CN116369907B - Ballistocardiogram signal positioning method - Google Patents
Ballistocardiogram signal positioning method Download PDFInfo
- Publication number
- CN116369907B CN116369907B CN202310273416.XA CN202310273416A CN116369907B CN 116369907 B CN116369907 B CN 116369907B CN 202310273416 A CN202310273416 A CN 202310273416A CN 116369907 B CN116369907 B CN 116369907B
- Authority
- CN
- China
- Prior art keywords
- cardiac cycle
- prediction
- signal
- peak
- ballistocardiogram
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 41
- 230000000747 cardiac effect Effects 0.000 claims abstract description 265
- 238000001514 detection method Methods 0.000 claims abstract description 29
- 230000033001 locomotion Effects 0.000 claims description 27
- 238000005070 sampling Methods 0.000 claims description 17
- 238000012545 processing Methods 0.000 claims description 13
- 230000002159 abnormal effect Effects 0.000 claims description 11
- 238000011176 pooling Methods 0.000 claims description 11
- 230000008569 process Effects 0.000 claims description 7
- 230000002457 bidirectional effect Effects 0.000 claims description 3
- 238000000605 extraction Methods 0.000 claims description 3
- 230000000877 morphologic effect Effects 0.000 claims description 2
- 230000000694 effects Effects 0.000 abstract description 4
- 238000010586 diagram Methods 0.000 description 9
- 238000004590 computer program Methods 0.000 description 7
- 239000012634 fragment Substances 0.000 description 7
- 208000010125 myocardial infarction Diseases 0.000 description 6
- 238000012549 training Methods 0.000 description 5
- 238000012935 Averaging Methods 0.000 description 4
- 230000006870 function Effects 0.000 description 4
- 238000001914 filtration Methods 0.000 description 3
- 238000007781 pre-processing Methods 0.000 description 3
- 230000011218 segmentation Effects 0.000 description 3
- 230000005540 biological transmission Effects 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000009191 jumping Effects 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000003776 cleavage reaction Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 230000004807 localization Effects 0.000 description 1
- 230000000241 respiratory effect Effects 0.000 description 1
- 230000029058 respiratory gaseous exchange Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 230000007017 scission Effects 0.000 description 1
- 238000012163 sequencing technique Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000001629 suppression Effects 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/103—Detecting, measuring or recording devices for testing the shape, pattern, colour, size or movement of the body or parts thereof, for diagnostic purposes
- A61B5/11—Measuring movement of the entire body or parts thereof, e.g. head or hand tremor, mobility of a limb
- A61B5/1102—Ballistocardiography
-
- 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
- A61B5/7221—Determining signal validity, reliability or quality
-
- 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
Abstract
The application provides a ballistocardiogram signal positioning method, which comprises the following steps: acquiring a ballistocardiogram signal of a target object; inputting the ballistocardiogram signals into a pre-trained target detection model to obtain a plurality of cardiac cycle prediction areas of the ballistocardiogram signals and prediction probabilities corresponding to the cardiac cycle prediction areas; the target detection model is used for detecting a cardiac cycle prediction area and a corresponding prediction probability of the ballistocardiogram signal; determining a cardiac cycle prediction area with highest prediction probability as an initial cardiac cycle template; and positioning the cardiac cycle prediction signal of the ballistocardiogram signal and the J peak position of the cardiac cycle prediction signal according to the signal morphology distance curve of the cardiac cycle template and the plurality of cardiac cycle prediction areas. The method and the device can improve the positioning accuracy of the J peak of the ballistocardiogram so as to improve the positioning effect of the ballistocardiogram.
Description
Technical Field
The application relates to the technical field of ballistocardiogram signal positioning, in particular to a ballistocardiogram signal positioning method.
Background
The existing ballistocardiogram signal positioning mode directly takes the range of a single J peak as a positioning target, however, because the J peak can obviously present the characteristics of high peak amplitude and large peak-to-valley value only in a typical ballistocardiogram, the J peak can be easily and accurately positioned only in the typical ballistocardiogram, however, when the ballistocardiogram is interfered, distortion can occur to form an atypical ballistocardiogram, and the technical defects of low positioning accuracy of the J peak and poor positioning effect of the ballistocardiogram can occur aiming at the atypical ballistocardiogram by taking the range of the single J peak as the positioning target.
Disclosure of Invention
The purpose of the application is to overcome the defects and shortcomings in the prior art, and provide a ballistocardiogram signal positioning method which can improve the positioning accuracy of J peaks of a ballistocardiogram so as to improve the positioning effect of the ballistocardiogram.
The embodiment of the application provides a ballistocardiogram signal positioning method, which comprises the following steps:
acquiring a ballistocardiogram signal of a target object;
inputting the ballistocardiogram signals into a pre-trained target detection model to obtain a plurality of cardiac cycle prediction areas of the ballistocardiogram signals and prediction probabilities corresponding to the cardiac cycle prediction areas; the target detection model is used for detecting a cardiac cycle prediction area and corresponding prediction probability of the ballistocardiogram signal;
determining the cardiac cycle prediction area with the highest prediction probability as an initial cardiac cycle template;
and positioning a cardiac cycle prediction signal of the ballistocardiogram signal and J peak positions of the cardiac cycle prediction signal according to the signal morphology distance curve of the cardiac cycle template and the plurality of cardiac cycle prediction areas.
Compared with the related art, the method and the device have the advantages that firstly, the prediction probabilities corresponding to the plurality of cardiac cycle prediction areas and the cardiac cycle prediction areas of the ballistocardiogram signal are obtained through the target detection model, then the cardiac cycle prediction area with the highest prediction probability is determined to be the initial cardiac cycle template, and then the J peak positions of the cardiac cycle prediction signal and the cardiac cycle prediction signal of the ballistocardiogram signal are positioned according to the signal form distance curve of the cardiac cycle template and the plurality of cardiac cycle prediction areas.
In order that the present application may be more clearly understood, specific embodiments thereof will be described below with reference to the accompanying drawings.
Drawings
Fig. 1 is a flowchart of a ballistocardiogram signal positioning method according to one embodiment of the present application.
Fig. 2 is a flowchart of steps S31-S33 of a ballistocardiogram signal positioning method according to one embodiment of the present application.
Fig. 3 is a schematic diagram of a cardiac cycle prediction area and a detection result of motion artifacts in a ballistocardiogram signal positioning method according to an embodiment of the present application.
Fig. 4 is a schematic diagram of a J-peak positioning result of a ballistocardiogram signal positioning method according to one embodiment of the present application.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the present application more apparent, the following detailed description of the embodiments of the present application will be given with reference to the accompanying drawings.
It should be understood that the described embodiments are merely some, but not all, of the embodiments of the present application. All other embodiments, which can be made by one of ordinary skill in the art without undue burden from the embodiments of the present application, are within the scope of the embodiments of the present application.
