Liver chronic disease screening method, apparatus, and storage medium

Abstract

The application relates to the technical field of medical treatment and discloses a liver chronic disease screening method, a device and a storage medium.The method comprises the following steps: a liver detection system collects liver stiffness data and liver fat attenuation parameters of a target object; a mobile terminal system receives the liver stiffness data and the liver fat attenuation parameters, combines the liver stiffness data and the liver fat attenuation parameters with basic information of a preset target object, generates acquisition characteristic data, carries out label marking matrix processing on pretreatment integrated data of the cloud service system, generates a label marking matrix, carries out fitting residual processing on the label marking matrix, obtains extended decision tree data, carries out discrete sampling processing on the extended decision tree data, obtains a sparse vector, carries out classification calculation processing on the sparse vector, obtains a classification result, and generates a disease screening report based on the classification result.In the embodiment of the application, through system architecture and cloud service system data processing, the early warning timeliness and accuracy of liver chronic disease screening are improved.

Description

Technical Field

[0001] This invention relates to the field of medical technology, and in particular to a method, device and storage medium for screening chronic liver diseases. Background Technology

[0002] Currently, liver disease screening mainly relies on a single FibroScan device, which uses non-invasive liver elastography (FibroScan) to measure liver tissue stiffness (LSM) and liver fat attenuation parameter (CAP) to assess the degree of liver fibrosis and fatty liver.

[0003] FibroScan devices are concentrated in large hospitals, with low adoption rates in primary healthcare institutions, where chronic disease screening largely relies on clinical examinations. However, while FibroScan results are mostly presented in report form, current screening methods lack predictive models based on historical data, making it difficult to classify and predict the progression of liver disease, resulting in poor intervention timeliness. Furthermore, traditional screening often relies solely on a single LSM or CAP value, leading to inaccurate analysis results due to dependence on a single indicator. Therefore, current technologies suffer from insufficient intervention timeliness and inaccurate analysis results in chronic liver disease screening, necessitating a new technology to address these issues. Summary of the Invention

[0004] The main objective of this invention is to solve the technical problems of insufficient early warning timeliness and inaccurate analysis results in chronic liver disease screening.

[0005] The first aspect of this invention provides a method for screening chronic liver diseases, which is applied in a chronic liver disease screening system. The chronic liver disease screening system includes: a liver detection system, a mobile terminal system, and a cloud service system. The chronic disease screening method includes:

[0006] The liver detection system collects liver stiffness data and liver fat attenuation parameters of the target object, and sends the liver stiffness data and liver fat attenuation parameters to the mobile terminal system;

[0007] The mobile terminal system receives the liver stiffness data and the liver fat attenuation parameter, combines the liver stiffness data and the liver fat attenuation parameter with the basic information of the preset target object to generate collected feature data, and displays the collected feature data.

[0008] When a confirmation command is received, the collected feature data is sent to the cloud service system.

[0009] The cloud service system receives the collected feature data and preprocesses the collected feature data to obtain preprocessed integrated data;

[0010] Based on a preset tag set, the preprocessed integrated data is subjected to tag matrix processing to generate a tag matrix;

[0011] According to the preset gradient boosting algorithm, the label matrix is ​​fitted with residuals to obtain extended decision tree data;

[0012] According to the preset sampling algorithm, the extended decision tree data is subjected to discrete sampling processing to obtain sparse vectors;

[0013] The sparse vectors are classified according to a preset logistic regression algorithm to obtain the classification result.

[0014] Based on the classification results, a disease screening report is generated and sent to the mobile terminal system.

[0015] Optionally, in a first implementation of the first aspect of the present invention, the step of performing tagging matrix processing on the preprocessed integrated data according to a preset tag set to generate a tagging matrix includes:

[0016] The preprocessed integrated data is then subjected to element-matrix processing to obtain an integrated data matrix;

[0017] Based on rows or columns, the integrated data matrix is ​​matched and labeled with a preset tag set to obtain the tag data of the rows or columns;

[0018] The label data of the rows or columns are merged with the integrated data matrix to generate a label marking matrix.

[0019] Optionally, in a second implementation of the first aspect of the present invention, the step of performing fitting residual processing on the label matrix according to a preset gradient boosting algorithm to obtain extended decision tree data includes:

[0020] The rows or columns of the corresponding tags in the tag marking matrix are globally scaled to obtain a global tag matrix;

[0021] Based on the corresponding tag elements of the global tag matrix, probability residuals are calculated on the rows or columns of the global tag matrix to obtain probability residuals;

[0022] Based on a preset CART regression tree, the rows or columns of the global label matrix are split to obtain extended decision tree data.

[0023] Optionally, in a third implementation of the first aspect of the present invention, the step of globally scaling the rows or columns of the tag matrix corresponding to the tags to obtain a global tag matrix includes:

[0024] F0(x)=log(p / (1-p)),

[0025] Where F0(x) is the global proportion of a single label, and p is the positive sample probability proportion of row or column labels.

