Sample data selection method and system
Through the processing and screening of heart shot data based on the group model, the target heart shot data is selected as sample data, which solves the problems of large number of labels and low accuracy in individualized training, and improves training efficiency and accuracy.
Patent Information
- Application Number
- CN202311100002.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-08-28
- Publication Date
- 2025-02-28
AI Technical Summary
In the field of heart shot type recognition, in the prior art, a large amount of heart shot sample data is required to be marked during individualized training, resulting in low efficiency and difficult to ensure accuracy.
By obtaining the ECG data of the modeled object, pre-processing and processing each heart beat data based on the group model, determining its probability data and heart beat type, thereby filtering out the target heart beat data as sample data and reducing the annotation amount.
It improves the efficiency of individualized training, reduces the amount of sample data annotation, and ensures the accuracy of individual models.
Smart Images
Figure CN119523490A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of sample screening, and in particular to a sample data selection method and system. Background Art
[0002] In the current field of heartbeat type recognition, group models (heartbeat type recognition models trained based on the heartbeat data of a large number of different users) are mostly used to identify the user's heartbeat type. This approach may lead to deviations in the recognition results, and then individual models for single users have emerged (using the heartbeat data of a single user as a sample, and performing transfer learning / individualized training on the group model to obtain a heartbeat type recognition model that is only applicable to the user). However, under the conventional individual model training method, a large number of heartbeat sample data of the user need to be labeled, which makes the aforementioned transfer learning process inefficient and the accuracy of the individual model still difficult to guarantee.
[0003] Based on this, we hope to propose a sample data selection method and system that can reduce the amount of individual heart rate sample data annotation in the individualized training process while ensuring the accuracy of the personal model, thereby improving the efficiency of individualized training. Summary of the invention
[0004] One or more embodiments of the present specification provide a sample data selection method. The sample data selection method includes: acquiring ECG data of a modeling object; preprocessing the ECG data to obtain preprocessed ECG data; for each heartbeat data, processing the heartbeat data based on a group model to determine probability data corresponding to the heartbeat data; determining the heartbeat type to which the heartbeat data belongs based on the probability data; for each heartbeat type, determining target heartbeat data from the heartbeat data corresponding to the heartbeat type, and determining the target heartbeat data corresponding to all heartbeat types as sample heartbeat data.
[0005] One or more embodiments of the present specification provide a sample data selection system. The sample data selection system includes: an acquisition module for acquiring ECG data of a modeling object; a preprocessing module for preprocessing the ECG data to obtain the preprocessed ECG data; a probability determination module for processing the ECG data based on a population model for each heartbeat data to determine the probability data corresponding to the heartbeat data; a type determination module for determining the heartbeat type to which the heartbeat data belongs based on the probability data; and a sample determination module for determining the target heartbeat data from the heartbeat data corresponding to each heartbeat type, and determining the target heartbeat data corresponding to all heartbeat types as sample heartbeat data.
[0006] One or more embodiments of the present specification provide a sample data selection device, including a processor, wherein the processor is configured to execute a sample data selection method.
[0007] One or more embodiments of the present specification provide a computer-readable storage medium, wherein the storage medium stores computer instructions. When a computer reads the computer instructions in the storage medium, the computer executes a sample data selection method.
[0008] One or more embodiments of the present specification provide an individual model training method. The individual model training method comprises: obtaining sample heartbeat data according to any one of claims 1 to 5; if the total number of sample heartbeat data is not zero, labeling the sample heartbeat data; and adjusting the group model based on the labeled sample heartbeat data to obtain an individual model. BRIEF DESCRIPTION OF THE DRAWINGS
[0009] This specification will be further described in the form of exemplary embodiments, which will be described in detail by the accompanying drawings. These embodiments are not restrictive, and in these embodiments, the same number represents the same structure, wherein:
[0010] Figure 1 is a schematic diagram of an application scenario of a sample data selection system according to some embodiments of this specification;
[0011] Figure 2 is an exemplary flow chart of a sample data selection method according to some embodiments of this specification;
[0012] Figure 3 is an exemplary flow chart of a pre-processing method according to some embodiments of this specification;
[0013] Figure 4 is an exemplary flow chart of a method for determining target heartbeat data according to some embodiments of this specification;
[0014] Figure 5 is an exemplary flow chart of a method for determining target heartbeat data based on a data group according to some embodiments of the present specification;
[0015] Figure 6 is a module diagram of a sample data selection system according to some embodiments of this specification;
[0016] Figure 7 It is an exemplary flowchart of the individual model training method shown in some embodiments of this specification. DETAILED DESCRIPTION
[0017] In order to more clearly illustrate the technical solutions of the embodiments of this specification, the following is a brief introduction to the drawings required for the description of the embodiments. Obviously, the drawings described below are only some examples or embodiments of this specification. For ordinary technicians in this field, this specification can also be applied to other similar scenarios based on these drawings without creative work. Unless it is obvious from the language environment or otherwise explained, the same reference numerals in the figures represent the same structure or operation.