When the following description refers to the accompanying drawings, the same numbers in different drawings refer to the same or similar elements, unless otherwise indicated. In the description of this application, it should be understood that the terms "first," "second," "third," and the like are used merely to distinguish between similar objects and are not necessarily used to describe a particular order or sequence, nor should they be construed to indicate or imply relative importance. The specific meaning of the terms in this application will be understood by those of ordinary skill in the art as the case may be. As used in this application and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. The word "if"/"if" as used herein may be interpreted as "at … …" or "at … …" or "in response to a determination".
Furthermore, in the description of the present application, unless otherwise indicated, "a plurality" means two or more. "and/or", describes an association relationship of an association object, and indicates that there may be three relationships, for example, a and/or B, and may indicate: a exists alone, A and B exist together, and B exists alone. The character "/" generally indicates that the context-dependent object is an "or" relationship.
Referring to fig. 1, a flowchart of a ballistocardiogram signal positioning method according to one embodiment of the present application includes:
s1: and acquiring a ballistocardiogram signal of the target object.
The heart attack signal is a signal which causes weak vibration of a human body by heart cycle diastole and systole, and is essentially a force signal, the heart attack signal of a target object can be acquired through a signal acquisition device based on piezoelectric sensing, and a heart attack image signal is generated, the signal acquisition device based on the piezoelectric sensing comprises a piezoelectric sensor module and a data processor, wherein the piezoelectric sensor module is used for acquiring data of the heart attack signal, and the data of the heart attack signal is digitally stored through a sampling frequency preset by the data processor so as to obtain the heart attack image signal. The preset sampling frequency may be 1000Hz.
Preferably, after obtaining the ballistocardiogram signal of the target object, the ballistocardiogram signal is further subjected to data preprocessing, wherein the data preprocessing comprises filtering interference noise of the ballistocardiogram signal, for example, filtering respiration and high-frequency interference noise through a band-pass filter, so as to reduce the interference signal in the ballistocardiogram signal. The bandpass filter may be a butterworth bandpass filter.
S2: inputting the ballistocardiogram signals into a pre-trained target detection model to obtain a plurality of cardiac cycle prediction areas of the ballistocardiogram signals and prediction probabilities corresponding to the cardiac cycle prediction areas; the target detection model is used for detecting a cardiac cycle prediction area and corresponding prediction probability of the ballistocardiogram signal.
The target detection model can be obtained by training the ballistocardiogram signal sample set marked with the cardiac cycle prediction area and the prediction probability, and when the target detection model is simultaneously used for identifying other signal areas of the ballistocardiogram signal except for the cardiac cycle prediction area, the training data marked with the other signal areas can be added to the ballistocardiogram signal sample set, or the other signal areas in the training data of the ballistocardiogram signal sample set are marked, and then the training of the target detection model is performed. The target training model can adopt a CenterNet, a common basic structure of the CenterNet comprises a backbone network, an up-sampling layer and a prediction layer, the backbone network can adopt a ResNet50, and the CenterNet firstly performs feature extraction on an input signal through the ResNet 50; then up-sampling is carried out by adopting three times of deconvolution, so as to obtain a high-resolution characteristic layer; and finally, inputting the high-resolution characteristic layer into a prediction layer to perform result prediction. In this embodiment, in order to improve the applicability of the central net to the center impact diagram signal, other structures may be added to the basic structure of the central net, so as to improve the recognition accuracy of the central net.
When the target detection model outputs the prediction region and the prediction probability, the prediction region can be screened according to the prediction probability, and the prediction region with the prediction probability higher than the preset probability threshold value is determined as a plurality of cardiac cycle prediction regions.
S3: and determining the cardiac cycle prediction area with the highest prediction probability as an initial cardiac cycle template.
S4: and positioning a cardiac cycle prediction signal of the ballistocardiogram signal and J peak positions of the cardiac cycle prediction signal according to the signal morphology distance curve of the cardiac cycle template and the plurality of cardiac cycle prediction areas.
The cardiac cycle template in step S4 may be an initial cardiac cycle template, and then the cardiac cycle template is updated according to the located signal data when the cardiac cycle prediction signal of the impulse map signal and the J peak position of the cardiac cycle prediction signal are relocated.
Compared with the related art, the method and the device have the advantages that firstly, the prediction probabilities corresponding to the plurality of cardiac cycle prediction areas and the cardiac cycle prediction areas of the ballistocardiogram signal are obtained through the target detection model, then the cardiac cycle prediction area with the highest prediction probability is determined to be the initial cardiac cycle template, and then the J peak positions of the cardiac cycle prediction signal and the cardiac cycle prediction signal of the ballistocardiogram signal are positioned according to the signal form distance curve of the cardiac cycle template and the plurality of cardiac cycle prediction areas.
In one possible embodiment, the object detection model includes a backbone network, a spatial pyramid pooling structure, an upsampling layer, a temporal feature convolution network, and a prediction layer;
the backbone network is used for extracting the characteristics of the input signals;
the input of the spatial pyramid pooling structure is connected with the output of the backbone network; the spatial pyramid pooling structure is used for carrying out multi-scale abstract processing on the characteristics of the input signals to obtain the characteristics of the input signals fused with the multi-scale signals to be output to the up-sampling layer;
the input of the up-sampling layer is connected with the input of the spatial pyramid pooling structure, and the up-sampling layer is used for up-sampling the characteristics of the input signals to obtain high-resolution characteristics;
the input of the time sequence characteristic convolution network is connected with the output of the up-sampling layer, the output of the time sequence characteristic convolution network is connected with the input of the prediction layer, and the time sequence characteristic convolution network is used for carrying out bidirectional time sequence characteristic extraction on high resolution characteristics to obtain the high resolution characteristics comprising the time sequence characteristics and outputting the high resolution characteristics to the prediction layer; the time sequence characteristic convolution network can also adopt a bidirectional time sequence characteristic convolution network.
And the input of the prediction layer is connected with the output of the time sequence characteristic convolution network, and the prediction layer is used for outputting a prediction result according to the high-resolution characteristic.
In this embodiment, the target detection model is a network model based on a central net basic structure and added with a spatial pyramid pooling structure and a time sequence feature convolution network, and the spatial pyramid pooling structure added after the backbone network can abstract the features of the input signal in different scales, so as to realize feature fusion in different scales, improve scale-invariance of the network model to the signal, and reduce overfitting (over-fitting) of the network model to the signal.
In one possible embodiment, the step S2: inputting the ballistocardiogram signal into a pre-trained target detection model to obtain a plurality of cardiac cycle prediction areas of the ballistocardiogram signal and prediction probabilities corresponding to the cardiac cycle prediction areas, wherein the method comprises the following steps of:
s21: and inputting the ballistocardiogram signals into a pre-trained target detection model to obtain a plurality of initial cardiac cycle prediction areas and corresponding prediction probabilities of the ballistocardiogram signals.
S22: and determining the initial cardiac cycle prediction area with the prediction probability higher than a preset probability threshold as the cardiac cycle prediction area.
S23: and obtaining a first interval threshold according to the average interval of the cardiac cycle prediction area and a preset threshold multiplying power.
