A method and system for constructing multimodal heterogeneous vital sign data

Abstract

The present invention discloses a method and system for constructing multimodal heterogeneous vital sign data, which relates to the technical field of data construction and solves the problem that the time characteristics between data are also relatively chaotic and the original multimodal heterogeneous vital sign data still needs to be strengthened in terms of construction. The present invention confirms the time parameters of different vital sign data and then performs correlation analysis of the time parameters on the related vital sign data belonging to the same mode, thereby determining the time data and the irregular data, and the time characteristics of the corresponding data are also confirmed. Based on the time characteristics of the corresponding vital sign data, the subsequent correlation classification and processing of the corresponding multimodal vital sign data can be effectively guaranteed, and different correlation classifications are performed for different types of vital sign data, so that the classified vital sign data can be more consistent in time parameters, and better classification processing can be achieved, which is convenient for the subsequent constructed association table, and the data inside it is more considered and its characteristics are more clustered.

Description

Technical Field

[0001] The present invention relates to the technical field of data construction, and in particular to a method and system for constructing multimodal heterogeneous vital sign data. Background Art

[0002] Multimodal heterogeneous vital sign data refers to the process of fusing and analyzing vital sign data from different sources and types in vital sign data processing, which involves multiple aspects such as data integration, conversion and fusion. With the development of technology, methods such as ensemble learning and deep learning are playing an increasingly important role in this field. Through the application of these advanced technologies, valuable information can be better extracted from multimodal vital sign data, thus providing support for applications in related fields.

[0003] In the process of constructing its multimodal heterogeneous vital sign data into relevant tables, a large amount of vital sign data of different modalities collected during the corresponding period is directly filled into the corresponding tables. The data in the tables are too chaotic and the capacity is too large, which will lead to a huge amount of viewing when relevant personnel manage and view such tables later, and the time characteristics between the data are also relatively chaotic. The construction of its original multimodal heterogeneous vital sign data needs to be strengthened. Summary of the invention

[0004] In view of the deficiencies in the prior art, the present invention provides a method and system for constructing multimodal heterogeneous vital sign data, which solves the problem that the temporal characteristics between data are relatively chaotic and the original multimodal heterogeneous vital sign data still needs to be strengthened in terms of construction.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solutions: A method for constructing multimodal heterogeneous vital sign data, comprising the following steps:

[0006] S1. Define a management period, perform preliminary processing on several groups of heterogeneous vital sign data associated within the management period, perform preliminary classification on vital sign data belonging to the same modality, and then determine the time features associated with the corresponding modality based on the different time features of different vital sign data within the same modality. The specific method is as follows:

[0007] S11, limiting the management period T, where T is a preset value, performing modality confirmation on several groups of vital sign data collected within the management period T, wherein vital sign data of different modalities are marked with different modality marks, and relevant vital sign data of the same modality are classified into the same category;

[0008] S12. Based on several groups of vital sign data associated in the same modality, confirm the time parameters carried by the vital sign data, and sort the several groups of time parameters according to the time sequence, confirm the time parameter sequence, and identify whether the time intervals between adjacent time parameters in the time parameter sequence are consistent:

[0009] If they are all consistent, it means that this type of vital sign data is collected regularly, and the vital sign data in this modality is marked as regularly collected data, and the corresponding time interval is used as its associated time feature;

[0010] If they are inconsistent, a set of time ranges is assigned, the values ​​of which can change gradually, and the time parameters within the time parameter sequence can belong to any position within this time range:

[0011] Determine the maximum time interval tmax generated by adjacent time parameters in its time parameter sequence, use T÷Tmax=GL to determine the associated value GL, and extract the integer value before the decimal point of this associated value GL as the division value H, divide this period T into H time ranges, and each time range is the same, correspond each different time parameter in the time parameter sequence to a group of time ranges, move the time line associated with the time range, and the time range can be synchronously reduced in the process of moving, stop changing when the time interval between each time range is consistent, use the time interval determined at this moment and the numerical value of the time range corresponding to each different time parameter as the time feature associated with this type of vital sign data, and calibrate the vital sign data in this modality as irregularly collected data;

