CN101427917A - ECG abnormal detection method based on inherent trend subsequence mode decomposition - Google Patents
ECG abnormal detection method based on inherent trend subsequence mode decomposition Download PDFInfo
- Publication number
- CN101427917A CN101427917A CNA2008100460067A CN200810046006A CN101427917A CN 101427917 A CN101427917 A CN 101427917A CN A2008100460067 A CNA2008100460067 A CN A2008100460067A CN 200810046006 A CN200810046006 A CN 200810046006A CN 101427917 A CN101427917 A CN 101427917A
- Authority
- CN
- China
- Prior art keywords
- subsequence
- trend
- sequence
- inherent
- ecg
- Prior art date
- Legal status (The legal status 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 status listed.)
- Granted
Links
Landscapes
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
The invention discloses an ECG anomaly detection method based on the natural trend subsequence mode decomposition, which comprises the steps as follows: (1) the method is defined; (2) a normal ECG signal sequence is processed through sequence decomposition and demixion; (3) a suspected ECG signal sequence is processed through sequence decomposition and demixion; (4) and matching detection is carried out. The provided natural trend subsequence mode reflects essential characteristics of cardiac activity in ECG signals, detects various abnormal ECG signal sequences accurately and effectively; and the method is simple and easy to implement and has excellent application prospect.
Description
Technical field
The present invention relates to biomedical signals measuring and signal processing technology field, be specifically related to a kind of ECG method for detecting abnormality that decomposes based on inherent trend subsequence pattern.
Background technology
Biomedical signals measuring and signal processing technology can be assisted the research of biological and physiological system, and assistance is diagnosed the patient and treated.In various biological signal measurings, it is very fast with the electrocardio to be that the physiological amount of the blood circulation of representative is measured development comparatively speaking, is one of method of clinical analysis, significant to the diagnosis of some disease especially cardiovascular disease.
Electrocardiogram (ECG) is the simple and practical method of record cardiac electrical activity, can reflect excited in heart communication process and the functional status of heart.If pathological changes takes place for the conducting system generation obstacle of heart or certain part cardiac muscle, the variation of electrocardio-activity can correctly be reflected on the electrocardiogram in time, shows as the ANOMALOUS VARIATIONS and the carrying out property evolution process of each waveform.Be applied to the electrocardiogram technology clinical from Beethoven in 1903, so far century-old, in a century, the electrocardiogram technology of sustainable development is human life and health, make huge contribution for biology, clinical medicine, become clinical indispensable, most important routine examination technology.In recent years, the M ﹠ M of cardiovascular and cerebrovascular disease rises year by year, heart is done made regular check on, and scents a hidden danger early, has become especially cardiac's needs of most of people.Electrocardiogram is competent assistants of diagnosis heart disease as the means of checking that heart disease is conventional and necessary, to its accurate interpretation for finding the state of an illness early and improving the national health level and quality of life has significance.Electrocardiogram has important function at aspects such as the variation of the diagnosis rhythm of the heart, myocardial ischemia, myocardial infarctions.But ill electrocardiogram is of a great variety, the variation very big, the different patients' of pathology of the same race electrocardiogram even same patient's ECG itself exists very big difference, to make accurate judgement to it, need the doctor to have rich knowledge and a large amount of clinical experiences of accumulation usually.In addition, very easily tired if the doctor is engaged in the identification work of a large amount of figures for a long time, omission easily, make mistakes, thereby Electrocardiographic automatic diagnosis becomes people's research focus.Electrocardiographic automatic diagnosis, not only alleviated clinicist's burden but also can further analyze, put in order by the eigenvalue that obtains, for medical research provides detailed research data, also be that electrocardiograph develops into the important foundation of home health care equipment in the future.
The electrocardiogram auto-check system is by handling electrocardiosignal, extract eigenvalue, and then disconnected conclusion, be that the doctor is an object with the crowd that the patient maybe may become the patient in clinical treatment work, the eigenvalue that reflects on electrocardiogram, diagnostic result are that the system summary of the disease cognitive process of content is summed up not the still information source of clinical medicine scientific research, also provide evidence for medical research.In the system patient's ECG data information with and objectively represented various Characteristics of electrocardiogram under specific disease or the various disease state by the eigenvalue that analysis obtains, the clinical diagnosis foundation that these are clear and definite, the ecg diagnostic criteria that helps some diseases (as myocardial infarction, cardiac hypertrophy etc.), and provide clinical information for inquiring into or estimating Electrocardiographic some parameter index.Simultaneously, the medical research personnel also can understand the effective information that the electrocardiogram detection technique can provide the diagnosis of some disease thus really, provide theoretical and support for electrocardiograph in the future develops into home health care equipment.
Early stage ECG analyzes and is finished by the doctor, this process waste time and energy and reliability not high.Computer assisted ECG analytical system starts from the end of the fifties.Dull, time-consuming, multiple artificial identification is replaced by computer generation, greatly reduces operator's omission situation that mental fatigue causes that works long hours.Along with the microcomputer function constantly strengthens and the development of modern microelectric technique, the function of ECG auto-check system strengthens greatly, has been widely used in the functional check of heart, the diagnosis and the prevention of cardiovascular disease, and many-side such as cardiac monitoring.
A normal ecg wave form is made up of P ripple, QRS wave group, T ripple etc.For concrete each ripple, all corresponding cardiomotility and electrophysiological moment.Under normal circumstances, these ripples periodically repeated according to the cycle of the excited pulse wave of sinuatrial node generation.If pathological changes takes place for the conducting system generation obstacle of heart or certain part cardiac muscle,, show as the ANOMALOUS VARIATIONS and the carrying out property evolution process of each waveform just the variation of electrocardio-activity can correctly be reflected on the electrocardiogram in time.Therefore each waveform that correctly detects in the ECG signal just becomes the basis of the ECG signal being carried out correct analyzing and processing.
By equipment such as computers the electrocardiosignal that collects is analyzed, be widely used in the functional check of heart, the diagnosis and the prevention of cardiovascular disease, and many-side such as cardiac monitoring.In the processing of electrocardiosignal, the most primary and crucial problem is exactly the detection of electrocardiosignal.The electrocardiograph signal detection method of the current overwhelming majority mainly was divided into for two steps: at first electrocardiosignal is carried out Filtering Processing, the main noise in the filtered signal (base floats, power frequency electricity, myoelectricity, apparatus displacement etc.) is strengthened the QRS wave group; Take certain criterion to determine threshold value then, detect required information.Will be a very loaded down with trivial details job with the multiple noise filtering in the electrocardiosignal, and this will make conversion speed reduce, operand increases.
Carry out because the interpretation of electrocardiogram output waveform still needs empirical expert, and their experience and level have directly had influence on the correctness of exporting interpretation of result, interpretation.In order to address this problem, occur adopting the output of the comprehensive and analysis Instrument equipment of expert system technology, obtained hat disease diagnoses and treatment specialist system, spectral electrocardiogram analysis expert system, the chronic pulmonary heart disease computer diagnosis expert system of the diagnostic result of auxiliary meaning.Set up an electrocardiographic diagnosis specialist system, the knowledge engineer will by with medical expert's opinion, obtain Professional knowledge about medical diagnostic techniqu, further summarize again, form notion and set up various relations between instrument output and diagnostic result, the experience that is about to the medical expert is carried out branch's classification by the logical reasoning relation, the flow chart of establishment analyzing and diagnosing, it is this according to medical knowledge and statistics to find the solution the inference machine of problem with the computer Implementation of Expert System then, imitation doctor thinking reasoning process, adopt linguistic variable, the system that structure meets the function of sight situation makes judgement meticulousr and be difficult for makeing mistakes.But not possessing the self study merit, specialist system can't from the process of continuous diagnosis, not carry out oneself's summary of experience.