The average distance between the cardiac cycle prediction areas may be obtained by removing the maximum distance between a plurality of cardiac cycle prediction areas and the minimum distance between a plurality of cardiac cycle prediction areas, and then averaging the distances between the remaining cardiac cycle prediction areas, for example: and arranging the intervals among the prediction areas of the cardiac cycle from large to small or from small to large, and removing the intervals among the prediction areas of the cardiac cycle of the first quarter and the intervals among the prediction areas of the cardiac cycle of the last quarter in a quarter bit mode.
S24: and if the current distance between the current cardiac cycle prediction area and the next cardiac cycle prediction area is larger than the first distance threshold, determining the current distance as an abnormal distance.
S25: and acquiring the midpoint distance of the adjacent initial cardiac cycle prediction areas.
Midpoint spacing refers to the distance between the midpoints of adjacent initial cardiac cycle prediction regions.
S26: and if the range of the abnormal interval is larger than the midpoint interval of a preset second interval threshold, determining the initial cardiac cycle prediction area with the largest prediction probability as the cardiac cycle prediction area according to the prediction probability of the initial cardiac cycle prediction area corresponding to the midpoint interval, and re-determining whether the abnormal interval is present or not and whether the range of the abnormal interval is larger than the midpoint interval of the second interval threshold or not.
In the present embodiment, considering that when there is an abnormal pitch between the cardiac cycle prediction regions, there may be cardiac cycle prediction regions that are ignored because the prediction probability does not reach the standard, the ignored cardiac cycle prediction regions can be found from the abnormal pitch in combination with the midpoint pitch of the initial cardiac cycle prediction regions.
In one possible embodiment, the step S3: the step of determining the cardiac cycle prediction region with the highest prediction probability as an initial cardiac cycle template comprises the following steps:
s31: and grouping the prediction areas of the cardiac cycle to obtain a plurality of prediction area groups.
For example, the predicted region is divided into a predicted region group every 20 cardiac cycles, or a predicted region group every 30 cardiac cycles, etc.
S32: and determining the cardiac cycle prediction region with the highest prediction probability in the first prediction region group as an initial cardiac cycle template.
S33: determining a peak conforming to the morphological characteristics of the J peak signal as a J peak from the initial cardiac cycle template to locate the J peak position of the initial cardiac cycle template; and if the peak which accords with the morphology characteristics of the J peak signal does not exist in the initial cardiac cycle template, determining the peak closest to the midpoint of the initial cardiac cycle template as the J peak so as to locate the J peak position of the initial cardiac cycle template.
In this embodiment, an initial cardiac cycle template and corresponding J-peak localization may be obtained from the first set of prediction regions.
In one possible embodiment, the step S4: positioning a cardiac cycle prediction signal of the ballistocardiogram signal and a J peak position of the cardiac cycle prediction signal according to a signal morphology distance curve of the cardiac cycle template and the plurality of cardiac cycle prediction areas, wherein the method comprises the following steps of:
s41: and grouping the prediction areas of the cardiac cycle to obtain a plurality of prediction area groups.
In this embodiment, if the cardiac cycle prediction regions are already grouped during the process of acquiring the initial cardiac cycle template in step S3, step S41 may be omitted, and the plurality of prediction region groups obtained in step S3 may be directly adopted as the plurality of prediction region groups obtained in step S41, and if the cardiac cycle prediction regions are not grouped during the process of acquiring the initial cardiac cycle template in step S3, step S41 is reserved and executed, for example, each 20 cardiac cycle prediction regions are divided into one prediction region group, or each 30 cardiac cycle prediction regions are divided into one prediction region group, etc.
S42: traversing each prediction region group according to the cardiac cycle template to locate cardiac cycle prediction signals of each prediction region group and J peak positions of the cardiac cycle prediction signals according to the signal morphology distance curve.
Wherein locating the J peak position of the cardiac cycle predicted signal for each of the set of predicted regions comprises:
s421: and obtaining the average interval of each prediction area group.
S422: and determining a peak of the maximum value of the prediction signal of the first cardiac cycle of the prediction area group as a J peak, and acquiring a peak point interval between the current J peak and the midpoint of the prediction signal of the next cardiac cycle.
S423: and acquiring the J peak of the predicted signal of the next cardiac cycle according to the average interval, the peak point interval and the position of the current J peak.
In this embodiment, the J peak of the predicted signal of the next cardiac cycle can be acquired more accurately according to the average interval, the peak-to-peak interval, and the position of the current J peak.
In a possible embodiment, the cardiac cycle template is updated according to the cardiac cycle prediction signals of the prediction area groups each time the J peak position of one of the prediction area groups is located during the positioning of the J peak positions of the cardiac cycle prediction signals of the respective prediction area groups.
In this embodiment, by updating the cardiac cycle template, the cardiac cycle template can be prevented from being single, which affects the positioning of the cardiac cycle prediction signals and the positioning of the J peak positions of each prediction region group, and the accuracy of the positioning of the cardiac cycle prediction signals and the positioning of the J peak positions of each prediction region group can be effectively improved.
In one possible embodiment, the step S421: the step of obtaining the average interval of each prediction area group comprises the following steps:
s4211: and calculating the initial average interval from the initial point coordinates of the current first cardiac cycle prediction region, the final point coordinates of the current last cardiac cycle prediction region and the current number of the cardiac cycle prediction regions of the prediction region group.
S4212: and deleting the first and last cardiac cycle prediction areas of the prediction area group after each initial average interval is calculated, and recalculating to obtain the next initial average interval until the prediction area group is emptied.
S4213: and obtaining the average interval of the prediction region group according to a plurality of initial average intervals corresponding to the prediction region group, and restoring all the deleted cardiac cycle prediction regions into the prediction region group.
In this embodiment, by repeatedly performing the two steps of calculating the initial average interval and deleting the cardiac cycle prediction region of the prediction region group, a plurality of initial average intervals when the prediction region group corresponds to the number of different cardiac cycle prediction regions can be obtained, and then the average interval of the prediction region group is calculated from the plurality of initial average intervals, so that the accuracy of the average interval can be improved.
In one possible embodiment, the step S423: the step of obtaining the J peak of the predicted signal of the next cardiac cycle according to the average interval, the peak point interval and the position of the current J peak comprises the following steps:
s4231: and if the quotient of the peak point interval and the average interval is not in a preset value range, determining the maximum value of the predicted signal of the next cardiac cycle as a J peak.
In this embodiment, when the quotient of the peak interval and the average interval is not within the preset value range, it indicates that the difference between the peak interval and the average interval is large, and at this time, the maximum value of the predicted signal of the next cardiac cycle is directly determined as the J peak, so as to avoid missing the positioning of the J peak.
In one possible embodiment, the step S423: the step of obtaining the J peak of the predicted signal of the next cardiac cycle according to the average interval, the peak point interval and the position of the current J peak comprises the following steps:
s4232: and if the quotient of the peak point interval and the average interval is not in the preset value range, obtaining the predicted point position of the J peak of the predicted signal of the next cardiac cycle according to the position of the current J peak and the average interval.
Specifically, the predicted point location of the J peak of the predicted signal of the next cardiac cycle is obtained by the following formula:
anchor＝J+MRRI；
Wherein, the anchor is the predicted point position of the J peak of the predicted signal of the next cardiac cycle; j is the position of the current J peak; MRRI is the average interval.
