CN112244855B - Electrocardio data-oriented data processing system - Google Patents

Abstract

The embodiment of the invention relates to a data processing system facing electrocardio data, which comprises: the system comprises a data integration module, an electrocardio database, a data cleaning module, a data segmentation module and a data conversion module. According to the data processing system provided by the embodiment of the invention, the collected electrocardio data is classified and stored according to the interference degree through the data integration module; the electrocardio data without training value is filtered through the data cleaning module, so that the quality of the electrocardio data is improved; the data segmentation module is used for segmenting the electrocardiogram data according to the data requirements of the electrocardiogram data calculation model, so that the applicability of the electrocardiogram data is widened; through the data conversion module, the selected section of the electrocardiogram data is converted into new training data through left-right shifting or subsection division, and the diversity of the electrocardiogram signal data is increased.

Description

A data processing system for ECG data

technical field

The invention relates to the technical field of data processing, in particular to a data processing system oriented to electrocardiographic data.

Background technique

With the promotion of big data, deep learning and other technologies, based on artificial intelligence (AI) models, ECG data calculation models for ECG data have begun to be applied in the field of traditional ECG data analysis. For the ECG data calculation model, before it is officially put into commercial use, it is necessary to use a large amount of ECG data for training iterations, and only when the analysis accuracy and stability of the training become reliable can it be officially output. During the training process, if the quality or quantity of the training data is insufficient, it will directly affect the iterative efficiency of the ECG data calculation model.

SUMMARY OF THE INVENTION

The purpose of the present invention is to aim at the defects of the prior art, to provide a data processing system for ECG data, to classify the collected ECG data according to the degree of interference through the data integration module, and to control the influence of the interference data; The data cleaning module filters the ECG data that does not have training value to improve the quality of the ECG data; through the data segmentation module, according to the data requirements of the ECG data calculation model, the ECG data is segmented to widen the range of ECG data. Applicability: Through the data conversion module, the selected segment of the ECG data can be converted to obtain new training data through the left-right shift of the ECG data, the left-right shift of the heartbeat, or the division of sub-segments, which increases the diversity of the ECG signal data.

In order to achieve the above purpose, an embodiment of the present invention provides a data processing system for ECG data, the data processing system includes: a data integration module, an ECG database, a data cleaning module, a data segmentation module and a data conversion module; The data processing system is respectively connected with the ECG data source, the first ECG data calculation model and the second ECG data calculation model;

The data integration module is respectively connected with the ECG data source and the ECG database; the data integration module is used to obtain the ECG data sequence and the corresponding ECG information set from the ECG data source; For the ECG information set, perform ECG data classification processing to generate classified data; and then store the ECG data sequence and the ECG information set into the corresponding data in the ECG database according to the classified data Subset;

The data cleaning module is respectively connected with the ECG database and the data segmentation module; the data cleaning module is used for preparing the calculation data of the first ECG data calculation model from the ECG database Obtain the ECG data sequence and the corresponding ECG information set; and perform available data labeling processing according to the set first cleaning type data and the ECG information set; The ECG data sequence and the corresponding ECG information set are sent to the data segmentation module;

The data segmentation module is connected with the first ECG data calculation model; the data segmentation module is used for the ECG data sequence and the ECG information set according to the set segmentation length data and sliding window length. data, perform ECG data segment division processing, and generate multiple ECG segment data sequences and corresponding ECG segment information sets; and then send the multiple ECG segment data sequences and the corresponding ECG segment information sets to the sending the first ECG data calculation model;

The data cleaning module is also connected with the data conversion module; the data cleaning module is further configured to acquire the ECG from the ECG database when preparing the calculation data of the second ECG data calculation model data sequence and the corresponding ECG information set; and according to the set second cleaning type data and the ECG information set, perform available data labeling processing on the ECG data sequence; The ECG data sequence of available data and the corresponding ECG information set are sent to the data conversion module;

The data conversion module is also connected with the second ECG data calculation model; the data conversion module is used to perform segment marking processing on the ECG data sequence to obtain selected segments; and then use the selected segments as Basically, according to the set conversion mode data and the ECG information set, perform data conversion processing to generate a conversion segment data sequence and a corresponding conversion segment ECG information set; then convert the conversion segment data sequence and the corresponding ECG information set. Convert the segment ECG information set, and send it to the second ECG data calculation model.

preferably,

The ECG data sequence includes a plurality of ECG data;

The ECG information set includes a heartbeat type information subset, an interference information subset and a low voltage information subset;

The heartbeat type information subset includes a plurality of first heartbeat type information sequences; the first heartbeat type information sequence includes first heartbeat type data and a plurality of first R point position data; the first R point position data, which is the index position information of the corresponding ECG data in the ECG data sequence;

The interference information subset includes a plurality of first interference information sequences; the first interference information sequence includes first interference type information and a plurality of first interference location information groups; the first interference type information is type 1 interference information , Type 2 interference information or Type 3 interference information, wherein the interference levels of the Type 1 interference information, the Type 2 interference information and the Type 3 interference information are sequentially increased; the first interference location information group includes the first Interference start information and first interference end information; the first interference start information and the first interference end information are respectively the index position information of the corresponding ECG data in the ECG data sequence ;

The subset of low voltage information includes a plurality of first low voltage information sequences; the first low voltage information sequence includes a plurality of first low voltage position information groups; the first low voltage position information groups include first low voltages Start information and first low-voltage end information; the first low-voltage start information and the first low-voltage end information are respectively the indexes of the corresponding ECG data in the ECG data sequence location information;

The ECG database includes a type 1 interference data subset, a type 2 interference data subset, a type 3 interference data subset, a low voltage data subset and a non-interference data subset.

preferably,

The data integration module is specifically configured to acquire the ECG data sequence and the corresponding ECG information set sent from the ECG data source;

and perform ECG data classification processing according to the low-voltage information subset and the interference information subset of the ECG information set to generate the classified data; When the total number of a low-voltage information sequence is 0, and the total number of valid first interference information sequences in the interference information subset is not 0, the first interference information sequence with the highest interference level among all the first interference information sequences One piece of interference type information is used as the highest level of interference information, and the classification data is set and processed according to the highest level of interference information; when the highest level of interference information is the first type of interference information, the classification data is set The data is a type-1 interference type; when the highest-level interference information is the type-2 interference information, the classified data is set to be a type-2 interference type; when the highest-level interference information is the type-3 interference information, Set the classification data to be the three-type interference type; when the total number of valid first low-voltage information sequences in the low-voltage information subset is not 0, set the classification data to be of the low-voltage type; When the total number of valid first low-voltage information sequences in the voltage information subset is 0, and the total number of valid first interference information sequences in the interference information subset is also 0, set the classification data as no interference type;

Then according to the classification data, the ECG data sequence and the corresponding ECG information set are stored and processed; when the classification data is the type 1 interference type, the ECG data sequence and the corresponding ECG information are stored and processed; The ECG information set is stored in the type 1 interference data subset of the ECG database; when the classified data is the type 2 interference type, the ECG data sequence and the corresponding The ECG information set is stored in the type 2 interference data subset of the ECG database; when the classified data is the type 3 interference type, the ECG data sequence and the corresponding ECG information are stored Collecting the three-type interference data subset stored in the ECG database; when the classified data is the low-voltage type, storing the ECG data sequence and the corresponding ECG information set in the ECG database the low-voltage data subset of the ECG database; when the classified data is the non-interference type, the ECG data sequence and the corresponding ECG information set are stored in the ECG database the undisturbed data subset.

preferably,

The data cleaning module is specifically configured to, when preparing the calculation data of the first ECG data calculation model, obtain from the data subset corresponding to the ECG database according to the calculation data requirements of the first ECG data calculation model. the ECG data sequence and the corresponding ECG information set;

Then according to the first cleaning type data and the ECG information set, the available data labeling processing is performed on the ECG data sequence; when the first cleaning type data is the fast heart rate cleaning type, according to the heart rate The heartbeat type information subset of the electrical information set, perform heart rate conversion processing to generate first heart rate data, and when the first heart rate data does not exceed the set heart rate upper limit, mark the electrocardiogram data sequence as the Available data; when the first cleaning type data is an interference cleaning type, perform three types of interference proportion calculation processing according to the interference information subset of the ECG information set to generate the first type 1 proportion data, The first and second types of proportion data and the first and third types of proportion data, when the first type of proportion data does not exceed the set upper limit of the first type of proportion, and the first and second types of proportion data does not exceed the set When the upper limit of the proportion of the second type is set, and the data of the first and third types of proportion does not exceed the set upper limit of the proportion of the third type, the ECG data sequence is marked as the available data; when the first cleaning When the type data is fast heart rate and interference cleaning type, first heart rate conversion processing is performed according to the heartbeat type information subset to generate second heart rate data, and when the second heart rate data does not exceed the heart rate upper limit, the second heart rate data is generated according to the heart rate The subset of interference information is obtained, and three types of interference proportion calculation processing are performed to generate the second type I proportion data, the second type II proportion data and the second and third type proportion data. When the second type I proportion data When the upper limit of the proportion of type 1 is not exceeded, the data of the proportion of the second type of type 2 does not exceed the upper limit of the proportion of type 2, and the proportion of data of the first type and the third type does not exceed the upper limit of the proportion of type 3 , marking the ECG data sequence as the available data.