Along with the appearance and the development of analysis of neural network algorithm, the multiple nerual network technique electrocardiogram classification of using has appearred both at home and abroad, carry out the research of ecg analysis.Wang Jicheng etc. have worked out the electrograph categorizing system based on symbolic neural network, the multistage neural network structure of H.Gholam Hosseini people such as (2001) research can electrocardiogram in ARR discrimination reach 88.3%.The rate of accuracy reached to 89.4% based on the myocardial infarction auto-check system of neural network and fuzzy logic of HL lug people such as (2000) research.Nerual network technique and tame system are different.Neutral net is formed by arranging stratified processing unit, do not need to design any mathematics type, only learn, very strong adaptive capacity is arranged and be convenient to realization on the hardware, can improve the speed of electrocardiographic diagnosis system greatly having noise or damaged input information by the experience of passing by.Specialist system lacked and these are united just.But compare with specialist system, neutral net also has some obvious defects, the study of neutral net and problem solving have the "black box" characteristic, and its solving result does not have interpretability, and " explanation " work is very important for the electrocardiographic diagnosis system.Be with the Electrocardiographic major issue of neutral net interpretation whether abundant reliable training data is arranged, be that data will be represented enough experiences, but in the document in past, selected training sample very little, the network that trains in this case can not clinical use because network not association handle the whole issue that may run in the clinical practice.And the learning cycle of neutral net is long, lacks effectively to append learning capacity, can cause temporal waste undoubtedly.Neutral net and traditional specialist system respectively have strong point and the deficiency of oneself, and a side strong point often is again the opposing party's a weakness, expert both domestic and external has begun one's study the two has organically been combined, and collaborative work reaches the purpose of " learning from other's strong points to offset one's weaknesses ".
The invention of Xue J.Q of GE, P.P Ai Erke " being used for analysis and editing ECG morphology and seasonal effect in time series method and apparatus " (referring to list of references [1]), on April 17th, 2007 applied for a patent to China national Department of Intellectual Property, open on October 24th, 2007, publication number is: publication number CN101057781A.A kind of method and apparatus of QT interval characteristics of the data that are used for the analysing ECG signal, this ECG signal data have a continuous wave that is produced by heartthrob.The ECG signal data obtains from the patient.Determine the R-R interval and the QT interval of the waveform of ECG signal data.The waveform of choosing the ECG signal data with stable heart-rate is used for determining the QT interval characteristics.Preferably, selected waveform is the waveform with minimum R-R separation standard deviation and minimum R-R interval deviation.Correct from calculating in the mode of front or according to the ECG signal data waveform of clinician's edit selection.The R-R interval, QT interval and the QTc that show the heartbeat of selected waveform for the purpose of analyzing and diagnose.The present invention also can be applied to obtain and show other cardiac datas from the ECG waveform with simulated mode.
In order to determine QT at interval and among the QTc, the method and apparatus of this inventive embodiments obtained the ECG signal data and determined in these data waveform R-R at interval and the QT interval.Selection demonstrates the part of the ECG signal data of relatively stable heart rate.Can be according to the interval between the R feature of ECG waveform, promptly R-R at interval, difficult this stable heart rate waveform of selecting of standard deviation.The R-R that also can use the waveform maximum at interval and the deviation of minimum R-R between at interval.
Particularly, the waveform of selected ECG signal data is those waveforms with minimum R-R separation standard deviation and minimum R-R interval deviation, can help to carry out the artificial selection by showing, calculate QTc according to selected ECG signal data with graphical association waveform R-R interval and QT novelty at interval.After this, to each selected heartbeat demonstration R-R interval of ECG signal data, QT interval and QTc diagnose to analyze the Q-T characteristic according to it.Although the ECG signal data has been described as long QT is analyzed at interval, but predictor as TdP, should be appreciated that also this analysis to be used for determining unusual short QT existence at interval, short QT is also relevant with life-threatening arrhythmia at interval.More generally, the technology of this invention can be used for obtaining and the video data group, this data set compares two or more situations relevant with heart for the useful mode of diagnostic purpose.
This scheme is obtained the ECG signal data by the traditional approach that click is added to person under inspection's health, and in the equipment of this invention, the signal of telecommunication in the electrode is exaggerated in preamplifier and is the analog form that comprises a series of continuous waves.Make this analogue signal in analog/digital converter, carry out analog digital conversion and be stored in the bin of computer, the CPU that its operation utilization comprises the algorithm of carrying out the inventive method is carried out, and computer comprises the screen that is used for showing with figure or written form information.
The method of this invention is obtained the ECG signal data of a series of continuous wave forms after beginning, or it is from person under inspection's reception, perhaps more typically comes from the data in the memorizer that is stored in computer.This takes place in step 42, this data show that obtains like this, for about 10 seconds typical sampling period, it comprises two helical pitches of ECG signal data, with lowest standard deviation STD at interval and R-R lowest difference at interval from waveform.Lowest standard deviation means that the signal data R-R gap size of selected portion centers on central value and closely assembles.Minimum deviation means that the difference between minimum and maximum is minimum.R-R at interval standard deviation and R-R at interval maximum-minimum deviation determine and have minimum R-R separation standard deviation and minimum R-R carry out in the step 60 that is chosen in method and 62 of a part of ECG signal data of deviation at interval.The clinician can be according to the ECG signal data waveform human-edited ECG signal data that will use.That part of ECG data that will use in the further step of this method (selecting and/or artificial selection in step 66 by the criterion of step 60 and 62) are set up and demonstration in step 70 in step 68.In step 72, to each definite acquisition of selecting waveform calculating QT interval or this QT from previous step 44, to carry out at interval of signal data.In step 74, determine QTc for the waveform of the ECG signal data of selected portion, QTc is illustrated in the effect of the heart rate on the QT interval.Usually a plurality of formula are used for this purpose.It comprises the Bazett formula, Friderica formula and equation of linear regression.This equation of linear regression usually is known as the Framingham formula.Carry out the calculating of QTc for each cardiac pulse waveform in the ECG signal data of selected portion.In step 76, for the selected cardiac pulse waveform of ECG signal data, with data show.For ECG characteristic time sequence analysis, the result of data show is a graph of a relation of clapping.This graph of a relation can utilize the R-R interval mode identical with the Q-T compartment analysis to use with combination.
This algorithm only is used for the QT characteristic of analysing ECG signal and calculates its R-R interval, QT interval and QTc, the key character that this is obtained is provided in the clinician then, auxiliary doctor is according to relevant feature, but this algorithm can not independently detect and judge whether occur unusually in the ECG signal, can not provide whether the ECG signal is comprised the preliminary judgement of heart change information and for the abnormality detection of ECG signal.
The invention of Zhao Yujing and Jin Yunyu " mobile diagnosis device ", on March 16th, 2006 applied for a patent and gets the Green Light to China national Department of Intellectual Property, and open on April 30th, 2008, publication number is: CN101170944A.
This invention relates to a kind of mobile diagnosis device, comprise the ECG unit that is used to write down the ECG signal, described ECG unit is connected in maybe can be connected in a plurality of ECG electrodes, also comprise the sphygmometer unit, be used for recording volume pulse signal simultaneously, described sphygmometer unit comprises that at least one light source and at least one are used for measuring with optical mode the optical sensor of blood concern of the vascular system of patient body tissue, also comprise programme controlled assessment unit, be used for analysing ECG signal and volume pulse signal.According to the present invention, this assessment unit is applicable to the extreme value in the identification volume pulse signal of son, and determines R peak value in the ECG signal and the time difference between the next extreme value in the volume pulse.In addition, assessment unit can be presented as the main peak value determined in the volume pulse signal and the time difference between the secondary peak value.
ECG is used for the most frequently used inspection form of diagnosis of cardiovascular diseases.By means of the ECG device, derived with two or more a plurality of ECG electrode from the signal of telecommunication of the patient's that will check health.The bio-voltage that takes place during obtaining expansion that ECG reflected heart like this and shrinking.ECG include many can be by the parameter of diagnostic assessment, it also is called the R peak value.In addition, ECG includes the so-called P ripple before the R peak value.R peak value back is so-called T ripple.Before the R peak value and afterwards the minima of moment is represented by Q and S respectively.Those are the persistent period of P ripple and amplitude, PQ persistent period, the persistent period of QRS complex wave, the persistent period at QT interval and the amplitude of T ripple at interval of P ripple for the relevant parameter of cardiovascular diagnosis.Can draw conclusion by the absolute value of described parameter and by the ratio of these parameters about the cardiovascular system health status.
