A physiological index detection system and method based on deep learning
Through a physiological index detection system based on deep learning, the problems of complexity and lack of intelligence in the detection of physiological indexes in the existing technology are solved, and the accurate identification and evaluation of physiological index data is achieved, which improves the intelligence and simplicity of the detection.
Patent Information
- Application Number
- CN202410625548.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-05-20
- Publication Date
- 2024-11-29
- Estimated Expiration
- 2044-05-20
Smart Images

Figure CN118340506B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of physiological index detection, and in particular to a physiological index detection system and method based on deep learning. Background Art
[0002] At present, physiological index detection can detect heart rate, blood sugar, blood pressure, etc., which is of great significance for the detection of human health status. Deep learning is a machine learning technology that imitates the neural network of the human brain. It realizes the mining and learning of complex data models through multi-level neural network learning features.
[0003] However, the detection of physiological indicators requires complex feature extraction, data monitoring, and analysis of physiological indicator data based on experience, which cannot adapt to the monitoring of physiological indicators in different situations, thereby increasing the complexity and cost of physiological indicator monitoring. Moreover, it is currently impossible to intelligently identify the type of physiological indicators, making the detection of physiological indicators not intelligent.
[0004] Therefore, the present invention provides a physiological index detection system and method based on deep learning. Summary of the invention
[0005] The present invention provides a physiological indicator detection system and method based on deep learning, which is used to obtain the indicator type of physiological indicators and the physiological indicator sample data sets corresponding to different indicator types, and perform the first deep learning and the second deep learning on the indicator type and the physiological indicator sample data sets respectively, so as to realize the accurate and effective construction of the physiological indicator recognition model and the physiological indicator evaluation model, so as to facilitate the accurate and effective analysis and evaluation of the indicator type of the patient's physiological indicator data and whether the indicator is normal. Secondly, the constructed physiological indicator recognition model and the physiological indicator evaluation model are encapsulated to obtain a physiological indicator detection mechanism, so as to realize the synchronous analysis of the indicator type and the indicator type through the integrated model, thereby improving the simplicity of the model. Finally, the real-time physiological indicator data of the patient is input into the physiological indicator detection mechanism for detection and analysis, so as to realize the accurate and effective detection of the patient's physiological indicator data, facilitate the patient and the doctor to intuitively read the detection results, and also facilitate the timely and effective formulation and correction of the patient's treatment plan according to the detection results, thereby realizing the intelligent detection of physiological indicators.
[0006] A physiological index detection system based on deep learning, comprising:
[0007] A data acquisition module, used to acquire several indicator types of physiological indicators and a physiological indicator sample data set of each indicator type;
[0008] A model building module is used to perform a first deep learning on several indicator types of physiological indicators to build a physiological indicator recognition model, and at the same time, perform a second deep learning on the physiological indicator sample data set of each indicator type to build a physiological indicator evaluation model;
[0009] The mechanism building module is used to encapsulate the physiological indicator recognition model and the physiological indicator evaluation model to build a physiological indicator detection mechanism;
[0010] The detection module is used to collect the patient's real-time physiological index data, input the real-time physiological index data into the physiological index detection mechanism for detection, and output the detection results of the real-time physiological index data.
[0011] Preferably, a physiological index detection system based on deep learning, a data acquisition module, comprises:
[0012] A database acquisition unit, used to acquire the physiological index detection field and retrieve the target medical database according to the physiological index detection field;
[0013] A type extraction unit, used to read the indicator types in the target medical database and obtain several indicator types of physiological indicators;
[0014] The physiological indicator sample data set extraction unit is used to generate a type identifier based on several indicator types, input the type identifier into the target medical database for identification, and extract the physiological indicator sample data set consistent with the type identifier in the target medical field database according to the identification result.
[0015] Preferably, a physiological index detection system based on deep learning, a model building module, comprises:
[0016] A first learning unit, configured to perform a first deep learning on a plurality of indicator types of physiological indicators, and to construct a physiological indicator recognition model according to the first deep learning result;
[0017] The second learning unit is used to perform second deep learning on the physiological indicator sample data set of each indicator type, and build a physiological indicator evaluation model according to the second deep learning results.
[0018] Preferably, a physiological index detection system based on deep learning, the first learning unit includes:
[0019] A parameter adjustment subunit, used to use the indicator type of the physiological indicator as a learning parameter, and adjust the parameters of the convolution layer according to the learning parameter to construct a deep learning network;
[0020] The physiological indicator recognition model construction subunit is used to input the indicator type of the physiological indicator into the deep learning network for the first deep learning, until the learning result reaches the preset convergence condition, complete the deep learning, and construct the physiological indicator recognition model according to the learning result.
[0021] Preferably, a physiological index detection system based on deep learning, the second learning unit comprises:
[0022] The node acquisition subunit is used to read the physiological indicator sample data set of each indicator type, and use the indicator type as the main node of deep learning. At the same time, the physiological indicator sample data set corresponding to each indicator type is used as the branch node of deep learning;
[0023] The interval acquisition subunit is used to read the branch nodes of deep learning, determine the data value distribution interval of the physiological indicator sample data set in the branch node, and call the preset health value judgment interval according to the current indicator type;
[0024] Divide into subunits for:
[0025] Dividing the data value distribution interval of the physiological indicator sample data set in the branch node according to the preset health value determination interval to obtain a first data value distribution interval, a second data distribution value interval and a third data value distribution interval;
[0026] The second data value distribution interval is consistent with the preset health determination interval, the first data value distribution interval is smaller than the second data value distribution interval, and the second data value distribution interval is smaller than the third data value distribution interval;
[0027] a learning factor determination subunit, configured to use the first data value interval as a first learning factor, the second data value interval as a second learning factor, and the third data value interval as a third learning factor;
[0028] A deep learning subunit is used to perform deep learning on the second learning factor to construct a healthy deep learning network in the branch node, perform deep learning on the first learning factor to construct a first unhealthy deep learning network in the branch node, and at the same time, perform deep learning on the third learning factor to construct a second unhealthy deep learning network in the branch node;
[0029] A sub-physiological index evaluation model construction sub-unit, used to construct a sub-physiological index evaluation model of a main node based on a healthy deep learning network of a branch node, a first unhealthy deep learning network, and a second unhealthy deep learning network;
[0030] The physiological index evaluation model construction sub-unit is used to integrate the sub-physiological index evaluation models of multiple main nodes to construct a physiological index evaluation model.