[0012] If the time parameter sequences are inconsistent and the time features associated with the vital sign data cannot be confirmed, a time confirmation error signal is directly generated and displayed;

[0013] S2. Based on the time features of the confirmed groups of vital sign data in the same modality, based on the corresponding time features, comprehensively classify the groups of multi-modal vital sign data, classify the vital sign data of different modalities with associated time features into the same classification, and generate multi-modal same-classification data packets, the specific method is as follows:

[0014] S21. Locking the time interval confirmed by the corresponding modality from the time characteristics of several groups of vital sign data in the same modality:

[0015] If there is only one mode with a set of time intervals at the maximum value, this mode is taken as the reference mode;

[0016] If there are multiple groups of modes with the same maximum time interval, the mode with the smallest time range is identified from the multiple groups of modes as the selected reference mode. When the time range does not exist in the corresponding mode, its time range takes the value of 0;

[0017] S22, if the several groups of vital sign data in the determined reference modality are periodically collected data, then based on the time interval in the reference modality as the dividing timeline, based on the time parameters of the vital sign data contained in all modalities, several groups of vital sign data of different modalities whose time parameters belong to the dividing timeline are comprehensively classified, and the multi-modal same-classification data packets corresponding to the dividing timeline are confirmed;

[0018] If several groups of vital sign data in the determined reference modality are irregularly collected data, the time interval and the corresponding time period range are identified from the time characteristics of the reference modality, and the time interval and the corresponding time period range are both used as the division time line. Then, in the same manner as above, based on the time parameters of the vital sign data contained in all modalities, several groups of vital sign data of different modalities whose time parameters belong to this division time line are comprehensively classified, and the multi-modal same-classification data packets corresponding to the division time line are confirmed. If the vital sign data to be divided are irregular data, the time period range of the vital sign data is confirmed, and the intersection area of ​​this time period range and the corresponding division time line is determined, and a group of division time lines with the largest intersection time area is selected as the classification method;

[0019] S3. For the determined multimodal data packets of the same category belonging to different divided timelines, a table is constructed for the vital sign data of a single group of multimodal data packets of the same category, a correlation table with the highest correlation degree is selected from the cloud database, and different multimodal vital sign data within such multimodal data packets of the same category are associated and filled into the association table; the specific method is:

[0020] S31, based on the modality types contained in the corresponding multi-modal same-classification data packets, identifying related tables with the same modality types from the cloud database, and marking the identified related tables as pending tables;

[0021] S32, if there are multiple groups of the calibrated pending tables, then confirm from the multiple groups of pending tables a group of pending tables whose number of modal types is closest to the number of modal types in the multi-modal same classification data packet as the associated table; if there is only one group of pending tables, then directly use this pending table as the associated table;

[0022] S33. Based on the corresponding modality classification in the association table, fill the relevant vital sign data of the same modality in the multi-modal same-classification data package to the corresponding modality classification position, complete the filling process of such association table, and display the filled association table.

[0023] Preferably, a multimodal heterogeneous vital sign data construction system comprises:

[0024] The time feature determination end defines a set of management cycles, and performs preliminary processing on several sets of heterogeneous vital sign data associated with the management cycle, performs preliminary classification on the vital sign data belonging to the same modality, and then determines the time feature associated with the corresponding modality based on the different time features of different vital sign data within the same modality;

[0025] The data packet determination end, based on the time features of the confirmed multiple groups of vital sign data in the same modality, comprehensively classifies the multiple groups of multi-modal vital sign data based on the corresponding time features, classifies the vital sign data of different modalities with associated time features into the same category, and generates a multi-modal same-classification data packet;

[0026] On the data construction side, for the determined multimodal data packets of the same category belonging to different divided time lines, the vital sign data of a single group of multimodal data packets of the same category are constructed into a table, the associated table with the highest correlation is selected from the cloud database, and the different multimodal vital sign data within such multimodal data packets are associated and filled into this associated table.