S4233: and determining the maximum value of the predicted signal of the next cardiac cycle as a J peak when the predicted point position is not in the range of the predicted area of the next cardiac cycle.
S4234: and when the prediction point position is in the range of the prediction area of the next cardiac cycle, acquiring the maximum two maxima of the prediction signal of the next cardiac cycle.
S4235: and updating the predicted signal of the next cardiac cycle according to the signal value of each point position of the predicted signal of the next cardiac cycle, the length of the predicted signal of the next cardiac cycle, the predicted point position and the maximum two maximum values.
Specifically, the next cardiac cycle prediction signal is updated by the following formula:
wherein Dit' n is the signal value of the nth signal point of the updated predicted signal of the next cardiac cycle; dit n is the signal value of the nth signal point of the predicted signal of the next cardiac cycle before updating; maxDit and MaxDit2 are the largest two maxima of the predicted signal of the next cardiac cycle; len (Dit) is the length of the predicted signal for the next cardiac cycle; max is the meaning of taking the maximum value of the two values.
S4236: and determining the maximum value of the updated prediction signal of the next cardiac cycle as a J peak.
In a possible embodiment, when the predicted point is not within the range of the predicted area of the next cardiac cycle, the maximum value of the predicted signal of the next cardiac cycle is directly determined as a J peak so as to avoid missing the positioning of the J peak, and when the predicted point is within the range of the predicted area of the next cardiac cycle, the constraint of the average interval is introduced to the predicted signal of the next cardiac cycle so as to update the signal value of the predicted signal of the next cardiac cycle, and then the J peak is positioned from the updated predicted signal of the next cardiac cycle so as to improve the accuracy of the positioning of the J peak.
In a possible embodiment, the object detection model is further used for detecting motion artifacts of the ballistocardiogram signal;
the S2: inputting the ballistocardiogram signal into a pre-trained target detection model to obtain a plurality of cardiac cycle prediction areas of the ballistocardiogram signal and prediction probabilities corresponding to the cardiac cycle prediction areas, wherein the method comprises the following steps of:
s201: and inputting the ballistocardiogram signals into a pre-trained target detection model to obtain a plurality of cardiac cycle prediction areas of the ballistocardiogram signals, prediction probabilities corresponding to the cardiac cycle prediction areas and a plurality of motion artifacts.
In step S201, the object detection model is trained from a ballistocardiogram signal sample set labeled with a cardiac cycle prediction region, a prediction probability, and motion artifacts.
The S3: the step of determining the cardiac cycle prediction region with the highest prediction probability as an initial cardiac cycle template comprises the following steps:
s301: dividing the ballistocardiogram signal into a plurality of signal segments according to the motion artifact.
S302: and determining the prediction area of the cardiac cycle with the highest prediction probability in each signal segment as an initial cardiac cycle template.
Since the steps S31-S33 described above have already disclosed the steps of determining the initial cardiac cycle template and locating the J peak position of the initial cardiac cycle template, and the step S302 may use the same steps to achieve the same technical content or achieve the same technical purpose, those skilled in the art may understand the technical content of the step S302 by referring to the steps S31-S33, so the description of the content of the step S302 will not be repeated here.
The S4: positioning a cardiac cycle prediction signal of the ballistocardiogram signal and a J peak position of the cardiac cycle prediction signal according to a signal morphology distance curve of the cardiac cycle template and the plurality of cardiac cycle prediction areas, wherein the method comprises the following steps of:
S401: and positioning a cardiac cycle prediction signal of the corresponding signal segment and J peak positions of the cardiac cycle prediction signal according to the signal morphology distance curve of the initial cardiac cycle template and the plurality of cardiac cycle prediction areas.
Wherein, since the steps S41-S42, S421-S423, S4211-S4213 and S4231-S4236 described above have already disclosed the step of locating the cardiac cycle prediction signal and the J peak position of the cardiac cycle prediction signal according to the signal morphology distance curve and the plurality of cardiac cycle prediction areas of the initial cardiac cycle template, and the step S401 may employ the same step to achieve the same technical content or achieve the same technical purpose, those skilled in the art may understand the technical content of the step S401 by referring to the steps S41-S42, S421-S423, S4211-S4213 and S4231-S4236, so the description of the content of the step S401 will not be repeated here.
Referring to fig. 3 to 4, in this embodiment, through the target detection model of this embodiment, the prediction probability and motion artifacts corresponding to a plurality of cardiac cycle prediction regions and each cardiac cycle prediction region of the ballistocardiogram signal (as shown in fig. 3, where a signal region in a largest box in the middle of fig. 3 is a motion artifact, and signal regions in the remaining boxes are cardiac cycle prediction regions), then the ballistocardiogram signal is divided into a plurality of signal segments according to the motion artifact, and then cardiac cycle prediction signal positioning and J peak position (J peak positioning results are shown in fig. 4) are performed on each signal segment of the ballistocardiogram signal, and J peaks of each cardiac cycle prediction region are marked with points or inverted triangle symbols), so that the influence of the motion artifact on the positioning of the ballistocardiogram signal can be reduced, which is beneficial to improving the positioning efficiency and accuracy of the cardiac cycle prediction signal and the corresponding J peak of each signal segment, thereby achieving the technical effect of improving the signal positioning accuracy of each signal segment of the ballistocardiogram signal.