[0018] It should be understood that the "system", "device", "unit" and / or "module" used herein are a method for distinguishing different components, elements, parts, portions or assemblies at different levels. However, if other words can achieve the same purpose, the words can be replaced by other expressions.
[0019] As shown in this specification and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "comprise" and "include" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.
[0020] Flowcharts are used in this specification to illustrate the operations performed by the system according to the embodiments of this specification. It should be understood that the preceding or following operations are not necessarily performed precisely in order. Instead, the steps may be processed in reverse order or simultaneously. At the same time, other operations may also be added to these processes, or one or more operations may be removed from these processes.
[0021] Figure 1 It is a schematic diagram of an application scenario of a sample data selection system according to some embodiments of this specification.
[0022] like Figure 1 As shown, in some embodiments, an application scenario 100 of a sample data selection system (hereinafter referred to as application scenario 100) may include a processing device 110, a network 120, a storage device 130, a data collection device 140, and a user terminal 150. The application scenario 100 may determine sample data by implementing the methods and / or processes disclosed in this specification, and annotate the sample data to adjust the group model and obtain an individual model.
[0023] The processing device 110 refers to a device for processing data and / or information from at least one component of the application scenario 100 or an external data source (e.g., a cloud data center). The processing device 110 can access data or information from the storage device 130, the data acquisition device 140, and / or the user terminal 150 through the network 120. The processing device 110 can also be directly connected to the storage device 130, the data acquisition device 140, and / or the user terminal 150 to access information and / or data. For example, the processing device 110 can obtain data of the modeling object from the data acquisition device 140 (e.g., real-time ECG data of the modeling object) and / or obtain data of the modeling object from the storage device 130 (e.g., historical ECG data of the modeling object). The processing device 110 can process the acquired data and / or information. For example, the processing device 110 can obtain preprocessed ECG data based on preprocessing the ECG data; for each heartbeat data, process the heartbeat data based on the group model to determine the probability data corresponding to the heartbeat data; determine the heartbeat type to which the heartbeat data belongs based on the probability data; for each heartbeat type, determine the target heartbeat data from the heartbeat data corresponding to the heartbeat type, and determine the target heartbeat data corresponding to all heartbeat types as sample heartbeat data. In some embodiments, the processing device 110 can be a single server or a server group. The processing device 110 can be local or remote. For more information about modeling objects, preprocessed ECG data, probability data, heartbeat types, and target heartbeat data, see Figure 2 and its related description.
[0024] The network 120 may connect the various components of the application scenario 100 and / or connect the application scenario 100 with external resources. In some embodiments, one or more components of the application scenario 100 (e.g., the processing device 110, the storage device 130, the data acquisition device 140, and / or the user terminal 150) may exchange information and / or data through the network 120.
[0025] In some embodiments, the network 120 may be any one or more of a wired network or a wireless network. In some embodiments, the network 120 may include one or more network access points. For example, the network 120 may include a wired or wireless network access point (e.g., a base station and / or a network switching point), through which one or more components of the application scenario 100 may be connected to the network 120 to exchange data and / or information.
[0026] The storage device 130 refers to a device for storing data, instructions and / or any other information. In some embodiments, the storage device 130 may store data and / or information obtained from the processing device 110, the data acquisition device 140, etc. For example, the storage device 130 may store the ECG data of the modeled object obtained from the data acquisition device 140. In some embodiments, the storage device 130 may include a large-capacity memory, a removable memory, etc., or any combination thereof.
[0027] The data acquisition device 140 refers to a device used to acquire the electrocardiogram data of the modeling object. For example, the data acquisition device 140 may be an electrocardiograph, a dynamic electrocardiograph, etc. used to acquire the electrocardiogram data of the modeling object.
[0028] The user terminal 150 refers to a terminal device used by a data annotation user. The data annotation user refers to a user who labels the sample heartbeat data selected by the sample data selection system, such as a doctor, an experiment organizer, etc. The user terminal 150 can be a mobile device, a tablet computer, a laptop computer, a desktop computer, or any combination thereof, having input and / or output functions.
[0029] Figure 2 is an exemplary flow chart of a sample data selection method according to some embodiments of this specification. Figure 2 As shown, the process 200 includes the following steps: In some embodiments, the process 200 may be executed by the sample data selection system 500 .
[0030] Step 210 , obtaining ECG data of the modeled object. In some embodiments, step 210 may be performed by the obtaining module 610 .
[0031] Modeling objects refer to people who need to build individual models, such as patients and experimenters who need to build individual models. For more information about individual models, please refer to the following text.
[0032] The ECG data refers to real-time ECG data and / or historical ECG data of the modeling object.
[0033] The ECG data refers to ECG waveform data of the modeled object within a preset time period, wherein the duration of the preset time period can be preset, for example, 5 minutes. The ECG data can be acquired based on a data acquisition device (for example, a data acquisition device 140) and / or acquired from a data storage device (for example, a storage device 130).
[0034] Step 220, preprocess the ECG data to obtain preprocessed ECG data, where the preprocessed ECG data includes at least one heartbeat data. In some embodiments, step 220 may be performed by the preprocessing module 620. It is understandable that the number of preprocessed ECG data may be more than one.
