A method, apparatus, device, and medium for fusing data specific to acute pancreatitis.

Abstract

This invention relates to the field of smart healthcare technology, and particularly to a method, apparatus, device, and medium for fusing specific data of acute pancreatitis. The method includes: collecting multi-source data from patients with acute pancreatitis; performing single-source coding processing on the multi-source data to obtain multi-source data coding vectors; constructing an inflammation stage determination function based on the inflammatory cascade response mechanism of pancreatitis to determine the stage weight coefficient vector of each data in the multi-source data; determining the key indicator mutation weights of the vital signs time-series data and the laboratory indicator dynamic data based on mutation identification; obtaining the fusion weights of each multi-source data based on the stage weight coefficient vectors and the key indicator mutation weights; and obtaining a fusion vector based on the fusion weights of each multi-source data and the multi-source data coding vectors, providing highly identifiable feature inputs for subsequent severe disease risk prediction models, thereby improving prediction accuracy and clinical adaptability.

Description

Technical Field

[0001] This invention relates to the field of smart healthcare technology, and in particular to a method, apparatus, device, and medium for fusing specific data of acute pancreatitis. Background Technology

[0002] Acute pancreatitis (AP) is a common acute abdominal condition in clinical practice. Its mortality rate is as high as 15% to 20% when it progresses to severe acute pancreatitis (SAP). Early identification of the risk of severe illness is the key to improving prognosis. Existing models for predicting the progression of acute pancreatitis to severe illness use multimodal data fusion algorithms. However, since the multi-source data are independent of each other and the fusion method is static fusion, it is difficult to obtain features that are strongly correlated with the disease, thus affecting the predictive performance.

[0003] Therefore, how to capture the dynamic characteristics of multi-source data is a technical problem that urgently needs to be solved. Summary of the Invention

[0004] In view of the above problems, the present invention provides a method, apparatus, device and medium for fusing acute pancreatitis-specific data to overcome or at least partially solve the above problems.

[0005] In a first aspect, the present invention provides a method for fusing data specific to acute pancreatitis, comprising:

[0006] Collect multi-source data from patients with acute pancreatitis, including: time-series data of vital signs, dynamic data of laboratory indicators, and quantitative data of imaging features;

[0007] The multi-source data are processed by single-source encoding to obtain multi-source data encoding vectors, which include: time-series vital signs data encoding vectors, dynamic laboratory indicator data encoding vectors, and quantitative imaging feature data encoding vectors.

[0008] Based on the inflammatory cascade response mechanism of acute pancreatitis, an inflammatory stage determination function is constructed to determine the stage weight coefficient vector of each data in multi-source data.

[0009] Based on mutation identification of time-series vital signs data and dynamic laboratory indicator data respectively, the mutation weights of key indicators in time-series vital signs data and dynamic laboratory indicator data are determined.

[0010] Based on the stage weight coefficient vector and the key indicator mutation weight, the fusion weights of the multi-source data are calculated through an attention mechanism.

[0011] Based on the fusion weights of the various sources of data and the encoding vectors of the various sources of data, a fusion vector is obtained.

[0012] Preferably, the time-series data of vital signs include: heart rate, blood oxygen saturation, blood pressure, and body temperature;

[0013] The dynamic data of the laboratory indicators include: the rate of change of amylase, lipase, and blood glucose within a preset time period;

[0014] The quantitative data of imaging features include: the extent of pancreatic edema, the area of ​​peripancreatic exudation, and necrotic lesions.

[0015] Preferably, the multi-source data are subjected to single-source encoding processing to obtain multi-source data encoding vectors. These multi-source data encoding vectors include: a vital signs time-series data encoding vector, a laboratory indicator dynamic data encoding vector, and an imaging feature quantification data encoding vector, comprising:

[0016] A bidirectional LSTM was used to extract time-series features from the vital signs time-series data to obtain the vital signs time-series data encoding vector;

[0017] A gated loop unit is used to perform dynamic feature encoding on the dynamic data of the laboratory indicators to obtain the dynamic data encoding vector of the laboratory indicators.

[0018] A multilayer perceptron is used to perform static feature encoding on the image feature quantization data to obtain the image feature quantization data encoding vector.

[0019] Preferably, based on the inflammatory cascade response mechanism of acute pancreatitis, an inflammatory stage determination function is constructed to determine the stage weight coefficient vector of each data point in the multi-source data, including:

[0020] The inflammation stage determination function is obtained according to the following formula:

[0021]

[0022] in, Any one of the multiple data sources This is the stage weight coefficient vector determined based on the inflammatory cascade response mechanism of acute pancreatitis. and These are all stage calibration parameters. The time of onset;

[0023] Based on the inflammation stage determination function, the stage weight coefficient vector of each data in the multi-source data is determined.