It is feasible catching and write down peripheral cardio-vascular parameters by so-called plethysmography.In the volume descriptive method, measure the volume fluctuation that is subjected to the blood flow influence in the external perihaemal canal.Recently, near-infrared photograph volume descriptive method has obtained application.Wherein applied diagnostic form is called sphygmometer for short.This sphygmometer generally comprises 2 light sources, and they are to the HONGGUANG or the infrared ray of patient's tissue emission different wave length.The light of emission is organized inscattering and is partially absorbed at patient body.Light sensors scattered light by means of suitable light cell form.On the other hand, commercial available sphygmometer generally uses the light in the 660nm wave-length coverage.In this scope, very big different of the light absorption of oxyhemoglobin and deoxyhemoglobin.Thereby by means of the scattered intensity of light sensors respectively according to changing for altogether the oxygen enrichment of the systemic blood of being checked and the degree of oxygen deprivation.On the other hand, use the interior light of 810nm wave-length coverage usually.This wavelength is in the scope that is called as the near-infrared line spectrum.Equal substantially in the light absorption of oxyhemoglobin and deoxyhemoglobin in this spectral region.The sphygmometer of prior art can produce the volume pulse signal, this signal reflected between heart beat period, can change and can produce the volume pulse signal by sphygmometer, this signal has reflected the volume of the blood that can change and can be passed through by the blood capillary system that sphygmometer is caught between heart beat period.When the different optical wavelength used in above-mentioned spectral region, can draw the conclusion of the oxygen content that is used for assessing blood by different light absorption.Common sphygmometer uses on patient's finger tip or on ear-lobe.This moment, the hemoperfusion from the interior blood capillary system in these zones of tissue produced the volume pulse signal.
The combination of known ECG and sphygmometer device allows at definite a plurality of cardio-vascular parameters.According to these data, the clinician can carry out comprehensive cardiovascular diagnosis.But the shortcoming of one type of prior art syringe is, it can not set up the tentative diagnosis of imminent or the cardiovascular disease that existed automatically.For this reason, one type of prior art syringe can not be easy to use by patient, can not automatic diagnosis.At this background, the purpose of this invention is, a kind of allow the to carry out state of cardiovascular system and the diagnostic equipment of trend diagnosis are provided.This device should be able to be indicated the early stage automatic diagnosis of cardiovascular disease to patient, and patient not proposed too much requirement according to a plurality of diagnosis coefficients of assessment.
This invention has solved this task according to the mobile diagnosis device of a kind of the above-mentioned type and characteristic, and wherein said assessment unit is arranged to 1, detects the R peak value in the ECG signal automatically; 2, the extreme value in the automatic detection volume pulse signal; 3 and determine the R peak value in the ECG signal and the time difference of the next extreme value support in the volume pulse signal.
The combination of ECG signal and volume pulse signal in the combined diagnosis device that is made of ECG unit and sphygmometer unit allows can discern R peak value in the ECG signal automatically to the assessment unit of the cardiovascular system diagnostic equipment.Thereby, can determine definite moment of heart beating automatically.In addition, based on programme-control of the present invention, assessment unit can be discerned the extreme value in the volume pulse signal, i.e. maximum or minima.According to the extreme value in the volume pulse signal, the due in by the extensive pulsating wave of heart beating is determined in the peripheral measuring position that can cover in the sphygmometer unit.Thereby finally can determine the R peak value of ECG signal and the next extreme value support in the volume impulse level with time correlation be the tolerance that is called as pulsating wave speed at interval.According to the pulsation wave propagation velocity, can carry out description about blood pressure.Because the reduction of pulsating wave speed is accompanied by the rising of blood pressure, and the reduction of blood pressure is represented in the rising of pulsating wave speed.But according to pulsating wave speed accurately estimated blood pressure be impossible, can only represent trend.In addition, pulsating wave speed is relevant with density of blood, and particularly relevant with the elasticity of blood vessel wall.Elasticity by blood vessel wall can draw again about there being arteriosclerotic conclusion.Pulsating wave speed also has the internal diameter of pulsation relevant.Therefore build constant elasticity and constant density of blood, can characterize the supply of blood at the relevant position of medical examination.By means of making up ECG signal and volume pulse signal in the assessment automatically, the diagnostic equipment of this invention can be carried out the functional assessment of patient's blood vessel system automatically.According to the signal of being assessed automatically, the diagnostic equipment of this invention is the section of bat patient's cardiovascular status roughly, and when arteriosclerotic any indicative sign, patient is produced enough caution signalss.Thereby patient can use the diagnostic equipment of the present invention to realize diagnosis voluntarily.Need be by the differential evaluation that in most of the cases patient is proposed the definite many cardio-vascular parameters of demanding device.Assessment unit also suitably is equipped with, its permission: determine the oxygen saturation of blood by the volume pulse signal, determine chamber property heart rate, determine that by the volume pulse signal volume describes heart rate by the ECG signal.
Blood oxygen concentration, chamber property heart rate, volume describe heart rate determine make it possible to cardiovascular system is carried out further more meticulous condition diagnosing.Utilize the diagnostic equipment of the present invention, can determine the absolute value of heart rate, the transmutability of heart rate and the corresponding arrhythmia of heart automatically.In this way, can determine arrhythmia, as sinus tachycardia, sinus bradycardia, hole cardiac arrest and so-called heart beating escape beat.According to the ECG signal, can also be described persistent period about atrial systole and persistent period time correlation, ventricular systole and persistent period time correlation and the ventricular diastole of a heart beating of heart etc.In addition, can also set up in the so-called heart electric stimulating holding wire obstruction and about the diagnosis in advance of disturbance of circulation or infarction.Can detect other scramblings in the pulse process according to the volume pulse signal.Blood oxygenation is also represented an important parameter in the condition diagnosing of cardiovascular system.Can draw efficient and adaptive conclusion according to blood oxygenation about cardiovascular system.
Advantageously, the assessment unit of the diagnostic equipment of this invention further suitably is equipped with, and makes it possible to discern automatically between main peak value and the secondary peak value in the volume pulse signal, the amplitude of determining main peak value and secondary peak value and definite main peak value and the minor peaks and interval time correlation.This makes it possible to the dicrotism of automatic check volume pulse signal.Dicrotism is that the stack by the recurrence pulse that begins progressive pulse from heart and formed by the pulse of the less vasoreflex of a part of blood vessel or elasticity causes.Know, elastic reduce and thereby the increase of arteriosclerosis degree make main peak value in plethysmographic signal and the interval between the secondary peak value diminish with time correlation, the second peaked intensity diminishes simultaneously.By determining the interval between main peak value and secondary peak value and determining the main peak value and the relative amplitude of secondary peak value, make it possible to obtain other important parameter, these parameters can be used to the diagnostic equipment of the present invention, so that detect arteriosclerotic sign automatically.In the volume pulse signal and second maximum mainly to be reflection by the pulse under the low extreme value cause.Therefore, between main peak value and secondary peak value mainly determine by aortal characteristic with interval time correlation.Thereby when relevant measurement point is determined first pulse wave velocity, can determine second pulse wave velocity, i.e. the speed of the pulse in aorta.Thereby, as illustrating in the above, utilize the diagnostic equipment of the present invention can advantageously determine two different pulse wave velocities and assess these speed so that diagnose the illness.The diagnostic equipment of this invention also comprises the pick off of the body temperature, ambient temperature or the air humidity that are used to measure patient easily.These parameters are even more important for the pulses measure unit of the diagnostic equipment of calibration ECG unit and this invention.Be used to be recorded in measure by means of assessment unit during determined parameter, date or time that storage is simultaneously measured.By means of memory element, can follow the tracks of and write down the process and the corresponding treatment effect of the disease of cardiovascular system on the one hand.On the other hand, the data that are stored in the memory element of the diagnostic equipment can be read out and be analyzed by the clinician, so that the doctor carries out detailed condition diagnosing to cardiovascular system.Easily, the diagnostic equipment of the present invention also comprises data transmission interface, be used for the transfer of data of the memory element that is stored in the diagnostic equipment to doctor's personal computer, this interface can be use always or wave point.
In addition, what meet purpose is, the diagnostic equipment of this invention comprises and suitably is equipped with the diagnosis unit of state that makes it possible to determine by means of the determined parameter of assessment unit patient's cardiovascular system.Thereby this diagnostic equipment has modular structure.Apparatus for evaluating only is responsible for the signal that assessment is caught, so that be identified for diagnosing in the above described manner needed those parameters.These parameters are handled by the diagnosis unit of the diagnostic equipment then, so that therefrom obtain the conclusion about the cardiovascular system state.Diagnosis unit also is responsible for the arteriosclerotic existence of identification automatically, and also patient is produced corresponding caution signals if necessary.
Advantageously, the diagnosis unit of the diagnostic equipment is further made local the outfit, makes it possible to the relevant trend of change of state by and patient's definite by means of parameter and change time correlation of cell stores cardiovascular system.In some cases, can not be by the conclusion that directly draws by the determined parameter of the assessment unit of the diagnostic equipment about the patient disease.But, the change of these parameters, for example continuing to increase of pulse wave velocity can be used to refer to the cardiovascular disease in the early-stage development.If patient is long this diagnostic equipment of time durations use repeatedly, by means of the parameter that the cell stores of the diagnostic equipment is determined automatically by assessment unit, this trend can be used for the diagnosis voluntarily of disease.The described diagnostic equipment comprises display unit, is used to show ECG signal, volume pulse signal and by means of the determined parameter of assessment unit.Use the patient and the clinician of the diagnostic equipment from display unit, conveniently to read all values, simultaneously, can check whether the diagnostic equipment correctly works.