[0035] Heartbeat data refers to data including all or part of the ECG waveform of a heartbeat. Preprocessing methods may include resampling, denoising, data segmentation, etc.
[0036] In some embodiments, the preprocessing module 620 can convert the format of the ECG data based on the first preset algorithm to obtain the first data; filter the first data to obtain the second data; for each second data, based on the second preset algorithm, mark the characteristic band in the second data to obtain the first candidate data; determine the second candidate data based on the first candidate data; determine the initial heartbeat data based on the second candidate data; determine the characteristic data of the initial heartbeat data based on the initial heartbeat data; determine the standardized characteristic data of the initial heartbeat data based on the characteristic data; determine the heartbeat data based on the initial heartbeat data and the standardized characteristic data. For more information about the above embodiments, please refer to Figure 3 and its related description.
[0037] Step 230, for each heartbeat data, the heartbeat data is processed based on the group model to determine the probability data corresponding to the heartbeat data, where the group model is a classification model trained based on the existing heartbeat data. The existing heartbeat data may include historical heartbeat data including historical electrocardiogram data collected from a large number of different people (e.g., patients or experimental subjects, etc.). In some embodiments, step 230 may be performed by the probability determination module 630.
[0038] In some embodiments, the input of the group model may be heartbeat data, and the output may be probability data corresponding to each heartbeat data.
[0039] Probability data refers to the classification result data output by the group model, reflecting the possibility that the heartbeat data belongs to various heartbeat types. In some embodiments, the probability data includes the probability value of the corresponding heartbeat data belonging to at least one heartbeat type. For example, the probability data may be "the probability that the heartbeat data belongs to the first type is 0.3, and the probability that it belongs to the second type is 0.7". It is understandable that the sum of the probability values in the probability data is 1.
[0040] The heartbeat types that the heartbeat data may belong to and the number of heartbeat types can be preset. For example, the heartbeat types can be preset to 7, namely normal heartbeat, left bundle branch block, right bundle branch block, ventricular premature beat, atrial premature beat, ventricular escape beat, R-on-T (ventricular premature contraction occurs too early and appears on the T wave of the previous heartbeat cycle), and the probability data is the probability value that the heartbeat data belongs to the above types respectively. For another example, the types of heartbeat data can also be preset to 2, namely the first type and the second type, wherein the first type includes ventricular premature beat, ventricular escape beat, R-on-T (that is, ventricular premature beat, ventricular escape beat, R-on-T are all considered as the first type), and the second type includes normal heartbeat, left bundle branch block, right bundle branch block, atrial premature beat (that is, normal heartbeat, left bundle branch block, right bundle branch block, atrial premature beat are all considered as the second type), and the probability data is the probability value that the heartbeat data belongs to the above two types respectively.
[0041] In some embodiments, the probability data may be data in the form of a vector, and the values of the elements of each dimension thereof respectively represent the probability that the heartbeat data belongs to each heartbeat type. For example, the group model may be a binary classification model (the corresponding heartbeat types include the first type and the second type), and the corresponding probability data may be (0.2, 0.8), which means "the probability that the heartbeat data belongs to the first type is 0.2, and the probability that it belongs to the second type is 0.8". For another example, the group model may be a multi-classification (the number of classification results is greater than 2) model (the corresponding heartbeat types include the first type, the second type, ..., the Nth type, where N is equal to the number of heartbeat types), and the corresponding probability data may be (0.2, 0.3, ..., 0.1), which means "the probability that the heartbeat data belongs to the first type is 0.2, the probability that it belongs to the second type is 0.3, ..., and the probability that it belongs to the Nth type is 0.1". It can be understood that the sum of the values of the elements of each dimension of the probability data is 1.
[0042] In some embodiments of the present specification, by setting the group model as a classification model, the probability of heartbeat data belonging to each category can be accurately and efficiently determined.
[0043] Step 240 , determining the heartbeat type to which the heartbeat data belongs based on the probability data. In some embodiments, step 240 may be performed by the type determination module 640 .
[0044] In some embodiments, the type determination module 640 can determine the heartbeat type to which the heartbeat data belongs based on the maximum value of the probability value in the probability data. For example, assuming that the number of types of heartbeat data is 3 (respectively the first type, the second type and the third type), the probability data of a certain heartbeat data is (0.13, 0.7, 0.17), wherein the element value of each dimension corresponds to the probability value of the heartbeat data belonging to the above three heartbeat types, then the maximum value of its probability value (hereinafter referred to as the maximum probability value) is 0.7 of the second dimension, indicating that the probability that the heartbeat data belongs to the second type is the largest, and the heartbeat type of the heartbeat data will be determined as the second type.
[0045] Step 250 , for each heartbeat type, determine target heartbeat data from the heartbeat data corresponding to the heartbeat type, and determine the target heartbeat data corresponding to all heartbeat types as sample heartbeat data. In some embodiments, step 250 may be performed by the sample determination module 650 .
[0046] Target heartbeat data refers to the heartbeat data that meets the preset conditions selected from the heartbeat data. For more information about the preset conditions, see Figure 5 and its related description.