[0024] Preferably, based on the mutation identification of time-series vital signs data and dynamic laboratory indicator data respectively, the mutation weights of key indicators in the time-series vital signs data and dynamic laboratory indicator data are determined, including:

[0025] The key indicator mutation detection function is obtained according to the following formula:

[0026]

[0027] in, This refers to the first-order difference of indicators in time-series vital signs data or the first-order difference of indicators in dynamic laboratory data. The mutation threshold, This is the sensitivity coefficient;

[0028] Based on the aforementioned key indicator mutation detection function, mutations are identified in both the time-series data of vital signs and the dynamic data of laboratory indicators, and the mutation weights of key indicators in the time-series data of vital signs and the dynamic data of laboratory indicators are determined.

[0029] Preferably, based on the stage weight coefficient vector and the key indicator mutation weight, the fusion weights of the multi-source data are calculated using an attention mechanism, including:

[0030] The fusion weights of the multi-source data are obtained according to the following formula:

[0031]

[0032]

[0033]

[0034] in, The learnable weight matrix corresponding to the time series data of vital signs. This is the learnable weight matrix corresponding to the dynamic data of laboratory indicators. The learnable weight matrix corresponding to the quantized data of image features. This is the stage weight coefficient vector corresponding to the time series data of vital signs. This is the vector of stage weight coefficients corresponding to the dynamic data of laboratory indicators. This is the stage weight coefficient vector corresponding to the quantized data of image features. The mutation weights of key indicators corresponding to the time series data of vital signs. The key indicator mutation weights corresponding to the dynamic data of laboratory indicators. Weights for the fusion of time-series vital signs data. For the fusion weights of dynamic data of laboratory indicators, Quantify the fusion weights of image feature data. This is the normalization function in the attention mechanism.

[0035] Preferably, a fusion vector is obtained based on the respective fusion weights of the multi-source data and the multi-source data encoding vector, including:

[0036] The fusion vector is obtained using the following formula:

[0037]

[0038] in, For the Hadamard product operator, The encoding vector for the aforementioned vital signs time-series data. The dynamic data encoding vector for the laboratory indicators, The image feature quantization data encoding vector is used.

[0039] Secondly, the present invention also provides a device for fusing data specific to acute pancreatitis, comprising:

[0040] The data acquisition module is used to collect multi-source data from patients with acute pancreatitis. The multi-source data includes: time-series data of vital signs, dynamic data of laboratory indicators, and quantitative data of imaging features.

[0041] The encoding processing module is used to perform single-source encoding processing on the multi-source data to obtain multi-source data encoding vectors. The multi-source data encoding vectors include: vital sign time series data encoding vectors, laboratory indicator dynamic data encoding vectors, and imaging feature quantification data encoding vectors.

[0042] A module is built to construct an inflammation stage determination function based on the inflammatory cascade response mechanism of acute pancreatitis, and to determine the stage weight coefficient vector of each data in multi-source data.

[0043] The determination module is used to determine the mutation weights of key indicators in the time series data of vital signs and the dynamic data of laboratory indicators based on mutation identification of the time series data of vital signs and the dynamic data of laboratory indicators, respectively.

[0044] The first obtaining module is used to calculate the fusion weights of the multi-source data based on the stage weight coefficient vector and the key indicator mutation weights through an attention mechanism.

[0045] The second obtaining module is used to obtain a fusion vector based on the fusion weights of the multi-source data and the multi-source data encoding vector.

[0046] Thirdly, the present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method described in the first aspect.

[0047] Fourthly, the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described in the first aspect.

[0048] One or more technical solutions in the embodiments of the present invention have at least the following technical effects or advantages:

[0049] This invention provides a method for fusing specific data of acute pancreatitis, comprising: collecting multi-source data from patients with acute pancreatitis, including: time-series data of vital signs, dynamic data of laboratory indicators, and quantitative data of imaging features; performing single-source encoding processing on the multi-source data to obtain multi-source data encoding vectors, the multi-source data encoding vectors including: encoding vectors for time-series data of vital signs, encoding vectors for dynamic data of laboratory indicators, and encoding vectors for quantitative data of imaging features; constructing an inflammation stage determination function based on the inflammatory cascade response mechanism of acute pancreatitis to determine the stage weight coefficient vector of each data in the multi-source data; and based on the single-source encoding processing of vital signs... The study identifies mutations in time-series data and dynamic laboratory indicator data, determining the mutation weights of key indicators in both data. Based on the stage weight coefficient vector and the mutation weights of key indicators, an attention mechanism is used to calculate the fusion weights of each multi-source data. Based on the fusion weights of each multi-source data and the encoding vectors of the multi-source data, a fusion vector is obtained. By combining the inflammatory cascade response mechanism of pancreatitis, a dynamic weight strategy is designed to accurately capture mutation nodes of key indicators, achieving efficient encoding and fusion of multi-source clinical data. This provides highly identifiable feature inputs for subsequent severe illness risk prediction models, improving prediction accuracy and clinical adaptability. Attached Figure Description

[0050] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0051] Figure 1 This invention illustrates a flowchart of the method for fusing specific data of acute pancreatitis in an embodiment of the present invention.