Further, the data cleaning module is further configured to perform data filtering processing on the ECG data sequence that is not marked as the available data.

preferably,

The data segmentation module is specifically used to set a first start index position and an end index position; and from the first start index position, the segmentation length data is used as the segmentation length, and the first index is divided from the ECG data sequence. each of the ECG segment data sequences, and perform the corresponding ECG segment information conversion processing according to the ECG information set to generate the corresponding first ECG segment information set, and then convert the first start index position The sum with the sliding window length data is used as the second starting index position; then from the second starting index position, the segment length data is used as the segment length, and the second is divided from the ECG data sequence. The ECG segment data sequence, and the corresponding ECG segment information conversion processing is performed according to the ECG information set to generate the corresponding second ECG segment information set, and then the second start index position The sum with the sliding window length data is used as the third starting index position; until the calculated starting index position exceeds the ending index position.

preferably,

The data conversion module is specifically configured to mark the selected segment in the ECG data sequence according to the set selected segment position; wherein, the selected segment position includes the selected segment start index data and the selected segment. Determining segment end index data; the selected segment start index data and the selected segment end index data are respectively the index position information of the corresponding ECG data in the ECG data sequence;

When the conversion mode data is the first conversion mode, randomly generate ECG left-shift data; and use the difference between the selected segment start index data and the ECG left-shift data as the first ECG start data Start index data, take the difference between the selected segment end index data and the ECG left-shift data as the first ECG end index data; and then extract the index position in the ECG data sequence from the ECG data sequence. The ECG data between the first ECG start index data and the first ECG end index data forms a first conversion segment data sequence; and then perform a corresponding first conversion segment according to the ECG information set information conversion processing, generating a corresponding first conversion segment ECG information set; wherein the first conversion segment ECG information set includes the first conversion segment heartbeat type information subset, the first conversion segment interference information subset and the first conversion segment. a subset of low-voltage information of the conversion segment;

When the conversion mode data is the second conversion mode, randomly generate ECG right-shift data; and use the sum of the selected segment start index data and the ECG right-shift data as the second ECG start start index data, take the sum of the selected segment end index data and the ECG right-shift data as the second ECG end index data; and then extract the index position in the ECG data sequence from the ECG data sequence. The ECG data between the second ECG start index data and the second ECG end index data forms a second conversion segment data sequence; and then perform a corresponding second conversion segment according to the ECG information set The information conversion process generates a corresponding second conversion segment ECG information set; wherein the second conversion segment ECG information set includes the second conversion segment heartbeat type information subset, the second conversion segment interference information subset and the first conversion segment. Two conversion segment low-voltage information subsets;

When the conversion mode data is the third conversion mode, the fourth conversion mode or the fifth conversion mode, sort all the first R point position data in the ECG information set to generate a first R point position data sequence; according to the first R point position data sequence and the first heartbeat type data corresponding to each of the first R point position data, perform statistical processing of heartbeat type and position information to generate a heartbeat Statistical data sequence; wherein the heartbeat statistical data sequence includes a plurality of heartbeat statistical data sets; the heartbeat statistical data sets include heartbeat type data, heartbeat start data and heartbeat end data; the heartbeat The start data and the heartbeat end data are respectively the index position information of the corresponding ECG data in the ECG data sequence;

When the conversion mode data is the third conversion mode, in the heartbeat statistical data sequence, the heartbeat start data and the heartbeat end data satisfy the selected segment start index data The heartbeat statistical data group is regarded as the first heartbeat data group; and the index position information corresponding to the first heartbeat data group in the heartbeat statistical data sequence is taken as the first heartbeat index data; then use the difference between the first heartbeat index data and the set heartbeat left shift data as the second heartbeat index data; and then extract the second heartbeat index data from the heartbeat statistical data sequence. The heartbeat start data of the heartbeat statistics data group corresponding to the heartbeat index data is used as the third ECG start index data; and then in the ECG data sequence, the third ECG start index data is used The starting index data is the starting position for extraction, and the data length of the selected segment is taken as the extraction length, and ECG data extraction processing is performed to form a third converted segment data sequence; Three conversion segment information conversion processing to generate a corresponding third converted segment ECG information set; wherein the third converted segment ECG information set includes a third converted segment heartbeat type information subset, a third converted segment interference information subset set and the third conversion segment low voltage information subset;

When the conversion mode data is the fourth conversion mode, in the heartbeat statistical data sequence, the heartbeat start data and the heartbeat end data satisfy the selected segment start index data The said heartbeat statistical data group is taken as the second heartbeat data group; and the index position information corresponding to the second heartbeat data group in the heartbeat statistical data sequence is taken as the third heartbeat index data; then use the sum of the third heartbeat index data and the set heartbeat right-shift data as the fourth heartbeat index data; and then extract and the fourth heartbeat index data from the heartbeat statistical data sequence The heartbeat start data of the heartbeat statistics data group corresponding to the heartbeat index data is used as the fourth ECG start index data; and then in the ECG data sequence, the fourth ECG start index data is used The starting index data is the starting position for extraction, and the data length of the selected segment is taken as the extraction length, and the ECG data extraction process is performed to form a fourth converted segment data sequence; Four-converted segment information conversion processing to generate a corresponding fourth converted segment ECG information set; wherein the fourth converted segment ECG information set includes a fourth converted segment heartbeat type information subset, a fourth converted segment interference information subset set and the fourth transition segment low voltage information subset;

When the conversion mode data is the fifth conversion mode, according to the preset sub-segment length, sub-segment division processing is performed on the selected segment to generate a plurality of sub-segments; and then according to the heartbeat statistical data sequence, For each of the sub-segments, perform statistical processing of the total length of the heartbeat of the same heartbeat type in the segment, generate a plurality of total length data, and use the heartbeat type corresponding to the maximum total length data as the corresponding sub-segment. Fragment heartbeat type data; and then all the sub-segments form a fifth converted fragment data sequence, and all the fragment heartbeat type data form a fifth converted fragment electrocardiographic information set.

The embodiment of the present invention provides a data processing system for ECG data. The collected ECG data is classified according to the degree of interference through the data integration module, and the influence of the interference data is controlled; the data cleaning module has no training value. The quality of the ECG data is improved by filtering the ECG data; the data segmentation module, according to the data requirements of the ECG data calculation model, segment the ECG data, which broadens the applicability of the ECG data; through data conversion The module converts the selected segment of the ECG data to obtain new training data through the left-right shift of the ECG data, the left-right shift of the heartbeat, or the division of sub-segments, which increases the diversity of the ECG signal data.

Description of drawings

FIG. 1 is a block diagram of a data processing system for ECG data according to an embodiment of the present invention.

Detailed ways

In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. . Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

An embodiment of the present invention provides a data processing system for ECG data. As shown in FIG. 1 , as shown in the module structure diagram of a data processing system for ECG data provided by an embodiment of the present invention, the data processing system 1 mainly includes : data integration module 11, ECG database 12, data cleaning module 13, data segmentation module 14 and data conversion module 15; the data processing system 1 is respectively connected with the ECG data source, the first ECG data calculation model and the second ECG data Computational model connections.

Here, the ECG data source is a collection or storage device for ECG signal data. After the ECG data source performs data sampling and analysis on the original ECG signal, it outputs a sampled ECG data sequence and a sequence corresponding to the ECG data sequence. ECG information collection; the first ECG data calculation model and the second ECG data calculation model are artificial intelligence calculation models for ECG data analysis, and both of them obtain ECG data for training through the data processing system 1 and the corresponding ECG information set used to verify the training results.

Here, before the modules of the data processing system 1 are further elaborated, the connection relationship of the data processing system 1 is first described in a unified manner, as shown in FIG. The source is connected with the internal ECG database 12; the ECG database 12 is also connected with the data cleaning module 13; the data cleaning module 13 is respectively connected with the data segmentation module 14 and the data conversion module 15; Model connection; the data conversion module 15 is connected with the external second ECG data calculation model.

After the description of the connection relationship is completed, the functions of the multiple modules mentioned in the embodiments of the present invention and the specific implementation of each module will be described in detail below.

The data integration module 11 is used to obtain the ECG data sequence and the corresponding ECG information set from the ECG data source; and perform the ECG data classification processing according to the ECG information set to generate classified data; The data sequence and the ECG information set are stored in the corresponding data subset in the ECG database 12 . The data subsets of the ECG database 12 specifically include a type 1 interference data subset 121 , a type 2 interference data subset 122 , a type 3 interference data subset 123 , a low voltage data subset 124 and a non-interference data subset 125 .

Here, the electrocardiographic data sequence includes a plurality of electrocardiographic data. Here, the ECG data sequence is a data sequence obtained from the above-mentioned ECG data source after sampling and processing the original ECG signal, and each ECG data included therein is a specific signal sample value.