[0027] The present invention provides a method and system for constructing multimodal heterogeneous vital sign data. Compared with the prior art, the method has the following beneficial effects:

[0028] After confirming the time parameters of different vital sign data, the present invention performs correlation analysis of the time parameters of the relevant vital sign data belonging to the same modality, thereby determining the regular data and irregular data, and the time characteristics of the corresponding data are also confirmed, and the time characteristics of the corresponding vital sign data are prioritized and correlation confirmation is performed, which can effectively guarantee the subsequent correlation classification and processing of the corresponding multi-modal vital sign data;

[0029] Through the relevant time lines divided by different vital sign data, the vital sign data of different modalities are associated and classified, and different associated classifications are performed for different types of vital sign data. This can make the classified vital sign data more consistent in time parameters, achieve better classification processing, and facilitate the subsequent construction of the association table. The data inside it is more considered and its characteristics are more clustered, which can achieve better data construction and management effects. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 It is a schematic diagram of the process of the present invention;

[0031] Figure 2 It is a schematic diagram of the principle framework of the present invention. DETAILED DESCRIPTION

[0032] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. Example 1

[0033] See also Figure 1 The present application provides a method for constructing multimodal heterogeneous vital sign data, comprising the following steps:

[0034] S1. Define a management cycle, perform preliminary processing on several groups of heterogeneous vital sign data associated within this management cycle, perform preliminary classification on vital sign data belonging to the same modality (that is, the same type, such as blood pressure, body temperature and other vital sign data of the same type), and then determine the time characteristics associated with the corresponding modality based on the different time characteristics of different vital sign data within the same modality. The so-called time characteristics mean that the acquisition frequencies of vital sign data of different modalities are different when they are collected. For example, electrocardiogram data is collected in real time, but body temperature data is collected at intervals, and the collection time also exists: if the interval is 2h, the collection time may be 5min. There is also a kind of data that is collected in a relatively loose collection method, such as an interval of 2h, and the collection can be performed at any time within the subsequent 1h. Then the time characteristics of such data are chaotic. Therefore, in order to better integrate and construct several vital sign data, it is necessary to confirm the time characteristics of each vital sign data associated with different modalities one by one;

[0035] Among them, the specific method of determining the time characteristics of the corresponding modal association is:

[0036] S11. Define a management period T, where T is a preset value, which is set by the operator based on experience, and is generally 24 hours or 72 hours. Perform modality confirmation on several groups of vital sign data collected within this management period T, and different modality vital sign data are marked with different modality tags (modality classification can also be performed based on the relevant headers of the corresponding vital sign data), and classify relevant vital sign data belonging to the same modality into the same category;

[0037] S12. Based on several groups of vital sign data associated in the same modality, confirm the time parameters carried by the vital sign data, and sort the several groups of time parameters according to the time sequence, confirm the time parameter sequence, and identify whether the time intervals between adjacent time parameters in the time parameter sequence are consistent:

[0038] If they are all consistent, it means that this type of vital sign data is collected regularly, and the vital sign data in this modality is marked as regularly collected data, and the corresponding time interval is used as its associated time feature;

[0039] If they are inconsistent, a set of time ranges is assigned, the values ​​of which can change gradually, and the time parameters within the time parameter sequence can belong to any position within this time range:

[0040] Determine the maximum time interval tmax generated by adjacent time parameters in the time parameter sequence, use T÷Tmax=GL to determine the associated value GL, and extract the integer value before the decimal point of this associated value GL as the division value H, divide this period T into H time ranges, and each time range is the same (the time range at this moment is in the maximum range, and can be reduced later), correspond each different time parameter in the time parameter sequence to a group of time ranges, and move the time line associated with the time range (that is, the time range is 0-5, and there is a group of time parameters of 2, then this time range can be moved from 0-5 to 1-6 and 1.5-6.5, as long as its time parameter belongs to this time range), and the time range can be reduced synchronously during the movement process. When the movement and range change to the time interval between each time period range is consistent, the change stops, and the time interval determined at this moment and the numerical value of the time period range corresponding to each different time parameter are used as the time feature associated with this type of vital sign data, and the vital sign data in this mode are calibrated as irregularly collected data (this situation belongs to some vital sign data collected at irregular intervals, such as data collected once a day, but can be collected at any time on the day of collection, then the determined time interval is one day, and the time period range is also one day. By determining the time period range, the time features related to this type of vital sign data can be effectively confirmed one by one, so when the multi-modal vital sign data is subsequently comprehensively managed, the overall effect of the comprehensive management can be fully guaranteed);

[0041] If the time parameter sequences are inconsistent and the time features associated with the vital sign data cannot be confirmed, a time confirmation error signal is directly generated and displayed (based on the displayed confirmation error signal, external operators need to analyze whether there is a missed acquisition or other abnormalities. Under normal circumstances, there will be no such error signal. Generally, it is the missed acquisition that will cause the time features of the vital sign data to be unable to be confirmed);

[0042] S2. Based on the time features of several groups of vital sign data in the same modality confirmed, and based on the corresponding time features, comprehensively classify several groups of multi-modal vital sign data, classify the vital sign data of different modalities with associated time features into the same classification, and generate multi-modal same-classification data packets. Specifically, firstly, although the multiple vital sign data belong to different modalities, during their collection process, the related data whose time features belong to the same type of features are the most related relative to the overall timeline, so some vital sign data with relatively clustered time features are associated and divided, which is convenient for the subsequent related overall, and the collected vital sign data are all for the same object;

[0043] The specific method of generating multi-modal data packets of the same category is as follows:

[0044] S21. Locking the time interval confirmed by the corresponding modality from the time characteristics of several groups of vital sign data in the same modality:

[0045] If there is only one set of modes with a maximum time interval, this mode is taken as the reference mode (that is, the mode with the largest time interval);

[0046] If there are multiple groups of modes with the same maximum time interval, the mode with the smallest time range is identified from the multiple groups of modes as the selected reference mode. When the time range does not exist in the corresponding mode, its time range takes the value of 0 (that is, the data is collected at regular intervals, and the time range does not exist in this data);

[0047] S22, if the several groups of vital sign data in the determined reference modality are periodically collected data, then based on the time interval in the reference modality as the dividing timeline, based on the time parameters of the vital sign data contained in all modalities, several groups of vital sign data of different modalities whose time parameters belong to the dividing timeline are comprehensively classified, and the multi-modal same-classification data packets corresponding to the dividing timeline are confirmed;

[0048] If several groups of vital sign data in the determined reference modality are irregularly collected data, their time intervals and corresponding time period ranges are identified from the time characteristics of the reference modality, and the time intervals and corresponding time period ranges are used as dividing time lines. The adjacent dividing time lines may be different (because the time intervals and corresponding time period ranges are not necessarily exactly the same), or they may be the same (that is, the time intervals and time period ranges are the same). Then, the same method as above is adopted, based on the time parameters of the vital sign data contained in all modalities, several groups of vital sign data of different modalities whose time parameters belong to the dividing time line are comprehensively classified, and the multi-modal same-classification data packets corresponding to the dividing time line are confirmed. If the vital sign data to be divided is irregular data, the time period range of the vital sign data is confirmed, and the intersection area of ​​the time period range and the corresponding dividing time line is determined. A group of dividing time lines with the largest intersection time area is selected as the classification method. For example, the time period range of a group of vital sign data is proposed to be: 15-30, and there are two groups of dividing time lines before and after it, namely A[0,20) and B[20,30), then the vital sign data is divided into the B time line;

[0049] In the specific division process, it is necessary to limit the relevant conditions of the time period endpoint values ​​of the division time line, that is, a set of time parameters cannot belong to two groups of division time lines. The proposed division time lines are 0-10, 10-20, and 20-30, and the three groups of time ranges associated with them are expressed as: [0, 10), [10, 20) and [20, 30). That is, if there is a vital sign data in the corresponding division time line that is consistent with the front-end endpoint time parameter, it belongs to this division time line, and if there is a vital sign data that is consistent with the back-end endpoint time parameter, it belongs to the range of the subsequent adjacent division time line;

[0050] S3. For the determined multimodal same-classified data packets belonging to different divided timelines, the vital sign data of a single group of multimodal same-classified data packets are table-constructed, the association table with the highest association degree is selected from the cloud database, and the different multimodal vital sign data within such multimodal same-classified data packets are associated and filled into the association table, thereby completing the overall construction and processing of different multimodal vital sign data;

[0051] The specific method of constructing the table is as follows:

[0052] S31, based on the modality types contained in the corresponding multi-modal same-classification data packets, identifying related tables with the same modality types from the cloud database, and marking the identified related tables as pending tables;

[0053] S32. If there are multiple groups of the calibrated pending tables, then confirm from the multiple groups of pending tables a group of pending tables whose number of modal types is closest to the number of modal types in the multi-modal same-classification data packet as the associated table (the modal types in the proposed pending tables are 4, 5 or 6, and the pending tables all contain modal types associated with the multi-modal same-classification data packet, then the modal type 4 belongs to the associated table and is the closest), if there is only one group of pending tables, then directly use this pending table as the associated table;

[0054] S33. Based on the corresponding modal classification in the association table, fill the relevant vital sign data of the same modality in the multi-modal same-classification data package to the corresponding modal classification position, complete the filling process of such association table, and display the filled association table to complete the construction and processing of different modal vital sign data. Example 2

[0055] Combination Figure 2 , a multimodal heterogeneous vital sign data construction system, comprising:

[0056] The time feature determination end defines a set of management cycles, and performs preliminary processing on several sets of heterogeneous vital sign data associated with the management cycle, performs preliminary classification on the vital sign data belonging to the same modality, and then determines the time feature associated with the corresponding modality based on the different time features of different vital sign data within the same modality;

[0057] The data packet determination end, based on the time features of the confirmed multiple groups of vital sign data in the same modality, comprehensively classifies the multiple groups of multi-modal vital sign data based on the corresponding time features, classifies the vital sign data of different modalities with associated time features into the same category, and generates a multi-modal same-classification data packet;

[0058] On the data construction side, for the determined multimodal data packets of the same category belonging to different divided time lines, the vital sign data of a single group of multimodal data packets of the same category are constructed into a table, the associated table with the highest correlation is selected from the cloud database, and the different multimodal vital sign data within such multimodal data packets are associated and filled into this associated table.

[0059] Some of the data in the above formulas are dimensionless and numerically calculated. Meanwhile, the contents not described in detail in this specification belong to the prior art known to those skilled in the art.

[0060] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A method for constructing multimodal heterogeneous vital sign data, characterized in that: The following steps are involved: S1. Define a management period, perform preliminary processing on several groups of heterogeneous vital sign data associated within the management period, perform preliminary classification on vital sign data belonging to the same modality, and then determine the time features associated with the corresponding modality based on the different time features of different vital sign data within the same modality. The specific method is as follows: S11, limiting the management period T, where T is a preset value, performing modality confirmation on several groups of vital sign data collected within the management period T, wherein vital sign data of different modalities are marked with different modality marks, and relevant vital sign data of the same modality are classified into the same category; S12. Based on several groups of vital sign data associated in the same modality, confirm the time parameters carried by the vital sign data, and sort the several groups of time parameters according to the time sequence, confirm the time parameter sequence, and identify whether the time intervals between adjacent time parameters in the time parameter sequence are consistent: If they are all consistent, it means that this type of vital sign data is collected regularly, and the vital sign data in this modality is marked as regularly collected data, and the corresponding time interval is used as its associated time feature; If they are inconsistent, a set of time ranges is assigned, the values ​​of which can change gradually, and the time parameters within the time parameter sequence can belong to any position within this time range: Determine the maximum time interval tmax generated by adjacent time parameters in its time parameter sequence, use T÷Tmax=GL to determine the associated value GL, and extract the integer value before the decimal point of this associated value GL as the division value H, divide this period T into H time ranges, and each time range is the same, correspond each different time parameter in the time parameter sequence to a group of time ranges, move the time line associated with the time range, and the time range can be synchronously reduced in the process of moving, stop changing when the time interval between each time range is consistent, use the time interval determined at this moment and the numerical value of the time range corresponding to each different time parameter as the time feature associated with this type of vital sign data, and calibrate the vital sign data in this modality as irregularly collected data; If the time parameter sequences are inconsistent and the time features associated with the vital sign data cannot be confirmed, a time confirmation error signal is directly generated and displayed; S2. Based on the time features of the confirmed groups of vital sign data in the same modality, based on the corresponding time features, comprehensively classify the groups of multi-modal vital sign data, classify the vital sign data of different modalities with associated time features into the same category, and generate multi-modal same-classification data packets; S3. For the determined multimodal data packets of the same category belonging to different divided timelines, a table is constructed for the vital sign data of a single group of multimodal data packets of the same category, a correlation table with the highest correlation degree is selected from the cloud database, and different multimodal vital sign data within such multimodal data packets of the same category are associated and filled into the association table, specifically in the following manner: S31, based on the modality types contained in the corresponding multi-modal same-classification data packets, identifying related tables with the same modality types from the cloud database, and marking the identified related tables as pending tables; S32, if there are multiple groups of the calibrated pending tables, then confirm from the multiple groups of pending tables a group of pending tables whose number of modal types is closest to the number of modal types in the multi-modal same classification data packet as the associated table; if there is only one group of pending tables, then directly use this pending table as the associated table; S33. Based on the corresponding modality classification in the association table, fill the relevant vital sign data of the same modality in the multi-modal same-classification data package to the corresponding modality classification position, complete the filling process of such association table, and display the filled association table.

2. A method for constructing multimodal heterogeneous vital sign data according to claim 1, characterized in that: In step S2, the specific method of generating a multi-modal data packet of the same category is: S21. Locking the time interval confirmed by the corresponding modality from the time characteristics of several groups of vital sign data in the same modality: If there is only one mode with a set of time intervals at the maximum value, this mode is taken as the reference mode; If there are multiple groups of modes with the same maximum time interval, the mode with the smallest time range is identified from the multiple groups of modes as the selected reference mode. When the time range does not exist in the corresponding mode, its time range takes the value of 0; S22, if the several groups of vital sign data in the determined reference modality are periodically collected data, then based on the time interval in the reference modality as the dividing timeline, based on the time parameters of the vital sign data contained in all modalities, several groups of vital sign data of different modalities whose time parameters belong to the dividing timeline are comprehensively classified, and the multi-modal same-classification data packets corresponding to the dividing timeline are confirmed; If several groups of vital sign data within the determined reference modality are irregularly collected data, their time intervals and corresponding time period ranges are identified from the time characteristics of this reference modality, and the time intervals and corresponding time period ranges are both used as dividing time lines. Then, the same method as above is adopted, based on the time parameters of the vital sign data contained in all modalities, several groups of vital sign data of different modalities whose time parameters belong to this dividing time line are comprehensively classified, and the multi-modal same-classified data packets corresponding to the dividing time line are confirmed. If the vital sign data to be divided are irregular data, the time period range of this vital sign data is confirmed, and the intersection area of ​​this time period range and the corresponding dividing time line is determined, and a group of dividing time lines with the largest intersection time area is selected as the classification method.

3. A multimodal heterogeneous vital sign data construction system, the system is operated according to a multimodal heterogeneous vital sign data construction method according to any one of claims 1-2, characterized in that: include: The time feature determination end defines a set of management cycles, and performs preliminary processing on several sets of heterogeneous vital sign data associated with the management cycle, performs preliminary classification on the vital sign data belonging to the same modality, and then determines the time feature associated with the corresponding modality based on the different time features of different vital sign data within the same modality; The data packet determination end, based on the time features of the confirmed multiple groups of vital sign data in the same modality, comprehensively classifies the multiple groups of multi-modal vital sign data based on the corresponding time features, classifies the vital sign data of different modalities with associated time features into the same category, and generates a multi-modal same-classification data packet; On the data construction side, for the determined multimodal data packets of the same category belonging to different divided time lines, the vital sign data of a single group of multimodal data packets of the same category are constructed into a table, the associated table with the highest correlation is selected from the cloud database, and the different multimodal vital sign data within such multimodal data packets are associated and filled into this associated table.

Images