[0052] Figure 2 The mutation detection function of key indicators in an embodiment of the present invention is shown as a function of the rate of change of the indicators. Trend chart of changes;

[0053] Figure 3 A schematic diagram of the structure of the fusion device for specific data of acute pancreatitis in an embodiment of the present invention is shown;

[0054] Figure 4 A schematic diagram of the structure of a computer device for implementing a method for fusing specific data of acute pancreatitis according to an embodiment of the present invention is shown. Detailed Implementation

[0055] Exemplary embodiments of the invention will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the invention are shown in the drawings, it should be understood that the invention may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the invention and to fully convey the scope of the invention to those skilled in the art.

[0056] Example 1:

[0057] Embodiments of the present invention provide a method for fusing data specific to acute pancreatitis, such as... Figure 1 As shown, it includes:

[0058] S101, collect multi-source data from patients with acute pancreatitis, including time-series data of vital signs, dynamic data of laboratory indicators, and quantitative data of imaging features;

[0059] S102, perform single-source coding processing on the multi-source data respectively to obtain the multi-source data coding vector, which includes: the coding vector of vital signs time series data, the coding vector of laboratory indicator dynamic data, and the coding vector of imaging feature quantification data.

[0060] S103, Based on the inflammatory cascade response mechanism of acute pancreatitis, an inflammatory stage determination function is constructed to determine the stage weight coefficient vector of each data in multi-source data;

[0061] S104, Based on the mutation identification of vital sign time series data and laboratory indicator dynamic data respectively, determine the mutation weight of key indicators in vital sign time series data and laboratory indicator dynamic data.

[0062] S105, based on the stage weight coefficient vector and the key indicator mutation weight, calculates the fusion weight of each of the multi-source data through the attention mechanism;

[0063] S106. Based on the fusion weights of the multi-source data and the encoding vectors of the multi-source data, the fusion vector is obtained.

[0064] The following is a detailed introduction to the multi-source data collected from patients with acute pancreatitis:

[0065] Time-series data of vital signs include: heart rate, blood oxygen saturation, blood pressure, and body temperature.

[0066] The dynamic data of laboratory indicators include the rate of change of amylase, lipase, and blood glucose within a preset time period.

[0067] The quantitative data of imaging features include: the extent of pancreatic edema, the area of ​​peripancreatic exudation, and necrotic lesions.

[0068] For example, vital sign time series data are The specific timing for collection is within 0-168 hours after the onset of the disease.

[0069] Dynamic data of laboratory indicators This included the rate of change of indicators such as amylase (AMY), lipase (LPS), and blood glucose (GLU) at 24h, 48h, and 72h. , It can be 24h, 48h, or 72h.

[0070] Quantitative data of imaging features The specific data for quantifying imaging features is obtained by using a CNN model to quantify and extract CT or MRI images, which will not be described in detail here.

[0071] Next, following S101, the multi-source data will be preprocessed:

[0072] Among them, the time series data of vital signs Missing values ​​were filled using linear interpolation, and 3 was used. The criteria identify outliers. Dynamic data of laboratory indicators. Z-Score standardization: (The result is missing from the original text.) ,in, The mean rate of change of this laboratory indicator dynamic data for patients with the same disease course. Standard deviation. Quantification data of imaging features. Perform Min-Max normalization: This is done by normalizing the data to map it to the interval [0, 1].

[0073] Next, S102 is executed to perform single-source encoding processing on the multi-source data to obtain multi-source data encoding vectors. These multi-source data encoding vectors include: time-series data encoding vectors of vital signs, dynamic data encoding vectors of laboratory indicators, and quantitative data encoding vectors of imaging features.

[0074] Specifically, a bidirectional LSTM is used to extract time-series features from vital signs time-series data to obtain the vital signs time-series data encoding vector;

[0075] A gated loop unit is used to perform dynamic feature encoding on the dynamic data of laboratory indicators to obtain the dynamic data encoding vector of laboratory indicators;

[0076] A multilayer perceptron is used to statically encode the quantized values ​​of image features to obtain the image feature quantization data encoding vector.