To facilitate an understanding of the present application by those skilled in the art, the following examples are provided to facilitate an understanding of one possible implementation of the technical content of the present application:
1. data preprocessing
Step1: the original signal is collected and marked as orgData, the original signal is sampled at intervals, one is collected every 5 points from the first data point, and a new signal set is generated after downsampling and marked as orgData_200Hz: orgData_200Hz= [ orgData (0), orgData (5), orgData (10), orgData (15), …, orgData (P) ], wherein p=0, 5,10,15, …, P//5, P is the total number of data points of the original signal, P//5 is integer division, and the whole is reduced;
step2: inputting the signal orgData_200Hz into a 4-order Butterworth band-pass filter with the frequency of 2-10 Hz for processing, wherein the processing comprises filtering a respiratory baseline, power frequency interference and Gaussian noise, and the processed signal can be marked as BCG (ballistocardiogram signal);
2. cardiac cycle identification
(1) Model processing
Step3: sampling the downsampled or processed signal set orgData_200Hz with the sampling frequency of 200Hz;
step4: the orgdata_200Hz is divided into 3000 sample points (15 seconds) in length and 2000 sample points in steps, and the result is denoted as X [ q ] = { X0, X1, …, XQ-1}; wherein X0, X1, … and XQ-1 each represent a cleavage fragment, and Q fragments are taken together;
Step5: initializing a prediction result to be prediction_all= { }, taking j=0, and X [ j ] as the input of a pretrained centrnet model, wherein X [ j ] represents a j-th object in X [ q ];
step6: the signal of X j is first extracted through the first 5 layers of network in ResNet50 to obtain the result of featuremap1; the method comprises the steps that the featuremap1 is processed in different scales through an SPP module (space pyramid pooling structure module) mainly through 3 convolution check features in different sizes, and finally the features in the 3 scales are fused, and the result is marked as featuremap2; then, the feature is restored to the length of 752 sample points through three-layer deconvolution operation of the featuremap2, and the length is recorded as featuremap3; inputting the featuremap3 into an LSTM network (time sequence feature convolution network), carrying out time sequence constraint processing on the features, and splicing the result with the featuremap3 before inputting the LSTM, and marking the result as featuremap4; finally, respectively sending the featuremap4 into three convolution layer branches to predict, and predicting the central point, the area length and the probability of being the target result of the target result;
step7: processing according to the result of Step6, converting it back to the region start-end coordinates of the original signal scale and the probability of being the target result, the result is recorded as the prediction = { (x_st) 0 ,x_ed 0 ,P 0 ),(x_st 1 ,x_ed 1 ,P 1 ),…,(x_st M-1 ,x_ed M-1 ,P M-1 ) X_st, where i Representing the start coordinates, x_ed, of the ith cardiac cycle (or motion artifact) i Representing the end coordinates, P, of the ith cardiac cycle (or motion artifact) i Representing the probability that the i-th region is a cardiac cycle (or motion artifact);
step8: non-maximum value suppression with the intersection ratio of 15% is carried out on the pre, namely when the intersection ratio of two areas is more than 15%, only the area with the highest probability is reserved, and then the pre is stored in the pre_all;
step9: let j=j+1 and repeat Step6-Step8 until all data in X [ q ] are traversed to obtain a prediction_all storing a plurality of predictions;
step10: classifying the data of the prediction_all to obtain an initial cardiac cycle prediction area and motion artifacts, wherein the initial cardiac cycle prediction area is marked as the prediction_J, and the motion artifacts are marked as the prediction_M;
(2) Cardiac cycle outcome selection
Step11: grouping the results in the prediction_j, wherein each 100 prediction regions are grouped, and the result is recorded as the prediction_group [ y ] = { prediction_j [0 ]: 100], predict_j [100:200], …, predictt_j [99 x 100:99 x 100+100] };
step12: initializing the prediction_final= { } to take t=0;
Step13：Predict＝Predict_group[t]；
step14: firstly, obtaining the result with probability higher than 0.7 from the previous, and marking the result as the previous_acc [ n ] ]＝{(pos 0 ,P 0 ),(pos 1 ,P 1 ),…,(pos n ,P n ) }, pos therein i Coordinates (x_st) representing the ith prediction result i ,x_ed i ) And P is i Is probability;
step15: calculating the regional midpoint according to the regional coordinates in the prediction and marking as Center [ y ]]＝{C 0 ,C 1 ,…,C 99 }；
Step16: calculating the midpoint of the region according to the region coordinates in the prediction_acc, and marking the midpoint as center_acc;
step17: differencing using Center acc, the result being noted Center acc diff;
step18: sorting the centers_acc_diff according to the size, removing the results of the front and rear quartiles, and then averaging, and marking as MCCI;
step19: calculating an interval anomaly threshold value by using the MCCI, and recording as th_ abn =mcci×1.5;
step20: judging whether the result in the center_acc_diff is larger than TH_ abn, and marking the result as pos_ abn = { (Cy, C) y+1 ) }, wherein C y Is a Center [ y ]]The y-th data of (C) y+1 Is a Center [ y ]]Y+1st data in (a);
step21: if pos_ abn is empty, jump to Step26;
step22: if Step22 is executed for the first time, jumping to Step26, otherwise, comparing whether pos_ abn and pos_ abn _last are consistent, if so, jumping to Step26;
step23: traversing pos_ abn, determining the value of the value in each abnormal section (C y ,C y+1 ) In Center [ y ]]Whether or not there is and distance C y And C y+1 The results with the distances larger than 200ms are stored in the center_acc with the highest probability in the results if the results are found;
Step24: ordering center_acc and marking pos_ abn as pos_ abn _last;
step25: repeating Step17-Step24;
step26: storing the result of the pretreatment corresponding to the center_acc into the pretreatment_final;
step27: let t=t+1, repeat Step13 to Step26 until traversing the prediction_group;
step28: taking the prediction_final as a model final cardiac cycle prediction result to be output;
3. motion artifact handling
Step29: initializing body motion as motion= { };
step30: firstly, obtaining a result with probability higher than 0.7 from the prediction_M, and storing the result into a Movement in the form of (Mov_STi, mov_EDi), wherein the final result is expressed as movement= { (Mov_ST0, mov_ED 0), (Mov_ST1, mov_ED 1), …, (Mov_STn, mov_EDn) }, wherein Mov_STi represents a starting position of body Movement, mov_EDi represents an ending position of body Movement, and n represents total n+1 individual movements;
step31: the signals orgData_200Hz and BCG are segmented according to the motion Movement, and the segmentation coordinate rule is as follows:
the segmentation results are denoted (orgDatai, BCGi), and all segmentation result sets are denoted Signal1: signal 1= { (orgData 0, BCG 0), (orgData 1, BCG 1), …, (orgDatam, BCGm) }, where orgDatai is the original Signal orgData_200Hz [ pos_cut [ i,0]: pos_cut [ i,1] ], BCGi is the filtered signal BCG [ pos_cut [ i,0]: pos_cut [ i,1] ], m represents a total of m+1 fragment production;
After the Signal is segmented, each pair of segmented fragments in Signal1 is used as a single Signal fragment to carry out the next J peak judgment, and then all J peak results are combined into a final positioning result.