This technical scheme, convenient and practical, determine the relevant key character of ECG signal and volume pulse signal by detecting ECG signal and volume pulse signal, the key character parameter value of cardiovascular system health status assessment needs is provided for the doctor.But, this invention mainly is to the monitoring of ECG signal and extracts some characteristic parameters, but lack analysis and the assessment of patient's heart being carried out intelligence according to these important ECG features, thereby further alleviate the function of doctor's diagnostic work, this equipment mainly is to concentrate on monitoring, and lacks important analysis and judgement.
Qiao Er Xue's .Q. of GE Medical Systems Information Technologies Inc. invention " method and apparatus that is used for the alternate data of definite ECG signal " (referring to list of references [3]), on April 13rd, 2005 applied for a patent and gets the Green Light to China national Department of Intellectual Property, open on October 19th, 2005, publication number is: CN 1682654A.The method and apparatus that is used for the alternate data of definite ECG signal.Described method can comprise: at least one value of at least one morphological characteristic that is identified for representing each pulsation of ECG signal; And produce data point set according to the total quantity of value and the total quantity of pulsation.Described data point can each comprise use determined first value of first mathematical function and use determined second value of second mathematical function.Described method can comprise that also several Preprocessing Algorithm improve signal to noise ratio.Described method comprises: data point is separated into first group of point and second group of point, and produces characteristic pattern by drawing first group of point and second group of point, so that estimate the numbering alternate mode.Can with statistical test come analytical characteristic figure, to determine in group and effective difference of bunch support.This invention relates to cardiology, and specific design is used for determining the alternative method and apparatus of ECG signal.It alternately is the variation between the meticulous pulsation in the repeating transmission pattern of ECG signal.Some researchs verified the individual for the high correlation between the sensitivity of ventricle arrhythmia and sudden cardiac death and the T ripple alternating pattern that in individual's ECG signal, has variation.Though the ECG signal normally has the amplitude that millivolt is measured, having a little, the variation alternating pattern of level amplitude may be important clinically.Therefore it is too little usually and can not be detected by visual the checking with its typical log resolution of ECG signal to change alternately group.On the contrary, the Digital Signal Processing of the alternating pattern of variation and quantification are necessary.The such signal processing and the quantification of the alternating pattern that changes, owing to exist the alignment precision of noise and variation or the restriction of physiological change to cause, but complicated.The current demand signal treatment technology that is used for detecting at the TWA pattern of the variation of ECG signal comprises spectral domain method and time domain approach.In view of above-mentioned, need a kind of technology that is used for detecting the TWA pattern that changes at the ECG signal, described technology provides the performance of improving, as independent technology with as the adnexa for other technologies.Therefore one or more embodiment of the present invention provides the method and apparatus of the alternate data that is used for definite ECG signal.In certain embodiments, described method can comprise: be identified for representing at least one value of at least one morphological characteristic of each pulsation of ECG signal, and produce one group of data point set according to the total quantity of value and the total quantity of pulsation.Described data point can each comprise use determined first value of first mathematical function and use second mathematical function to determine second value.Described method also can comprise: data point is separated into first group of point and second group of point, and by drawing the generation characteristic pattern of first group of point and second group of point, so that estimate to change alternative pattern.A kind of method that is used for determining the alternate data of ECG signal, described method comprises: at least one value of at least one morphological characteristic that is identified for representing each pulsation of ECG signal; Produce data point set according to the total amount of value and the total quantity of pulsation, each comprises described data point use determined first value of first mathematical function and uses determined second value of second mathematical function; Data point is separated into first group of point and second group of point; And produce characteristic pattern by drawing first group of point and second group of point, so that estimate to change alternate mode, determine the estimation amplitude of variation alternate mode simultaneously.The first nodal point by specified data point and second central point and determine distance between the first nodal point and second central point are determined the range value of estimating.Produce eigenmatrix according to the total quantity of value and the total quantity of pulsation, and use principal component analysis to come processing feature to put to the proof, described principal component analysis produces principal component vector sum principal component, described data point corresponding to the principal component vector at least one.Use characteristic figure visually determines whether to exist the variation alternate mode.Produce second data point set, described second data point set comprises use determined the 3rd value of the 3rd mathematical function and uses determined the 4th value of the 4th mathematical function, and second data point set is separated into the 3rd group of point and the 4th group of point.And by draw the 3rd group the point and the 4th group of point produce second characteristic pattern.Produce the 3rd data point set, described the 3rd data point set comprises the 5th value that use the 5th mathematical function lock is determined and uses determined the 6th value of the 6th mathematical function, the 3rd data point set is separated into the 5th group of point and the 6th group of point, and produces the 3rd characteristic pattern by drawing the 5th group of point and the 6th group of point.Use and bunch to come the analyzing and processing data point, analyze for described bunch and produce first bunch of point and second bunch of point, compare first bunch of point and first group of point, relatively second bunch of point and second group of point, and the value that is used to be illustrated in the match point between match point between first bunch of point and the first group of point and second bunch of point and the second group of point.
This mainly is that alternate data to the ECG signal detects, and comes to provide relevant alternate data information for doctor's clinical diagnosis heart disease by the detection for ECG signal alternate data.But this invention only provides for the detection method of alternate data and device, and do not provide pathology intellectual analysis and tentative diagnosis based on the ECG alternate data, thereby, and can find heart of patient to have potentially dangerous timely and the prevention and the processing of being correlated with for the doctor provides further diagnostic message alleviating doctor's burden.
Summary of the invention
Technical problem to be solved by this invention is how a kind of ECG method for detecting abnormality that decomposes based on inherent trend subsequence pattern is provided, and this detection method can further improve ECG and detect accuracy rate and diagnostic reliability, has overcome the defective of prior art.
Technical problem proposed by the invention is to solve like this: a kind of ECG method for detecting abnormality that decomposes based on inherent trend subsequence pattern is provided, it is characterized in that, may further comprise the steps:
1. definition:
ECG signal sequence: the data set that the electrocardiogram element is arranged according to time sequencing;
Inherent trend subsequence pattern: given ECG signal sequence, if the variation tendency of all subsequences of certain subsequence is identical with its variation tendency, and in given ECG signal sequence, do not exist the subsequence identical with its trend to comprise it, this subsequence is called as inherent trend subsequence pattern;
Layer: in the ECG signal sequence, the inherent trend subsequence pattern with similar Trend value is formed a layer;
Sequence is decomposed: the ECG signal sequence is decomposed into some inherent trend subsequence patterns, and the inherent trend subsequence with similar Trend value forms corresponding layer;
2. normal ECG signal sequence being carried out sequence decomposes and layering:
The ECG signal sequence is decomposed into an inherent trend subsequence set of patterns, then to its inherent trend subsequence pattern according to its variation tendency parameter value layering, the subsequence that the variation tendency parameter value is close is divided into one deck, and its trend range of parameter values of every layer of usefulness is sectioned out;
3. doubtful ECG signal sequence being carried out sequence decomposes and layering:
Doubtful ECG signal is carried out sequence decompose the inherent trend subsequence pattern that also obtains, the back to its inherent trend subsequence pattern according to its variation tendency parameter value layering, the subsequence that the variation tendency parameter value is close is divided into one deck, and its trend range of parameter values of every layer of usefulness is sectioned out;
4. matching detection:
Respectively according to the trend range of parameter values of each layer, the inherent trend pattern subsequence pattern of above-mentioned both equivalent layers is mated, if the trend parameter value of doubtful inherent trend subsequence outside the trend range of parameter values of normal-sub sequence, this doubtful ECG abnormal signal then; If within scope, then calculate the distance of all the normal-sub sequences in this subsequence and the equivalent layer, get minima and be its distance, if the distance of this subsequence and its respective layer is greater than the threshold value of presetting with respective layer, then this subsequence is unusual subsequence, i.e. this doubtful ECG abnormal signal.
According to the ECG method for detecting abnormality that decomposes based on inherent trend subsequence pattern provided by the present invention, it is characterized in that the excavation step of described inherent trend subsequence is as follows:
1. find out all maximums and minima in the ECG signal sequence, and the position of their correspondences in the ECG signal sequence;
2. in the ECG signal sequence with maximum and minima as the subsequence of starting point or terminal point as inherent trend subsequence pattern.