[0077] Specifically, single-source encoding is performed on the vital signs time-series data. This involves using a bidirectional LSTM (Bi-LSTM) to extract temporal features from the vital signs time-series data, outputting a vital signs time-series data encoding vector.

[0078]

[0079] in, for The number of hidden layer neurons in the model This is the feature dimension, which defaults to 64. for The number of stacked levels in the model, from which this is obtained It is a vital signs time-series data encoding vector, which is a vector containing... An ordered array of real numbers is used to facilitate subsequent model input.

[0080] A gated cyclic unit is used to perform dynamic feature encoding on the dynamic data of laboratory indicators, resulting in the encoded vector of the dynamic data of laboratory indicators:

[0081]

[0082] in, The standardized values ​​of the dynamic data of laboratory indicators. The number of hidden layer neurons in the gated recurrent unit. The number of stacked levels in the gated loop unit is used to obtain... Encode the dynamic data of laboratory indicators into vectors.

[0083] A multilayer perceptron is used to perform static feature encoding on the image feature quantization data to obtain the image feature quantization data encoding vector:

[0084]

[0085] in, The normalized values ​​of the image feature quantification data. This represents the number of hidden layer neurons in a multilayer perceptron. This indicates that the multilayer perceptron uses The function processes the output, thus obtaining... Encode vectors for quantized data of imaging features.

[0086] After obtaining the multi-source data encoding vector, step S103 is executed. Based on the inflammatory cascade response mechanism of acute pancreatitis, an inflammatory stage determination function is constructed to determine the stage weight coefficient vector of each data point in the multi-source data. Specifically:

[0087] The inflammation stage determination function is obtained according to the following formula:

[0088] Specifically, the Sigmoid function is used to achieve a smooth transition between stages, that is...

[0089]

[0090] This is the Sigmoid function.

[0091] in, Any one of the multiple data sources This is the stage weight coefficient vector determined based on the inflammatory cascade response mechanism of acute pancreatitis. and These are all stage calibration parameters. The time of onset;

[0092] Based on this inflammation stage determination function, the stage weight coefficient vector of each data in the multi-source data is determined.

[0093] The inflammation stage determination function is specifically based on the onset time. Identify the inflammation stage, and then output the stage weight coefficient vector:

[0094] Within 0-24 hours of onset (inflammatory initiation period): , , Therefore, dynamic data of laboratory indicators were determined to be the core.

[0095] Within 24-72 hours of onset (the period of inflammatory progression): , , Therefore, it was determined that both time-series data of vital signs and dynamic data of laboratory indicators should be given equal importance.

[0096] If the onset of illness is more than 72 hours later (the period of high risk for severe illness): , , Therefore, the core focus was determined to be the time series data of vital signs.

[0097] Next, S104 is executed, based on the mutation identification of the time-series data of vital signs and the dynamic data of laboratory indicators, to determine the mutation weights of key indicators in the time-series data of vital signs and the dynamic data of laboratory indicators, including:

[0098] The key indicator mutation detection function is obtained according to the following formula:

[0099]

[0100] in, This refers to the first-order difference of indicators in time-series vital signs data or the first-order difference of indicators in dynamic laboratory data. The mutation threshold, This is the sensitivity coefficient;

[0101] Based on the key indicator mutation detection function, mutations are identified in both the time-series data of vital signs and the dynamic data of laboratory indicators, and the mutation weights of key indicators in the time-series data of vital signs and the dynamic data of laboratory indicators are determined.

[0102] This mutation detection only checks time-series data, thus outputting the mutation weights of key indicators. Specifically, for At that time, for amylase, =0.5; for heart rate, These data were all obtained through statistical analysis of clinical data. The default value is 10.

[0103] exist hour, This strengthens the characteristics of the mutation node.

[0104] when hour, At this point, the normal weights are maintained.

[0105] This yields the mutation weight of the key indicator corresponding to each data point. For example... Figure 2 As shown, the mutation detection function of key indicators varies with the rate of change of the indicators. The trend chart.

[0106] Next, step S105 is executed. Based on the stage weight coefficient vector and the key indicator mutation weights, the fusion weights of the multi-source data are calculated using an attention mechanism, including:

[0107] The fusion weights of the multi-source data are obtained according to the following formula:

[0108]

[0109]

[0110]

[0111] in, The learnable weight matrix corresponding to the time series data of vital signs. This is the learnable weight matrix corresponding to the dynamic data of laboratory indicators. The learnable weight matrix corresponding to the quantized data of image features. This is the stage weight coefficient vector corresponding to the time series data of vital signs. This is the vector of stage weight coefficients corresponding to the dynamic data of laboratory indicators. This is the stage weight coefficient vector corresponding to the quantized data of image features. The mutation weights of key indicators corresponding to the time series data of vital signs. The key indicator mutation weights corresponding to the dynamic data of laboratory indicators. Weights for the fusion of time-series vital signs data. For the fusion weights of dynamic data of laboratory indicators, Quantify the fusion weights of imaging feature data.