[0031] Preferably, a physiological index detection system based on deep learning, the mechanism building module includes:
[0032] A model input and output port acquisition unit, used to acquire the input end of the physiological index recognition model and the output end of the physiological index recognition model, and at the same time, acquire the input end of the physiological index evaluation model and the output end of the physiological index evaluation model;
[0033] Package unit for:
[0034] Constructing a data window, and making a first connection between the output end of the data window and the input end of the physiological indicator recognition model, and at the same time, making a second connection between the output end of the physiological indicator recognition model and the input end of the physiological indicator evaluation model;
[0035] Complete the encapsulation of the physiological indicator recognition model and the physiological indicator evaluation model based on the first connection result and the second connection result;
[0036] The physiological indicator detection mechanism determination unit is used to generate a physiological indicator detection mechanism based on the encapsulation result.
[0037] Preferably, in a physiological indicator detection system based on deep learning, the input end of the physiological indicator detection mechanism is the input end of the output window, and the output end of the physiological indicator detection mechanism is the output end of the physiological indicator evaluation model.
[0038] Preferably, a physiological index detection system based on deep learning, the detection module includes:
[0039] A data reading unit, used to read the patient's real-time physiological index data;
[0040] The detection and analysis unit is used to input the patient's real-time physiological index data into the physiological index detection mechanism for detection. The specific process is as follows:
[0041] The real-time physiological index data is received based on the input end of the data window in the physiological index detection mechanism, and the real-time physiological index data is processed based on the data window to obtain the target physiological index data to be detected;
[0042] Transmitting the target physiological index data to be detected to the input end of the physiological index recognition model based on the output end of the data window;
[0043] Based on the physiological indicator recognition model, the type of the target physiological indicator data to be detected is identified, and the target indicator type of the target physiological indicator data to be detected is output according to the output end of the physiological indicator recognition model;
[0044] Inputting the target indicator type and the target physiological indicator output to be detected into the physiological indicator evaluation model at the same time, performing a health evaluation on the target physiological indicator to be detected according to the target indicator type based on the physiological indicator evaluation model, and outputting the health evaluation result at the output end of the physiological indicator evaluation model;
[0045] Among them, the health assessment result is the detection result of the real-time physiological indicator data output by the physiological indicator detection mechanism.
[0046] Preferably, a physiological index detection system based on deep learning, the detection module includes:
[0047] A result processing unit, used to perform text conversion on the detection results of the real-time physiological index data, obtain the detection text corresponding to the detection result, and perform vocabulary and sentence splitting on the detection text based on grammatical rules;
[0048] A text feature acquisition unit, used to determine sentence structure features and grammatical objects of text words in the detected text based on the word and sentence splitting results, and to determine text features of the detection results based on the sentence structure features and grammatical objects;
[0049] A physiological index database construction unit, used for classifying the test results based on text features, and constructing a physiological index database based on the classification results;
[0050] A data level acquisition unit is used to acquire the data level structure of the physiological indicator database and acquire the authorized account of the physiological indicator database. At the same time, it acquires the authorized account information of the physiological indicator database and determines the account level of the authorized account;
[0051] Data splitting unit, used for:
[0052] Matching the account level of the authorized account with the data level structure of the physiological indicator database, and outputting the data range that can be read by each authorized account according to the matching result;
[0053] According to the data range that can be read by each authorized account, the data in the physiological indicator database is divided into intervals to obtain data sets corresponding to multiple data intervals;
[0054] Cryptographic unit for:
[0055] Performing a first encryption on the data interval based on the authorized account to obtain a first encryption result;
[0056] Performing a second encryption on the data set corresponding to each data interval based on a preset encryption key to obtain a second encryption result;
[0057] The physiological indicator database is encrypted based on the first encryption result and the second encryption result.
[0058] A physiological index detection method based on deep learning, comprising:
[0059] Step 1: Obtain several indicator types of physiological indicators and a physiological indicator sample data set of each indicator type;
[0060] Step 2: Perform a first deep learning on several indicator types of physiological indicators to build a physiological indicator recognition model. At the same time, perform a second deep learning on the physiological indicator sample data set of each indicator type to build a physiological indicator evaluation model.
[0061] Step 3: Encapsulate the physiological indicator recognition model and the physiological indicator evaluation model to build a physiological indicator detection mechanism;
[0062] Step 4: Collect the patient's real-time physiological index data, input the real-time physiological index data into the physiological index detection mechanism for detection, and output the detection result of the real-time physiological index data.
[0063] Compared with the prior art, the present invention has the following beneficial effects:
[0064] 1. By obtaining the indicator types of physiological indicators and the physiological indicator sample data sets corresponding to different indicator types, and performing the first deep learning and the second deep learning on the indicator types and the physiological indicator sample data sets respectively, the physiological indicator recognition model and the physiological indicator evaluation model can be accurately and effectively constructed, so as to facilitate the accurate and effective analysis and evaluation of the indicator types of the patient's physiological indicator data and whether the indicators are normal. Secondly, the constructed physiological indicator recognition model and the physiological indicator evaluation model are encapsulated to obtain the physiological indicator detection mechanism, so as to realize the synchronous analysis of the indicator type and the indicator type through the integrated model, thereby improving the simplicity of the model. Finally, the patient's real-time physiological indicator data is input into the physiological indicator detection mechanism for detection and analysis, so as to realize accurate and effective detection of the patient's physiological indicator data, facilitate patients and doctors to intuitively read the test results, and also facilitate the timely and effective formulation and correction of the patient's treatment plan according to the test results, so as to realize the intelligent detection of physiological indicators.