[0047] In some embodiments, the sample determination module 650 can set a data group corresponding to the heartbeat type based on the heartbeat type; for each heartbeat data, divide the heartbeat data into a corresponding data group based on the heartbeat type to which the heartbeat data belongs; for each data group, determine the target heartbeat data from the heartbeat data included in the data group. For more information about the above embodiments, please refer to Figure 4 and its related description.
[0048] The sample heartbeat data refers to heartbeat data used to adjust the group model.
[0049] In some embodiments of the present specification, the sample data required to establish an individual model can be obtained quickly and efficiently through the above-mentioned sample data selection method, and the amount of sample data is small, which can effectively reduce the time consumed by labelers in subsequent labeling work while ensuring the accuracy of the individual model.
[0050] Figure 3 is an exemplary flow chart of the pre-processing method according to some embodiments of this specification. Figure 3 As shown, the process 300 includes the following steps: In some embodiments, the process 300 may be executed by the pre-processing module 620 .
[0051] Step 310: Based on the first preset algorithm, convert the format of the ECG data to obtain the first data. For more information about ECG data, see Figure 2 and its related description.
[0052] The first preset algorithm refers to an algorithm for adjusting the format of the ECG data, for example, interpolation, deconvolution, random downsampling, etc. In some embodiments, the first preset algorithm may be a spline sampling algorithm.
[0053] The format of the ECG data refers to the sampling frequency of the ECG data. For example, the format of the ECG data may be 200 Hz.
[0054] The first data refers to ECG data with a unified format obtained by format conversion of the ECG data. The unified format means that different ECG data are in the same format after conversion. Preferably, the format of the first data can be 250Hz. It can be understood that the first data can correspond to the ECG data one by one.
[0055] In some embodiments, the first preset algorithm may be used multiple times (eg, 3 times) to adjust the same ECG data to obtain the first data.
[0056] Step 320: filter the first data to obtain second data.
[0057] Filtering refers to the process of removing wave points in a specific frequency range in the first data, wherein the specific frequency range can be preset, for example, the specific frequency range can be 0.05 Hz to 100 Hz. In some embodiments, the first data can be filtered based on a Butterworth bandpass.
[0058] The second data refers to the electrocardiogram data obtained after filtering the first data. It can be understood that the second data can correspond to the first data one by one.
[0059] Step 330 : for each second data, based on a second preset algorithm, mark the characteristic bands in the second data to obtain first candidate data.
[0060] The second preset algorithm refers to an algorithm for identifying and marking characteristic bands in the second data. Characteristic bands refer to special bands with high recognition in the second data (for example, R waves in the ECG waveform). For example, the second preset algorithm may be a PT (Pan-Tompkins) algorithm, which can identify and mark R waves in ECG data.
[0061] The first candidate data refers to the ECG band data marked based on the second preset algorithm in the second data. It can be understood that one second data can determine multiple first candidate data, and the first candidate data can correspond to the characteristic band one by one.
[0062] Step 340: Determine at least one second candidate data based on the at least one first candidate data.
[0063] The second candidate data refers to the ECG band data obtained after data cleaning based on the first candidate data. It can be understood that the second candidate data can correspond to the first candidate data one by one.
[0064] Data cleaning refers to the process of removing redundant data in the first candidate data. The redundant data may include repeated marking of the same characteristic band by the second preset algorithm and the annotation information accompanying the marking (for example, information marking the duration of the characteristic band).
[0065] Step 350: Determine initial heartbeat data based on the second candidate data.
[0066] The initial heartbeat data refers to the electrocardiogram band data obtained by data segmentation based on the second candidate data. It can be understood that the initial heartbeat data can correspond to the second candidate data one by one.
[0067] Data segmentation refers to the process of intercepting part of the data of the preset time length for the second candidate data and removing the rest of the data. For example, taking the characteristic band as R wave and the preset time length as 0.7 seconds as an example, data segmentation can be based on the R wave peak point (the wave point with the highest central electric potential value of the R wave), retaining the wave points within 0.25 seconds before the peak point and 0.45 seconds after the peak point, and removing the other wave points of the second candidate data to obtain the waveform data as the initial heartbeat data.
[0068] Step 360: determining feature data of the initial heartbeat data based on the initial heartbeat data.
[0069] The characteristic data includes waveform characteristic data and time characteristic data of the initial heartbeat data.
[0070] The waveform characteristic data refers to data reflecting the waveform information of the initial heartbeat data. In some embodiments, the waveform characteristic data may be data in the form of a vector. For example, the waveform characteristic data may be (0.17, 0.17, ..., 0.18, ..., 0.23, ..., 0.18), which means that in the waveform of the initial heartbeat data, the electrocardiographic potential values corresponding to each wave point are 0.17mV, 0.17mV, ..., 0.18mV, ..., 0.23mV, ..., 0.18mV respectively.
[0071] The waveform characteristic data can be directly read and determined based on the waveform diagram of the initial heartbeat data and the time points corresponding to each wave point in the waveform.