Here, the ECG information set includes a heartbeat type information subset, an interference information subset and a low voltage information subset;

The heartbeat type information subset includes a plurality of first heartbeat type information sequences; the first heartbeat type information sequence includes first heartbeat type data and a plurality of first R point position data; the first R point position data, which is The index position information of the corresponding ECG data in the ECG data sequence;

The interference information subset includes a plurality of first interference information sequences; the first interference information sequence includes first interference type information and a plurality of first interference location information groups; the first interference type information is type 1 interference information, type 2 interference information or Type 3 interference information, wherein the interference levels of Type 1 interference information, Type 2 interference information and Type 3 interference information increase in sequence; the first interference location information group includes first interference start information and first interference end information; The start information and the first interference end information are respectively the index position information of the corresponding ECG data in the ECG data sequence;

The low-voltage information subset includes a plurality of first low-voltage information sequences; the first low-voltage information sequence includes a plurality of first low-voltage position information groups; the first low-voltage position information group includes first low-voltage start information and a first Low-voltage end information; the first low-voltage start information and the first low-voltage end information are the index position information of the corresponding ECG data in the ECG data sequence, respectively.

Here, the ECG information set is a data set used to count the characteristic information of the corresponding ECG data sequence:

1) The heartbeat type information subset is used to count the heartbeat type corresponding to the heartbeat cycle where each R point in the ECG data sequence is located; in the subset, each heartbeat type information sequence corresponds to a heartbeat type data , and the R point position data of all R points that satisfy the heartbeat type; each R point position data is the index position of the ECG data identified as the R point in the ECG data sequence;

2) The interference information subset is used to count the types and positions of the data identified as interference signals in the ECG data sequence. an interference position information group consisting of the start and end positions of continuous ECG data; the interference start and end information of each interference position information group is the start and end position of a piece of interference signal data; each interference signal data The start and end positions of , are the index positions of a specific ECG data in the ECG data sequence; here, the interference type information is divided into type 1 interference information, type 2 interference information or type 3 interference information; for type 1 interference, It means that although an ECG signal is interfered and has baseline drift during the acquisition process, the complete heart cycle can be identified, that is, the QRS wave is visible; for type 2 interference, it means that an ECG signal is interfered during the acquisition process. Larger, the start and end positions of the heartbeat cycle are difficult to identify, but the maximum peak point of the heartbeat cycle can still be identified, that is, the R point is visible; for type III interference, it means that an ECG signal is disturbed during the collection process is very large, neither the heartbeat cycle nor the position of point R can be correctly identified; therefore, from the perspective of the severity of the interference, that is, the interference level, the type 1 interference is weaker than the type 2 interference, and the type 2 interference is weaker than the type 3 interference. interference;

3) The low-voltage information subset is used to count the positions identified as low-voltage signal data in the ECG data sequence; in the subset, the low-voltage start and end information of each low-voltage position information group corresponds to a segment of low-voltage The start and end positions of the signal data; the start and end positions of each low-voltage signal data are the index positions of a specific ECG data in the ECG data sequence; here, the low-voltage threshold is conventionally used as the judgment limit , ECG data that is continuously lower than this threshold will be regarded as low voltage data.

In a specific implementation provided by this embodiment, the data integration module 11 is specifically configured to acquire the ECG data sequence and the corresponding ECG information set sent from the ECG data source;

And according to the low-voltage information subset and the interference information subset of the ECG information set, the ECG data is classified and processed to generate classified data:

When the total number of valid first low-voltage information sequences in the low-voltage information subset is 0, and the total number of valid first interference information sequences in the interference information subset is not 0, the one with the highest interference level among all the first interference information sequences The first interference type information is used as the highest-level interference information, and the classified data is set and processed according to the highest-level interference information; when the highest-level interference information is type-1 interference information, the classified data is set as type-1 interference type; When the level interference information is type 2 interference information, set the classified data as type 2 interference type; when the highest level interference information is type 3 interference information, set the classified data as type 3 interference type;

When the total number of valid first low-voltage information sequences in the low-voltage information subset is not 0, setting the classification data as a low-voltage type;

When the total number of valid first low-voltage information sequences in the low-voltage information subset is 0, and the total number of valid first interference information sequences in the interference information subset is also 0, the classification data is set as a non-interference type;

Here, the total number of valid first low-voltage information sequences in the low-voltage information subset is 0, which means that there is no low-voltage signal data in the acquired ECG data sequence, and the total number of valid first interference information sequences in the interference information subset is not 0 means that there is interference signal data in the obtained ECG data sequence; count the highest-level interference type in the ECG data sequence as the highest-level interference information, and then set the classification data according to the specific value of the highest-level interference information: Type 1, 2, or 3 interference types;

Here, the fact that the total number of valid first low-voltage information sequences in the low-voltage information subset is not 0 means that the acquired ECG data sequence contains low-voltage signal data. The embodiment of the present invention stipulates that as long as the low-voltage signal data is included, no low-voltage signal data is included. For further interference type identification, the classification data are uniformly set as low voltage type;

Here, the total number of valid first low-voltage information sequences in the low-voltage information subset is 0, which means that there is no low-voltage signal data in the acquired ECG data sequence, and the total number of valid first interference information sequences in the interference information subset is also 0 means that there is no interference signal data in the obtained ECG data sequence, then the embodiment of the present invention determines that the ECG data sequence is undisturbed ECG data, and its classification data is set to a non-interference type;

Then according to the classified data, the electrocardiographic data sequence and the corresponding electrocardiographic information set are stored and processed; when the classified data is a type of interference type, the electrocardiographic data sequence and the corresponding electrocardiographic information set are stored in the electrocardiographic database 12. Type 1 interference data subset 121; when the classified data is type 2 interference type, the ECG data sequence and the corresponding ECG information set are stored in the type 2 interference data subset 122 of the ECG database 12; when the classified data is type 3 interference data When the type of interference type is classified, the ECG data sequence and the corresponding ECG information set are stored in the type-3 interference data subset 123 of the ECG database 12; when the classified data is of low voltage type, the ECG data sequence and the corresponding The electrical information set is stored in the low-voltage data subset 124 of the ECG database 12; when the classified data is a non-interference type, the ECG data sequence and the corresponding ECG information set are stored in the non-interference data subset of the ECG database 12 125.

Here, there are multiple data subsets in the ECG database 12 according to the embodiment of the present invention, the purpose of which is to classify and store the obtained ECG data sequence and the corresponding ECG information set; or three types of interference type, it means that the acquired ECG data sequence contains the new type of interference data but does not contain low-voltage signal data, the data integration module 11 will store the ECG data sequence and the ECG information set in the ECG database 12 Corresponding type 1 interference data subset 121, type 2 interference data subset 122 and type 3 interference data subset 123; when the classified data is of low voltage type, it means that the obtained ECG data sequence contains low voltage signal data, The data integration module 11 will store the ECG data sequence and the ECG information set in the low-voltage data subset 124 of the ECG database 12; when the classified data is of a non-interference type, the data integration module 11 will store the ECG data sequence and the ECG data set. The set of information is stored in the undisturbed data subset 125 of the ECG database 12 .

To sum up, the embodiment of the present invention uses the data integration module 11 to classify and store the collected ECG data according to the degree of interference, thereby discriminating the data type and quality of the original data, and improving the accuracy of the training data.

The data processing system 1 of the embodiment of the present invention serves for the calculation model of ECG data. On the one hand, it is used to collect ECG data for training the ECG data calculation model, and on the other hand, it is also used to calculate the input data of the model according to the ECG data. Requirement prepares calculation data for it. In the first ECG data calculation model in the embodiment of the present invention, the input data used for training is a plurality of ECG segment data obtained by dividing the original data, and the reference information used for verification of the training results comes from the corresponding ECG data A collection of fragment information. In order to achieve this purpose, the data processing system 1 uses the data segmentation module 14 to segment the ECG data sequence obtained from the ECG database 12, and correspondingly generates an ECG segment information set; before the data segmentation module 14 performs data segmentation, In order to further ensure the quality of the training data, it is also necessary to use the data cleaning module 13 to clean the ECG data sequence first. The specific implementation is described below.

The data cleaning module 13 is used to obtain the ECG data sequence and the corresponding ECG information set from the ECG database 12 when preparing the calculation data of the first ECG data calculation model; The ECG information set is processed for labeling the available data; and the ECG data sequence marked as available data and the corresponding ECG information set are sent to the data segmentation module 14 .

Here, the first cleaning type data is a preset value for selecting a cleaning process, and the first cleaning type data includes three cleaning types: fast heart rate cleaning type, interference cleaning type, and fast heart rate and interference cleaning type. Here, the operation process of the available data labeling processing is the specific data cleaning process. In the end, the qualified data will be marked as available data, and the unqualified data will be filtered out.

In another specific implementation provided by this embodiment, the data cleaning module 13 is specifically configured to, when preparing the calculation data of the first ECG data calculation model, In the data subset corresponding to the database 12, obtain the ECG data sequence and the corresponding ECG information set;

Then according to the first cleaning type data and the ECG information set, the available data annotation processing is performed on the ECG data sequence:

When the first cleaning type data is the fast heart rate cleaning type, heart rate conversion processing is performed according to the heartbeat type information subset of the ECG information set to generate the first heart rate data, and when the first heart rate data does not exceed the set heart rate upper limit, Mark the ECG data sequence as available data;

Here, the heart rate upper limit is the heart rate threshold used to identify whether it is fast heart rate data;

Here, when the first cleaning type data is the fast heart rate cleaning type, it means that the fast heart rate data needs to be identified for the ECG data sequence; the so-called fast heart rate data routinely refers to the data whose heart rate exceeds a certain heart rate upper limit, and the fast heart rate data will be It is filtered as unqualified data; the operation process of heart rate conversion processing may be to convert the heart rate to obtain the first heart rate data by counting the total number of the first R point position data in the subset of heartbeat type information, or to obtain the first heart rate data according to the first R point position data. Calculate the mean value of the heartbeat interval, and convert the heart rate with the mean value of the heartbeat interval to obtain the first heart rate data;

For example, the upper limit of the heart rate is set to 300, the current ECG data sequence is a 2-second period of ECG data, the heartbeat type information subset contains 6 first R point position data, and the first heart rate data is obtained by heart rate conversion = 6* (60/2)=180 (times/minute), because the first heart rate data is less than the upper limit of the heart rate (180<300), the current ECG data sequence is marked as available data;

For another example, the upper limit of the heart rate is set to 300, the current ECG data sequence is a 2-second period of ECG data, the heartbeat type information subset contains 6 first R point position data, and all the first R point position data are sorted. The first R point position data sequence is obtained, and the difference value calculation is performed on the adjacent first R point position data in the first R point position data sequence to obtain the first R point difference value data sequence including 5 difference data, Because the position data of the first R point is the index data of the ECG data, then the difference data is multiplied by the known sampling frequency of the ECG data to obtain the duration of the heartbeat interval between adjacent R points. The average value of the heartbeat interval is calculated by calculating the mean value of the heartbeat interval. The first heart rate data = 60/the mean value of the heartbeat interval can be obtained by performing the heart rate conversion with the heartbeat interval mean value. If the first heart rate data is less than the upper limit of the heart rate, the current heart rate Electrical data series are marked as available data;

When the first cleaning type data is of the interference cleaning type, according to the interference information subset of the ECG information set, three types of interference proportion calculation processing are performed to generate the first type 1 proportion data, the first type 2 proportion data and the third type Type 1 and Type 3 ratio data, when Type 1 Type 1 ratio data does not exceed the set Type 1 ratio upper limit, and Type 1 and Type 2 ratio data does not exceed the set Type 2 ratio upper limit, and Type 1 and Type 3 ratio data does not exceed the set Type 2 ratio upper limit When the proportion data does not exceed the set upper limit of the three-type proportion, the ECG data sequence is marked as available data;

Here, the upper limit of the proportion of type 1 is the set threshold for judging whether the data length of the type 1 interference signal in the ECG data sequence is too long; The set threshold for whether the ratio is too long; the upper limit of the three-type ratio is the set threshold for judging whether the length ratio of the three-type interference signal data in the ECG data sequence is too long;

Here, when the first cleaning type data is the interference cleaning type, it means that it is necessary to identify whether the interference length of the ECG data sequence is lower than the normal threshold; if the length of the interference signal in the ECG data sequence exceeds the set threshold, it means that The over-disturbance of the ECG data sequence cannot be used for normal data training; if the length of the interference signal in the ECG data sequence does not exceed the set threshold, it is considered that the quality meets the standard and can be regarded as available data; the proportion of three types of interference The operation process of the calculation processing is to separately calculate the interference proportion data of three types of interference: the first type 1 proportion data, the first type 2 proportion data and the first 3 type proportion data; for one of the specific interference types The calculation process of the proportion data is: in the interference information subset, use the first interference end and starting position information of all the first interference position information groups corresponding to this type of interference type to calculate the length of each interference signal, and then count all interferences. After the sum of the lengths of the signals, the ratio of the sum of the lengths to the total number of ECG data in the ECG data sequence is used as the interference proportion data of this type of interference;

For example, the upper limit of the proportion of type 1 is 20%, the upper limit of type 2 is 20%, and the upper limit of type 3 is 20%. The current ECG data sequence includes 100 ECG data, and the corresponding interference information of the ECG information set The subset includes 3 interference information sequences, and each interference information sequence has 3 interference location information groups, as shown in Table 1, then, the first type ratio data = ((5-1+1)+(10 -7+1)+(13-11+1))/100=12%, the first and second type proportion data=((16-14+1)+(19-17+1)+(25-21+ 1))/100=11%, the proportion of the first three types = ((43-40+1)+(47-45+1)+(50-49+1))/100=9%, because 12 %<20%, so the proportion of the first type is less than the upper limit of the proportion of the first type, because 11%<20%, so the proportion of the first and second type is less than the upper limit of the proportion of the second type, because 9%<20%, so The proportion of the first and third types of data is less than the upper limit of the proportion of the third type; because the proportion of the first and third types of data is smaller than the upper limit of the proportion of the first and third types, and the proportion of the first and second types of data is less than the upper limit of the proportion of the second and third types, and the proportion of the first and third types of The ratio data is less than the upper limit of the proportion of three types, so the current ECG data sequence is marked as available data;

Table I

When the first cleaning type data is fast heart rate and interference cleaning type, first, according to the heartbeat type information subset, heart rate conversion processing is performed to generate the second heart rate data, and when the second heart rate data does not exceed the upper limit of the heart rate, the second heart rate data is generated according to the heartbeat type information subset. Set, carry out the calculation and processing of the three types of interference proportions, and generate the second type of proportion data, the second type of proportion data and the second and third types of proportion data. When the second type of proportion data does not exceed the first type of proportion data When the upper limit is reached, the second type II proportion data does not exceed the second type proportion upper limit, and the first and third type proportion data does not exceed the third type proportion upper limit, the ECG data sequence is marked as available data.

Here, when the first cleaning type data is fast heart rate and interference cleaning type, the fast heart rate data is first identified on the ECG data sequence, and if it is identified as not fast heart rate data, then the interference length is lower than the normal threshold. The identification is actually a superimposed processing flow of the above two cleaning methods, which will not be further described here.

The data cleaning module 13 is further configured to perform data filtering processing on the ECG data sequences that are not marked as available data.

Here, it means that if the heart rate of the current ECG data sequence exceeds the set threshold or the length of the interference signal in the current ECG data sequence exceeds the set threshold during the cleaning process above, the current ECG data sequence should be used as a filter The data is filtered out. The specific data filtering process can be to discard the current ECG data sequence and continue to clean the next acquired ECG data sequence, or it can stop the current data preparation process and report the data according to the actual result of the cleaning. mistake.

The data segmentation module 14 is used for the ECG data sequence and the ECG information set, according to the set segmentation length data and the sliding window length data, to perform ECG data segment division processing, and generate a plurality of ECG segment data sequences and corresponding cardiac segments. electrical segment information set; and then send the multiple electrocardiographic segment data sequences and the corresponding electrocardiographic segment information set to the first electrocardiographic data calculation model.

Here, the division length data and the sliding window length data are preset segment division parameters.

Here, when the segmentation length data is equal to the sliding window length data, the ECG data sequence is divided equally, and there is no data overlap between the adjacent ECG segment data sequences; when the segmentation length data is greater than the sliding window length data, it is divided into The ECG data sequence overlaps segments, and there is data overlap between adjacent ECG segment data sequences. In addition, while generating the ECG segment data sequence, the ECG information set should also be fragmented to generate the corresponding ECG segment information set, because the index of the ECG data sequence of the ECG segment data sequence is relative to the index of the original ECG data sequence If adjustment occurs, the information corresponding to the index position of the ECG data in the ECG segment information set also needs to be adjusted accordingly.

In another specific implementation provided by this embodiment, the data segmentation module 14 is specifically configured to set a first start index position and an end index position; and from the first start index position, the segmentation length data is used as the segmentation length, and The first ECG segment data sequence is divided into the electrical data sequence, and the corresponding ECG segment information conversion processing is performed according to the ECG information set to generate the corresponding first ECG segment information set, and then the first starting index position is set. The sum with the sliding window length data is used as the second starting index position; then from the second starting index position, the segment length data is used as the segmentation length, and the second ECG segment data sequence is divided from the ECG data sequence, And according to the ECG information set, the corresponding ECG segment information conversion processing is performed to generate a corresponding second ECG segment information set, and then the sum of the second starting index position and the sliding window length data is used as the third starting index. position; until the calculated start index position exceeds the end index position.

Here, the first start index position and the end index position are preset data index information of two ECGs. Here, by modifying the set values of the first start index position and end index position, part of the original data can be selected for segment division. When the first start index position and end index position respectively correspond to the first one of the original ECG data sequence and the last ECG data, it means that the original ECG data sequence is divided into full sequence segments; when the first start index position and end index position are not the first and last ECG data of the original ECG data sequence When the data is used, it means that a segment of intermediate data is intercepted from the original ECG data sequence, and then segment division processing is performed on the segment of intermediate data.

For example, the current ECG data sequence includes 100 ECG data, the first start index position is 1, the end index position is 100, the segmentation length data and the sliding window length data are both 50, and the ECG corresponding to the current ECG data sequence In the information set, the total number of low-voltage information sequences in the low-voltage information subset is 0, the total number of interference information sequences in the interference information subset is 0, and the heartbeat type information subset has two heartbeat type information sequences. The heartbeat type data of one heartbeat type information sequence is the first heartbeat type, including 5 R point position data, and the heartbeat type data of the second heartbeat type information sequence is the second heartbeat type, including The position data of the 4 R points are shown in Table 2; after the ECG data segment division processing, the ECG data index of the first ECG segment data sequence obtained is 1 to 50, corresponding to the current ECG data sequence The 1st to 50th ECG data, the total number of low-voltage information sequences of the low-voltage information subset of the first ECG segment information set is 0, and the total number of interference information sequences of the interference information subset is 0. There is only one heartbeat type information sequence in the type information subset, as shown in Table 3; the ECG data index of the second ECG segment data sequence is 1 to 50, corresponding to the 51st to 100th of the current ECG data sequence ECG data, the total number of low-voltage information sequences in the low-voltage information subset of the second ECG segment information set is 0, the total number of interference information sequences in the interference information subset is 0, and the heartbeat type information subset is only 1 The heartbeat type information sequence is shown in Table 4; among them, the heartbeat type information sequences in Tables 3 and 4 are less, because the ECG data corresponding to the first heartbeat type R point in the current ECG data sequence belongs to The first ECG segment data sequence and the ECG data corresponding to point R of the second heartbeat type belong to the second ECG segment data sequence; the position of point R of the first heartbeat type in Table 3 has not changed, because the first The ECG data index in each ECG segment data sequence does not need to be reset; the change in the position data of point R in Table 4 is because the ECG data index number in the second ECG segment data sequence is reset, and the new index Number = old index number - sliding window length data * (ECG segment data sequence index number - 1) = old index number - 50 * (2-1) = old index number - 50, then the corresponding second heartbeat type R The position of the points should also be reset accordingly: the first R point position data of the second heartbeat type is changed from 60 minus 50 to 10, the second R point position data is changed from 66 to 16, the third R point position data is changed from 73 becomes 23, and the 4th R point position data changes from 80 to 30.

Heartbeat Type Information Sequence Index heartbeat type data R point position data 1 first heartbeat type 17, 26, 34, 42, 50 2 second heartbeat type 60, 66, 73, 80

Table II

Heartbeat Type Information Sequence Index heartbeat type data R point position data 1 first heartbeat type 17, 26, 34, 42, 50

Table 3

Heartbeat Type Information Sequence Index heartbeat type data R point position data 1 second heartbeat type 10, 16, 23, 30

Table 4

To sum up, in the embodiment of the present invention, the quality of the ECG data is improved by using the data cleaning module 13 to clean the original data of the calculation data required by the first ECG data calculation model; the data segmentation module 14 provides a flexible segmentation method, The different segmentation requirements of the first ECG data calculation model for the calculation data are met, and the applicability of the ECG data is broadened.

In the second ECG data calculation model in the embodiment of the present invention, the input data used for training is a selected segment of the original data, and the reference information used for verification of the training result comes from the corresponding ECG information set of the selected segment. In order to achieve this purpose, the data processing system 1 only needs to intercept the data segment according to the position of the selected segment. However, in practical applications, due to the total amount of ECG data (especially the ECG data of some rare ECG characteristics) and the The large data volume requirements of model training cannot be fully met, so in the embodiment of the present invention, the data conversion module 15 provides the left and right displacement of the ECG data (the first and second conversion modes), the left and right displacement of the heartbeat data (third, Four conversion modes) and other methods to achieve the purpose of increasing the diversity of ECG signal data, the structure of the ECG information set involved in these four conversion methods is inherited from the ECG information set of the original ECG data sequence, and its conversion rules are the same as The previous segmentation and conversion are similar; in addition, the data conversion module 15 also provides a special conversion method (the fifth conversion mode), which divides the original ECG data sequence into segments first, and then obtains the segment heartbeat type of each sub-segment, and converts the segment by segment heart rate. The beat type data constitutes a new segment ECG information set. The segment ECG information set generated in this conversion mode has no inheritance relationship with the ECG information set of the original ECG data sequence. Before the data conversion module 15 performs data conversion, in order to further ensure the quality of the training data, the data cleaning module 13 needs to be used to clean the ECG data sequence. The specific implementation is described below.

The data cleaning module 13 is also used to obtain the ECG data sequence and the corresponding ECG information set from the ECG database 12 when preparing the calculation data of the second ECG data calculation model; and according to the set second cleaning type data and the ECG information set, perform available data labeling processing on the ECG data sequence; and then send the ECG data sequence marked as available data and the corresponding ECG information set to the data conversion module 15 .

Here, the cleaning process flow of the data cleaning module 13 before data segmentation is similar to that described above, and will not be further described here.

The data conversion module 15 is used to perform segment marking processing on the ECG data sequence to obtain the selected segment; and then based on the selected segment, perform data conversion processing according to the set conversion mode data and the ECG information set to generate the converted segment The data sequence and the corresponding converted segment ECG information set; and then the converted segment data sequence and the corresponding converted segment ECG information set are sent to the second ECG data calculation model.

In another specific implementation provided by this embodiment, the data conversion module 15 is specifically configured to mark the selected segment in the ECG data sequence according to the set selected segment position;

Wherein, the position of the selected segment includes the start index data of the selected segment and the end index data of the selected segment; the start index data of the selected segment and the end index data of the selected segment are respectively the corresponding ECG data in the ECG data sequence index location information;

Here, the position of the selected segment is the set position information for marking the selected segment, and the start index data of the selected segment is the index of the segment start ECG data of the selected segment in the ECG data sequence. The segment end index data is the index of the segment end ECG data in the ECG data sequence of the selected segment; for example, the current ECG data sequence includes 100 ECG data, and the selected segment start index data at the selected segment position is 10 , the selected segment end index data is 80, then the selected segment is the data sequence between the 10th to the 80th ECG data in the current ECG data sequence;

When the conversion mode data is the first conversion mode, the ECG left-shift data is randomly generated; and the difference between the selected segment start index data and the ECG left-shift data is used as the first ECG start index data to select The difference between the segment end index data and the ECG left-shift data is determined as the first ECG end index data; and from the ECG data sequence, the index positions are extracted between the first ECG start index data and the first ECG end index data. The ECG data between the index data forms a first conversion segment data sequence; and then perform the corresponding first conversion segment information conversion processing according to the ECG information set to generate a corresponding first conversion segment ECG information set; wherein, the first The conversion segment ECG information set includes a first conversion segment heartbeat type information subset, a first conversion segment interference information subset, and a first conversion segment low voltage information subset;

Here, when the conversion mode data is the first conversion mode, the position of the selected segment is taken as the original position, and the length of the segment is kept unchanged, and the start position and the end position are shifted to the left by a length respectively, and this length is the center The data is shifted to the left, and a new segment position marked by the first ECG start index data and the first ECG end index data is generated, and the segment extraction is performed again according to the new segment position from the ECG data sequence. The first converted segment data sequence with the same length as the selected segment and similar features but not overlapping can be obtained, and then the index position of the ECG information set is adjusted similarly to the segmentation process, and the new ECG information of the first converted segment can be obtained. gather;

Here, the ECG left-shift data is randomly generated. Conventionally, random number calculation is performed in a small range, for example, the displacement length is randomly generated between 1 and 10. The purpose of this is to increase the conversion segment through the randomness of the displacement. diversity;

When the conversion mode data is the second conversion mode, the ECG right-shift data is randomly generated; and the sum of the selected segment start index data and the ECG right-shift data is used as the second ECG start index data to select The sum of the end index data of the fixed segment and the right-shifted ECG data is used as the second ECG end index data; and then from the ECG data sequence, the index positions are extracted from the second ECG start index data and the second ECG end index data. The ECG data between the index data forms a second conversion segment data sequence; and then perform the corresponding second conversion segment information conversion processing according to the ECG information set to generate a corresponding second conversion segment ECG information set; wherein, the second The conversion segment ECG information set includes a second conversion segment heartbeat type information subset, a second conversion segment interference information subset and a second conversion segment low voltage information subset;

Here, when the conversion mode data is the second conversion mode, the position of the selected segment is taken as the original position, and the length of the segment is kept unchanged, and the start position and the end position are respectively shifted to the right by a length, and this length is the center The data is shifted to the right, and a new segment position marked by the second ECG start index data and the second ECG end index data is generated, and the segment extraction is performed again according to the new segment position from the ECG data sequence. A second converted segment data sequence with the same length as the selected segment and similar features but not overlapping can be obtained, and then the index position of the ECG information set is adjusted like a segmentation process, and a new second converted segment ECG information can be obtained. gather;

Here, the ECG right-shift data is also randomly generated, which is consistent with the purpose of the ECG left-shift data, which is also to increase the diversity of conversion segments through the randomness of displacement;

When the conversion mode data is the third conversion mode, the fourth conversion mode or the fifth conversion mode, sort all the first R point position data in the ECG information set to generate a first R point position data sequence; The R point position data sequence, and the first heartbeat type data corresponding to each first R point position data, perform statistical processing of the heartbeat type and position information to generate a heartbeat statistical data sequence;

The heartbeat statistical data sequence includes multiple heartbeat statistical data groups; the heartbeat statistical data group includes heartbeat type data, heartbeat start data, and heartbeat end data; the heartbeat start data and heartbeat end data are respectively The index position information of the corresponding ECG data in the ECG data sequence;

Here, when the third conversion mode, the fourth conversion mode or the fifth conversion mode is converted, it is necessary to refer to the heartbeat position relationship of the ECG data sequence, so by sorting and counting the R points of the original ECG data sequence in advance, Obtain the heartbeat statistical data sequence; when implementing the statistical processing of heartbeat type and position information, a percentage data can be set between the adjacent R point positions, and this percentage data can be used to determine the heartbeat cycle corresponding to the R point For example, if the percentage data is 80% and there are 4 R points: 1, 2, 3, and 4R points, the starting position of the first heart cycle is in the direction from the 1R point to the 2R point. 80% position, the end position of the first heart cycle is at the 80% position in the direction from point 2R to point 3R; the start position of the second heart cycle is the same as the end position of the first heart cycle, the second heart The end position of the stroke cycle is 80% of the direction from point 3R to point 4R;

When the conversion mode data is the third conversion mode, in the heartbeat statistical data sequence, the heartbeat statistical data group whose heartbeat start data and heartbeat end data satisfy the selected segment start index data is taken as the first heartbeat data group; take the index position information corresponding to the first heartbeat data group in the heartbeat statistical data sequence as the first heartbeat index data; and then use the first heartbeat index data and the set heartbeat left shift data The second heartbeat index data is used as the second heartbeat index data; and the heartbeat start data of the heartbeat statistical data group corresponding to the second heartbeat index data is extracted from the heartbeat statistical data sequence, and used as the third ECG start index data; then in the ECG data sequence, take the third ECG start index data as the extraction start position, and use the data length of the selected segment as the extraction length, perform ECG data extraction processing, and form a third converted segment data sequence and then perform the corresponding third conversion segment information conversion processing according to the ECG information set to generate a corresponding third conversion segment ECG information set; wherein, the third conversion segment ECG information set includes the third conversion segment heartbeat type information element. set, a third conversion segment interference information subset, and a third conversion segment low voltage information subset;

Here, when the conversion mode data is the third conversion mode, in the heartbeat statistical data sequence, the heartbeat start data and the heartbeat end data of each heartbeat statistical data group form a heartbeat range, and the heartbeat range includes the selected heartbeat range. The heartbeat statistical data group at the starting position of the segment is the first heartbeat data group corresponding to the selected segment, and its index position in the heartbeat statistical data sequence is the first heartbeat index data; The heartbeat data group starts to shift the heartbeat data to the left, and the second heartbeat index data is obtained; then according to the second heartbeat index data, the first position of the corresponding heartbeat statistical data group in the heartbeat statistical data sequence is inquired to obtain the second heartbeat index data. Three ECG start index data; then, on the premise of ensuring that the data segment remains unchanged, start from the third ECG start index data of the ECG data sequence, and re-extract the segment to obtain the same length and similar characteristics as the selected segment. However, for the data sequence of the third conversion segment that does not overlap, the index position of the ECG information set is adjusted similarly to the segmentation process, and a new ECG information set of the third conversion segment can be obtained;

Here, the left-shift data of the heartbeat is preset, because the number of heartbeats that may be included in an ECG data sequence is not many. To set, the purpose of generating new training data by left-shifting the heartbeat is also to increase the diversity of conversion segments;

When the conversion mode data is the fourth conversion mode, in the heartbeat statistical data sequence, the heartbeat statistical data group whose heartbeat start data and heartbeat end data satisfy the selected segment start index data is used as the second heartbeat data group; and use the index position information corresponding to the second heartbeat data group in the heartbeat statistical data sequence as the third heartbeat index data; and then use the third heartbeat index data and the set heartbeat right shift data and, as the fourth heartbeat index data; then, from the heartbeat statistical data sequence, extract the heartbeat start data of the heartbeat statistical data group corresponding to the fourth heartbeat index data, and use it as the fourth ECG start index data; in the ECG data sequence, the fourth ECG start index data is used as the extraction start position, and the data length of the selected segment is used as the extraction length, and the ECG data extraction process is performed to form a fourth conversion segment data sequence and then perform the corresponding fourth conversion segment information conversion processing according to the ECG information set to generate a corresponding fourth conversion segment ECG information set; wherein, the fourth conversion segment ECG information set includes the fourth conversion segment heartbeat type information element set, a fourth conversion segment interference information subset, and a fourth conversion segment low voltage information subset;

Here, when the conversion mode data is the fourth conversion mode, the heartbeat start data and the heartbeat end data of each heartbeat statistical data group form a heartbeat range, and the heartbeat range includes the heartbeat at the start position of the selected segment. The heartbeat statistics data group is the second heartbeat data group corresponding to the selected segment, and its index position in the heartbeat statistics data sequence is the third heartbeat index data; and then the heartbeat is shifted to the right from the initial heartbeat Shift the data to the right to obtain the fourth heartbeat index data; then according to the fourth heartbeat index data, obtain the fourth ECG start index data by querying the starting position of the corresponding heartbeat statistical data group in the heartbeat statistical data sequence; On the premise of ensuring that the data segment remains unchanged, starting from the fourth ECG start index data of the ECG data sequence, segment extraction is performed again to obtain a fourth converted segment with the same length as the selected segment and similar features but not overlapping. data sequence, and then adjust the index position of the ECG information set similar to the segmentation process, and then a new fourth conversion segment ECG information set can be obtained;

Here, the heartbeat right-shift data is preset, which is similar to the heartbeat left-shift data, and new training data is generated by the heartbeat right-shift to increase the diversity of conversion segments;

When the conversion mode data is the fifth conversion mode, according to the preset sub-segment length, sub-segment division processing is performed on the selected segment to generate multiple sub-segments; The total length of the heartbeat of the same heartbeat type is statistically processed to generate multiple total length data, and the heartbeat type corresponding to the maximum total length data is used as the segment heartbeat type data corresponding to the sub-segment; The five converted segment data sequences are composed of all segments of heartbeat type data to form a fifth converted segment ECG information set.

Here, the sub-segment length is preset for the sub-segment division. The difference between the sub-segment division here and the previous segment division is that the sub-segment division here is to divide the selected segment in sequence, and there is no sliding step. long parameter.

For example, the current ECG data sequence includes 100 pieces of ECG data. In the ECG information set corresponding to the current ECG data sequence, the total number of low-voltage information sequences in the low-voltage information subset is 0, and the interference information sequences in the interference information subset are 0. The total number is 0, and the heartbeat type information subset has 2 heartbeat type information sequences, as shown in Table 5.

Heartbeat Type Information Sequence Index heartbeat type data R point position data 1 first heartbeat type 15, 30, 45 2 second heartbeat type 60, 75

Table 5

If the percentage data of the calculated heartbeat position is 80%, then the heartbeat statistical data sequence corresponding to the current electrocardiographic data sequence includes 3 heartbeat statistical data groups, as shown in Table 6.

Heartbeat Statistics Group Index heartbeat type data heart attack data end-of-beat data 1 first heartbeat type 27 42 2 first heartbeat type 42 57 3 second heartbeat type 57 72

Table 6

The selected segment is a data sequence composed of the 11th to 80th ECG data. The selected segment includes 70 ECG data, and the sub-segment length is 35. The selected segment is divided into two sub-segments: 1st, 2nd Sub-segments, where the first sub-segment corresponds to the 11th to 45th ECG data of the current ECG data sequence, the second sub-segment corresponds to the 46th to 80th ECG data of the current ECG data sequence, and the second The five-transition segment data sequence consists of the first and second sub-segments.

In the first sub-segment, referring to Table 6, there is only one heartbeat type: the first heartbeat type, so the total heartbeat length of the first heartbeat type of the first sub-segment is 45-10=35, the first heartbeat type The segment heartbeat type data of sub-segment 1 is the first heartbeat type; in the second sub-segment, referring to Table 6, there are two heartbeat types: the first heartbeat type and the second heartbeat type, where the first heartbeat type The total length of the heartbeat of the second heartbeat type is 57-45=12, and the total length of the heartbeat of the second heartbeat type is 72-57=15. Here, 15>12, so the maximum total length data is the heartbeat of the second heartbeat type. is the total length of the stroke, and the segment heartbeat type data of the second sub-segment is the second heartbeat type; thus, a fifth converted segment ECG information set is obtained, as shown in Table 7.

subsegment index Fragment beat type data 1 first heartbeat type 2 second heartbeat type

Table 7

To sum up, the embodiment of the present invention improves the quality of the ECG data by using the data cleaning module 13 to clean the original data of the calculation data required by the second ECG data calculation model; the data conversion module 15 provides a variety of data conversion processing The process satisfies the diversity requirements of the second ECG data calculation model for the calculation data.

It should be noted that it should be understood that the division of each module of the above system is only a division of logical functions, and may be fully or partially integrated into a physical entity in actual implementation, or may be physically separated. And these modules can all be implemented in the form of software calling through processing elements; they can also all be implemented in hardware; some modules can also be implemented in the form of calling software through processing elements, and some modules can be implemented in hardware. For example, the data integration module may be a separately established processing element, or it may be integrated into a certain chip of the above-mentioned system to realize, in addition, it may also be stored in the memory of the above-mentioned system in the form of program code, and processed by one of the above-mentioned systems. The element calls and executes the functions of the above-determined modules. The implementation of other modules is similar. In addition, all or part of these modules can be integrated together, and can also be implemented independently. The processing element described herein may be an integrated circuit with signal processing capabilities. In the implementation process, each step of the above-mentioned method or each of the above-mentioned modules can be completed by an integrated logic circuit of hardware in the processor element or an instruction in the form of software.

For example, the above modules may be one or more integrated circuits configured to implement the above methods, such as: one or more specific integrated circuits (Application Specific Integrated Circuit, ASIC), or one or more digital signal processors ( Digital Signal Processor, DSP), or, one or more Field Programmable Gate Array (Field Programmable Gate Array, FPGA), etc. For another example, when one of the above modules is implemented in the form of a processing element scheduling program code, the processing element may be a general-purpose processor, such as a central processing unit (Central Processing Unit, CPU) or other processors that can invoke program codes. For another example, these modules can be integrated together and implemented in the form of a System-on-a-chip (SOC).

In the above-mentioned embodiments, it may be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented in software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the procedures or functions described in accordance with the embodiments of the present invention are generated in whole or in part. The aforementioned computer may be a general purpose computer, a special purpose computer, a computer network, or other programmable devices. The above-mentioned computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the above-mentioned computer instructions may be transmitted from a website site, computer, server, or data center via wired communication. (eg coaxial cable, optical fiber, Digital Subscriber Line (DSL)) or wireless (eg infrared, wireless, bluetooth, microwave, etc.) means to another website site, computer, server or data center. The above-mentioned computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server, a data center, etc. that includes one or more available media integrated. The above-mentioned usable media may be magnetic media (eg, floppy disks, hard disks, magnetic tapes), optical media (eg, DVD), or semiconductor media (eg, solid state disk (SSD)), and the like.

The embodiment of the present invention provides a data processing system for ECG data. The collected ECG data is classified according to the degree of interference through the data integration module, and the influence of the interference data is controlled; the data cleaning module has no training value. The quality of ECG data is improved by filtering the ECG data; through the data segmentation module, according to the needs of different ECG data calculation models, the ECG data is divided twice, which broadens the applicability of ECG data; The conversion module converts the ECG data through left and right ECG data shift, left and right heart beat shift or sub-segment division to obtain new training data, which increases the diversity of ECG signal data.

Professionals should be further aware that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. Interchangeability, the above description has generally described the components and steps of each example in terms of function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Skilled artisans may implement the described functionality using different methods for each particular application, but such implementations should not be considered beyond the scope of the present invention.

The steps of a method or algorithm described in connection with the embodiments disclosed herein may be implemented in hardware, a software module executed by a processor, or a combination of the two. A software module can be placed in random access memory (RAM), internal memory, read only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other in the technical field. in any other known form of storage medium.

The specific embodiments described above further describe the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above descriptions are only specific embodiments of the present invention, and are not intended to limit the scope of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims (5)

1. A data processing system for electrocardiographic data, the data processing system comprising: the system comprises a data integration module, an electrocardiogram database, a data cleaning module, a data segmentation module and a data conversion module; the data processing system is respectively connected with the electrocardiogram data source, the first electrocardiogram data calculation model and the second electrocardiogram data calculation model;

the data integration module is respectively connected with the electrocardio data source and the electrocardio database; the data integration module is used for acquiring an electrocardiogram data sequence and a corresponding electrocardiogram information set from the electrocardiogram data source; according to the electrocardio information set, carrying out electrocardio data classification processing to generate classification data; then according to the classified data, storing the electrocardio data sequence and the electrocardio information set into a corresponding data subset in the electrocardio database;

the data cleaning module is respectively connected with the electrocardio database and the data segmentation module; the data cleaning module is used for acquiring the electrocardiogram data sequence and the corresponding electrocardiogram information set from the electrocardiogram database when preparing the calculation data of the first electrocardiogram data calculation model; marking available data according to the set first cleaning type data and the electrocardio information set; then sending the electrocardiogram data sequence marked as available data and the corresponding electrocardiogram information set to the data segmentation module;

the data segmentation module is connected with the first electrocardiogram data calculation model; the data segmentation module is used for carrying out segmentation processing on the electrocardio data sequences and the electrocardio information set according to set segmentation length data and sliding window length data to generate a plurality of electrocardio fragment data sequences and corresponding electrocardio fragment information sets; then sending the plurality of electrocardiogram fragment data sequences and the corresponding electrocardiogram fragment information sets to the first electrocardiogram data calculation model;

the data cleaning module is also connected with the data conversion module; the data cleaning module is further used for acquiring the electrocardiogram data sequence and the corresponding electrocardiogram information set from the electrocardiogram database when preparing the calculation data of the second electrocardiogram data calculation model; according to the set second cleaning type data and the electrocardio information set, carrying out available data labeling processing on the electrocardio data sequence; then sending the electrocardiogram data sequence marked as the available data and the corresponding electrocardiogram information set to the data conversion module;

the data conversion module is also connected with the second electrocardiogram data calculation model; the data conversion module is used for carrying out fragment marking processing on the electrocardiogram data sequence to obtain a selected fragment; then based on the selected segment, carrying out data conversion processing according to set conversion mode data and the electrocardio information set to generate a conversion segment data sequence and a corresponding conversion segment electrocardio information set; then sending the conversion segment data sequence and the corresponding conversion segment electrocardio information set to the second electrocardio data calculation model;

wherein the electrocardiogram data sequence comprises a plurality of electrocardiogram data;

the electrocardio information set comprises a heart beat type information subset, an interference information subset and a low voltage information subset;

the heart beat type information subset comprises a plurality of first heart beat type information sequences; the first heart beat type information sequence comprises first heart beat type data and a plurality of first R point position data; the first R point position data is index position information of the corresponding electrocardiogram data in the electrocardiogram data sequence;

the subset of interference information comprises a plurality of first interference information sequences; the first interference information sequence comprises first interference type information and a plurality of first interference position information groups; the first interference type information is type I interference information, type II interference information or type III interference information, wherein the interference levels of the type I interference information, the type II interference information and the type III interference information are sequentially increased; the first interference position information group comprises first interference starting information and first interference ending information; the first interference starting information and the first interference ending information are respectively index position information of the corresponding electrocardiogram data in the electrocardiogram data sequence;

the subset of low voltage information comprises a plurality of first low voltage information sequences; the first low voltage information sequence includes a plurality of first low voltage position information groups; the first low-voltage position information group includes first low-voltage start information and first low-voltage end information; the first low-voltage starting information and the first low-voltage ending information are respectively index position information of the electrocardiogram data corresponding to the first low-voltage starting information and the first low-voltage ending information in the electrocardiogram data sequence;

the electrocardio database comprises a first type interference data subset, a second type interference data subset, a third type interference data subset, a low voltage data subset and an interference-free data subset;

the data integration module is specifically used for acquiring the electrocardiogram data sequence sent from the electrocardiogram data source and the corresponding electrocardiogram information set;

according to the low-voltage information subset and the interference information subset of the electrocardio information set, carrying out electrocardio data classification processing to generate classification data; when the total number of the first low-voltage information sequences effective in the low-voltage information subset is 0 and the total number of the first interference information sequences effective in the interference information subset is not 0, taking the first interference type information with the highest interference level in all the first interference information sequences as highest-level interference information, and setting and processing the classification data according to the highest-level interference information; when the highest-level interference information is the type-I interference information, setting the classification data as a type-I interference type; when the highest-level interference information is the second-type interference information, setting the classified data as a second-type interference type; when the highest-level interference information is the type III interference information, setting the classification data as a type III interference type; setting the classification data to a low voltage type when a total number of the first low voltage information sequences that are valid in the low voltage information subset is not 0; when the total number of the first low-voltage information sequences effective in the low-voltage information subset is 0 and the total number of the first interference information sequences effective in the interference information subset is also 0, setting the classification data as an interference-free type;

then, according to the classification data, storing and processing the electrocardio data sequence and the corresponding electrocardio information set; when the classified data is the type I interference type, storing the electrocardio data sequence and the corresponding electrocardio information set into the type I interference data subset of the electrocardio database; when the classified data is the type II interference type, storing the electrocardio data sequence and the corresponding electrocardio information set into the type II interference data subset of the electrocardio database; when the classified data is of the triple-type interference type, storing the electrocardio data sequence and the corresponding electrocardio information set into the triple-type interference data subset of the electrocardio database; when the classified data is the low-voltage type, storing the electrocardio data sequence and the corresponding electrocardio information set into the low-voltage data subset of the electrocardio database; and when the classified data is the non-interference type, storing the electrocardio data sequence and the corresponding electrocardio information set into the non-interference data subset of the electrocardio database.

2. The system for processing electrocardiographic data according to claim 1,

the data cleaning module is specifically used for acquiring the electrocardiogram data sequence and the corresponding electrocardiogram information set from the data subset corresponding to the electrocardiogram database according to the calculation data requirement of the first electrocardiogram data calculation model when the calculation data of the first electrocardiogram data calculation model is prepared;

then according to the first cleaning type data and the electrocardio information set, carrying out available data labeling processing on the electrocardio data sequence; when the first cleaning type data is a fast heart rate cleaning type, performing heart rate conversion processing according to the heart beat type information subset of the electrocardio information set to generate first heart rate data, and when the first heart rate data does not exceed a set heart rate upper limit, marking the electrocardio data sequence as the available data; when the first cleaning type data is an interference cleaning type, performing three types of interference proportion calculation processing according to the interference information subset of the electrocardio information set to generate first type proportion data, first type proportion data and first type proportion data, and marking the electrocardio data sequence as the available data when the first type proportion data does not exceed a set first type proportion upper limit, the first type proportion data does not exceed a set second type proportion upper limit, and the first type proportion data does not exceed a set third type proportion upper limit; when the first cleaning type data is a fast heart rate and interference cleaning type, firstly performing heart rate conversion processing according to the heart beat type information subset to generate second heart rate data, when the second heart rate data does not exceed the upper limit of the heart rate, then performing three types of interference ratio calculation processing according to the interference information subset to generate second type one ratio data, second type two ratio data and second type three ratio data, and when the second type one ratio data does not exceed the upper limit of the type one ratio, the second type two ratio data does not exceed the upper limit of the type two ratio, and the first type three ratio data does not exceed the upper limit of the type three ratio, marking the electrocardiogram data sequence as the available data.

3. The system for processing electrocardiographic data according to claim 2,

and the data cleaning module is also used for carrying out data filtering processing on the electrocardio data sequence which is not marked as the available data.

4. The system for processing electrocardiographic data according to claim 1,

the data segmentation module is specifically used for setting a first starting index position and an ending index position; dividing the 1 st electrocardiogram fragment data sequence from a first initial index position by taking the division length data as a division length, performing corresponding electrocardiogram fragment information conversion processing according to the electrocardiogram information set to generate a corresponding 1 st electrocardiogram fragment information set, and taking the sum of the first initial index position and the sliding window length data as a second initial index position; dividing a 2 nd electrocardio segment data sequence from a second initial index position by taking the division length data as a division length, carrying out corresponding electrocardio segment information conversion processing according to the electrocardio information set to generate a corresponding 2 nd electrocardio segment information set, and taking the sum of the second initial index position and the sliding window length data as a third initial index position; until the calculated start index position exceeds the end index position.

5. The system for processing electrocardiographic data according to claim 1,

the data conversion module is specifically used for marking the selected segment in the electrocardiogram data sequence according to the set position of the selected segment; wherein the selected segment position comprises selected segment start index data and selected segment end index data; the selected segment start index data and the selected segment end index data are respectively index position information of the corresponding electrocardiogram data in the electrocardiogram data sequence;

when the conversion mode data is a first conversion mode, randomly generating electrocardiogram left shift data; taking the difference between the selected segment initial index data and the electrocardiogram left shift data as first electrocardiogram initial index data, and taking the difference between the selected segment end index data and the electrocardiogram left shift data as first electrocardiogram end index data; extracting the electrocardiogram data with an index position between the first electrocardiogram starting index data and the first electrocardiogram ending index data from the electrocardiogram data sequence to form a first conversion segment data sequence; then, carrying out corresponding first conversion fragment information conversion processing according to the electrocardio information set to generate a corresponding first conversion fragment electrocardio information set; wherein the first conversion segment cardiac electrical information set comprises a first conversion segment heart beat type information subset, a first conversion segment interference information subset and a first conversion segment low voltage information subset;

when the conversion mode data is a second conversion mode, randomly generating right electrocardio shift data; taking the sum of the selected segment initial index data and the electrocardiogram right shift data as second electrocardiogram initial index data, and taking the sum of the selected segment ending index data and the electrocardiogram right shift data as second electrocardiogram ending index data; extracting the electrocardiogram data with an index position between the second electrocardiogram starting index data and the second electrocardiogram ending index data from the electrocardiogram data sequence to form a second conversion fragment data sequence; then, carrying out corresponding second conversion fragment information conversion processing according to the electrocardio information set to generate a corresponding second conversion fragment electrocardio information set; wherein the second conversion segment electrocardio information set comprises a second conversion segment heart beat type information subset, a second conversion segment interference information subset and a second conversion segment low voltage information subset;

when the conversion mode data is a third conversion mode, a fourth conversion mode or a fifth conversion mode, sorting all the first R point position data in the electrocardiogram information set to generate a first R point position data sequence; performing heart beat type and position information statistical processing according to the first R point position data sequence and the first heart beat type data corresponding to each first R point position data to generate a heart beat statistical data sequence; wherein the heart beat statistics sequence comprises a plurality of heart beat statistics groups; the heartbeat statistic data group comprises heartbeat type data, heartbeat starting data and heartbeat ending data; the heartbeat starting data and the heartbeat ending data are respectively index position information of the corresponding electrocardiogram data in the electrocardiogram data sequence;

when the conversion mode data is the third conversion mode, taking the heartbeat statistical data group in which the heartbeat start data and the heartbeat end data meet the selected segment start index data as a first heartbeat data group in the heartbeat statistical data sequence; index position information corresponding to the first heart beat data group in the heart beat statistical data sequence is used as first heart beat index data; then taking the difference between the first heart beat index data and the set heart beat left shift data as second heart beat index data; extracting the heart beat starting data of the heart beat statistic data group corresponding to the second heart beat index data from the heart beat statistic data sequence to be used as third electrocardiogram starting index data; then in the electrocardiogram data sequence, taking the third electrocardiogram starting index data as an extraction starting position and the data length of the selected segment as an extraction length, and carrying out electrocardiogram data extraction processing to form a third conversion segment data sequence; then, carrying out corresponding third conversion fragment information conversion processing according to the electrocardio information set to generate a corresponding third conversion fragment electrocardio information set; wherein the third conversion segment electrocardio information set comprises a third conversion segment heart beat type information subset, a third conversion segment interference information subset and a third conversion segment low voltage information subset;

when the conversion mode data is the fourth conversion mode, the heartbeat statistical data group of which the heartbeat starting data and the heartbeat ending data meet the selected segment starting index data is used as a second heartbeat data group in the heartbeat statistical data sequence; index position information corresponding to the second heart beat data group in the heart beat statistic data sequence is used as third heart beat index data; taking the sum of the third heartbeat index data and the set heartbeat right shift data as fourth heartbeat index data; then extracting the heart beat initial data of the heart beat statistical data group corresponding to the fourth heart beat index data from the heart beat statistical data sequence to be used as fourth electrocardio initial index data; then in the electrocardiogram data sequence, taking the fourth electrocardiogram starting index data as an extraction starting position and the data length of the selected segment as an extraction length, and carrying out electrocardiogram data extraction processing to form a fourth conversion segment data sequence; then, performing corresponding fourth conversion fragment information conversion processing according to the electrocardiogram information set to generate a corresponding fourth conversion fragment electrocardiogram information set; the fourth conversion segment electrocardio information set comprises a fourth conversion segment heart beat type information subset, a fourth conversion segment interference information subset and a fourth conversion segment low-voltage information subset;

when the conversion mode data is the fifth conversion mode, performing sub-segment division processing on the selected segment according to the preset sub-segment length to generate a plurality of sub-segments; then according to the heart beat statistical data sequence, carrying out total heart beat length statistical processing on the same heart beat type in each sub-segment to generate a plurality of total length data, and taking the heart beat type corresponding to the maximum total length data as the segment heart beat type data corresponding to the sub-segment; and then all the sub-segments form a fifth conversion segment data sequence, and all the segment heartbeat type data form a fifth conversion segment electrocardio information set.

Images