[0026] Optionally, in a fourth implementation of the first aspect of the present invention, the step of performing discrete sampling processing on the extended decision tree data according to a preset sampling algorithm to obtain sparse vectors includes:

[0027] Based on the preset OneHotEncoder() function, discrete features will be extracted from the extended decision tree data to obtain sparse vectors.

[0028] Optionally, in a fifth implementation of the first aspect of the present invention, the collected feature data includes: liver stiffness, liver fat attenuation parameter, BMI parameter, and gender parameter; the step of preprocessing the collected feature data to obtain preprocessed integrated data includes:

[0029] The liver stiffness and liver fat attenuation parameters are sorted from smallest to largest to obtain sorted liver stiffness and liver fat sorting parameters.

[0030] Calculate the difference between the 75th percentile and 25th percentile data of the sorted liver stiffness and liver fat sorting parameters to obtain the liver stiffness analysis value and liver fat analysis value.

[0031] Based on the liver stiffness analysis values, the sorted liver stiffness is truncated by upper and lower limits to obtain the corrected liver stiffness.

[0032] Based on the liver fat analysis values, the upper and lower limits of the liver fat ranking parameters are truncated to obtain the corrected liver fat attenuation parameters.

[0033] The BMI parameters are normalized according to the preset Z-score algorithm to obtain the corrected BMI parameters;

[0034] According to the preset mapping rules, the gender parameter is mapped to obtain a gender mapping value;

[0035] The modified liver stiffness, the modified liver fat attenuation parameter, the modified BMI parameter, and the gender mapping value are combined to obtain preprocessed integrated data.

[0036] Optionally, in a sixth implementation of the first aspect of the present invention, after the step of displaying the acquired feature data, the method further includes:

[0037] When an error feedback instruction is received, the feedback input instruction is received and error description data is generated.

[0038] The abnormality description data and the collected feature data are sent to the liver detection system in order to adjust the abnormal data in the collected feature data.

[0039] Optionally, in a seventh implementation of the first aspect of the present invention, after the steps of generating a disease screening report based on the classification results and sending the disease screening report to the mobile terminal system, the method further includes:

[0040] The disease screening report is sent to the liver detection system, and the clinical feedback data returned by the liver detection system is received.

[0041] Based on the clinical feedback data, the label set is adjusted to obtain a new label set.

[0042] A second aspect of the present invention provides a liver chronic disease screening device, comprising: a memory and at least one processor, wherein the memory stores instructions, and the memory and the at least one processor are interconnected via a circuit; the at least one processor invokes the instructions in the memory to cause the liver chronic disease screening device to perform the above-described liver chronic disease screening method.

[0043] A third aspect of the present invention provides a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the above-described method for screening chronic liver diseases.

[0044] In this embodiment of the invention, a three-system architecture is established, comprising a liver detection system, a mobile terminal system, and a cloud service system. The liver detection system collects liver tissue stiffness values ​​and liver fat attenuation parameters, which, combined with basic information from the target individual's electronic medical record or health record, are sent to the mobile terminal system. After user confirmation on the mobile terminal system, the collected feature data is sent to the cloud service system. The cloud service system performs feature analysis and regression calculations on the collected feature data to generate high-precision disease screening results, which are then sent back to the mobile terminal system. This improves the timeliness and accuracy of early warning for chronic liver diseases, solving the technical problems of insufficient early warning timeliness and inaccurate analysis results for chronic liver diseases. Attached Figure Description

[0045] Figure 1 This is a schematic diagram of one embodiment of the liver chronic disease screening method in this invention;

[0046] Figure 2 This is a schematic diagram of a specific embodiment of the 103 steps of the liver chronic disease screening method in this invention;

[0047] Figure 3This is a schematic diagram of a specific embodiment of the 105 steps of the liver chronic disease screening method in this invention;

[0048] Figure 4 This is a schematic diagram of a specific embodiment of the 106 steps of the liver chronic disease screening method in this invention;

[0049] Figure 5 This is a schematic diagram of one embodiment of the liver chronic disease screening device in this invention. Detailed Implementation

[0050] This invention provides a method, device, and storage medium for screening chronic liver diseases.

[0051] The embodiments of the present invention will now be described in more detail with reference to the accompanying drawings. While some embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of the present disclosure. It should be understood that the accompanying drawings and embodiments are for illustrative purposes only and are not intended to limit the scope of protection of the present invention.

[0052] In the description of the embodiments disclosed in this invention, the term "comprising" and similar terms should be understood as open-ended inclusion, i.e., "including but not limited to". The term "based on" should be understood as "at least partially based on". The term "one embodiment" or "the embodiment" should be understood as "at least one embodiment". The terms "first", "second", etc., may refer to different or the same objects. Other explicit and implicit definitions may also be included below.