[0072] Time characteristic data refers to the time relationship data between the initial heartbeat data and other initial heartbeat data adjacent to it in time sequence (i.e., the previous initial heartbeat data and the next initial heartbeat data). The time characteristic data may include the time interval between the initial heartbeat data and the previous initial heartbeat data (hereinafter referred to as the previous time interval) and the time interval between the initial heartbeat data and the next initial heartbeat data (hereinafter referred to as the next time interval). For example, the time characteristic data may be "the previous time interval is 0.75s, and the next time interval is 0.72s".
[0073] In some embodiments, the time characteristic data of the initial heartbeat data can be determined based on the reference point of the characteristic band, wherein the reference point of the characteristic band is a wave point of the characteristic band selected manually. For example, still taking the characteristic band as the R wave and the reference point as the R wave peak point as an example, the front time interval can be the time interval between the R wave peak point of the initial heartbeat data and the R wave peak point of the previous initial heartbeat data, and the rear time interval can be the time interval between the R wave peak point of the initial heartbeat data and the R wave peak point of the subsequent initial heartbeat data.
[0074] Step 370: Determine standardized feature data of the initial heartbeat data based on the feature data.
[0075] The standardized characteristic data refers to data obtained after the characteristic data is standardized. The standardized characteristic data may include standardized waveform characteristic data and standardized time characteristic data, wherein the standardized time characteristic data may include a time interval before standardization and a time interval after standardization.
[0076] In some embodiments, the feature data may be normalized based on the following formula (1):
[0077] Where i = 1 or 2 or 3. When i = 1, X1 represents the standardized waveform feature data, X 10 represents the waveform feature data, μ1 represents the mean of the waveform feature data of all initial heartbeat data, σ1 represents the variance of the waveform feature data of all initial heartbeat data; when i = 2, X2 represents the time interval before standardization, X 20 represents the previous time interval, μ2 represents the mean of the previous time interval of all initial heartbeat data, σ2 represents the variance of the previous time interval of all initial heartbeat data; when i = 3, X3 represents the standardized time interval, X 30 represents the post-time interval, μ3 represents the mean of the post-time intervals of all initial heartbeat data, and σ3 represents the variance of the post-time intervals of all initial heartbeat data.
[0078] Step 380, determining the heartbeat data based on the initial heartbeat data and the standardized feature data.
[0079] In some embodiments, the initial heartbeat data may be combined with the corresponding standardized feature data to obtain the corresponding heartbeat data. It is understandable that the heartbeat data may correspond one-to-one with the initial heartbeat data.
[0080] In some embodiments of the present specification, the heartbeat data can be determined efficiently and accurately through the above method for determining the heartbeat data based on preprocessing.
[0081] It should be noted that the above description of process 300 is only for illustration and explanation, and does not limit the scope of application of this specification. For those skilled in the art, various modifications and changes can be made to process 300 under the guidance of this specification. However, these modifications and changes are still within the scope of this specification. For example, the meaning of σ in formula (1) is changed from variance to standard deviation.
[0082] Figure 4 is an exemplary flow chart of a method for determining target heartbeat data according to some embodiments of this specification. Figure 4 As shown, the process 400 includes the following steps. In some embodiments, the process 400 may be executed by the sample determination module 650 .
[0083] Step 410: Based on the heartbeat type, set the data group corresponding to the heartbeat type. For more information about heartbeat types, see Figure 2 and its related description.
[0084] A data group refers to a data set for collecting heartbeat data. It can be understood that before the heartbeat data is divided into a data group, the data group is an empty set. For more information about heartbeat data, see Figure 2 and / or Figure 3 and its related description.
[0085] In some embodiments, the data group can be set based on the probability value range. For example, multiple data groups can be set, and the corresponding probability value ranges are (0.9, 1], (0.8, 0.9], (0.7, 0.8], ...
[0086] In some embodiments, data groups can be set based on heartbeat types, and the number of data groups can be set to be the same as the number of heartbeat types, that is, as many data groups as there are heartbeat types, and there can be a one-to-one correspondence between data groups and heartbeat types.
[0087] Step 420 , for each heartbeat data, classify the heartbeat data into corresponding data groups based on the heartbeat type to which the heartbeat data belongs.
[0088] Exemplarily, assuming that there are three heartbeat types (namely the first type, the second type and the third type), and there are also three corresponding data groups (namely the first data group, the second data group and the third data group, corresponding to the first type, the second type and the third type respectively), if the heartbeat type to which a certain heartbeat data belongs is the second type, then the heartbeat data will be classified into the second data group.
[0089] In some embodiments, the sample determination module 650 can also determine the probability value range where the maximum probability value is located for any one of the heartbeat data based on the maximum probability value in the probability data corresponding to the heartbeat data, and classify the heartbeat data into a data group corresponding to the probability value range. For example, if the maximum probability value of a certain heartbeat data is 0.69, then the heartbeat data will be classified into a data group with a corresponding probability value range of (0.6, 0.7]. For more information about probability data, see Figure 2 and / or Figure 3 and its related description.
[0090] Step 430: for each data group, determine the target heartbeat data from the heartbeat data included in the group data.