According to the ECG method for detecting abnormality that decomposes based on inherent trend subsequence pattern provided by the present invention, it is characterized in that the mining algorithm step of described inherent trend subsequence is as follows:
1. definition:
Subsequence trend: in ECG signal sequence T, S={s
1... s
m, being the inherent trend subsequence pattern among the T, the trend Trend of S (S) is (s
1-s
m)/(1-m);
Maximum MAX: at sequence T={t
1... t
n, t
iIf ∈ T is t
I-1<t
iAnd t
iT
I+1, t then
iBe a maximum MAX among the sequence T, and all maximums among the sequence T are formed maximum S set MAX, SMAX={e|e is a maximum among the sequence T };
Minimum value MIN: at sequence T={t
1... t
n, t
iIf ∈ T is t
I-1T
iAnd t
i<t
I+1, t then
iBe a minimum value MIN among the sequence T, and all among the sequence T are worth most forms minima S set MIN, SMIN={e|e is a minima among the sequence T };
Stretching distance: subsequence Q={q
1..., q
Mq, C={c
1..., c
Mc, and mq≤mc, Q
s={ q
s 1, q
s 2..., q
s sFor Q is the stretching sequence of coefficient with mq/s,
Q, the stretching distance d of C
u(Q C) is:
R is a range parameter of determining extreme point, and D is the range parameter of extreme value each data point difference in the R scope;
2. algorithm: list entries T, range parameter R and difference parameter D
Find?Intrinsic?Subsequences(T,R,D)
5) maximum and minimizing set MM are empty, and element position set Loc is empty among the MM, and ITS is empty, and TR is empty;
6) if t
i∈ T, t
iBe set { t
I-R..., t
I+RMiddle maximum or minima, and in set, exist element and its poor absolute value more than or equal to D, then with t
iJoin among the set MM, and i is joined among the Loc;
7) each element of pair set T repeats above 1 successively), 2);
8) each adjacent maximum and the subsequence between the minimum point are inherent trend subsequence pattern, and all inherent trend subsequence patterns are added ITS, and the Trend value that it is corresponding adds TR.
3. export inherent trend subsequence set of modes ITS and its corresponding trend set TR.
According to the ECG method for detecting abnormality that decomposes based on inherent trend subsequence pattern provided by the present invention, it is characterized in that the abnormality detection algorithm steps is as follows:
Input: the inherent trend subsequence set of patterns TITS of the inherent trend subsequence set of patterns NITS of normal sequence and doubtful sequence
Output: unusual subsequence set
Anormaly?Detection(NITS,TITS)
4) the inherent trend subsequence pattern layering of normal ECG signal sequence set Nlayer is empty;
5) with the subsequence among the NITS according to its Trend value layering, the approaching sequence of Trend value is formed one deck, and layering is joined among the set Nlayer;
6) S ∈ TITS if there be not the layer not approaching with the Trend value of S among the Nlayer, then joins AS with S; If the layer approaching with the Trend value of S arranged, then calculate the stretching distance of all subsequences of S and this layer, and stretching distance obtained average distance divided by the length of S, being averaged minimum value and value is the distance D that S arrives this layer, if D〉η, then S is unusual subsequence, and S is joined AS, otherwise, forward 4 to);
7) all subsequences of pair set TITS repeat 3), return unusual subsequence set AS at last.
The present invention is for the abnormal signal in can identification electrocardiogram more accurately and efficiently, help the various pathological changes of expert diagnosis heart, the algorithm that proposition is decomposed based on inherent trend subsequence mode sequences detects unusual in the ECG signal, thereby provides diagnostic recommendations for expert diagnosis.This programme has proposed the notion of inherent trend subsequence pattern according to the feature of ECG signal, inherent trend pattern subsequence is in a sequence, it comprises the maximum subsequence of the subsequence that trend is identical with it, it is to occur with the constant integral body of trend sequence as can be seen from the definition of natural mode, has very stable tendency, inherent trend subsequence pattern also has practical significance in the ECG signal, because Electrocardiographic reacting condition the activity of heart, in the time of under heart is in certain activity, electrocardiogram has certain steady change trend, take care after the moving change of dirty work, electrocardiogram then reflects the trend of another steady change, and the sequence of this stable tendency has then met in this programme the definition for inherent trend subsequence pattern.
The ECG signal sequence that the abnormality detection of ECG signal obtains body surface is decomposed into different layers, each layer is made up of the close inherent trend subsequence pattern of trend, by the difference of doubtful ECG sequence and normal ECG sequence layering comparison inherent trend subsequence pattern is judged whether doubtful ECG sequence is unusual.Because the ECG signal has reflected the situation of change of cardiomotility, when heart is in certain operational phase, its ECG signal presents certain stable trend in this range of activity inside, when heart is in another kind of active node, then its ECG signal trend can change, and changes that stable tendency of another scope into.Because the continous-stable trend characteristic of ECG signal when certain heart is in certain range of activity, the ECG sequence of its generation has more consistent tendency.When carrying out the decomposition of inherent trend subsequence pattern, this sequence has consistent variation tendency, so it can reflect the cardiomotility of certain phase.In contrast be that noise in the ECG sequence is mixed and disorderly unordered, it is the overall variation trend that can not influence signal in decomposing carrying out inherent trend subsequence pattern, so noise is very big may be broken down into each different layer.Noise is because of decomposing weakened unusual ECG signal because it has reflected the operational phase of heart abnormality, and its variation tendency is strong, and majority can be broken down into the subsequence with certain trend, and therefore it can be not weakened in the process of decomposing.After the ECG sequence is decomposed, can at first find whether occurred unusual trend in the test data, then the difference of the inherent trend pattern subsequence under comparing same trend or close trend (promptly with the subsequence in the layer).So at first can find tangible abnormal heart activity, in addition, only compare the subsequence in the identical layer, dwindle relatively space, thereby improve efficient and accuracy rate that the ECG abnormal signal detects.
The specific embodiment
The present invention is further described below in conjunction with embodiment.
Heart is sanguimotor power source, just as an electromotor that never stops, being accompanied by human all one's life.Heart mainly is made up of myocardial cell, and any activity of myocardial cell all is accompanied by the variation of physiology current potential, and this variation is known as cardiac electrical activity, is called for short electrocardio.The mechanical movement of the bio electricity process of heart and the Biochemical processes of heart tissue, heart and the nervous system of relevant controlling cardiomotility have confidential relation.Cardiac electrical activity passes to body surface through tissue, makes each position of body surface at each cardiac cycle potential change clocklike take place all also.Measurement electrode is placed on body surface, and the continuous change curve that the potential change that body surface took place shows after amplifying or traces is electrocardiogram, is called for short ECG, and electrocardiogram is comprising abundant cardiac rhythm information and physiology, pathological information.
Nowadays, electrocardiogram (ECG) check oneself through be clinically with health examination in one of four big routine examinations.Because ECG checks harmless, and is easy and quick, therefore obtained extensive use clinically, become a kind of most important means that cardiovascular disease is diagnosed.Development along with technology such as computer and artificial intelligences, the ECG automatic analysis technology is also constantly ripe, but the computer automatic analysis of ECG, diagnostic system the popularity of using and acceptable aspect also do not reach gratifying degree, the ratio shared in the ecg analysis field is especially little.Therefore, Chinese scholars is all very active all the time to the research of electrocardiographic wave detection and Identification algorithm, further improves the research that EGC detects accuracy rate and diagnostic reliability, still has very big realistic meaning.
This programme proposes to detect unusual subsequence based on the algorithm that inherent trend subsequence pattern is decomposed, thereby detects unusual cardiomotility, for the clinician provides reference information according to the substitutive characteristics of ECG signal sequence.Given sequence T if the variation tendency of all subsequences of certain subsequence is identical with its variation tendency, and in sequence T, does not exist the subsequence identical with its trend to comprise it, and then this subsequence is called as inherent trend subsequence pattern.The definition of inherent trend subsequence pattern has actual implication, and when heart is under a certain activity, the subsequence that its ECG signal shows certain steady change tendency in this range of activity, this subsequence are exactly inherent trend subsequence pattern.Intrinsic subsequence set of patterns with similar variation tendency constitutes a layer.When the ECG abnormal signal detects, the ECG signal sequence at first is broken down into an inherent trend subsequence set of patterns, then to normal inherent trend subsequence pattern according to its variation tendency parameter value layering, the sequence that the variation tendency parameter value is close is divided into one deck, and its trend range of parameter values of every layer of usefulness is marked.Secondly, to the inherent trend subsequence pattern that doubtful ECG signal decomposition obtains, at first calculate its trend parameter value whether in the Trend value scope of normal-sub sequence, if outside scope, then this sequence is unusual sequence; If within scope, then calculate the distance of all the normal-sub sequences in this subsequence and the equivalent layer, get the distance of minima for itself and respective layer.If the distance of this subsequence and its respective layer is greater than a threshold value, then this subsequence is unusual subsequence.