[0112] In specific implementations, the fusion weights of vital sign time-series data are determined based on the corresponding learnable weight matrix, stage weight coefficient vector, and key indicator mutation weights; the fusion weights of laboratory indicator dynamic data are determined based on the corresponding learnable weight matrix, stage weight coefficient vector, and key indicator mutation detection function; and the fusion weights of imaging feature vectorization data are determined based on the corresponding learnable weight matrix and stage weight coefficient vector.

[0113] This involves converting a set of arbitrary real numbers into real numbers representing a probability distribution. This allows us to obtain the fusion weights for each of the multi-source data.

[0114] Specifically, the Hadamard product is used to achieve precise matching between weights and multi-source data encoding vectors, and finally, a fused vector is output. :

[0115] The fusion vector is obtained according to the following formula. :

[0116]

[0117] in, For the Hadamard product operator, Encode vectors for time-series data of vital signs. Encoding vectors for dynamic data of laboratory indicators. The image feature quantization data encoding vector is used.

[0118] By encoding multi-source data, combining it with specific weights for acute pancreatitis, and then linking it with clinical logic (such as time-series data of vital signs and dynamic data of laboratory indicators during the inflammation progression phase), the interpretability of such encoding results is better than that of general fusion algorithms, which facilitates the clinical implementation of subsequent prediction models.

[0119] One or more technical solutions in the embodiments of the present invention have at least the following technical effects or advantages:

[0120] This invention provides a method for fusing specific data of acute pancreatitis, comprising: collecting multi-source data from patients with acute pancreatitis, including: time-series data of vital signs, dynamic data of laboratory indicators, and quantitative data of imaging features; performing single-source encoding processing on the multi-source data to obtain multi-source data encoding vectors, the multi-source data encoding vectors including: time-series data encoding vectors of vital signs, dynamic data of laboratory indicators, and quantitative data of imaging features; constructing an inflammation stage determination function based on the inflammatory cascade response mechanism of acute pancreatitis to determine the stage weight coefficient vector of each data in the multi-source data; and based on the time-series data of vital signs, a single-source encoding processing method is used to obtain multi-source data encoding vectors. Mutation identification of time-series vital signs data and dynamic laboratory indicator data determines the mutation weights of key indicators in both data. Based on the stage weight coefficient vector and the mutation weights of key indicators, an attention mechanism is used to calculate the fusion weights of each multi-source data. Based on the fusion weights of each multi-source data and the encoding vector of the multi-source data, a fusion vector is obtained. By combining the inflammatory cascade response mechanism of pancreatitis, a dynamic weight strategy is designed to accurately capture mutation nodes of key indicators, achieving efficient encoding and fusion of multi-source clinical data. This provides highly identifiable feature inputs for subsequent severe illness risk prediction models, improving prediction accuracy and clinical adaptability.

[0121] Example 2:

[0122] Based on the same inventive concept, the present invention also provides a device for fusing data specific to acute pancreatitis, such as... Figure 3 As shown, it includes:

[0123] The acquisition module 301 is used to acquire multi-source data from patients with acute pancreatitis. The multi-source data includes: time-series data of vital signs, dynamic data of laboratory indicators, and quantitative data of imaging features.

[0124] The encoding processing module 302 is used to perform single-source encoding processing on the multi-source data to obtain multi-source data encoding vectors. The multi-source data encoding vectors include: vital sign time series data encoding vectors, laboratory indicator dynamic data encoding vectors, and imaging feature quantification data encoding vectors.

[0125] Module 303 is used to construct an inflammation stage determination function based on the inflammatory cascade response mechanism of acute pancreatitis, and to determine the stage weight coefficient vector of each data in multi-source data.

[0126] The determination module 304 is used to determine the key indicator mutation weights of the vital signs time series data and the laboratory indicator dynamic data based on mutation identification of the vital signs time series data and the laboratory indicator dynamic data, respectively.

[0127] The first module 305 is used to calculate the fusion weights of the multi-source data based on the stage weight coefficient vector and the key indicator mutation weights through an attention mechanism.

[0128] The second obtaining module 306 is used to obtain a fusion vector based on the respective fusion weights of the multi-source data and the multi-source data encoding vector.

[0129] In one optional implementation, the vital signs time-series data includes: heart rate, blood oxygen saturation, blood pressure, and body temperature;

[0130] The dynamic data of the laboratory indicators include: the rate of change of amylase, lipase, and blood glucose within a preset time period;

[0131] The quantitative data of imaging features include: the extent of pancreatic edema, the area of ​​peripancreatic exudation, and necrotic lesions.

[0132] In one optional implementation, the encoding processing module 302 is configured to:

[0133] A bidirectional LSTM was used to extract time-series features from the vital signs time-series data to obtain the vital signs time-series data encoding vector;

[0134] A gated loop unit is used to perform dynamic feature encoding on the dynamic data of the laboratory indicators to obtain the dynamic data encoding vector of the laboratory indicators.

[0135] A multilayer perceptron is used to perform static feature encoding on the image feature quantization data to obtain the image feature quantization data encoding vector.

[0136] In one alternative implementation, the construction module 303 is configured to:

[0137] The inflammation stage determination function is obtained according to the following formula:

[0138]

[0139] in, Any one of the multiple data sources This is the stage weight coefficient vector determined based on the inflammatory cascade response mechanism of acute pancreatitis. and These are all stage calibration parameters. The time of onset;

[0140] Based on the inflammation stage determination function, the stage weight coefficient vector of each data in the multi-source data is determined.

[0141] In one alternative implementation, the determining module 304 is configured to:

[0142] The key indicator mutation detection function is obtained according to the following formula:

[0143]

[0144] in, This refers to the first-order difference of indicators in time-series vital signs data or the first-order difference of indicators in dynamic laboratory data. The mutation threshold, This is the sensitivity coefficient;

[0145] Based on the aforementioned key indicator mutation detection function, mutations are identified in both the time-series data of vital signs and the dynamic data of laboratory indicators, and the mutation weights of key indicators in the time-series data of vital signs and the dynamic data of laboratory indicators are determined.

[0146] In one alternative implementation, the first receiving module 305 is configured to:

[0147] The fusion weights of the multi-source data are obtained according to the following formula:

[0148]

[0149]

[0150]

[0151] in, The learnable weight matrix corresponding to the time series data of vital signs. This is the learnable weight matrix corresponding to the dynamic data of laboratory indicators. The learnable weight matrix corresponding to the quantized data of image features. This is the stage weight coefficient vector corresponding to the time series data of vital signs. This is the vector of stage weight coefficients corresponding to the dynamic data of laboratory indicators. This is the stage weight coefficient vector corresponding to the quantized data of image features. The mutation weights of key indicators corresponding to the time series data of vital signs. The key indicator mutation weights corresponding to the dynamic data of laboratory indicators. Weights for the fusion of time-series vital signs data. For the fusion weights of dynamic data of laboratory indicators, Quantify the fusion weights of image feature data. This is the normalization function in the attention mechanism.

[0152] In one alternative implementation, the second obtaining module is used for:

[0153] The fusion vector is obtained using the following formula:

[0154]

[0155] in, For the Hadamard product operator, Encode vectors for time-series data of vital signs. Encoding vectors for dynamic data of laboratory indicators. The image feature quantization data encoding vector is used.

[0156] Example 3:

[0157] Based on the same inventive concept, embodiments of the present invention provide a computer device, such as... Figure 4 As shown, it includes a memory 404, a processor 402, and a computer program stored in the memory 404 and executable on the processor 402. When the processor 402 executes the program, it implements the steps of the above-described method for fusing specific data of acute pancreatitis.

[0158] Among them, Figure 4 In this document, a bus architecture (represented by bus 400) is used. Bus 400 may include any number of interconnected buses and bridges, linking various circuits including one or more processors represented by processor 402 and memory represented by memory 404. Bus 400 may also link various other circuits such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and therefore will not be described further herein. Bus interface 406 provides an interface between bus 400 and receiver 401 and transmitter 403. Receiver 401 and transmitter 403 may be the same element, i.e., a transceiver, providing a unit for communicating with various other devices over a transmission medium. Processor 402 is responsible for managing bus 400 and general processing, while memory 404 can be used to store data used by processor 402 during operation.

[0159] Example 4:

[0160] Based on the same inventive concept, embodiments of the present invention provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method for fusing specific data of acute pancreatitis.

[0161] The algorithms and displays provided herein are not inherently related to any particular computer, virtual system, or other device. Various general-purpose systems can also be used in conjunction with the teachings herein. The required structure for constructing such systems is apparent from the above description. Furthermore, this invention is not directed to any particular programming language. It should be understood that the contents of the invention described herein can be implemented using various programming languages, and the above description of specific languages ​​is for the purpose of disclosing the best mode of implementation of the invention.

[0162] Numerous specific details are set forth in the specification provided herein. However, it will be understood that embodiments of the invention may be practiced without these specific details. In some instances, well-known methods, structures, and techniques have not been shown in detail so as not to obscure the understanding of this specification.

[0163] Similarly, it should be understood that, in order to simplify the invention and aid in understanding one or more of the various inventive aspects, in the above description of exemplary embodiments of the invention, various features of the invention are sometimes grouped together in a single embodiment, figure, or description thereof. However, this disclosure should not be construed as reflecting an intention that the claimed invention requires more features than are explicitly recited in each embodiment. Rather, as reflected in each embodiment, inventive aspects lie in fewer than all features of the single foregoing disclosed embodiment. Therefore, the claims, following the detailed description, are hereby expressly incorporated into this detailed description, wherein each claim itself is a separate embodiment of the invention.

[0164] Those skilled in the art will understand that modules in the device of the embodiments can be adaptively changed and placed in one or more devices different from that embodiment. Modules, units, or components in the embodiments can be combined into a single module, unit, or component, and further, they can be divided into multiple sub-modules, sub-units, or sub-components. Except where at least some of such features and / or processes or units are mutually exclusive, any combination can be used to combine all features disclosed in this specification (including the accompanying claims, abstract, and drawings) and all processes or units of any method or device so disclosed. Unless expressly stated otherwise, each feature disclosed in this specification (including the accompanying claims, abstract, and drawings) may be replaced by an alternative feature that serves the same, equivalent, or similar purpose.

[0165] Furthermore, those skilled in the art will understand that although some embodiments herein include certain features included in other embodiments but not others, combinations of features from different embodiments are meant to be within the scope of the invention and form different embodiments. For example, in the specific implementation, any of the claimed embodiments can be used in any combination.

[0166] The various component embodiments of the present invention can be implemented in hardware, or as software modules running on one or more processors, or a combination thereof. Those skilled in the art will understand that microprocessors or digital signal processors (DSPs) can be used in practice to implement some or all of the functions of some or all of the components in the acute pancreatitis-specific data fusion apparatus or computer device according to embodiments of the present invention. The present invention can also be implemented as a device or apparatus program (e.g., a computer program and computer program product) for performing part or all of the methods described herein. Such programs implementing the present invention can be stored on a computer-readable medium or can be in the form of one or more signals. Such signals can be downloaded from an Internet website, provided on a carrier signal, or provided in any other form.

[0167] It should be noted that the above embodiments are illustrative of the invention and not restrictive, and that those skilled in the art can devise alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses should not be construed as limiting the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The invention can be implemented by means of hardware comprising several different elements and by means of a suitably programmed computer. In the unit claims enumerating several means, several of these means may be embodied by the same item of hardware. The use of the words first, second, and third, etc., does not indicate any order. These words can be interpreted as names.

Claims

1. A method for fusing specific data of acute pancreatitis, characterized in that, include: Collect multi-source data from patients with acute pancreatitis, including: time-series data of vital signs, dynamic data of laboratory indicators, and quantitative data of imaging features; The multi-source data are processed by single-source encoding to obtain multi-source data encoding vectors, which include: time-series vital signs data encoding vectors, dynamic laboratory indicator data encoding vectors, and quantitative imaging feature data encoding vectors. Based on the inflammatory cascade response mechanism of acute pancreatitis, an inflammatory stage determination function is constructed to determine the stage weight coefficient vector of each data point in multi-source data, including: The inflammation stage determination function is obtained according to the following formula: ； in, Any one of the multiple data sources This is the stage weight coefficient vector determined based on the inflammatory cascade response mechanism of acute pancreatitis. and These are all stage calibration parameters. The time of onset; Based on the inflammation stage determination function, determine the stage weight coefficient vector of each data in the multi-source data; Based on mutation identification of time-series vital signs data and dynamic laboratory indicator data, the mutation weights of key indicators in the time-series vital signs data and dynamic laboratory indicator data are determined, including: The key indicator mutation detection function is obtained according to the following formula: ； in, This refers to the first-order difference of indicators in time-series vital signs data or the first-order difference of indicators in dynamic laboratory data. The mutation threshold, This is the sensitivity coefficient; Based on the aforementioned key indicator mutation detection function, mutations are identified in the time series data of vital signs and the dynamic data of laboratory indicators, respectively, and the key indicator mutation weights of the time series data of vital signs and the dynamic data of laboratory indicators are determined. Based on the stage weight coefficient vector and the key indicator mutation weight, the fusion weights of the multi-source data are calculated through an attention mechanism. Based on the fusion weights of the various sources of data and the encoding vectors of the various sources of data, a fusion vector is obtained.

2. The method as described in claim 1, characterized in that, The vital signs time-series data include: heart rate, blood oxygen saturation, blood pressure, and body temperature; The dynamic data of the laboratory indicators include: the rate of change of amylase, lipase, and blood glucose within a preset time period; The quantitative data of imaging features include: the extent of pancreatic edema, the area of ​​peripancreatic exudation, and necrotic lesions.

3. The method as described in claim 1, characterized in that, The multi-source data is processed by single-source encoding to obtain multi-source data encoding vectors. These multi-source data encoding vectors include: time-series vital signs data encoding vectors, dynamic laboratory indicator data encoding vectors, and quantified imaging feature data encoding vectors. A bidirectional LSTM was used to extract temporal features from the vital signs time-series data to obtain the vital signs time-series data encoding vector; A gated loop unit is used to perform dynamic feature encoding on the dynamic data of the laboratory indicators to obtain the dynamic data encoding vector of the laboratory indicators. A multilayer perceptron is used to perform static feature encoding on the image feature quantization data to obtain the image feature quantization data encoding vector.

4. The method as described in claim 1, characterized in that, Based on the stage weight coefficient vector and the key indicator mutation weight, the fusion weights of the multi-source data are calculated through an attention mechanism, including: The fusion weights of the multi-source data are obtained according to the following formula: ； ； ； in, The learnable weight matrix corresponding to the time series data of vital signs. This is the learnable weight matrix corresponding to the dynamic data of laboratory indicators. The learnable weight matrix corresponding to the quantized data of image features. This is the stage weight coefficient vector corresponding to the time series data of vital signs. This is the vector of stage weight coefficients corresponding to the dynamic data of laboratory indicators. This is the stage weight coefficient vector corresponding to the quantized data of image features. The mutation weights of key indicators corresponding to the time series data of vital signs. The key indicator mutation weights corresponding to the dynamic data of laboratory indicators. Weights for the fusion of time-series vital signs data. For the fusion weights of dynamic data of laboratory indicators, Quantify the fusion weights of image feature data. This is the normalization function in the attention mechanism.

5. The method as described in claim 4, characterized in that, Based on the fusion weights of the various multi-source data and the multi-source data encoding vectors, a fusion vector is obtained, including: The fusion vector is obtained using the following formula: ； in, For the Hadamard product operator, Encoding vector for the aforementioned vital signs time-series data. The dynamic data encoding vector for the laboratory indicators, The image feature quantization data encoding vector is used.

6. A device for fusing pancreatitis-specific data, characterized in that, include: The data acquisition module is used to collect multi-source data from patients with acute pancreatitis. The multi-source data includes: time-series data of vital signs, dynamic data of laboratory indicators, and quantitative data of imaging features. The encoding processing module is used to perform single-source encoding processing on the multi-source data to obtain multi-source data encoding vectors. The multi-source data encoding vectors include: vital sign time series data encoding vectors, laboratory indicator dynamic data encoding vectors, and imaging feature quantification data encoding vectors. A construction module is used to construct an inflammatory stage determination function based on the inflammatory cascade response mechanism of acute pancreatitis, and to determine the stage weight coefficient vector of each data point in multi-source data. This construction module is used for: The inflammation stage determination function is obtained according to the following formula: ； in, Any one of the multiple data sources This is the stage weight coefficient vector determined based on the inflammatory cascade response mechanism of acute pancreatitis. and These are all stage calibration parameters. The time of onset; Based on the inflammation stage determination function, determine the stage weight coefficient vector of each data in the multi-source data; The determination module is used to determine the mutation weights of key indicators in the time-series data of vital signs and the dynamic data of laboratory indicators based on mutation identification of vital signs time-series data and laboratory indicator dynamic data, respectively. The determination module is used to: The key indicator mutation detection function is obtained according to the following formula: ； in, This refers to the first-order difference of indicators in time-series vital signs data or the first-order difference of indicators in dynamic laboratory data. The mutation threshold, This is the sensitivity coefficient; Based on the aforementioned key indicator mutation detection function, mutations are identified in the time series data of vital signs and the dynamic data of laboratory indicators, respectively, and the key indicator mutation weights of the time series data of vital signs and the dynamic data of laboratory indicators are determined. The first obtaining module is used to calculate the fusion weights of the multi-source data based on the stage weight coefficient vector and the key indicator mutation weights through an attention mechanism. The second obtaining module is used to obtain a fusion vector based on the fusion weights of the multi-source data and the multi-source data encoding vector.

7. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the method as described in any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1 to 5.

Images