[0091] In some embodiments, the sample determination module 650 can determine candidate data from the heartbeat data included in the data group based on the probability data and the preset conditions for each data group; sort the candidate data based on the probability data; and determine the target heartbeat data from the sorted candidate data based on the data selection strategy. For more information about the above embodiments, please refer to Figure 5 and its related description.
[0092] In some embodiments of the present specification, grouping heartbeat data based on type can make the target heartbeat data subsequently selected based on the data group more diverse and have a wider coverage, and also make the user's labeling of sample heartbeat data more convenient and efficient.
[0093] Figure 5 is an exemplary flow chart of a method for determining target heartbeat data based on a data set according to some embodiments of this specification. Figure 5 As shown, the process 500 includes the following steps. In some embodiments, the process 500 can be executed by the sample determination module 650.
[0094] Step 510: Determine candidate data from the heartbeat data included in the data group based on the probability data and preset conditions.
[0095] The preset condition means that the maximum probability value of the probability data of the heartbeat data is less than the probability threshold. The probability threshold can be preset, for example, preferably, the probability threshold can be preset to 0.75. For more information about probability data, see Figure 2 and its related description.
[0096] Candidate data refers to heartbeat data in the data group that meets the preset conditions. For example, assuming that the probability data of a heartbeat data in the second data group is (0.15, 0.55, 0.3), and the probability threshold is preset to 0.75, since the maximum probability value of the heartbeat data is 0.55 (less than 0.75), the heartbeat data meets the preset conditions. Based on this, the heartbeat data can be determined as a candidate data. For more information about data groups, see Figure 4 and its related description.
[0097] Step 520, sorting the candidate data based on the probability data.
[0098] In some embodiments, the candidate data may be sorted in descending order of maximum probability value based on the maximum probability value of the probability data of the candidate data.
[0099] Step 530: Determine the target heartbeat data from the sorted candidate data based on the data selection strategy. Figure 2 and its related description.
[0100] In some embodiments, the sample determination module 650 can, for each data group: determine the number of alternative data in the data group; in response to the number of alternative data being greater than the target number, determine the target heart beat data from the alternative data based on the data selection strategy; in response to the number being less than or equal to the target number and not zero, determine all the alternative data as target heart beat data; in response to the number being zero, do not determine the target heart beat data from the data group.
[0101] The data selection strategy refers to a method for selecting a target number of target heartbeat data from the sorted candidate data. The target number can be preset, for example, the target number can be 5.
[0102] In some embodiments, the data selection strategy may include uniform interval selection and / or non-uniform interval selection with a bias toward low probability.
[0103] Uniform interval selection means that when selecting the sorted candidate data, the number of candidate data intervals between the selected candidate data is equal. For example, assuming that there are 15 sorted candidate data in total and the target number is 5, the 2nd, 5th, 8th, 11th, and 14th candidate data (the number of candidate data intervals between the selected candidate data is 2) can be selected as the target heartbeat data.
[0104] Non-uniform interval selection with a bias toward low probability means that when selecting the sorted candidate data, the larger the maximum probability value of the probability data of the candidate data, the more candidate data are selected with intervals between the selected candidate data. For example, assuming that there are 15 sorted candidate data in total, the target number is 5, and the candidate data is sorted from large to small, then the 1st, 8th, 12th, 14th, and 15th candidate data can be selected (the number of candidate data intervals between the selected candidate data decreases as the maximum probability value decreases) as the target heartbeat data.
[0105] In some embodiments of the present specification, based on the above-mentioned method for determining target heartbeat data based on a data group and the data selection strategy, the coverage of the selected sample data can be made wider, making the individual model trained subsequently more accurate.
[0106] In some embodiments of the present specification, by determining alternative data, the selected data contains more valuable information, which can improve the adaptability and accuracy of the individual model trained subsequently; by sorting the alternative data before selecting, the targetedness of the selected data to the modeling object can be enhanced, and after sorting, alternative data with different amounts of information (the probability data of the alternative data is different, and the amount of information is different, the smaller the maximum probability value, the greater the amount of information) are clear at a glance, and data containing more information can be accurately selected, which is more beneficial to the subsequent training of individual models.
[0107] Figure 6 It is a module diagram of a sample data selection system according to some embodiments of the present specification.
[0108] In some embodiments, Figure 6 As shown, the sample data selection system 600 may include an acquisition module 610 , a preprocessing module 620 , a probability determination module 630 , a type determination module 640 and a sample determination module 650 .
[0109] The acquisition module 610 can be used to acquire the ECG data of the modeled object. For more information about the functions of the acquisition module 610, see Figure 2 and its related description.
[0110] The preprocessing module 620 can be used to preprocess the ECG data to obtain preprocessed ECG data. For more information about the functions of the preprocessing module 620, see Figure 2 and / or Figure 3 and its related description.
[0111] The probability determination module 630 can be used to process each heartbeat data based on the group model to determine the probability data corresponding to the heartbeat data. For more information about the function of the probability determination module 630, see Figure 2 and its related description.
[0112] The type determination module 640 may be used to determine the heartbeat type to which the heartbeat data belongs based on the probability data. For more information about the functions of the type determination module 640, see Figure 2 and its related description.