[0065] 2. By taking the indicator type of the physiological indicator as the main node of deep learning, and taking the physiological indicator sample data set corresponding to each indicator type as the branch node of deep learning, it is convenient to perform standardized and reliable training according to the correspondence between the main node and the branch node. Secondly, by determining the data value distribution interval of the branch node, and dividing the data value distribution interval according to the preset health value judgment interval, it is convenient to accurately and effectively limit the value of the physiological indicator under different circumstances, and construct the deep learning network corresponding to the different data value intervals according to the division results, so as to realize the accurate and effective construction of the judgment tool of the physiological indicator content under different circumstances. Finally, the deep learning network corresponding to each main node is first summarized and the sub-physiological indicator evaluation models corresponding to different main nodes are second summarized, so as to realize the accurate and effective construction of the physiological indicator evaluation model, thereby ensuring the accuracy and reliability of the evaluation of the physiological indicator data, and facilitating doctors and patients to accurately and effectively understand their own conditions.
[0066] 3. By analyzing the test results, the text features of the test results can be accurately and effectively determined, and the test results of different patients can be classified according to the text features, so that the physiological indicator database can be accurately and effectively constructed according to the classification results. Secondly, by determining the data hierarchy structure of the physiological indicator database and the corresponding authorized account, the readable data range of different authorized accounts can be accurately and effectively determined according to the matching results of the data hierarchy structure of the physiological indicator database and the account level of the authorized account. Finally, the physiological indicator database is divided into intervals according to the data range that each authorized account can read, and the data range of different authorized accounts is locked. The data interval and the test results contained in the data interval are encrypted according to the authorized account, ensuring the security and reliability of the test results of different patients.
[0067] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures specifically pointed out in this application document.
[0068] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS
[0069] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:
[0070] Figure 1 This is a structural diagram of a physiological index detection system based on deep learning in an embodiment of the present invention;
[0071] Figure 2 This is a structural diagram of a data acquisition module in a physiological index detection system based on deep learning in an embodiment of the present invention;
[0072] Figure 3 This is a flow chart of a physiological index detection method based on deep learning in an embodiment of the present invention. DETAILED DESCRIPTION
[0073] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.
[0074] Embodiment 1:
[0075] This embodiment provides a physiological index detection system based on deep learning, such as Figure 1 As shown, including:
[0076] A data acquisition module, used to acquire several indicator types of physiological indicators and a physiological indicator sample data set of each indicator type;
[0077] A model building module is used to perform a first deep learning on several indicator types of physiological indicators to build a physiological indicator recognition model, and at the same time, perform a second deep learning on the physiological indicator sample data set of each indicator type to build a physiological indicator evaluation model;
[0078] The mechanism building module is used to encapsulate the physiological indicator recognition model and the physiological indicator evaluation model to build a physiological indicator detection mechanism;
[0079] The detection module is used to collect the patient's real-time physiological index data, input the real-time physiological index data into the physiological index detection mechanism for detection, and output the detection results of the real-time physiological index data.
[0080] In this embodiment, the several indicator types of the physiological indicators may be types included in the physiological indicators, such as blood pressure, blood sugar, heart rate and other data.
[0081] In this embodiment, the physiological indicator sample data set of each indicator type includes: all data values of different indicator types in different situations, for example, when the indicator type is heart rate, it includes the value ranges corresponding to normal, too fast and slow heart rate.
[0082] In this embodiment, the first deep learning may be to train different indicator types of physiological indicators, with the purpose of training a model that can identify the indicator type of physiological indicator sample data, that is, a physiological indicator recognition model.
[0083] In this embodiment, the second deep learning can be to train the physiological indicator sample data set corresponding to each indicator type, and learn the content characteristics of different physiological indicator sample data sets, so as to facilitate the construction of a model that can identify the specific value conditions of different physiological indicator data, that is, a physiological indicator evaluation model.
[0084] In this embodiment, the physiological indicator detection mechanism may be a comprehensive model that can be directly applied after integrating the physiological indicator recognition model and the physiological indicator evaluation model, that is, a model that can simultaneously analyze the type and content of the physiological indicator.
[0085] In this embodiment, the test result may be the patient's current physiological index value obtained after analysis by the physiological index detection mechanism, and when the test result is abnormal, the test result is higher or lower than the normal result value range.
[0086] The working principle and beneficial effects of the above technical solution are: by obtaining the indicator types of physiological indicators and the physiological indicator sample data sets corresponding to different indicator types, and performing the first deep learning and the second deep learning on the indicator types and the physiological indicator sample data sets respectively, the physiological indicator recognition model and the physiological indicator evaluation model are accurately and effectively constructed, so as to facilitate the accurate and effective analysis and evaluation of the indicator types of the patient's physiological indicator data and whether the indicators are normal. Secondly, the constructed physiological indicator recognition model and the physiological indicator evaluation model are encapsulated to obtain a physiological indicator detection mechanism, so as to realize the synchronous analysis of the indicator type and the indicator type through the integrated model, thereby improving the simplicity of the model. Finally, the patient's real-time physiological indicator data is input into the physiological indicator detection mechanism for detection and analysis, so as to realize accurate and effective detection of the patient's physiological indicator data, facilitate patients and doctors to intuitively read the test results, and facilitate timely and effective formulation and correction of the patient's treatment plan according to the test results, thereby realizing intelligent detection of physiological indicators.
[0087] Embodiment 2:
[0088] Based on Example 1, this embodiment provides a physiological index detection system based on deep learning, such as Figure 2 As shown, the data acquisition module includes:
[0089] A database acquisition unit, used to acquire the physiological index detection field and retrieve the target medical database according to the physiological index detection field;
[0090] A type extraction unit, used to read the indicator types in the target medical database and obtain several indicator types of physiological indicators;
[0091] The physiological indicator sample data set extraction unit is used to generate a type identifier based on several indicator types, input the type identifier into the target medical database for identification, and extract the physiological indicator sample data set consistent with the type identifier in the target medical field database according to the identification result.