4.J peak decision
Step32: grouping the results in the prediction_final, wherein each 20 prediction results are a group, and the result is recorded as the prediction_group [ n ] = { prediction_final [0 ]: 20], predict_final [20:40], …, predictfinalt [ n x 20: n is 20+20] };
step33: initializing jpeaks_final= [ ], q=0;
step34: taking the prefix=prefix_group [ q ];
step35: intercepting the filtered signal according to the maximum coordinate and the minimum coordinate in the prediction, and marking the signal as WBCG;
step36: if q= 0 goes to Step37, otherwise go to Step43;
(1) Initial template acquisition
Step37: acquiring the regional coordinates of the cardiac cycle with the highest probability of the prediction result in the first prediction, intercepting WBCG according to the coordinates, taking the intercepted result as an initial Template, and marking the initial Template as Template;
step38: firstly, obtaining the coordinates of all wave Peaks and wave troughs in a Template, marking the coordinates as Peaks and Valley s, and respectively marking the corresponding amplitudes as peaks_amp and Valley_amp;
step39: judging whether the amplitude of the J peak is obvious: firstly, taking the maximum value in the Peaks_amp and the corresponding sitting marks as (P_amp 0, P_pos 0), taking the next-largest value in the Peaks_amp as P_amp1, and calculating the difference between the maximum value and the next-largest value amplitude as A=P_amp 0-P_amp1; then calculating the average value of all peak amplitudes except the maximum value, and recording the average value as M= (sum (peaks_amp) -P_amp 0)/(n-1), wherein n represents the number of Peaks; finally, calculating the ratio R=A/M, if R > =0.5, judging that the J peak is obvious, marking P_pos0 as the coordinates of the template J peak, turning to Step42, and if R <0.5, turning to Step40;
Step40: judging whether the J peak-to-valley value is obvious: searching the minimum trough as the corresponding trough within the range of 200ms forward of the peak [ i ], if no trough exists within 200ms, taking the nearest trough as the corresponding trough, marking as valley [ J ], calculating peak-valley values as PV=Peaks_amp [ i ] -valley_amp [ J ], traversing Peaks, calculating all peak-valley values as PVs, taking PVs as input, judging whether the peak-valley values are obvious according to the flow of Step39, marking the peak corresponding to the maximum value in the PVs as a template J peak if the peak-valley values are obvious, turning to Step42, otherwise turning to Step41;
step41: if the J peak of the template is not obvious in either amplitude or peak valley, calculating the midpoint of the template, marking the midpoint as C, and taking the peak closest to C in Peaks as the J peak of the template;
step42: outputting Template and Template J peak;
(2) Template update
Step43: updating and optimizing the template according to the judging result after 20J peaks are judged each time;
step44: calculating the distance from J_T to two ends of the Template, and recording as W1 and W2;
step45: BCG was truncated according to jpeaks= { J0, J1, …, J19} and the fragment pool was noted as winbcgs= { BCG [ J0-W1: J0+W2], BCG [ J1-W1 ]: J1+W2], …, BCG [ J19-W1: j19+w2] };
Step46: averaging all fragments in WinBCGs together with the template, and outputting the result as a new template;
(3) Average interval acquisition
Step47: initializing MRRIs= [ ]
Step48: taking the initial point coordinate of the first prediction area in the prediction, and marking the initial point coordinate as ST;
step49: taking the end point coordinate of the last prediction area in the prediction, and marking as ED;
step50: calculating the number of the prediction results in the prediction, and marking the number as N;
step51: calculating an average interval, which is denoted as MRRI' = (ED-ST)/N, and storing the average interval in MRRIs;
step52: deleting the first and last prediction results in the prediction;
step53: repeating Step 48-Step 52 until the prediction is empty;
step54: sequencing the MRRIs, removing the maximum and minimum values, and averaging, wherein the result is a final average interval which is recorded as the MRRI;
step55: reducing the Prect to be the Prect_group [ q ], and outputting MRRI;
(4) J peak judgment
Step56: the Euclidean distance is calculated by using a Template and the filtered signal BCG, and the result is recorded as BCGdit [ n ];
step57: for convenience in processing, BCGdit is turned over, so that bcgdit=max (BCGdit) -BCGdit;
step58: traversing the prediction, and intercepting BCGdit according to the prediction area inside, wherein the result is recorded as BCGdit_group= { BCGdit [ pos0], BCGdit [ pos1], …, BCGdit [ pos19] };
Step59: calculating a regional midpoint according to regional coordinates in the prediction, and marking the regional midpoint as a Center;
step60: initializing a J peak complex to jpeaks= [ ]
Step61: maximizing the first result in bcgdit_group, which is considered as a J peak, and storing it in jeaks;
step62: calculating the number of Jpeaks, namely K, taking the (K+1) th midpoint, and namely C=center [ K+1];
step63: calculating the interval between the last result J=Jpeaks [ K ] and C in Jpeaks, and marking as L=C-J;
step64: taking dit=bcgdit_group [ k+1];
step65: judging that the L/MRRI is less than 0.5 and less than 1.5, otherwise, turning to Step70;
step66: taking an anchor=j+mrri;
step67: judging whether the anchor falls in a prediction area of the prediction [ K+1], if not, turning to Step70;
step68: taking the maximum and the secondary maximum in Dit, and respectively marking the maximum and the secondary maximum as MaxDat and MaxDat 2;
step69: traversing Dit, processing Dit according to the following formula (note: the purpose is to introduce constraints on the average interval);
step70: taking the maximum value in Dit as a J peak, and storing the J peak into Jpeaks;
step71: repeating Step 62-Step 70 until all BCGdit_groups are traversed;
step72: adding the minimum coordinate value in the predicts to the Jpeaks, storing the Jpeaks_final, and updating q=q+1;
Step73: and repeating Step 34-Step 72 until all the predictjgroups are traversed, and outputting the Jpeaks_final as a final J peak positioning result.
It will be appreciated by those skilled in the art that embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks. These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks and/or block diagram block or blocks.
In one typical configuration, a computing device includes one or more processors (CPUs), input/output interfaces, network interfaces, and memory.
The memory may include volatile memory in a computer-readable medium, random Access Memory (RAM) and/or nonvolatile memory, etc., such as Read Only Memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.
Computer readable media, including both non-transitory and non-transitory, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of storage media for a computer include, but are not limited to, phase change memory (PRAM), static Random Access Memory (SRAM), dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), read Only Memory (ROM), electrically Erasable Programmable Read Only Memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), digital Versatile Discs (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices, or any other non-transmission medium, which can be used to store information that can be accessed by a computing device. Computer-readable media, as defined herein, does not include transitory computer-readable media (transmission media), such as modulated data signals and carrier waves.
It should also be noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article or apparatus that comprises an element.
The foregoing is merely exemplary of the present application and is not intended to limit the present application. Various modifications and changes may be made to the present application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc. which are within the spirit and principles of the present application are intended to be included within the scope of the claims of the present application.
Claims (8)
1. A ballistocardiogram signal positioning method, comprising:
acquiring a ballistocardiogram signal of a target object;
inputting the ballistocardiogram signals into a pre-trained target detection model to obtain a plurality of cardiac cycle prediction areas of the ballistocardiogram signals and prediction probabilities corresponding to the cardiac cycle prediction areas;
Determining the cardiac cycle prediction area with the highest prediction probability as an initial cardiac cycle template;
positioning a cardiac cycle prediction signal of the ballistocardiogram signal and J peak positions of the cardiac cycle prediction signal according to the signal morphology distance curve of the cardiac cycle template and the plurality of cardiac cycle prediction areas;
the step of inputting the ballistocardiogram signal into a pre-trained target detection model to obtain a plurality of cardiac cycle prediction areas of the ballistocardiogram signal and prediction probabilities corresponding to the cardiac cycle prediction areas comprises the following steps:
inputting the ballistocardiogram signals into a pre-trained target detection model to obtain a plurality of initial cardiac cycle prediction areas and corresponding prediction probabilities of the ballistocardiogram signals;
determining the initial cardiac cycle prediction region with the prediction probability higher than a preset probability threshold as the cardiac cycle prediction region;
obtaining a first interval threshold according to the average interval of the cardiac cycle prediction area and a preset threshold multiplying power;
if the current distance between the current cardiac cycle prediction area and the next cardiac cycle prediction area is larger than the first distance threshold, determining the current distance as an abnormal distance;
Acquiring midpoint spacing of adjacent initial cardiac cycle prediction areas;
if the range of the abnormal interval is larger than the midpoint interval of a preset second interval threshold, determining the initial cardiac cycle prediction area with the largest prediction probability as the cardiac cycle prediction area according to the prediction probability of the initial cardiac cycle prediction area corresponding to the midpoint interval, and re-determining whether the abnormal interval is present or not and whether the range of the abnormal interval is larger than the midpoint interval of the second interval threshold or not;
the step of locating the cardiac cycle predicted signal of the ballistocardiogram signal and the J-peak position of the cardiac cycle predicted signal according to the signal morphology distance curve of the cardiac cycle template and the plurality of cardiac cycle predicted areas comprises the following steps:
grouping the prediction areas of the cardiac cycle to obtain a plurality of prediction area groups;
traversing each prediction region group according to the cardiac cycle template to locate cardiac cycle prediction signals of each prediction region group and J peak positions of the cardiac cycle prediction signals according to the signal morphology distance curve;
wherein locating the J peak position of the cardiac cycle predicted signal for each of the set of predicted regions comprises:
Acquiring the average interval of each prediction area group;
determining a peak of the maximum value of the prediction signal of the first cardiac cycle of the prediction area group as a J peak, and acquiring a peak point interval between the current J peak and the midpoint of the prediction signal of the next cardiac cycle;
and acquiring the J peak of the predicted signal of the next cardiac cycle according to the average interval, the peak point interval and the position of the current J peak.