[0113] The sample determination module 650 can be used to determine the target heartbeat data from the heartbeat data corresponding to each heartbeat type, and determine the target heartbeat data corresponding to all heartbeat types as the sample heartbeat data. For more information about the functions of the sample determination module 650, see Figure 2 and / or Figure 4 and its related description.
[0114] It should be noted that the above description of the sample data selection system 600 and its modules is only for convenience of description and does not limit the present specification to the scope of the embodiments. It is understandable that, after understanding the principle of the system, those skilled in the art may arbitrarily combine the modules or form a subsystem to connect with other modules without deviating from the principle. In some embodiments, Figure 6 The acquisition module 610, preprocessing module 620, probability determination module 630, type determination module 640 and sample determination module 650 disclosed in the specification can be different modules in a system, or a module can realize the functions of two or more modules mentioned above. For example, each module can share a storage module, or each module can have its own storage module. Such variations are all within the protection scope of this specification.
[0115] Figure 7 is an exemplary flow chart of the individual model training method according to some embodiments of this specification. Figure 7 As shown, process 700 includes the following steps.
[0116] Step 710: Obtain sample heartbeat data according to any one of the methods of claims 1 to 5. For more information about the above method, see Figure 2 and its related description.
[0117] Step 720: If the total number of sample heartbeat data is not zero, mark the sample heartbeat data.
[0118] The total number of sample heartbeat data refers to the sum of the number of sample heartbeat data corresponding to all heartbeat types, and its value is equal to the sum of the number of target heartbeat data determined from each data group. For more information about heartbeat types, data groups, and target heartbeat data, see Figure 2 and its related description.
[0119] Labeling refers to labeling sample heartbeat data, where the label is the heartbeat type to which the sample heartbeat data belongs. The labeling process can be performed by a data labeling user (eg, a doctor, an experiment organizer, etc.) through a user terminal (eg, user terminal 150).
[0120] Step 730: adjust the group model based on the labeled sample heartbeat data to obtain an individual model.
[0121] Adjustment refers to the use of labeled sample heartbeat data to train the group model (i.e., individualized training) through the method of transfer learning. After the training, the group model becomes an individual model.
[0122] The individual model refers to a personal model obtained by adjusting the group model based on the labeled sample heartbeat data for the modeling object. The individual model can be used to process the heartbeat data corresponding to the electrocardiogram data of the modeling object and determine the heartbeat type corresponding to the aforementioned heartbeat data. The input of the individual model can be the aforementioned heartbeat data, and the output can be the heartbeat type corresponding to the aforementioned heartbeat data.
[0123] For more information about the group model, see Figure 2 and its related description.
[0124] In some embodiments, if the total number of sample heartbeat data is zero, the group model is not adjusted and is determined as the individual model.
[0125] The basic concepts have been described above. Obviously, for those skilled in the art, the above detailed disclosure is only for example and does not constitute a limitation of this specification. Although not explicitly stated here, those skilled in the art may make various modifications, improvements and corrections to this specification. Such modifications, improvements and corrections are suggested in this specification, so such modifications, improvements and corrections still belong to the spirit and scope of the exemplary embodiments of this specification.
[0126] At the same time, this specification uses specific words to describe the embodiments of this specification. For example, "one embodiment", "an embodiment", and / or "some embodiments" refer to a certain feature, structure or characteristic related to at least one embodiment of this specification. Therefore, it should be emphasized and noted that "one embodiment" or "an embodiment" or "an alternative embodiment" mentioned twice or more in different positions in this specification does not necessarily refer to the same embodiment. In addition, certain features, structures or characteristics in one or more embodiments of this specification can be appropriately combined.
[0127] In addition, unless explicitly stated in the claims, the order of the processing elements and sequences described in this specification, the use of alphanumeric characters, or the use of other names are not intended to limit the order of the processes and methods of this specification. Although the above disclosure discusses some invention embodiments that are currently considered useful through various examples, it should be understood that such details are only for illustrative purposes, and the attached claims are not limited to the disclosed embodiments. On the contrary, the claims are intended to cover all modifications and equivalent combinations that are consistent with the essence and scope of the embodiments of this specification. For example, although the system components described above can be implemented by hardware devices, they can also be implemented only by software solutions, such as installing the described system on an existing server or mobile device.
[0128] Similarly, it should be noted that in order to simplify the description disclosed in this specification and thus help understand one or more embodiments of the invention, in the above description of the embodiments of this specification, multiple features are sometimes combined into one embodiment, figure or description thereof. However, this disclosure method does not mean that the features required by the subject matter of this specification are more than the features mentioned in the claims. In fact, the features of the embodiments are less than all the features of the single embodiment disclosed above.