[0092] In this embodiment, the physiological indicator detection field may be a disease field, such as a cardiovascular and cerebrovascular disease field, and the corresponding indicator type may be determined based on the cardiovascular and cerebrovascular disease and the corresponding physiological indicator sample data may be retrieved.
[0093] In this embodiment, the target medical database may be a database corresponding to the field of physiological index detection, including specific physiological index type composition and corresponding specific index content.
[0094] In this embodiment, the type identifier may be a marking symbol that can represent different indicator types, and different types of indicator types can be distinguished through the type identifier.
[0095] The working principle and beneficial effects of the above technical solution are: by determining the physiological indicator detection field, the corresponding target medical database can be accurately and effectively determined according to the physiological indicator detection field; secondly, by reading the indicator type in the target medical database, a variety of different indicator types of physiological indicators can be accurately and effectively determined; finally, according to the type identifier of the indicator type, the corresponding physiological indicator sample data set is retrieved from the target medical database, so as to achieve accurate and effective acquisition of the indicator type of the physiological indicator and the physiological indicator sample data set, which provides convenience for building a physiological indicator detection mechanism.
[0096] Embodiment 3:
[0097] Based on Example 1, this embodiment provides a physiological index detection system based on deep learning, and the model building module includes:
[0098] A first learning unit, configured to perform a first deep learning on a plurality of indicator types of physiological indicators, and to construct a physiological indicator recognition model according to the first deep learning result;
[0099] The second learning unit is used to perform second deep learning on the physiological indicator sample data set of each indicator type, and build a physiological indicator evaluation model according to the second deep learning results.
[0100] The working principle and beneficial effect of the above technical solution are: by learning several indicator types of physiological indicators and physiological indicator sample data sets of each indicator type respectively, the physiological indicator recognition model and the physiological indicator evaluation model can be accurately and effectively constructed to ensure that the patient's real-time physiological indicator data can be accurately and effectively analyzed.
[0101] Embodiment 4:
[0102] On the basis of Example 3, this embodiment provides a physiological index detection system based on deep learning, wherein the first learning unit includes:
[0103] A parameter adjustment subunit, used to use the indicator type of the physiological indicator as a learning parameter, and adjust the parameters of the convolution layer according to the learning parameter to construct a deep learning network;
[0104] The physiological indicator recognition model construction subunit is used to input the indicator type of the physiological indicator into the deep learning network for the first deep learning, until the learning result reaches the preset convergence condition, complete the deep learning, and construct the physiological indicator recognition model according to the learning result.
[0105] In this embodiment, the deep learning network may be a model structure obtained by training the convolutional layer by the indicator type of the physiological indicator, and is a part of the physiological indicator recognition model finally obtained.
[0106] In this embodiment, the preset convergence condition is known in advance, that is, it is a reference basis for achieving the learning purpose and a reference indicator for determining whether to jump out of the learning stage.
[0107] In this embodiment, inputting the indicator type of the physiological indicator into the deep learning network for the first deep learning can be analyzing the indicator types of different physiological indicators through the deep learning network, with the aim of ensuring that the deep learning network can accurately identify different physiological indicator types.
[0108] The working principle and beneficial effects of the above technical solution are: by adjusting the parameters of the known convolutional layer by taking the indicator type of the physiological indicator as the learning parameter, the deep learning network can be accurately and effectively acquired. Secondly, the indicator type of the physiological indicator is first deeply learned by the obtained deep learning network, and the first deep learning process is detected according to the preset convergence condition, so as to ensure that the physiological indicator recognition model is accurately and effectively constructed when the preset convergence condition is met, thereby ensuring the accuracy of the recognition of the physiological indicator type.
[0109] Embodiment 5:
[0110] On the basis of Example 3, this embodiment provides a physiological index detection system based on deep learning, wherein the second learning unit includes:
[0111] The node acquisition subunit is used to read the physiological indicator sample data set of each indicator type, and use the indicator type as the main node of deep learning. At the same time, the physiological indicator sample data set corresponding to each indicator type is used as the branch node of deep learning;
[0112] The interval acquisition subunit is used to read the branch nodes of deep learning, determine the data value distribution interval of the physiological indicator sample data set in the branch node, and call the preset health value judgment interval according to the current indicator type;
[0113] Divide into subunits for:
[0114] Dividing the data value distribution interval of the physiological indicator sample data set in the branch node according to the preset health value determination interval to obtain a first data value distribution interval, a second data distribution value interval and a third data value distribution interval;
[0115] The second data value distribution interval is consistent with the preset health determination interval, the first data value distribution interval is smaller than the second data value distribution interval, and the second data value distribution interval is smaller than the third data value distribution interval;
[0116] a learning factor determination subunit, configured to use the first data value interval as a first learning factor, the second data value interval as a second learning factor, and the third data value interval as a third learning factor;
[0117] A deep learning subunit is used to perform deep learning on the second learning factor to construct a healthy deep learning network in the branch node, perform deep learning on the first learning factor to construct a first unhealthy deep learning network in the branch node, and at the same time, perform deep learning on the third learning factor to construct a second unhealthy deep learning network in the branch node;
[0118] A sub-physiological index evaluation model construction sub-unit, used to construct a sub-physiological index evaluation model of a main node based on a healthy deep learning network of a branch node, a first unhealthy deep learning network, and a second unhealthy deep learning network;
[0119] The physiological index evaluation model construction sub-unit is used to integrate the sub-physiological index evaluation models of multiple main nodes to construct a physiological index evaluation model.
[0120] In this embodiment, the main node may be an indicator type of a physiological indicator, and the branch node may be a physiological indicator sample data set. The main node may be, for example, heart rate, blood pressure, etc., and the branch node may be, for example, specific indicator content corresponding to blood pressure and heart rate.
[0121] In this embodiment, the data value distribution interval may be the value range of the physiological indicator sample data set under different master nodes.
[0122] In this embodiment, the preset healthy value determination interval is known in advance and is used to characterize the value range corresponding to the physiological indicator data set of each master node under normal circumstances.