2. The ballistocardiogram signal positioning method according to claim 1, wherein the target detection model comprises a backbone network, a spatial pyramid pooling structure, an up-sampling layer, a time sequence characteristic convolution network and a prediction layer;
the backbone network is used for extracting the characteristics of the input signals;
the input of the spatial pyramid pooling structure is connected with the output of the backbone network; the spatial pyramid pooling structure is used for carrying out multi-scale abstract processing on the characteristics of the input signals to obtain the characteristics of the input signals fused with the multi-scale signals to be output to the up-sampling layer;
the input of the up-sampling layer is connected with the input of the spatial pyramid pooling structure, and the up-sampling layer is used for up-sampling the characteristics of the input signals to obtain high-resolution characteristics;
The input of the time sequence characteristic convolution network is connected with the output of the up-sampling layer, the output of the time sequence characteristic convolution network is connected with the input of the prediction layer, and the time sequence characteristic convolution network is used for carrying out bidirectional time sequence characteristic extraction on high resolution characteristics to obtain the high resolution characteristics comprising the time sequence characteristics and outputting the high resolution characteristics to the prediction layer;
and the input of the prediction layer is connected with the output of the time sequence characteristic convolution network, and the prediction layer is used for outputting a prediction result according to the high-resolution characteristic.
3. The ballistocardiogram signal positioning method according to claim 1, wherein the step of obtaining the average interval of each of the prediction area groups comprises:
calculating the starting point coordinates of the current first cardiac cycle prediction region, the ending point coordinates of the current last cardiac cycle prediction region and the current number of the cardiac cycle prediction regions of the prediction region group to obtain an initial average interval;
deleting the first cardiac cycle prediction area and the last cardiac cycle prediction area of the prediction area group after each initial average interval is calculated, and recalculating to obtain a next initial average interval until the prediction area group is emptied;
And obtaining the average interval of the prediction region group according to a plurality of initial average intervals corresponding to the prediction region group, and restoring all the deleted cardiac cycle prediction regions into the prediction region group.
4. The ballistocardiogram signal positioning method according to claim 1, wherein the step of acquiring the J peak of the predicted signal of the next cardiac cycle based on the average interval, the peak point interval, and the position of the current J peak comprises:
and if the quotient of the peak point interval and the average interval is not in a preset value range, determining the maximum value of the predicted signal of the next cardiac cycle as a J peak.
5. The ballistocardiogram signal positioning method according to claim 1, wherein the step of acquiring the J peak of the predicted signal of the next cardiac cycle based on the average interval, the peak point interval, and the position of the current J peak comprises:
if the quotient of the peak point interval and the average interval is not in a preset value range, obtaining a predicted point position of a J peak of the predicted signal of the next cardiac cycle according to the position of the current J peak and the average interval;
determining a maximum value of a prediction signal of a next cardiac cycle as a J peak when the prediction point is not within a range of the prediction region of the next cardiac cycle;
When the prediction point position is in the range of the prediction area of the next cardiac cycle, acquiring the maximum two maxima of the prediction signal of the next cardiac cycle;
updating the predicted signal of the next cardiac cycle according to the signal value of each point position of the predicted signal of the next cardiac cycle, the length of the predicted signal of the next cardiac cycle, the predicted point position and the maximum two maximum values;
and determining the maximum value of the updated prediction signal of the next cardiac cycle as a J peak.
6. The ballistocardiogram signal positioning method according to claim 1, wherein: in the process of positioning the J peak positions of the cardiac cycle prediction signals of the prediction region groups, the cardiac cycle template is updated according to the cardiac cycle prediction signals of the prediction region groups every time the J peak positions of one prediction region group are positioned.
7. The ballistocardiogram signal positioning method according to claim 1, wherein the step of determining the cardiac cycle prediction area with the highest prediction probability as an initial cardiac cycle template comprises:
grouping the prediction areas of the cardiac cycle to obtain a plurality of prediction area groups;
Determining the cardiac cycle prediction region with the highest prediction probability in the first prediction region group as an initial cardiac cycle template;
determining a peak conforming to the morphological characteristics of the J peak signal as a J peak from the initial cardiac cycle template to locate the J peak position of the initial cardiac cycle template; and if the peak which accords with the morphology characteristics of the J peak signal does not exist in the initial cardiac cycle template, determining the peak closest to the midpoint of the initial cardiac cycle template as the J peak so as to locate the J peak position of the initial cardiac cycle template.