[0129] In some embodiments, numbers describing the number of components and attributes are used. It should be understood that such numbers used in the description of the embodiments are modified by the modifiers "about", "approximately" or "substantially" in some examples. Unless otherwise specified, "about", "approximately" or "substantially" indicate that the numbers are allowed to vary by ±20%. Accordingly, in some embodiments, the numerical parameters used in the specification and claims are approximate values, which may change according to the required features of individual embodiments. In some embodiments, the numerical parameters should take into account the specified significant digits and adopt the general method of retaining digits. Although the numerical domains and parameters used to confirm the breadth of their range in some embodiments of this specification are approximate values, in specific embodiments, the setting of such numerical values is as accurate as possible within the feasible range.
[0130] Each patent, patent application, patent application publication, and other materials, such as articles, books, specifications, publications, documents, etc., cited in this specification are hereby incorporated by reference in their entirety. Except for application history documents that are inconsistent with or conflicting with the contents of this specification, documents that limit the broadest scope of the claims of this specification (currently or later attached to this specification) are also excluded. It should be noted that if the descriptions, definitions, and / or use of terms in the materials attached to this specification are inconsistent or conflicting with the contents described in this specification, the descriptions, definitions, and / or use of terms in this specification shall prevail.
[0131] Finally, it should be understood that the embodiments described in this specification are only used to illustrate the principles of the embodiments of this specification. Other variations may also fall within the scope of this specification. Therefore, as an example and not a limitation, alternative configurations of the embodiments of this specification may be considered consistent with the teachings of this specification. Accordingly, the embodiments of this specification are not limited to the embodiments explicitly introduced and described in this specification.
Claims
1. A sample data selection method, characterized in that: include: Obtaining ECG data of the modeled object; Preprocessing the ECG data to obtain preprocessed ECG data, wherein the preprocessed ECG data includes at least one heartbeat data; For each of the heartbeat data, the heartbeat data is processed based on a group model to determine probability data corresponding to the heartbeat data, wherein the group model is a classification model trained based on existing heartbeat data; Determine the heartbeat type to which the heartbeat data belongs based on the probability data; For each of the heartbeat types, target heartbeat data is determined from the heartbeat data corresponding to the heartbeat type, and the target heartbeat data corresponding to all of the heartbeat types are determined as sample heartbeat data.
2. The method according to claim 1, characterized in that For each of the heartbeat types, determining target heartbeat data from the heartbeat data corresponding to the heartbeat type includes: Based on the heartbeat type, setting a data group corresponding to the heartbeat type; For each of the heartbeat data, based on the heartbeat type to which the heartbeat data belongs, the heartbeat data is divided into corresponding data groups; For each of the data groups, the target heartbeat data is determined from the heartbeat data included in the group of data.
3. The method according to claim 1, characterized in that The probability data includes a probability value of the corresponding heartbeat data belonging to at least one heartbeat type, and determining the heartbeat type to which the heartbeat data belongs based on the probability data includes: The heartbeat type to which the heartbeat data belongs is determined based on a maximum value of the probability values in the probability data.
4. The method according to claim 2, characterized in that For each of the data groups, determining the target heartbeat data from the heartbeat data included in the group of data includes: Based on the probability data and a preset condition, determining candidate data from the heartbeat data included in the data group; sorting the candidate data based on the probability data; Based on the data selection strategy, the target heartbeat data is determined from the sorted candidate data.
5. The method according to claim 4, characterized in that The data selection strategy includes uniform interval selection and / or non-uniform interval selection with a bias toward low probability. The step of determining the target heartbeat data from the sorted candidate data based on the data selection strategy includes: determining the number of the candidate data in the data group; In response to the number of the candidate data being greater than the target number, determining the target heartbeat data from the candidate data based on a data selection strategy; In response to the number of the candidate data being less than or equal to the target number and not being zero, determining all of the candidate data as the target heartbeat data; In response to the number of the candidate data being zero, the target heartbeat data is not determined from the data group.
6. A sample data selection system, characterized in that: include: An acquisition module, used for acquiring the ECG data of the modeling object; A preprocessing module, used for preprocessing the ECG data to obtain preprocessed ECG data, wherein the preprocessed ECG data includes at least one heartbeat data; A probability determination module, for processing each heartbeat data based on a group model to determine probability data corresponding to the heartbeat data, wherein the group model is a classification model trained based on existing sample heartbeat data; A type determination module, which determines the heartbeat type to which the heartbeat data belongs based on the probability data; The sample determination module is used to determine target heartbeat data from the heartbeat data corresponding to each heartbeat type, and determine the target heartbeat data corresponding to all the heartbeat types as sample heartbeat data.
7. A sample data selection device, comprising a storage medium and a processor, wherein the storage medium stores computer instructions, and the processor is configured to execute at least part of the computer instructions to implement the method according to any one of claims 1 to 5.
8. A storage medium for storing computer instructions, when at least a part of the computer instructions is executed by a processor, the method according to any one of claims 1 to 5 is implemented.
9. An individual model training method, characterized in that: include: Acquiring sample heartbeat data according to the method of any one of claims 1 to 5; If the total amount of the sample heartbeat data is not zero, marking the sample heartbeat data; The group model is adjusted based on the labeled sample heartbeat data to obtain an individual model.
10. The method according to claim 9, characterized in that Also includes: If the total number of the sample heartbeat data is zero, the group model is not adjusted, and the group model is determined as the individual model.