Electrocardiogram (ECG) is the certain position that measurement electrode is placed on human body surface, the heart physiological electricity change curve that record comes out, the process that reflecting myocardium cells physiological current potential produces in conducting system of heart, conducts and recover, and all show as the waveform of surface electrocardiogram on the physiology potential change macroscopic view of myocardial cell.Cardiac muscle belongs to excitable tissue, produces potential change in process of excitation, promptly has bio-electric phenomenon and electrophysiological characteristics.The bio electricity of myocardial cell is divided into the action potential of resting potential and excitation time.Under quiescent condition, the current potential of myocardial surface each point equates to have only the inside and outside formed transmembrane potential of ion concentration difference of cell membrane.Predicted by microelectrode, resting potential is about 90mv, and shows as the outer positively charged of film, electronegative polarized state in the film.When myocardial cell is upset, under the effect of the permselectivity of potassium, sodium, calcium channel on ion concentration gradient, electric-force gradient, the cell membrane and active transport mechanism etc., by controlling the transfer of each ion inside and outside film, cause the distribution of exterior charging in the cell membrane to change, finish depolarization and process of repolarization, form action potential.
The electrocardiographic wave of record is different though respectively lead, and all comprises a P ripple basically, a QRS wave group and a T ripple.The implication of each ripple, wave band, interval is as follows:
(1) P ripple: the potential change of representing the left atrium process of depolarization.Starting point represents that right atrium begins depolarization, and on behalf of two atrium depolarizations, terminal point finish.P wave-wave shape is little and have circle blunt, and slight incisura can be arranged, and lasts .008~.011 second, and amplitude is no more than 0.25mv.
(2) PR section (PR Segment): it is one section equipotential line behind the P ripple, to QRS wave group starting point, represents excitement to conduct in atrioventricular junction, atrioventricular bundle and part bundle branch from P ripple terminal point.The composition that wherein contains atrial T wave (Ta ripple) reflects not obviously because of voltage is faint, be a horizontal line on electrocardiogram.
(3) PR interval (PR Interval): refer to that the interval of P ripple starting point between the QRS wave group starting point, representative begin time of beginning to sequences of ventricular depolarization from the atrium depolarization.Normal adult is 0.12-0.205, but can be variant between the individuality.The P-R interval, changed with heart rate and age, and the age is big more or heart rate is slow more, and its P-R interval is long more.
(4) QRS wave group (QRS Complex): the potential change of representing left and right two sequences of ventricular depolarization processes.In normal adult, the QRS wave group lasts 0.06-0.105, and is general many about 0.085.To the spectrum analysis of the electrocardiosignal center frequency-band (this frequency is also referred to as the characteristic frequency of QRS ripple) about 17Hz of QRS ripple as can be known, bandwidth is about 10Hz.Typical QRS complex wave comprises three ripples that closely link to each other, and first downward ripple is the Q ripple, the height that makes progress thereafter and point be the R ripple, a downward ripple after the R ripple is the S ripple.But in the electrocardiogram _ core of leading and being write down with difference, these three ripples differ and establish a capital appearance, and the variation of its waveform and amplitude is also bigger.
(5) ST section (ST Segment): one section horizontal line from the terminal point (J point) of QRS wave group to T ripple starting point.It represents the process of the slow multipole of ventricle, is limited to 0.05~0.155 when normal.Normal TS section shows as one section equipotential line, also has slight skew sometimes.
(6) T ripple: represent the potential change in the ventricle rapid repolarization process.The waveform circle is blunt, and lifting is propped up and be not exclusively symmetrical, and preceding of waveform is then propped up shorter than length.Normal T ripple the time be limited to 0.05-0.255, but the T wave amplitude is higher, its time limit is longer.The T wave line of propagation is consistent with the main ripple direction of QRS wave group, is in leading of master at the R ripple, and the voltage amplitude of T ripple should not be lower than with leading 1/10 of R wave amplitude.
(8) QT interval (QT Interval): the time-histories from QRS wave group starting point to T ripple terminal point, represent ventricular muscles depolarization and required altogether time of multipole overall process.Its length and heart rate have substantial connection, and heart rate is fast more, and the Q-T interval is short more.Heart rate is in the 60-100 timesharing, and be .032~0.445 normal range of Q-T interval.
The method for digging of inherent trend subsequence pattern is as follows:
Definition 1: subsequence trend: in sequence T, S={s
1... s
m, being the inherent trend subsequence pattern among the T, the trend Trend of S (S) is (s
1-s
m)/(1-m)
Definition 2: maximum MAX: at sequence T={t
1... t
n, t
iIf ∈ T is t
I-1<t
iAnd t
iT
I+1, t then
iBe a maximum MAX among the sequence T.And all maximums among the sequence T are formed maximum S set MAX, and SMAX={e|e is a maximum among the sequence T }.
Definition 3: minimum value MIN: at sequence T={t
1... t
n, t
iIf ∈ T is t
I-1T
iAnd t
i<t
I+1, t then
iIt is a minimum value MIN among the sequence T.And all among the sequence T are worth most forms minima S set MIN, and SMIN={e|e is a minima among the sequence T }.
Because inherent trend subsequence pattern has different length usually, therefore define the distance that the stretching distance function calculates two different length subsequences here.
Definition 4: stretching distance: subsequence Q={q
1..., q
Mq, C={c
1..., c
Mc, and mq≤mc, Q
s={ q
s 1, q
s 2..., q
s sFor Q is the stretching sequence of coefficient with mq/s,
, Q, the stretching distance d of C
u(Q C) is:
Adopt this method computed range mainly to be based on following two reasons: at first, inherent trend subsequence pattern has different length, in the process of decomposing, even identical inherent trend subsequence pattern also has different length.Secondly, exist certain noise data in the data of reality, noise data also can cause decomposing the inherent trend subsequence modal length that obtains and difference occur.Therefore the distance between the subsequence that needs to adopt stretching distance to calculate different length.In addition, before the computed range, this programme is at first to each inherent trend pattern subsequence normalized, because be nonsensical for the data comparison distance with different side-play amounts and amplitude.
This programme excavates inherent trend subsequence pattern according to following step: (1) find out maximums all among the sequence T and minima with and in T corresponding position; (2) find out among the sequence T with maximum and minima as the subsequence of starting point or terminal point as inherent trend subsequence pattern.
Owing in real ECG signal sequence, exist a large amount of noise datas, therefore exist some small fluctuations, therefore not maximum and the minima of finding out under the strict difinition in algorithm, but look within the specific limits maximum and minima, thereby the influence that can avoid minor fluctuations that overall trend is judged.
Inherent trend subsequence mode excavation algorithm is as follows: wherein R is for determining the range parameter of extreme point, D is the range parameter of extreme value each data point difference in the R scope, for each element among the sequence T, if it is in the R scope, with all other element differences less than D, then this element is maximum or minimum.Find after the maximum and minimum all among the T, the subsequence between adjacent maximum and the minimum then is an inherent trend subsequence pattern.
Input: sequence T, range parameter R and difference parameter D
Output: inherent trend subsequence set of modes ITS and its corresponding trend set TR
Find?Intrinsic?Subsequences(T,R,D)
1) maximum and minimizing set MM are empty, and element position set Loc is empty among the MM, and ITS is empty, and TR is empty;
2) if t
i∈ T, t
iBe set { t
I-R..., t
I+RMiddle maximum or minima, and in set, exist element and its poor absolute value more than or equal to D, then with t
iJoin among the set MM, and i is joined among the Loc;
3) each element of pair set T repeats above 1 successively), 2);
4) each adjacent maximum and the subsequence between the minimum point are inherent trend subsequence pattern, and all inherent trend subsequence patterns are added ITS, and the Trend value that it is corresponding adds TR.
The abnormality detection algorithm is as follows:
When the ECG signal sequence occurs when unusual, there will be two kinds of situations usually: (1) occurs having the trend subsequence that did not occur in the normal ECG signal sequence; (2) subsequence of new trend do not occur having, but the subsequence of the same layer that trend is close with it is widely different in some subsequence and the normal ECG signal sequence.Our ECG signal sequence abnormality detection algorithm is that inherent trend subsequence pattern is decomposed into layer according to its Trend value, the doubtful sequence that will detect normal sequence part then and not detect is partly by layer comparison and calculate its interfloor distance, if distance judges then that greater than a threshold value η it is unusual.