8. The ballistocardiogram signal positioning method according to claim 1, wherein the object detection model is further used for detecting motion artifacts of the ballistocardiogram signal;
the step of inputting the ballistocardiogram signal into a pre-trained target detection model to obtain a plurality of cardiac cycle prediction areas of the ballistocardiogram signal and prediction probabilities corresponding to the cardiac cycle prediction areas comprises the following steps:
inputting the ballistocardiogram signals into a pre-trained target detection model to obtain a plurality of cardiac cycle prediction areas of the ballistocardiogram signals, prediction probabilities corresponding to the cardiac cycle prediction areas and a plurality of motion artifacts;
The step of determining the cardiac cycle prediction region with the highest prediction probability as an initial cardiac cycle template comprises the following steps:
dividing the ballistocardiogram signal into a plurality of signal segments according to the motion artifact;
determining the cardiac cycle prediction area with the highest prediction probability in each signal segment as an initial cardiac cycle template;
positioning a cardiac cycle prediction signal of the ballistocardiogram signal and a J peak position of the cardiac cycle prediction signal according to a signal morphology distance curve of the cardiac cycle template and the plurality of cardiac cycle prediction areas, wherein the method comprises the following steps of:
and positioning a cardiac cycle prediction signal of the corresponding signal segment and J peak positions of the cardiac cycle prediction signal according to the signal morphology distance curve of the initial cardiac cycle template and the plurality of cardiac cycle prediction areas.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310273416.XA CN116369907B (en) | 2023-03-20 | 2023-03-20 | Ballistocardiogram signal positioning method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310273416.XA CN116369907B (en) | 2023-03-20 | 2023-03-20 | Ballistocardiogram signal positioning method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116369907A CN116369907A (en) | 2023-07-04 |
| CN116369907B true CN116369907B (en) | 2024-02-13 |
Family
ID=86960756
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310273416.XA Active CN116369907B (en) | 2023-03-20 | 2023-03-20 | Ballistocardiogram signal positioning method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116369907B (en) |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104783787A (en) * | 2015-04-24 | 2015-07-22 | 太原理工大学 | J-wave detecting method based on neural network |
| CN107569226A (en) * | 2017-09-27 | 2018-01-12 | 广州中科新知科技有限公司 | HRV method and application is obtained based on piezoelectric sensing |
| CN110420019A (en) * | 2019-07-29 | 2019-11-08 | 西安电子科技大学 | A kind of depth recurrence heart rate estimation method of ballistocardiography signal |
| CN110507318A (en) * | 2019-08-16 | 2019-11-29 | 武汉中旗生物医疗电子有限公司 | A kind of electrocardiosignal QRS wave group localization method and device |
| WO2020114068A1 (en) * | 2018-12-07 | 2020-06-11 | 上海数创医疗科技有限公司 | Electrocardiosignal st section automatic determination method and apparatus based on artificial intelligence technology |
| CN114010186A (en) * | 2022-01-11 | 2022-02-08 | 华南师范大学 | Ballistocardiogram signal positioning method and computer equipment |
| CN114041786A (en) * | 2022-01-11 | 2022-02-15 | 华南师范大学 | Ballistocardiogram signal detection method, ballistocardiogram signal detection device and ballistocardiogram signal detection equipment |
| CN114098721A (en) * | 2022-01-25 | 2022-03-01 | 华南师范大学 | Ballistocardiogram signal extraction method, ballistocardiogram signal extraction device and ballistocardiogram signal extraction equipment |
| CN114287903A (en) * | 2021-12-31 | 2022-04-08 | 佳禾智能科技股份有限公司 | Heart rate detection method and device based on piezoelectric sensor and storage medium |
| CN114983374A (en) * | 2022-06-09 | 2022-09-02 | 中物云信息科技(无锡)有限公司 | Method for extracting characteristics of BCG signal and calculating heart rate in complex environment |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10188305B2 (en) * | 2015-07-09 | 2019-01-29 | Drägerwerk AG & Co. KGaA | Locating J-points in electrocardiogram signals |
| CN111035367B (en) * | 2019-12-31 | 2021-05-18 | 华南师范大学 | Signal detection system for judging sleep apnea |
-
2023
- 2023-03-20 CN CN202310273416.XA patent/CN116369907B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104783787A (en) * | 2015-04-24 | 2015-07-22 | 太原理工大学 | J-wave detecting method based on neural network |
| CN107569226A (en) * | 2017-09-27 | 2018-01-12 | 广州中科新知科技有限公司 | HRV method and application is obtained based on piezoelectric sensing |
| WO2020114068A1 (en) * | 2018-12-07 | 2020-06-11 | 上海数创医疗科技有限公司 | Electrocardiosignal st section automatic determination method and apparatus based on artificial intelligence technology |
| CN110420019A (en) * | 2019-07-29 | 2019-11-08 | 西安电子科技大学 | A kind of depth recurrence heart rate estimation method of ballistocardiography signal |
| CN110507318A (en) * | 2019-08-16 | 2019-11-29 | 武汉中旗生物医疗电子有限公司 | A kind of electrocardiosignal QRS wave group localization method and device |
| CN114287903A (en) * | 2021-12-31 | 2022-04-08 | 佳禾智能科技股份有限公司 | Heart rate detection method and device based on piezoelectric sensor and storage medium |
| CN114010186A (en) * | 2022-01-11 | 2022-02-08 | 华南师范大学 | Ballistocardiogram signal positioning method and computer equipment |
| CN114041786A (en) * | 2022-01-11 | 2022-02-15 | 华南师范大学 | Ballistocardiogram signal detection method, ballistocardiogram signal detection device and ballistocardiogram signal detection equipment |
| CN114098721A (en) * | 2022-01-25 | 2022-03-01 | 华南师范大学 | Ballistocardiogram signal extraction method, ballistocardiogram signal extraction device and ballistocardiogram signal extraction equipment |
| CN114983374A (en) * | 2022-06-09 | 2022-09-02 | 中物云信息科技(无锡)有限公司 | Method for extracting characteristics of BCG signal and calculating heart rate in complex environment |
Non-Patent Citations (3)
| Title |
|---|
| J peak extraction from non-standard ballistocardiography data: A preliminary study;Xin Li，Ye Li;2016 38th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC);688-691 * |
| 基于混合神经网络的脑电情感识别;蔡冬丽，钟清华，朱永升，张涵;华南师范大学学报;109-118 * |
| 心冲击伪迹抑制关键方法研究;刘亚丽;信息科技;23-50页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116369907A (en) | 2023-07-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| EP3506158B1 (en) | Method and apparatus for determining lane line on road | |
| CN110175630B (en) | Method and system for approximating deep neural networks for anatomical object detection | |
| CN109685768B (en) | Pulmonary nodule automatic detection method and system based on pulmonary CT sequence | |
| CN109685060B (en) | Image processing method and device | |
| JP6455113B2 (en) | Object tracking method and apparatus | |
| WO2018021576A1 (en) | Method for detecting object in image and objection detection system | |
| JP6341265B2 (en) | Accumulated object recognition method and apparatus | |
| US8035687B2 (en) | Image processing apparatus and program | |
| FR2549259A1 (en) | CHARACTER RECOGNIZING METHOD AND DEVICE | |
| CN109584266B (en) | Target detection method and device | |
| US20220114724A1 (en) | Image processing model generation method, image processing method and device, and electronic device | |
| CN111160065A (en) | Remote sensing image ship detection method, device, equipment and storage medium thereof | |
| CN114170212A (en) | False positive detection method and system based on small lung nodule in CT image | |
| Modasshir et al. | Enhancing coral reef monitoring utilizing a deep semi-supervised learning approach | |
| US20220230076A1 (en) | Data processing method and apparatus for machine learning | |
| CN111091101A (en) | High-precision pedestrian detection method, system and device based on one-step method | |
| CN111652825A (en) | Edge tracking straight line segment rapid detection device and method based on gradient direction constraint | |
| CN108460118A (en) | Time series data restorative procedure based on neighbour and device | |
| CN108038826A (en) | The bearing calibration of the shelf image of perspective distortion and device | |
| CN116369907B (en) | Ballistocardiogram signal positioning method | |
| US20120220855A1 (en) | Method and System for MR Scan Range Planning | |
| CN109948515B (en) | Object class identification method and device | |
| CN111724365A (en) | Interventional instrument detection method, system and device for endovascular aneurysm repair operation | |
| CN113288132B (en) | Method, apparatus, storage medium, and processor for predicting blood glucose level | |
| JP5078669B2 (en) | Target detection apparatus, target detection method, and target detection program |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |