施例的心电信号处理方法的流程图，用于示例性地对图1所示的步骤S130的实施过程进行说明。
如图3所示，步骤S130基于符号动力学，将间期序列转换为符号值序列的过程可以包括如下子步骤S131～S133。
在子步骤S131，计算间期序列中每个时间间隔所针对的瞬时心率，以得到瞬时心率序列。
然后，在子步骤S132，对瞬时心率序列中的多个瞬时心率分别进行第一编码，以得到多个编码符号，并由该多个编码符号组成符号序列。
接着，在子步骤S133，对符号序列中的多个编码符号分别进行第二编码，以得到多个符号值，并由多个符号值组成符号值序列。
例如，对于间期序列{RR 1，RR 2，…，RR N}，计算时间间隔RR i所针对的瞬时心率HR i＝60/RR i，单位为bpm(beats per minute)，表示每分钟心搏次数，i为大于等于1且小于等于N的整数。由此得到瞬时心率序列{HR 1，HR 2，…，HR N}。
示例性地，上述对瞬时心率序列中的多个瞬时心率分别进行编码，以得到多个编码符号的过程可以按照如下方式进行。
对于所述瞬时心率序列中的任一瞬时心率，基于该任一瞬时心率的数值，对该任一瞬时心率进行分类。当该任一瞬时心率属于第一分类时，将预定符号作为该任一瞬时心率的编码符号。当该任一瞬时心率属于第二分类时，计算该任一瞬时心率与第一数值的比值，并将所得到的比值作为该任一瞬时心率的编码符号。
例如，上述基于该任一瞬时心率的数值，对该任一瞬时心率进行分类的过程可以是：当该瞬时心率的数值大于等于预定数值时，确定该瞬时心率属于第一分类，当该瞬时心率的数值小于预定数值时，确定该瞬时心率属于第二分类。可以根据公式(6)对瞬时心率序列{HR 1，HR 2，…，HR N}中的瞬时心率HR i进行第一编码，以得到HR i的编码符号SY i，本例中通过两位数字表示该编码符号SY i。
Figure PCTCN2020117707-appb-000006
其中，f为上述对每个瞬时心率进行分类的预定数值，g为第一数值，预定数值与第一数值之间的比值为本例中的预定符号。上述参数可以根据实际需要进行设置。本例中，f＝315，g＝5，预定符号为63。[]表示四舍五入取整操作，0≤SY i≤63。经过上述第一编码过程，将瞬时心率序列{HR 1，HR 2，…，HR N}转换为符号序列{SY 1，SY 2，…，SY N}。
示例性地，上述对符号序列中的多个编码符号分别进行第二编码，以得到多个符号值的过程可以按照如下方式进行：对于符号序列中的任一编码符号，计算该任一编码符号的在前编码符号、该任一编码符号、以及该任一编码符号的在后编码符号的加权和，并将所得到的加权和作为针对该任一编码符号的符号值。
例如，可以根据公式(7)对符号序列{SY 1，SY 2，…，SY N}中的编码符号SY i进行 第二编码，以得到编码符号SY i所针对的符号值SYV i。该第二编码过程也可称为符号值模板化过程。
SYV i＝SY i-1×h 1+SY i×h 2+SY i+1×h 3
                                            公式(7)
公式(7)中，i为大于0且小于N的整数。h 1、h 2和h 3分别表示第一权重、第二权重和第三权重，可以根据实际需要对h 1、h 2和h 3进行设置。本例中，h 1＝2 12，h 2＝2 6，h 3＝1。例如对于符号序列{01，20，13}，编码符号“20”所针对的符号值等于SYV＝01×4096+20×64+13，0≤SYV i≤262143。经过上述第二编码过程，将符号序列{SY 1，SY 2，…，SY N}转换为符号值序列{SYV 1，SYV 2，…，SYV N}。
图4示意性示出了根据本公开实施例的间期序列转换为符号值序列的过程的示例示意图。
如图4所示，4-1为间期序列的示例示意图，展示了从第1个时间间隔至第61个时间间隔的幅值变化。4-2为瞬时心率序列的示例示意图，展示了从第1个瞬时心率至第61个瞬时心率的幅值变化。4-3为符号序列的示例示意图，展示了从第1个编码符号至第61个编码符号的幅值变化。4-4为符号值序列的示例示意图，展示了从第2个符号值至第60个符号值的幅值变化。
由图4可以看出从间期序列到符号值序列的符号化转换过程，其中时间间隔的取值范围是连续的，而符号值的取值范围是有限的。其基本思想是将无关细节信息去除，将数据在离散值上进行分类，把本来具有很多可能性的时间间隔数据转换为有限个数的符号值。从而保留心电信号的大尺度特征，丢弃了无关噪声的干扰，有利于后续基于符号值序列的香农熵对心电信号进行分类的进行。
根据本公开的实施例，上述确定符号值序列的香农熵的过程可以按照如下方式进行。
预设多个符号值区间。然后，对于多个符号值区间中的任一符号值区间，计算符号值序列中落入该任一符号值区间的符号值的数量与符号值序列中符号值的数量之间的比例，并将该比例作为符号值序列关于该任一符号值区间的分布概率。接着，基于符号值序列关于预设的多个符号值区间各自的分布概率，确定符号值序列的香农熵。
例如，基于上文中的公式(7)，符号值的最小取值为0，最大取值为262143。将[0，262143]的取值空间称为符号值空间(SYV空间)中，以2048为间隔可以将该SYV空间划分为128个符号值区间，作为预设的128个符号值区间。如第一个符号值区间为[0，2047]， 第二个符号值区间为[2048，4095]，以此类推，不再赘述。或者，在另一个例子中，对于SYV空间[0，262143]，以4096为间隔可以将该SYV空间划分为64个符号值区间，作为预设的64个符号值区间。随着符号值空间范围的不同，划分间隔的不同，可以预设不同数量的符号值区间，在此不做限制。
例如，预设Q个符号值区间，符号值序列{SYV 1，SYV 2，…，SYV N}中落入第j个符号值区间的符号值的数量为n j，符号值序列{SYV 1，SYV 2，…，SYV N}所包含的符号值总数量为N，则计算比例p j＝n j/N，将p j作为该符号值序列关于第j个符号值区间的分布概率。其中，j为大于等于1且小于等于Q的整数，n j≤N。然后可以基于符号值序列{SYV 1，SYV 2，…，SYV N}关于预设的Q个符号值区间各自的分布概率，确定该符号值序列{SYV 1，SYV 2，…，SYV N}的香农熵。
根据本公开的实施例，为了更准确地适配于心电信号的分类问题，对香农熵的定义进行了修正。示例性地，上述基于符号值序列关于多个符号值区间各自的分布概率，确定符号值序列的香农熵的过程可以按照如下方式进行。
一方面，基于符号值序列关于多个符号值区间各自的分布概率，计算符号值序列关于多个符号值区间的分布期望。另一方面，基于符号值序列中符号值的数量和多个符号值区间的数量，确定增益系数。然后，基于上述分布期望和增益系数，确定符号值序列的香农熵。
仍以上文中的示例为例，符号值区间的数量为Q，符号值序列{SYV 1，SYV 2，…，SYV N}关于第j个符号值区间的分布概率为p j，可以根据公式(8)计算该符号值序列{SYV 1，SYV 2，…，SYV N}的香农熵SE。
Figure PCTCN2020117707-appb-000007
其中，
Figure PCTCN2020117707-appb-000008
为增益系数，
Figure PCTCN2020117707-appb-000009
为符号值序列{SYV 1，SYV 2，…，SYV N}关于Q个符号值区间的分布期望。
根据本公开的实施例，在计算得到符号值序列的香农熵之后，可以通过比较该香农熵的数值和第一预定阈值ThSe之间的大小关系来确定心电信号的类别。
示例性地，上述基于香农熵的数值，对心电信号进行分类的过程可以包括：当香农熵的数值大于第一预定阈值时，确定心电信号属于第一预定类别。对于心电信号来说， 当针对该心电信号的符号值序列的香农熵较小时，表征该心电信号的波动较为稳定，针对该心电信号的符号值序列的香农熵越大，表征该心电信号的波动越不稳定，通常情况下，不稳定的心电波动可能是由一些疾病导致的。本实施例中以第一预定阈值(也可称为“香农熵阈值”)为分界，判定心电信号是否属于第一预定类别。示例性地，第一预定类别可以与某类疾病具有相关性，例如第一预定类别与房颤(atrial fibrillation，AF)具有相关性。
符号值序列的长度N和香农熵阈值ThSe这两个值的设置会直接影响分类结果的准确度。香农熵阈值过低会导致将一些正常心电信号划分至第一预定类别，香农熵过高则会导致对于一些异常心电信号的不敏感。符号值序列的长度选取不当不仅会造成分类结果不准确，符号值序列太长还会影响算法的实时性和复杂度。经过参数调优，例如可以将香农熵阈值ThSe设置在0.2～0.5的范围内，符号值序列的长度N可以设置在15～60的范围内，此时能兼顾灵敏度和准确度。并且，可以仅利用一分钟左右的心电信号数据即可实现对于心电信号的分类，以获知心电信号与某类疾病(例如房颤)的相关性，从而为临床监护和远程医疗等场景提供协助。
进一步地，在确定所获取的心电信号属于第一预定类别之后，还可以通过如下投票决策机制更精确地定位心电信号中属于第一预定类别的心搏位置所在。
示例性地，根据本公开实施例的心电信号处理方法还可以包括以下方式。
首先，设置滑动窗口，滑动窗口的长度需要小于符号值序列的长度。将滑动窗口在符号值序列中从前至后进行滑动，以从符号值序列中提取出多个符号值子序列。其中，滑动窗口的每次滑动步长为1个符号值。然后，对于提取出的多个符号值子序列中的任一符号值子序列，确定该任一符号值子序列的香农熵，当该任一符号值子序列的香农熵大于第一预定阈值时，将该任一符号值子序列中的每个符号值的评分增加预定分值，该预定分值可以根据需要进行设置，在此不做限制。接着，在滑动窗口滑动结束后，确定符号值序列中的每个符号值的总评分，并确定总评分大于第二预定阈值的符号值属于第一预定类别。
图5示意性示出了根据本公开实施例的投票决策的过程的示例示意图。
如图5所示，本例中滑动窗口的长度为20，即对应于20个符号值的长度。符号值序列{SYV 1，SYV 2，…，SYV N}的长度N大于20。将滑动窗口在符号值序列中从前至后进行滑动，提取出多个符号值子序列。如第1个符号值子序列为{SYV 1，SYV 2，…，SYV 20}， 第2个符号值子序列为{SYV 2，SYV 2，…，SYV 21}，第3个符号值子序列为{SYV 3，SYV 2，…，SYV 22}，以此类推，不再赘述。
分别计算所提取出的每个符号值子序列的香农熵，计算方式与上文中计算符号值序列的香农熵的方式原理相同，上文中已详细说明，在此不再赘述。例如预定分值为1。例如，计算第1个符号值子序列{SYV 1，SYV 2，…，SYV 20}的香农熵为SE 1，如果SE 1大于第一预定阈值ThSe，则将该符号值子序列{SYV 1，SYV 2，…，SYV 20}中的每个符号值的评分加1。其中，符号值SYV i的评分以AF i表示。如果SE 1小于等于第一预定阈值ThSe，则该符号值子序列{SYV 1，SYV 2，…，SYV 20}中的每个符号值的评分维持不变。在计算完成所有符号值子序列的香农熵并进行评分后，可以得到符号值序列{SYV 1，SYV 2，…，SYV N}中每个符号值的总评分。当一个符号值的总评分大于第二预定阈值时，确定该符号值属于第一预定类别，表征该符号值所针对的心搏属于第一预定类别，例如该心搏与房颤具有相关性。本例中，第二预定阈值设置为12，在其他例子中，可以根据需要对第二预定阈值进行设置，原则上保证第二预定阈值小于等于滑动窗口的长度即可。
本领域技术人员可以理解，根据本公开实施例的心电信号处理方法不仅可以对心电信号进行分类，还可以通过投票决策机制更加精确地定位心电信号中属于第一预定类别的心搏位置所在，对于医学诊断和监测等领域均具有重大意义。
一些情况下，由于人体运动以及人体与外部联系等，会造成心电信号被噪声所干扰，给心电信号处理带来困难。例如，由被测对象的呼吸、电极移动等低频干扰所引起的基线漂移，导致实际测量中心电信号偏离正常的基线位置，出现上下缓慢波动变化的现象。心电信号本身含有丰富的低频成分，基线漂移会掩盖有用信息，影响针对心电信号的分析、识别、分类、定位等的准确性。根据本公开的实施例，在获取到心电信号之后，可以去除心电信号中的基线漂移噪声。可以采用各种方式进行基线漂移噪声的去除，例如自适应滤波法、卡尔曼滤波法、小波变换法等等，在此不做限制。本例中采用中值滤波法来快速有效地抑制基线漂移。
根据本公开实施例的心电信号处理方法，可针对单导联心电信号进行处理从而得到关于心搏是否异常的分类结果，不必基于多导联心电信号中的复杂信号进行处理，十分简单可靠。
图6A示意性示出了根据本公开实施例的心电信号处理过程的示例示意图。图6B示意性示出了根据本公开另一实施例的心电信号处理过程的示例示意图。需要说明的是， 图6A和图6B所示的心电信号处理过程仅作为示例以利于理解本公开，本公开对此不做限定。
如图6A和图6B所示，该心电信号处理过程可以包括步骤S601～S610。
在步骤S601，获取原始心电信号。
在步骤S602，对原始心电信号进行预处理，以去除基线漂移噪声，从而得到预处理后的心电信号。
在步骤S603，识别心电信号中的R波。
在步骤S604，提取初始RR间期序列。
在步骤S605，处理初始RR间期序列中的异位心搏，以得到RR间期序列。
在步骤S606，计算瞬时心率序列。
在步骤S607，进行符号转换，以得到符号序列。
在步骤S608，进行符号值模板化，以得到符号值序列。
在步骤S609，计算符号值序列的香农熵。
在步骤S610，基于第一预定阈值，对香农熵进行阈值判别，从而根据判别结果确定心电信号所属的类别，并进行标注，例如第一预定类别与房颤相关，则对于确定属于第一预定类别的心电信号可标注为“房颤相关”。然后再重复步骤S601，直至全部心电信号标注完毕。
如图6B所示，心电处理过程在步骤S610后，还进一步包括步骤S611，进行投票决策，以在属于第一预定类别的心电信号内部定位异常心搏，从而进行更具体的标注。例如在确定属于第一预定类别的心电信号内的一个或多个心搏位置处可标注为“房颤相关”。投票决策在上文中已详细说明，在此不再赘述。然后再重复步骤S601，直至全部心电信号标注完毕。
可以理解，本公开实施例在对心电信号进行处理的过程中结合和符号动力学和自定义的香农熵，对于与某类疾病相关的心电特征较为敏感，且算法复杂度不高，无需训练分类器，有效提高了心电信号监测过程中对于心电信号的分类、以及对于心电信号内部异常心搏的定位的准确性和实时性。
图7示意性示出了根据本公开实施例的心电信号处理装置的框图。
如图7所示，心电信号处理装置700可以包括：获取模块710、确定模块720、转换 模块730、熵计算模块740以及分类模块750。
获取模块710用于获取心电信号，心电信号包括多个指定波。
确定模块720用于确定针对心电信号中多个指定波的间期序列。
转换模块730用于基于符号动力学，将间期序列转换为符号值序列。
熵计算模块740用于确定符号值序列的香农熵。
分类模块750用于基于香农熵的数值，对心电信号进行分类。
需要说明的是，上述装置部分实施例中各模块/单元/子单元等的实施方式、解决的技术问题、实现的功能、以及达到的技术效果分别与方法部分实施例中各对应的步骤的实施方式、解决的技术问题、实现的功能、以及达到的技术效果相同或类似，在此不再赘述。
根据本公开的实施例的模块、子模块、单元、子单元中的任意多个、或其中任意多个的至少部分功能可以在一个模块中实现。根据本公开实施例的模块、子模块、单元、子单元中的任意一个或多个可以被拆分成多个模块来实现。根据本公开实施例的模块、子模块、单元、子单元中的任意一个或多个可以至少被部分地实现为硬件电路，例如现场可编程门阵列(FPGA)、可编程逻辑阵列(PLA)、片上系统、基板上的系统、封装上的系统、专用集成电路(ASIC)，或可以通过对电路进行集成或封装的任何其他的合理方式的硬件或固件来实现，或以软件、硬件以及固件三种实现方式中任意一种或以其中任意几种的适当组合来实现。或者，根据本公开实施例的模块、子模块、单元、子单元中的一个或多个可以至少被部分地实现为计算机程序模块，当该计算机程序模块被运行时，可以执行相应的功能。
图8示意性示出了根据本公开的实施例的适于实现上文描述的方法的电子设备的框图。图8示出的电子设备仅仅是一个示例，不应对本公开实施例的功能和使用范围带来任何限制。
如图8所示，电子设备800包括一个或多个处理器810以及计算机可读存储介质820。该电子设备800可以执行根据本公开实施例的方法。
例如，处理器810例如可以包括通用微处理器、指令集处理器和/或相关芯片组和/或专用微处理器(例如，专用集成电路(ASIC))，等等。处理器810还可以包括用于缓存用途的板载存储器。处理器810可以是用于执行根据本公开实施例的方法流程的不同动作的单一处理单元或者是多个处理单元。
计算机可读存储介质820，例如可以是非易失性的计算机可读存储介质，具体示例包括但不限于：磁存储装置，如磁带或硬盘(HDD)；光存储装置，如光盘(CD-ROM)；存储器，如随机存取存储器(RAM)或闪存；等等。
计算机可读存储介质820可以包括计算机程序821，该计算机程序821可以包括代码/计算机可执行指令，其在由处理器810执行时使得处理器810执行根据本公开实施例的方法或其任何变形。
计算机程序821可被配置为具有例如包括计算机程序模块的计算机程序代码。例如，在示例实施例中，计算机程序821中的代码可以包括一个或多个程序模块，例如包括821A、模块821B、……。应当注意，模块的划分方式和个数并不是固定的，本领域技术人员可以根据实际情况使用合适的程序模块或程序模块组合，当这些程序模块组合被处理器810执行时，使得处理器810可以执行根据本公开实施例的方法或其任何变形。
本公开还提供了一种计算机可读存储介质，该计算机可读存储介质可以是上述实施例中描述的设备/装置/系统中所包含的；也可以是单独存在，而未装配入该设备/装置/系统中。上述计算机可读存储介质承载有一个或者多个程序，当上述一个或者多个程序被执行时，实现根据本公开实施例的方法。
根据本公开的实施例，计算机可读存储介质可以是非易失性的计算机可读存储介质，例如可以包括但不限于：便携式计算机磁盘、硬盘、随机访问存储器(RAM)、只读存储器(ROM)、可擦式可编程只读存储器(EPROM或闪存)、便携式紧凑磁盘只读存储器(CD-ROM)、光存储器件、磁存储器件、或者上述的任意合适的组合。在本公开中，计算机可读存储介质可以是任何包含或存储程序的有形介质，该程序可以被指令执行系统、装置或者器件使用或者与其结合使用。
尽管已经参照本公开的特定示例性实施例示出并描述了本公开，但是本领域技术人员应该理解，在不背离所附权利要求及其等同物限定的本公开的精神和范围的情况下，可以对本公开进行形式和细节上的多种改变。因此，本公开的范围不应该限于上述实施例，而是应该不仅由所附权利要求来进行确定，还由所附权利要求的等同物来进行限定。

Claims (18)

  1. 一种心电信号处理方法，包括：
    获取心电信号，所述心电信号包括多个指定波；
    确定针对所述多个指定波的间期序列；
    基于符号动力学，将所述间期序列转换为符号值序列；
    确定所述符号值序列的香农熵；以及
    基于所述香农熵的数值，对所述心电信号进行分类。
  2. 根据权利要求1所述的方法，其中，所述基于所述香农熵的数值，对所述心电信号进行分类包括：
    当所述香农熵的数值大于第一预定阈值时，确定所述心电信号属于第一预定类别。
  3. 根据权利要求1所述的方法，其中，所述基于符号动力学，将所述间期序列转换为符号值序列包括：
    计算所述间期序列中每个时间间隔所针对的瞬时心率，以得到瞬时心率序列；
    对所述瞬时心率序列中的多个瞬时心率分别进行第一编码，以得到多个编码符号，并由所述多个编码符号组成符号序列；以及
    对所述符号序列中的多个编码符号分别进行第二编码，以得到多个符号值，并由所述多个符号值组成所述符号值序列。
  4. 根据权利要求3所述的方法，其中，所述对所述瞬时心率序列中的多个瞬时心率分别进行第一编码，以得到多个编码符号包括：
    对于所述瞬时心率序列中的任一瞬时心率，基于所述任一瞬时心率的数值，对所述任一瞬时心率进行分类；
    当所述任一瞬时心率属于第一分类时，将预定符号作为所述任一瞬时心率的编码符号；以及
    当所述任一瞬时心率属于第二分类时，计算所述任一瞬时心率与第一数值的比值，并将所述比值作为所述任一瞬时心率的编码符号。
  5. 根据权利要求3所述的方法，其中，所述对所述符号序列中的多个编码符号分别进行第二编码，以得到多个符号值包括：
    对于所述符号序列中的任一编码符号，计算所述任一编码符号的在前编码符号、所述任一编码符号、以及所述任一编码符号的在后编码符号的加权和，并将所述加权和作为针对所述任一编码符号的符号值。
  6. 根据权利要求1所述的方法，其中，所述确定所述符号值序列的香农熵包括：
    预设多个符号值区间；
    对于所述多个符号值区间中的任一符号值区间，计算所述符号值序列中落入所述任一符号值区间的符号值的数量与所述符号值序列中符号值的数量之间的比例，并将所述比例作为所述符号值序列关于所述任一符号值区间的分布概率；以及
    基于所述符号值序列关于所述多个符号值区间各自的分布概率，确定所述符号值序列的香农熵。
  7. 根据权利要求6所述的方法，其中，所述基于所述符号值序列关于所述多个符号值区间各自的分布概率，确定所述符号值序列的香农熵包括：
    基于所述符号值序列关于所述多个符号值区间各自的分布概率，计算所述符号值序列关于所述多个符号值区间的分布期望；
    基于所述符号值序列中符号值的数量和所述多个符号值区间的数量，确定增益系数；以及
    基于所述分布期望和所述增益系数，确定所述符号值序列的香农熵。
  8. 根据权利要求1所述的方法，还包括：
    设置滑动窗口，所述滑动窗口的长度小于所述符号值序列的长度；
    将所述滑动窗口在所述符号值序列中从前至后进行滑动，以从所述符号值序列中提取出多个符号值子序列，其中，所述滑动窗口的每次滑动步长为1个符号值；
    对于所述多个符号值子序列中的任一符号值子序列，确定所述任一符号值子序列的香农熵，当所述任一符号值子序列的香农熵大于第一预定阈值时，将所述任一符号值子序列中的每个符号值的评分加1；
    在所述滑动窗口滑动结束后，确定所述符号值序列中的每个符号值的总评分；以及
    确定总评分大于第二预定阈值的符号值属于第一预定类别。
  9. 根据权利要求1所述的方法，其中，所述确定针对所述多个指定波的间期序列包括：
    识别所述多个指定波各自的波峰；以及
    基于所述多个指定波各自的波峰中每两个相邻波峰之间的时间间隔，确定所述间期序列。
  10. 根据权利要求9所述的方法，其中，所述多个指定波包括M+1个指定波：
    所述基于所述多个指定波各自的波峰中每两个相邻波峰之间的时间间隔，确定所述间期序列包括：
    由M+1个指定波各自的波峰中每两个相邻波峰之间的时间间隔，组成包含M个时间间隔的初始间期序列；以及
    对初始间期序列进行预处理，以得到包含N个时间间隔的所述间期序列；
    其中，M和N均为正整数，M大于等于N。
  11. 根据权利要求10所述的方法，其中，所述对初始间期序列进行预处理，以得到包含N个时间间隔的所述间期序列包括：
    确定所述M个时间间隔的平均值；
    对于所述M个时间间隔中的任一时间间隔，基于所述任一时间间隔与所述平均值之间的差异确定所述任一时间间隔是否属于第二预定类别；以及
    如果确定所述任一时间间隔属于第二预定类别，则从所述初始间期序列中将所述任一时间间隔以及所述任一时间间隔的在后时间间隔筛除。
  12. 根据权利要求10所述的方法，其中，所述对初始间期序列进行预处理，以得到包含N个时间间隔的所述间期序列包括：
    确定所述M个时间间隔的平均值；
    对于所述M个时间间隔中的任一时间间隔，对所述任一时间间隔与所述任一时间间隔的在后时间间隔进行求和以得到合并间隔；
    基于所述合并间隔与所述平均值之间的差异确定所述任一时间间隔是否属于第二预定类别；以及
    如果所述任一时间间隔属于第二预定类别，则将所述任一时间间隔以及所述任一时间间隔的在后时间间隔合并为一个时间间隔。
  13. 根据权利要求10所述的方法，其中，所述对初始间期序列进行预处理，以得到包含N个时间间隔的所述间期序列包括以下步骤中的至少一个：
    对于所述M个时间间隔中的任一时间间隔，如果所述任一时间间隔与所述任一时间间隔的在前时间间隔的比值大于第二数值、且所述任一时间间隔的在后时间间隔与所述任一时间间隔的比值大于第三数值，则从所述初始间期序列中将所述任一时间间隔以及所述任一时间间隔的在后时间间隔筛除；以及
    对于所述任一时间间隔，如果所述任一时间间隔与所述任一时间间隔的在前时间间隔的比值大于第四数值、且所述任一时间间隔的在后时间间隔与所述任一时间间隔的比值小于第五数值，则从所述初始间期序列中将所述任一时间间隔以及所述任一时间间隔的在后时间间隔筛除；
    其中，所述第二数值小于所述第三数值，所述第四数值大于所述第五数值。
  14. 根据权利要求1所述的方法，还包括：
    在所述获取心电信号之后，去除所述心电信号中的基线漂移噪声。
  15. 根据权利要求1所述的方法，其中，所述心电信号为单导联心电信号。
  16. 一种心电信号处理装置，包括：
    获取模块，用于获取心电信号，所述心电信号包括多个指定波；
    确定模块，用于确定针对所述多个指定波的间期序列；
    转换模块，用于基于符号动力学，将所述间期序列转换为符号值序列；
    熵计算模块，用于确定所述符号值序列的香农熵；以及
    分类模块，用于基于所述香农熵的数值，对所述心电信号进行分类。
  17. 一种电子设备，包括：
    存储器，配置为存储指令；
    至少一个处理器：
    所述至少一个处理器执行存储在存储器中的指令，以实现根据权利要求1～15之一所述的方法。
  18. 一种非暂时性计算机存储介质，存储计算机可读指令，其中，当所述计算机可读指令由计算机执行时可以执行根据权利要求1～15之一所述的方法。
PCT/CN2020/117707 2019-11-29 2020-09-25 心电信号处理方法、心电信号处理装置和电子设备 WO2021103796A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US17/289,082 US20220304610A1 (en) 2019-11-29 2020-09-25 Method and apparatus for processing an electrocardiogram signal and electronic device

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201911211864.7A CN110840443B (zh) 2019-11-29 2019-11-29 心电信号处理方法、心电信号处理装置和电子设备
CN201911211864.7 2019-11-29

Publications (1)

Publication Number Publication Date
WO2021103796A1 true WO2021103796A1 (zh) 2021-06-03

Family

ID=69607588

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/117707 WO2021103796A1 (zh) 2019-11-29 2020-09-25 心电信号处理方法、心电信号处理装置和电子设备

Country Status (3)

Country Link
US (1) US20220304610A1 (zh)
CN (1) CN110840443B (zh)
WO (1) WO2021103796A1 (zh)

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110840443B (zh) * 2019-11-29 2022-06-10 京东方科技集团股份有限公司 心电信号处理方法、心电信号处理装置和电子设备
CN113499082B (zh) * 2020-03-23 2023-12-19 疆域康健创新医疗科技成都有限公司 Qrs波群检测方法、心电检测装置及可读存储介质
CN113712570B (zh) * 2020-05-12 2024-03-08 深圳市科瑞康实业有限公司 一种长间歇心电信号数据预警方法
CN114521901B (zh) * 2021-12-28 2024-03-19 宁波慈溪生物医学工程研究所 一种心电特征提取方法、装置及存储介质

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20120016249A1 (en) * 2010-07-13 2012-01-19 Jie Lian Method and device for noise detection in physiological signals
CN105496402A (zh) * 2015-11-20 2016-04-20 北京理工大学 基于散点图和符号动力学的心电特征分析方法
US20170224238A1 (en) * 2016-02-10 2017-08-10 Regents Of The University Of Minnesota Graphically mapping rotors in a heart
CN109452938A (zh) * 2018-12-29 2019-03-12 中国矿业大学 一种基于多尺度多重分形的hfecg信号特征频率检测方法
CN109480819A (zh) * 2017-09-11 2019-03-19 南京大学 一种利用多路体表心电预测房颤发生的方法
CN110363177A (zh) * 2019-07-23 2019-10-22 上海图灵医疗科技有限公司 一种人体生物电信号混沌特征的提取方法
CN110840443A (zh) * 2019-11-29 2020-02-28 京东方科技集团股份有限公司 心电信号处理方法、心电信号处理装置和电子设备

Family Cites Families (17)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7025729B2 (en) * 2001-09-14 2006-04-11 Biancamed Limited Apparatus for detecting sleep apnea using electrocardiogram signals
US8475387B2 (en) * 2006-06-20 2013-07-02 Adidas Ag Automatic and ambulatory monitoring of congestive heart failure patients
US8666483B2 (en) * 2007-10-24 2014-03-04 Siemens Medical Solutions Usa, Inc. System for cardiac medical condition detection and characterization
US8568330B2 (en) * 2011-03-08 2013-10-29 Pulsaw Informatics, Inc. Composite human physiological stress index based on heart beat and sleep and/or activity history data including actigraphy
CN102284138B (zh) * 2011-08-03 2014-03-05 复旦大学 基于二阶导数编码的符号序列熵自动判别室速室颤的体外除颤器
US9020583B2 (en) * 2013-03-13 2015-04-28 Siemens Medical Solutions Usa, Inc. Patient signal analysis and characterization
EP2786704B1 (en) * 2013-04-02 2016-10-05 Georg Schmidt Device and method for assessing mortality risk of a cardiac patient
US10258249B2 (en) * 2015-12-08 2019-04-16 Regents Of The University Of Minnesota Graphically mapping rotors in a heart using Shannon entropy
US20200375480A1 (en) * 2017-03-24 2020-12-03 Beth Israel Deaconess Medical Center, Inc. Non-Invasive Cardiovascular Risk Assessment Using Heart Rate Variability Fragmentation
CN107832737B (zh) * 2017-11-27 2021-02-05 上海优加利健康管理有限公司 基于人工智能的心电图干扰识别方法
US11717686B2 (en) * 2017-12-04 2023-08-08 Neuroenhancement Lab, LLC Method and apparatus for neuroenhancement to facilitate learning and performance
CA3091209C (en) * 2018-03-01 2021-08-31 Polyvagal Science LLC Systems and methods for modulating physiological state
US11918352B2 (en) * 2018-05-15 2024-03-05 Isbrg Corp. Non-invasive determination of a physiological state of interest in a subject
CN109864736A (zh) * 2019-03-22 2019-06-11 深圳市理邦精密仪器股份有限公司 心电信号的处理方法、装置、终端设备及介质
KR20230168891A (ko) * 2022-06-08 2023-12-15 조선대학교산학협력단 컨볼루션 신경망 및 장단기 메모리 네트워크를 이용한 비가압적 혈압 예측 장치, 방법 및 컴퓨터로 판독 가능한 저장 매체
WO2023242016A1 (en) * 2022-06-13 2023-12-21 Koninklijke Philips N.V. Detecting user infection using wearable sensor data cross-reference to related applications
US20230397829A1 (en) * 2022-06-13 2023-12-14 Medical Predictive Science Corporation Method, System and Device for Individualized Supplemental Oxygen Therapy for Preterm Infants

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20120016249A1 (en) * 2010-07-13 2012-01-19 Jie Lian Method and device for noise detection in physiological signals
CN105496402A (zh) * 2015-11-20 2016-04-20 北京理工大学 基于散点图和符号动力学的心电特征分析方法
US20170224238A1 (en) * 2016-02-10 2017-08-10 Regents Of The University Of Minnesota Graphically mapping rotors in a heart
CN109480819A (zh) * 2017-09-11 2019-03-19 南京大学 一种利用多路体表心电预测房颤发生的方法
CN109452938A (zh) * 2018-12-29 2019-03-12 中国矿业大学 一种基于多尺度多重分形的hfecg信号特征频率检测方法
CN110363177A (zh) * 2019-07-23 2019-10-22 上海图灵医疗科技有限公司 一种人体生物电信号混沌特征的提取方法
CN110840443A (zh) * 2019-11-29 2020-02-28 京东方科技集团股份有限公司 心电信号处理方法、心电信号处理装置和电子设备

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
CHUNHUA BIAN, QIANLI MA, JUNFENG SI, XUHUI WU, JUN SHAO, XINBAO NING , DONGJIN WANG: "Sign Series Entropy Analysis of Short-Term Heart Rate Variability", CHINESE SCIENCE BULLETIN, vol. 54, no. 3, 20 June 2009 (2009-06-20), pages 340 - 344, XP009528349, ISSN: 0023-074X *
MU, YUANHUI: "ECG Arrhythmia Feature Extraction And Analysis by Multi-scale Time-frequency Method", CHINESE MASTER’S THESES FULL-TEXT DATABASE, 1 June 2015 (2015-06-01), pages 1 - 75, XP055816668 *
PARLITZ U.; BERG S.; LUTHER S.; SCHIRDEWAN A.; KURTHS J.; WESSEL N.: "Classifying cardiac biosignals using ordinal pattern statistics and symbolic dynamics", COMPUTERS IN BIOLOGY AND MEDICINE, vol. 42, no. 3, 31 March 2012 (2012-03-31), pages 319 - 327, XP028896032, ISSN: 0010-4825, DOI: 10.1016/j.compbiomed.2011.03.017 *

Also Published As

Publication number Publication date
CN110840443A (zh) 2020-02-28
US20220304610A1 (en) 2022-09-29
CN110840443B (zh) 2022-06-10

Similar Documents

Publication Publication Date Title
WO2021103796A1 (zh) 心电信号处理方法、心电信号处理装置和电子设备
US20210275046A1 (en) Peak Detection System and Method for Calculation of Signal-Derived Metrics
US9042973B2 (en) Apparatus and method for measuring physiological signal quality
US8755876B2 (en) Methods and systems for atrial fibrillation detection
US8718753B2 (en) Motion and noise artifact detection for ECG data
Zhao et al. A robust QRS detection and accurate R-peak identification algorithm for wearable ECG sensors
US20150105666A1 (en) Narrow band feature extraction from cardiac signals
US20120209126A1 (en) Method and system for detecting cardiac arrhythmia
US20200107775A1 (en) Methods and Systems for Monitoring Sleep Apnea
CN107622259B (zh) 一种t波检测方法、心电数据分析方法及装置
Liu et al. Refined generalized multiscale entropy analysis for physiological signals
CN111557658B (zh) Ppg实时心率信号质量评估方法及装置、存储介质
Lee et al. A real-time abnormal beat detection method using a template cluster for the ECG diagnosis of IoT devices
Bhoi et al. QRS Complex Detection and Analysis of Cardiovascular Abnormalities: A Review.
Pimentel et al. Probabilistic estimation of respiratory rate using Gaussian processes
EP3203900A1 (en) Weaning readiness indicator, sleeping status recording device, and air providing system applying nonlinear time-frequency analysis
Alvarez-Estevez et al. Spectral Heart Rate Variability analysis using the heart timing signal for the screening of the Sleep Apnea–Hypopnea Syndrome
WO2021249493A1 (zh) T波起点的确定方法、装置、存储介质及计算机程序产品
US10064560B2 (en) System for cardiac pathology detection and characterization
US9020583B2 (en) Patient signal analysis and characterization
US9277889B2 (en) Patient signal analysis based on actiniform segmentation
KR101515725B1 (ko) 심장 질환 진단 장치 및 방법
Zhang et al. Evaluation of single-lead ECG signal quality with different states of motion
EP3014501B1 (en) Measuring respiration
Apandi et al. Noise Reduction Method based on Autocorrelation for Threshold-Based Heartbeat Detection

Legal Events

Date Code Title Description
NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20892497

Country of ref document: EP

Kind code of ref document: A1

122 Ep: pct application non-entry in european phase

Ref document number: 20892497

Country of ref document: EP

Kind code of ref document: A1

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 14/02/2023)

122 Ep: pct application non-entry in european phase

Ref document number: 20892497

Country of ref document: EP

Kind code of ref document: A1