[0123] In this embodiment, the first learning factor is the first data value interval, the second learning factor is the second data value interval, and the third learning factor is the third data value interval, wherein the first learning factor, the second learning factor and the third learning factor are respectively training parameters for constructing a physiological indicator evaluation model.
[0124] In this embodiment, the health deep learning network can be constructed after deep learning of the second learning factor, and is a tool for determining whether the value of the physiological indicator data is within a preset health value determination interval.
[0125] In this embodiment, the first unhealthy deep learning network can be constructed based on the first learning factor, and is a tool for making a judgment when the value of the physiological indicator data is less than a preset healthy value judgment interval.
[0126] In this embodiment, the second unhealthy deep learning network can be constructed based on the third learning factor, and is a tool for making a judgment when the value of the physiological indicator data is greater than a preset healthy value judgment interval.
[0127] In this embodiment, the sub-physiological indicator evaluation model can be obtained by summarizing the healthy deep learning network, the first unhealthy deep learning network and the second unhealthy deep learning network of the branch nodes corresponding to each main node, that is, a model that can comprehensively analyze the indicator type and indicator content of the physiological indicators corresponding to the corresponding main node.
[0128] The working principle and beneficial effects of the above technical solution are: by taking the indicator type of the physiological indicator as the main node of deep learning, and taking the physiological indicator sample data set corresponding to each indicator type as the branch node of deep learning, it is convenient to perform standardized and reliable training according to the correspondence between the main node and the branch node. Secondly, by determining the data value distribution interval of the branch node, and dividing the data value distribution interval according to the preset health value judgment interval, it is convenient to accurately and effectively limit the value of the physiological indicator under different circumstances, and construct the deep learning network corresponding to the different data value intervals according to the division results, so as to realize the accurate and effective construction of the judgment tool of the physiological indicator content under different circumstances. Finally, the deep learning network corresponding to each main node is first summarized and the sub-physiological indicator evaluation models corresponding to different main nodes are second summarized, so as to realize the accurate and effective construction of the physiological indicator evaluation model, thereby ensuring the accuracy and reliability of the evaluation of the physiological indicator data, and facilitating doctors and patients to accurately and effectively understand their own conditions.
[0129] Embodiment 6:
[0130] Based on Example 1, this embodiment provides a physiological index detection system based on deep learning, and the mechanism building module includes:
[0131] A model input and output port acquisition unit, used to acquire the input end of the physiological index recognition model and the output end of the physiological index recognition model, and at the same time, acquire the input end of the physiological index evaluation model and the output end of the physiological index evaluation model;
[0132] Package unit for:
[0133] Constructing a data window, and making a first connection between the output end of the data window and the input end of the physiological indicator recognition model, and at the same time, making a second connection between the output end of the physiological indicator recognition model and the input end of the physiological indicator evaluation model;
[0134] Complete the encapsulation of the physiological indicator recognition model and the physiological indicator evaluation model based on the first connection result and the second connection result;
[0135] The physiological indicator detection mechanism determination unit is used to generate a physiological indicator detection mechanism based on the encapsulation result.
[0136] In this embodiment, the physiological indicator detection mechanism includes: a data window, which is used to receive the patient's real-time physiological indicator data; a physiological indicator recognition model, which is used to recognize the target indicator type corresponding to the real-time physiological indicator data; and a physiological indicator evaluation model, which is used to evaluate the real-time physiological indicator data based on the target indicator type, thereby realizing the detection of the real-time physiological indicator data.
[0137] In this embodiment, the input end of the physiological indicator detection mechanism is the input end of the output window, and the output end of the physiological indicator detection mechanism is the output end of the physiological indicator evaluation model.
[0138] The working principle and beneficial effects of the above technical solution are: by determining the data window, the reception of real-time physiological indicator data can be effectively realized, thereby effectively ensuring that the real-time physiological indicator data can be accurately input into the physiological indicator recognition model, and by making a first connection between the output end of the data window and the input end of the physiological indicator recognition model, and at the same time, making a second connection between the output end of the physiological indicator recognition model and the input end of the physiological indicator evaluation model, the effectiveness and accuracy of encapsulating the physiological indicator recognition model and the physiological indicator evaluation model can be effectively guaranteed, thereby improving the efficiency of real-time physiological indicator data detection.
[0139] Embodiment 7:
[0140] Based on Example 1, this embodiment provides a physiological index detection system based on deep learning, and the detection module includes:
[0141] A data reading unit, used to read the patient's real-time physiological index data;
[0142] The detection and analysis unit is used to input the patient's real-time physiological index data into the physiological index detection mechanism for detection. The specific process is as follows:
[0143] The real-time physiological index data is received based on the input end of the data window in the physiological index detection mechanism, and the real-time physiological index data is processed based on the data window to obtain the target physiological index data to be detected;
[0144] Transmitting the target physiological index data to be detected to the input end of the physiological index recognition model based on the output end of the data window;
[0145] Based on the physiological indicator recognition model, the type of the target physiological indicator data to be detected is identified, and the target indicator type of the target physiological indicator data to be detected is output according to the output end of the physiological indicator recognition model;
[0146] Inputting the target indicator type and the target physiological indicator output to be detected into the physiological indicator evaluation model at the same time, performing a health evaluation on the target physiological indicator to be detected according to the target indicator type based on the physiological indicator evaluation model, and outputting the health evaluation result at the output end of the physiological indicator evaluation model;
[0147] Among them, the health assessment result is the detection result of the real-time physiological indicator data output by the physiological indicator detection mechanism.
[0148] In this embodiment, the data window processes the real-time physiological indicator data, which may include operations such as denoising the real-time physiological indicator data and removing erroneous data, so as to obtain the target physiological indicator data to be detected, thereby ensuring the validity and accuracy of the target physiological indicator data to be detected and avoiding interference from irrelevant data.
[0149] In this embodiment, the health assessment result may include: whether the real-time physiological indicator data corresponding to the indicator type is within the preset health judgment interval; if so, a health mark is output; otherwise, when the real-time physiological indicator data is greater than the preset health judgment interval, a second unhealthy mark is output; when the real-time physiological indicator data is less than the preset health judgment interval, a first unhealthy mark is output.
[0150] The working principle and beneficial effects of the above technical solution are: by reading the patient's real-time physiological indicator data, the physiological indicator data is input into the data window, and by processing the real-time physiological indicator data, the purity of the real-time physiological indicator data can be effectively guaranteed, which is conducive to ensuring the effectiveness of the data analysis of the physiological indicator recognition model and the physiological indicator evaluation model, and then effectively outputting the health assessment results, ensuring the effectiveness and accuracy of the detection results of the real-time physiological indicator data output by the physiological indicator detection mechanism, and the intelligence of the identification and analysis of the real-time physiological indicator data.
[0151] Embodiment 8:
[0152] Based on Example 1, this embodiment provides a physiological index detection system based on deep learning, and the detection module includes:
[0153] A result processing unit, used to perform text conversion on the detection results of the real-time physiological index data, obtain the detection text corresponding to the detection result, and perform vocabulary and sentence splitting on the detection text based on grammatical rules;
[0154] A text feature acquisition unit, used to determine sentence structure features and grammatical objects of text words in the detected text based on the word and sentence splitting results, and to determine text features of the detection results based on the sentence structure features and grammatical objects;
[0155] A physiological index database construction unit, used for classifying the test results based on text features, and constructing a physiological index database based on the classification results;
[0156] A data level acquisition unit is used to acquire the data level structure of the physiological indicator database and acquire the authorized account of the physiological indicator database. At the same time, it acquires the authorized account information of the physiological indicator database and determines the account level of the authorized account;
[0157] Data splitting unit, used for:
[0158] Matching the account level of the authorized account with the data level structure of the physiological indicator database, and outputting the data range that can be read by each authorized account according to the matching result;
[0159] According to the data range that can be read by each authorized account, the data in the physiological indicator database is divided into intervals to obtain data sets corresponding to multiple data intervals;
[0160] Cryptographic unit for:
[0161] Performing a first encryption on the data interval based on the authorized account to obtain a first encryption result;
[0162] Performing a second encryption on the data set corresponding to each data interval based on a preset encryption key to obtain a second encryption result;
[0163] The physiological indicator database is encrypted based on the first encryption result and the second encryption result.
[0164] In this embodiment, text conversion may be converting the detection result into corresponding text content, thereby facilitating determination of data features corresponding to the detection result.
[0165] In this embodiment, the grammatical rules are known in advance, that is, the defined rules that can accurately separate words and sentences.
[0166] In this embodiment, the sentence structure feature may be the association relationship between different sentences in the detection result.
[0167] In this embodiment, the grammatical object may be the role of different text words in the detection text, for example, it may be a name, a medical record diagnosis result, etc.
[0168] In this embodiment, the text feature may be the type corresponding to the physiological index detection result contained in the detection result, etc.
[0169] In this embodiment, the data hierarchy structure may be a standard for limiting access rights to the physiological indicator database, that is, the types of users who can access the physiological indicator database may be determined through the data hierarchy structure.
[0170] In this embodiment, the account level corresponds to the data level structure, and the physiological indicator database can be accessed only when the account level satisfies the data level structure.
[0171] In this embodiment, the data range that can be read by each authorized account is output based on the matching results. The data that can be read by different authorized accounts in the physiological indicator database can be limited according to the matching situation between the account level of the authorized account and the data hierarchy structure of the physiological indicator database, thereby ensuring that different authorized accounts can only read the test results under the corresponding authority. For example, the authorized account corresponding to the doctor can read the test results of all patients, and the patient's authorized account can only read its own test results.
[0172] In this embodiment, the first encryption may be to encrypt the corresponding data interval according to the authorized account, that is, to ensure that the corresponding data interval can only be accessed by authorized users.
[0173] In this embodiment, the preset encryption key is set in advance and is a parameter required for encryption.
[0174] In this embodiment, the second encryption may be to decrypt the data set corresponding to each data interval, that is, to encrypt the specific detection result of each data interval.
[0175] The working principle and beneficial effects of the above technical solution are: by analyzing the test results, the text features of the test results can be accurately and effectively determined, and the test results of different patients can be classified according to the text features, so as to accurately and effectively construct the physiological indicator database according to the classification results; secondly, by determining the data hierarchy structure of the physiological indicator database and the corresponding authorized account, the readable data range of different authorized accounts can be accurately and effectively determined according to the matching results of the data hierarchy structure of the physiological indicator database and the account level of the authorized account; finally, the physiological indicator database is divided into intervals according to the data range that each authorized account can read, so as to lock the data range of different authorized accounts, and encrypt the data interval and the test results contained in the data interval according to the authorized account, so as to ensure the security and reliability of the test results of different patients.
[0176] Embodiment 9:
[0177] This embodiment provides a physiological index detection method based on deep learning, such as Figure 3 As shown, including:
[0178] Step 1: Obtain several indicator types of physiological indicators and a physiological indicator sample data set of each indicator type;
[0179] Step 2: Perform a first deep learning on several indicator types of physiological indicators to build a physiological indicator recognition model. At the same time, perform a second deep learning on the physiological indicator sample data set of each indicator type to build a physiological indicator evaluation model.
[0180] Step 3: Encapsulate the physiological indicator recognition model and the physiological indicator evaluation model to build a physiological indicator detection mechanism;
[0181] Step 4: Collect the patient's real-time physiological index data, input the real-time physiological index data into the physiological index detection mechanism for detection, and output the detection result of the real-time physiological index data.
[0182] The working principle and beneficial effects of the above technical solution are: by obtaining the indicator types of physiological indicators and the physiological indicator sample data sets corresponding to different indicator types, and performing the first deep learning and the second deep learning on the indicator types and the physiological indicator sample data sets respectively, the physiological indicator recognition model and the physiological indicator evaluation model are accurately and effectively constructed, so as to facilitate the accurate and effective analysis and evaluation of the indicator types of the patient's physiological indicator data and whether the indicators are normal. Secondly, the constructed physiological indicator recognition model and the physiological indicator evaluation model are encapsulated to obtain a physiological indicator detection mechanism, so as to realize the synchronous analysis of the indicator type and the indicator type through the integrated model, thereby improving the simplicity of the model. Finally, the patient's real-time physiological indicator data is input into the physiological indicator detection mechanism for detection and analysis, so as to realize accurate and effective detection of the patient's physiological indicator data, facilitate patients and doctors to intuitively read the test results, and facilitate timely and effective formulation and correction of the patient's treatment plan according to the test results, thereby realizing intelligent detection of physiological indicators.
[0183] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.
Claims
1. A physiological index detection system based on deep learning, characterized in that: include: A data acquisition module, used to acquire several indicator types of physiological indicators and a physiological indicator sample data set of each indicator type; A model building module is used to perform a first deep learning on several indicator types of physiological indicators to build a physiological indicator recognition model, and at the same time, perform a second deep learning on the physiological indicator sample data set of each indicator type to build a physiological indicator evaluation model; The mechanism building module is used to encapsulate the physiological indicator recognition model and the physiological indicator evaluation model to build a physiological indicator detection mechanism; The detection module is used to collect the patient's real-time physiological index data, input the real-time physiological index data into the physiological index detection mechanism for detection, and output the detection result of the real-time physiological index data; The data acquisition module includes: A database acquisition unit, used to acquire the physiological index detection field and retrieve the target medical database according to the physiological index detection field; A type extraction unit, used to read the indicator types in the target medical database and obtain several indicator types of physiological indicators; A physiological indicator sample data set extraction unit is used to generate a type identifier based on a number of indicator types, input the type identifier into a target medical database for identification, and extract a physiological indicator sample data set consistent with the type identifier from the target medical field database according to the identification result; Among them, the model building module includes: A first learning unit, configured to perform a first deep learning on a plurality of indicator types of physiological indicators, and to construct a physiological indicator recognition model according to the first deep learning result; A second learning unit is used to perform a second deep learning on the physiological indicator sample data set of each indicator type, and to build a physiological indicator evaluation model according to the second deep learning result; Among them, the mechanism building module includes: A model input and output port acquisition unit, used to acquire the input end of the physiological index recognition model and the output end of the physiological index recognition model, and at the same time, acquire the input end of the physiological index evaluation model and the output end of the physiological index evaluation model; Package unit for: Constructing a data window, and making a first connection between the output end of the data window and the input end of the physiological indicator recognition model, and at the same time, making a second connection between the output end of the physiological indicator recognition model and the input end of the physiological indicator evaluation model; Complete the encapsulation of the physiological indicator recognition model and the physiological indicator evaluation model based on the first connection result and the second connection result; A physiological indicator detection mechanism determination unit, used for generating a physiological indicator detection mechanism based on the encapsulation result; The detection module includes: A data reading unit, used to read the patient's real-time physiological index data; The detection and analysis unit is used to input the patient's real-time physiological index data into the physiological index detection mechanism for detection. The specific process is as follows: The real-time physiological index data is received based on the input end of the data window in the physiological index detection mechanism, and the real-time physiological index data is processed based on the data window to obtain the target physiological index data to be detected; Transmitting the target physiological index data to be detected to the input end of the physiological index recognition model based on the output end of the data window; Based on the physiological indicator recognition model, the type of the target physiological indicator data to be detected is identified, and the target indicator type of the target physiological indicator data to be detected is output according to the output end of the physiological indicator recognition model; Inputting the target indicator type and the target physiological indicator output to be detected into the physiological indicator evaluation model at the same time, performing a health evaluation on the target physiological indicator to be detected according to the target indicator type based on the physiological indicator evaluation model, and outputting the health evaluation result at the output end of the physiological indicator evaluation model; Among them, the health assessment result is the detection result of the real-time physiological indicator data output by the physiological indicator detection mechanism; The detection module includes: A result processing unit, used to perform text conversion on the detection results of the real-time physiological index data, obtain the detection text corresponding to the detection result, and perform vocabulary and sentence splitting on the detection text based on grammatical rules; A text feature acquisition unit, used to determine sentence structure features and grammatical objects of text words in the detected text based on the word and sentence splitting results, and to determine text features of the detection results based on the sentence structure features and grammatical objects; A physiological index database construction unit, used for classifying the test results based on text features, and constructing a physiological index database based on the classification results; A data level acquisition unit is used to acquire the data level structure of the physiological indicator database and acquire the authorized account of the physiological indicator database. At the same time, it acquires the authorized account information of the physiological indicator database and determines the account level of the authorized account. Data splitting unit, used for: Matching the account level of the authorized account with the data level structure of the physiological indicator database, and outputting the data range that can be read by each authorized account according to the matching result; According to the data range that can be read by each authorized account, the data in the physiological indicator database is divided into intervals to obtain data sets corresponding to multiple data intervals; Cryptographic unit for: Performing a first encryption on the data interval based on the authorized account to obtain a first encryption result; Performing a second encryption on the data set corresponding to each data interval based on a preset encryption key to obtain a second encryption result; The physiological indicator database is encrypted based on the first encryption result and the second encryption result.
2. A physiological index detection system based on deep learning according to claim 1, characterized in that: The first learning unit includes: A parameter adjustment subunit, used to use the indicator type of the physiological indicator as a learning parameter, and adjust the parameters of the convolution layer according to the learning parameter to construct a deep learning network; The physiological indicator recognition model construction subunit is used to input the indicator type of the physiological indicator into the deep learning network for the first deep learning, until the learning result reaches the preset convergence condition, complete the deep learning, and construct the physiological indicator recognition model according to the learning result.
3. A physiological index detection system based on deep learning according to claim 1, characterized in that: The second learning unit includes: The node acquisition subunit is used to read the physiological indicator sample data set of each indicator type, and use the indicator type as the main node of deep learning. At the same time, the physiological indicator sample data set corresponding to each indicator type is used as the branch node of deep learning; The interval acquisition subunit is used to read the branch nodes of deep learning, determine the data value distribution interval of the physiological indicator sample data set in the branch node, and call the preset health value judgment interval according to the current indicator type; Divide into subunits for: Dividing the data value distribution interval of the physiological indicator sample data set in the branch node according to the preset health value determination interval to obtain a first data value distribution interval, a second data distribution value interval and a third data value distribution interval; The second data value distribution interval is consistent with the preset health determination interval, the first data value distribution interval is smaller than the second data value distribution interval, and the second data value distribution interval is smaller than the third data value distribution interval; a learning factor determination subunit, configured to use the first data value interval as a first learning factor, the second data value interval as a second learning factor, and the third data value interval as a third learning factor; A deep learning subunit is used to perform deep learning on the second learning factor to construct a healthy deep learning network in the branch node, perform deep learning on the first learning factor to construct a first unhealthy deep learning network in the branch node, and at the same time, perform deep learning on the third learning factor to construct a second unhealthy deep learning network in the branch node; A sub-physiological index evaluation model construction sub-unit, used to construct a sub-physiological index evaluation model of a main node based on a healthy deep learning network of a branch node, a first unhealthy deep learning network, and a second unhealthy deep learning network; The physiological index evaluation model construction sub-unit is used to integrate the sub-physiological index evaluation models of multiple main nodes to construct a physiological index evaluation model.
4. The physiological index detection system based on deep learning according to claim 1, characterized in that: The input end of the physiological index detection mechanism is the input end of the data window, and the output end of the physiological index detection mechanism is the output end of the physiological index evaluation model.
5. A physiological index detection method based on deep learning, characterized in that: include: Step 1: Obtain several indicator types of physiological indicators and a physiological indicator sample data set of each indicator type; Step 2: Perform a first deep learning on several indicator types of physiological indicators to build a physiological indicator recognition model. At the same time, perform a second deep learning on the physiological indicator sample data set of each indicator type to build a physiological indicator evaluation model. Step 3: Encapsulate the physiological indicator recognition model and the physiological indicator evaluation model to build a physiological indicator detection mechanism; Step 4: Collect the patient's real-time physiological index data, input the real-time physiological index data into the physiological index detection mechanism for detection, and output the detection result of the real-time physiological index data; Wherein, step 1 includes: Acquire the physiological index detection field, and retrieve the target medical database according to the physiological index detection field; Read the indicator types in the target medical database to obtain several indicator types of physiological indicators; Generate a type identifier based on several indicator types, input the type identifier into a target medical database for identification, and extract a physiological indicator sample data set consistent with the type identifier from the target medical field database according to the identification result; Wherein, step 2 includes: Performing a first deep learning on several indicator types of physiological indicators, and constructing a physiological indicator recognition model according to the first deep learning result; Performing a second deep learning on the physiological indicator sample data set of each indicator type respectively, and building a physiological indicator evaluation model according to the second deep learning result; Wherein, step 3 includes: Acquire the input end and the output end of the physiological index recognition model, and at the same time, acquire the input end and the output end of the physiological index evaluation model; Constructing a data window, and making a first connection between the output end of the data window and the input end of the physiological indicator recognition model, and at the same time, making a second connection between the output end of the physiological indicator recognition model and the input end of the physiological indicator evaluation model; Complete the encapsulation of the physiological indicator recognition model and the physiological indicator evaluation model based on the first connection result and the second connection result; Generate physiological index detection mechanism based on encapsulation results; Wherein, step 4 includes: Read the patient's real-time physiological index data; The patient's real-time physiological index data is input into the physiological index detection mechanism for detection. The specific process is as follows: The real-time physiological index data is received based on the input end of the data window in the physiological index detection mechanism, and the real-time physiological index data is processed based on the data window to obtain the target physiological index data to be detected; Transmitting the target physiological index data to be detected to the input end of the physiological index recognition model based on the output end of the data window; Based on the physiological indicator recognition model, the type of the target physiological indicator data to be detected is identified, and the target indicator type of the target physiological indicator data to be detected is output according to the output end of the physiological indicator recognition model; Inputting the target indicator type and the target physiological indicator output to be detected into the physiological indicator evaluation model at the same time, performing a health evaluation on the target physiological indicator to be detected according to the target indicator type based on the physiological indicator evaluation model, and outputting the health evaluation result at the output end of the physiological indicator evaluation model; Among them, the health assessment result is the detection result of the real-time physiological indicator data output by the physiological indicator detection mechanism; Wherein, step 4 further includes: Perform text conversion on the test results of real-time physiological index data to obtain the test text corresponding to the test results, and perform vocabulary and sentence splitting on the test text based on grammatical rules; Determine sentence structure features and grammatical objects of text vocabulary in the detected text based on the vocabulary and sentence splitting results, and determine text features of the detection results based on the sentence structure features and grammatical objects; Classify the test results based on text features, and build a physiological index database based on the classification results; Obtaining the data hierarchy structure of the physiological indicator database, and obtaining the authorized account of the physiological indicator database, and at the same time, obtaining the authorized account information of the physiological indicator database, and determining the account level of the authorized account; Matching the account level of the authorized account with the data level structure of the physiological indicator database, and outputting the data range that can be read by each authorized account according to the matching result; According to the data range that can be read by each authorized account, the data in the physiological indicator database is divided into intervals to obtain data sets corresponding to multiple data intervals; Performing a first encryption on the data interval based on the authorized account to obtain a first encryption result; Performing a second encryption on the data set corresponding to each data interval based on a preset encryption key to obtain a second encryption result; The physiological indicator database is encrypted based on the first encryption result and the second encryption result.
Citation Information
Patent Citations
Central signal segregation system
CN109478427A
Pathological condition detector using kernel methods and oximeters
US20080081974A1
Prediction of Health Status from Physiological Data
US20160302671A1
glove
US20200085373A1