If first kind of situation is unusual, such unusual subsequence can at first be detected owing to its trend anomaly; If second kind of situation is unusual, such subsequence with its comparison procedure with the normal-sub sequence of layer in, can be detected, therefore second class unusually also can be detected.Simultaneously because subsequence is compared in layering, data volume is relatively dwindled, some is very faint unusual in whole long sequence, because after the layering refinement relatively, it is more obvious then can to seem, thereby our algorithm can more sensitively detect faint unusually, and amount of calculation has relatively been dwindled in layering simultaneously, and the efficient of abnormality detection algorithm will improve greatly.In the real heart activity; when abnormal movement occurring; tend to occur new variation tendency in the ECG sequence; perhaps its trend does not change; but itself has taken place change in sequence; all represented heart new movable or active state to occur variation has taken place, the abnormal conditions of this and heart are corresponding.Cardiomotility occurs when unusual, often or new cardiomotility trend occurred or unusual situation has appearred in original activity.
This abnormality detection algorithm at first respectively with the inherent trend subsequence pattern of normal and doubtful sequence according to its trend parameter value layering, then respectively according to the trend range of parameter values of each layer, the inherent trend pattern subsequence pattern of equivalent layer is mated, calculate between normal and doubtful each layer of sequence corresponding sequence minimum value and value as the distance of normal and doubtful sequence at this layer, if this distance greater than certain thresholding η, thinks that then it is unusual.
The abnormality detection algorithm is as follows:
Input: the inherent trend subsequence set of patterns TITS of the inherent trend subsequence set of patterns NITS of normal sequence and doubtful sequence
Output: unusual subsequence set
Anormaly?Detection(NITS,TITS)
8) the inherent trend subsequence pattern layering of normal ECG signal sequence set Nlayer is empty;
9) with the subsequence among the NITS according to its Trend value layering, the approaching sequence of Trend value is formed one deck, and layering is joined among the set Nlayer;
10) S ∈ TITS if there be not the layer not approaching with the Trend value of S among the Nlayer, then joins AS with S; If the layer approaching with the Trend value of S arranged, then calculate the stretching distance of all subsequences of S and this layer, and stretching distance obtained average distance divided by the length of S, being averaged minimum value and value is the distance D that S arrives this layer, if D〉η, then S is unusual subsequence, and S is joined AS, otherwise, forward 4 to);
11) all subsequences of pair set TITS repeat 3), return unusual subsequence set AS at last.
Claims (4)
1, a kind of ECG method for detecting abnormality that decomposes based on inherent trend subsequence pattern is characterized in that, may further comprise the steps:
1. definition:
ECG signal sequence: the data set that the electrocardiogram element is arranged according to time sequencing;
Inherent trend subsequence pattern: given ECG signal sequence, if the variation tendency of all subsequences of certain subsequence is identical with its variation tendency, and in given ECG signal sequence, do not exist the subsequence identical with its trend to comprise it, this subsequence is called as inherent trend subsequence pattern;
Layer: in the ECG signal sequence, the inherent trend subsequence pattern with similar Trend value is formed a layer;
Sequence is decomposed: the ECG signal sequence is decomposed into some inherent trend subsequence patterns, and the inherent trend subsequence with similar Trend value forms corresponding layer;
2. normal ECG signal sequence being carried out sequence decomposes and layering:
The ECG signal sequence is decomposed into an inherent trend subsequence set of patterns, then to normal inherent trend subsequence pattern according to its variation tendency parameter value layering, the subsequence that the variation tendency parameter value is close is divided into one deck, and its trend range of parameter values of every layer of usefulness is sectioned out;
3. doubtful ECG signal sequence being carried out sequence decomposes and layering:
Doubtful ECG signal is carried out sequence decompose the inherent trend subsequence pattern that also obtains, the back to its inherent trend subsequence pattern according to its variation tendency parameter value layering, the subsequence that the variation tendency parameter value is close is divided into one deck, and its trend range of parameter values of every layer of usefulness is sectioned out;
4. matching detection:
Respectively according to the trend range of parameter values of each layer, the inherent trend pattern subsequence pattern of above-mentioned both equivalent layers is mated, if the trend parameter value of doubtful inherent trend subsequence outside the trend range of parameter values of normal-sub sequence, this doubtful ECG abnormal signal then; If within scope, then calculate the distance of all the normal-sub sequences in this subsequence and the equivalent layer, get minima and be its distance, if the distance of this subsequence and its respective layer is greater than the threshold value of presetting with respective layer, then this subsequence is unusual subsequence, i.e. this doubtful ECG abnormal signal.
2, the ECG method for detecting abnormality that decomposes based on inherent trend subsequence pattern according to claim 1 is characterized in that the excavation step of described inherent trend subsequence is as follows:
1. find out all maximums and minima in the ECG signal sequence, and the position of their correspondences in the ECG signal sequence;
2. in the ECG signal sequence with maximum and minima as the subsequence of starting point or terminal point as inherent trend subsequence pattern.
3, the ECG method for detecting abnormality that decomposes based on inherent trend subsequence pattern according to claim 2 is characterized in that the mining algorithm step of described inherent trend subsequence is as follows:
1. definition:
Subsequence trend: in ECG signal sequence T, S={s
1... s
m, being the inherent trend subsequence pattern among the T, the trend Trend of S (S) is (s
1-s
m)/(1-m);
Maximum MAX: at sequence T={t
1... t
n, t
iIf ∈ T is t
I-1<t
iAnd t
iT
I+1, t then
iBe a maximum MAX among the sequence T, and all maximums among the sequence T are formed maximum S set MAX, SMAX={e|e is a maximum among the sequence T };
Minimum value MIN: at sequence T={t
1... t
n, t
iIf ∈ T is t
I-1T
iAnd t
i<t
I+1, t then
iBe a minimum value MIN among the sequence T, and all among the sequence T are worth most forms minima S set MIN, SMIN={e|e is a minima among the sequence T };
Stretching distance: subsequence Q={q
1..., q
Mq, C={c
1..., c
Mc, and mq≤mc, Q
s={ q
s 1, q
s 2..., q
s sFor Q is the stretching sequence of coefficient with mq/s,
Q, the stretching distance d of C
u(Q C) is:
R is a range parameter of determining extreme point, and D is the range parameter of extreme value each data point difference in the R scope;
2. algorithm: list entries T, range parameter R and difference parameter D
Find?Intrinsic?Trend?Subsequences(T,R,D)
1) maximum and minimizing set MM are empty, and element position set Loc is empty among the MM, and ITS is empty, and TR is empty;
2) if t
i∈ T, t
iBe set { t
I-R..., t
I+RMiddle maximum or minima, and in set, deposit
At the absolute value of element and its difference more than or equal to D, then with t
iJoin among the set MM, and i is joined among the Loc;
3) each element of pair set T repeats above 1 successively), 2);
4) each adjacent maximum and the subsequence between the minimum point are inherent trend subsequence pattern, and all inherent trend subsequence patterns are added ITS, and the Trend value that it is corresponding adds TR.
3. export inherent trend subsequence set of modes ITS and its corresponding trend set TR.
4, the ECG method for detecting abnormality that decomposes based on inherent trend subsequence pattern according to claim 1 is characterized in that the abnormality detection algorithm steps is as follows:
Input: the inherent trend subsequence set of patterns TITS of the inherent trend subsequence set of patterns NITS of normal sequence and doubtful sequence
Output: unusual subsequence set
Anormaly?Detection(NITS,TITS)
1) the inherent trend subsequence pattern layering of normal ECG signal sequence set Nlayer is empty;
2) with the subsequence among the NITS according to its Trend value layering, the approaching sequence of Trend value is formed one deck, and layering is joined among the set Nlayer;
3) S ∈ TITS if there be not the layer not approaching with the Trend value of S among the Nlayer, then joins AS with S; If the layer approaching with the Trend value of S arranged, then calculate the stretching distance of all subsequences of S and this layer, and stretching distance obtained average distance divided by the length of S, being averaged minimum value and value is the distance D that S arrives this layer, if D〉η, then S is unusual subsequence, and S is joined AS, otherwise, forward 4 to);
All subsequences of pair set TITS repeat 3), return unusual subsequence set AS at last.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2008100460067A CN101427917B (en) | 2008-09-08 | 2008-09-08 | ECG abnormal acquiring method based on inherent trend subsequence mode decomposition |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2008100460067A CN101427917B (en) | 2008-09-08 | 2008-09-08 | ECG abnormal acquiring method based on inherent trend subsequence mode decomposition |
Publications (2)
Publication Number | Publication Date |
---|---|
CN101427917A true CN101427917A (en) | 2009-05-13 |
CN101427917B CN101427917B (en) | 2011-02-09 |
Family
ID=40643699
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN2008100460067A Expired - Fee Related CN101427917B (en) | 2008-09-08 | 2008-09-08 | ECG abnormal acquiring method based on inherent trend subsequence mode decomposition |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN101427917B (en) |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
TWI461177B (en) * | 2010-12-28 | 2014-11-21 | Ind Tech Res Inst | System, method of detecting abnormal movement of a physical object by periodic signal analysis and computer readable storage medium thereof |
CN105279367A (en) * | 2014-07-15 | 2016-01-27 | 富士胶片株式会社 | Medical assistance device, operation method for medical assistance device, and medical assistance system |
CN105512500A (en) * | 2015-12-30 | 2016-04-20 | 广州市润杰医疗器械有限公司 | Method for rapidly screening characteristic of schizophrenia and evaluating therapeutic effect |
CN106228002A (en) * | 2016-07-19 | 2016-12-14 | 北京工业大学 | A kind of high efficiency exception time series data extracting method based on postsearch screening |
CN107296598A (en) * | 2017-06-22 | 2017-10-27 | 无锡力芯微电子股份有限公司 | Heart rate measurement method and device based on photoelectric sensor |
CN107944493A (en) * | 2017-11-30 | 2018-04-20 | 天津大学 | A kind of the first lead electrocardiogram beat classification method based on convolutional neural networks |
CN108367156A (en) * | 2015-12-02 | 2018-08-03 | 心脏起搏器股份公司 | Filtering in cardiac rhythm management apparatus automatically determining and selecting |
CN111557661A (en) * | 2020-05-15 | 2020-08-21 | 深圳清华大学研究院 | Electrocardiosignal processing method and device |
CN112243387A (en) * | 2018-06-05 | 2021-01-19 | 美敦力公司 | Cardiac signal T-wave detection |
CN112352289A (en) * | 2018-04-25 | 2021-02-09 | 皮斯投资公司 | Method and system for providing ECG analysis interface |
CN118236072A (en) * | 2024-05-29 | 2024-06-25 | 大连智驱科技有限公司 | Patient sign state monitoring method for anesthesia process |
Family Cites Families (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7194298B2 (en) * | 2002-10-02 | 2007-03-20 | Medicale Intelligence Inc. | Method and apparatus for trend detection in an electrocardiogram monitoring signal |
US7072709B2 (en) * | 2004-04-15 | 2006-07-04 | Ge Medical Information Technologies, Inc. | Method and apparatus for determining alternans data of an ECG signal |
US20070244402A1 (en) * | 2006-02-17 | 2007-10-18 | Brockway Brian P | System and method of monitoring physiological signals |
US7813792B2 (en) * | 2006-04-17 | 2010-10-12 | General Electric Company | Method and apparatus for analyzing and editing ECG morphology and time series |
CN100415159C (en) * | 2006-07-17 | 2008-09-03 | 浙江大学 | Dynamic characteristic analysis method of real-time tendency of heart state |
-
2008
- 2008-09-08 CN CN2008100460067A patent/CN101427917B/en not_active Expired - Fee Related
Cited By (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
TWI461177B (en) * | 2010-12-28 | 2014-11-21 | Ind Tech Res Inst | System, method of detecting abnormal movement of a physical object by periodic signal analysis and computer readable storage medium thereof |
CN105279367A (en) * | 2014-07-15 | 2016-01-27 | 富士胶片株式会社 | Medical assistance device, operation method for medical assistance device, and medical assistance system |
CN105279367B (en) * | 2014-07-15 | 2019-06-04 | 富士胶片株式会社 | Diagnosis and treatment assisting system and its working method, diagnosis support system and storage medium |
CN108367156A (en) * | 2015-12-02 | 2018-08-03 | 心脏起搏器股份公司 | Filtering in cardiac rhythm management apparatus automatically determining and selecting |
CN108367156B (en) * | 2015-12-02 | 2021-08-17 | 心脏起搏器股份公司 | Automatic determination and selection of filtering in cardiac rhythm management devices |
CN105512500A (en) * | 2015-12-30 | 2016-04-20 | 广州市润杰医疗器械有限公司 | Method for rapidly screening characteristic of schizophrenia and evaluating therapeutic effect |
CN106228002A (en) * | 2016-07-19 | 2016-12-14 | 北京工业大学 | A kind of high efficiency exception time series data extracting method based on postsearch screening |
CN106228002B (en) * | 2016-07-19 | 2021-11-26 | 北京工业大学 | High-efficiency abnormal time sequence data extraction method based on secondary screening |
CN107296598A (en) * | 2017-06-22 | 2017-10-27 | 无锡力芯微电子股份有限公司 | Heart rate measurement method and device based on photoelectric sensor |
CN107944493A (en) * | 2017-11-30 | 2018-04-20 | 天津大学 | A kind of the first lead electrocardiogram beat classification method based on convolutional neural networks |
CN112352289A (en) * | 2018-04-25 | 2021-02-09 | 皮斯投资公司 | Method and system for providing ECG analysis interface |
CN112243387A (en) * | 2018-06-05 | 2021-01-19 | 美敦力公司 | Cardiac signal T-wave detection |
CN111557661A (en) * | 2020-05-15 | 2020-08-21 | 深圳清华大学研究院 | Electrocardiosignal processing method and device |
CN111557661B (en) * | 2020-05-15 | 2024-03-08 | 深圳清华大学研究院 | Electrocardiosignal processing method and device |
CN118236072A (en) * | 2024-05-29 | 2024-06-25 | 大连智驱科技有限公司 | Patient sign state monitoring method for anesthesia process |
Also Published As
Publication number | Publication date |
---|---|
CN101427917B (en) | 2011-02-09 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN101427917B (en) | ECG abnormal acquiring method based on inherent trend subsequence mode decomposition | |
US8233972B2 (en) | System for cardiac arrhythmia detection and characterization | |
CN100515327C (en) | Detector methods and apparatus of cardiovascular system combining with variability guideline | |
CN107837082A (en) | Electrocardiogram automatic analysis method and device based on artificial intelligence self study | |
CN107714023A (en) | Static ecg analysis method and apparatus based on artificial intelligence self study | |
Ravish et al. | Heart function monitoring, prediction and prevention of heart attacks: Using artificial neural networks | |
CN107951485A (en) | Ambulatory ECG analysis method and apparatus based on artificial intelligence self study | |
CN101057781B (en) | Method and apparatus for analyzing and editing ECG morphology and time series | |
US20100217144A1 (en) | Diagnostic and predictive system and methodology using multiple parameter electrocardiography superscores | |
CN109691994A (en) | A kind of rhythm of the heart analysis method based on electrocardiogram | |
Zhang et al. | Localization of myocardial infarction with multi-lead bidirectional gated recurrent unit neural network | |
Chocron et al. | Remote atrial fibrillation burden estimation using deep recurrent neural network | |
Ranjan et al. | A unified approach of ECG signal analysis | |
Ansari et al. | Noise detection in electrocardiography signal for robust heart rate variability analysis: A deep learning approach | |
CN105877739A (en) | Clinical examination method of electrocardio intelligent analyzing system | |
Zhang et al. | Deep learning-based signal quality assessment for wearable ECGs | |
Tung et al. | Multi-lead ECG classification via an information-based attention convolutional neural network | |
CN1915166A (en) | Method for examining property of heart recovery based on cardiogram of flat motion | |
Zou et al. | A generalizable and robust deep learning method for atrial fibrillation detection from long-term electrocardiogram | |
Huo et al. | ECG segmentation algorithm based on bidirectional hidden semi-Markov model | |
Hori et al. | Arrhythmia detection based on patient-specific normal ECGs using deep learning | |
Zhang et al. | Evaluation of capacitive ECG for unobtrusive atrial fibrillation monitoring | |
Kovalchuk et al. | A novel feature vector for ECG classification using deep learning | |
Al-Zaiti et al. | The role of automated 12-lead ECG interpretation in the diagnosis and risk stratification of cardiovascular disease | |
Chen et al. | Beat-to-beat heart rate detection based on seismocardiogram using BiLSTM network |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
C17 | Cessation of patent right | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20110209 Termination date: 20120908 |