[0053] For ease of understanding, the specific process of the embodiments of the present invention is described below. Please refer to [link / reference]. Figure 1 This invention provides an embodiment of a liver chronic disease screening method, which is applied in a liver chronic disease screening system. The liver chronic disease screening system includes: a liver detection system, a mobile terminal system, and a cloud service system. The chronic disease screening method includes:

[0054] 101. The liver detection system collects liver stiffness data and liver fat attenuation parameters of the target object, and sends the liver stiffness data and the liver fat attenuation parameters to the mobile terminal system;

[0055] In this embodiment, the liver detection system's acquisition module uses the FibroScan device to collect data. FibroScan is a non-invasive liver detection device based on transient elastography technology. It can collect liver stiffness data and liver fat attenuation parameters from the target object without puncture or surgery. Finally, the collected liver stiffness data and liver fat attenuation parameters are sent to the mobile terminal system via the network.

[0056] 102. The mobile terminal system receives the liver stiffness data and the liver fat attenuation parameter, combines the liver stiffness data and the liver fat attenuation parameter with the basic information of the preset target object to generate collected feature data, and displays the collected feature data.

[0057] In this embodiment, the liver detection system sends the collected feature data to a mobile system, which can be a smartphone app. The app stores basic information about the target subject, such as age, BMI, drinking history, metabolic syndrome-related indicators, and family history of liver disease. Alternatively, after receiving liver stiffness data and liver fat attenuation parameters, the app fills in its own basic information, structures and combines the collected information to generate collected feature data, and displays the collected feature data so that the user can confirm the relevant collected feature data.

[0058] 103. When a confirmation command is received, the collected feature data is sent to the cloud service system;

[0059] In this embodiment, after the APP on the mobile system receives the confirmation command for the collected feature data, it sends the collected feature data to the cloud service system for feature analysis.

[0060] Furthermore, after "displaying the collected feature data" in step 102, the following is also included:

[0061] 1031. When an abnormal feedback instruction is received, the feedback input instruction is received and abnormal description data is generated.

[0062] 1032. Send the abnormality description data and the collected feature data to the liver detection system so as to adjust the abnormal data in the collected feature data.

[0063] In steps 1031-1032, an anomaly feedback command for the collected feature data is received from the mobile system. This command can include a descriptive text for the anomaly, such as missing liver stiffness data or liver fat attenuation parameters exceeding the normal range. The anomaly description and the collected feature data are then sent to the liver detection system. The system administrators, based on the anomaly description and the target object's basic information in the collected feature data, adjust the anomaly data to promptly resolve any abnormalities in the measurement data collected. An encrypted communication protocol is used during data transmission to ensure the security of the test results and personal information.

[0064] 104. The cloud service system receives the collected feature data and preprocesses the collected feature data to obtain preprocessed integrated data;

[0065] In this embodiment, the cloud service system receives collected feature data. The structure of the collected feature data can be a combination of liver stiffness data (LSM), liver fat attenuation parameter (LSM), BMI, age, gender, lifestyle, etc. For example, a sorted structure of [[liver stiffness data', liver fat attenuation parameter', age', gender, BMI', lifestyle']] by column combination. Of course, it can also be a sorted structure of row combination, which is just a transpose of the above data.

[0066] For details, please refer to Figure 2 , Figure 2 This is a specific embodiment of step 104 of the liver chronic disease screening method in this invention. The collected feature data includes: liver stiffness, liver fat attenuation parameter, BMI parameter, and gender parameter. Step 104, "preprocessing the collected feature data to obtain preprocessed integrated data," includes the following specific implementation:

[0067] 1041. Sort the liver stiffness and liver fat attenuation parameters from smallest to largest to obtain sorted liver stiffness and liver fat sorting parameters.

[0068] 1042. Calculate the difference between the 75th percentile and 25th percentile data of the sorted liver stiffness and liver fat sorting parameters to obtain the liver stiffness analysis value and liver fat analysis value.

[0069] 1043. Based on the liver stiffness analysis values, the sorted liver stiffness is truncated to the upper and lower limits to obtain the corrected liver stiffness.

[0070] 1044. Based on the liver fat analysis values, the upper and lower limits of the liver fat ranking parameters are truncated to obtain the corrected liver fat attenuation parameters.

[0071] 1045. The BMI parameters are normalized according to the preset Z-score algorithm to obtain the corrected BMI parameters;

[0072] 1046. According to the preset mapping rules, the gender parameter is mapped to obtain a gender mapping value;

[0073] 1047. The modified liver stiffness, the modified liver fat attenuation parameter, the modified BMI parameter, and the gender mapping value are combined and processed to obtain preprocessed integrated data.

[0074] In steps 1041-1047, the preprocessing methods for liver stiffness and liver fat attenuation parameters are similar. Taking liver stiffness as an example, the processing in steps 1041-1044 is explained. The liver fat attenuation parameters are based on the preprocessing of liver stiffness.

[0075] For example, the liver stiffness test data of a certain batch is [4.2, 5.1, 6.3, 7.5, 8.2, 25.6, 9.1] (unit: kPa), and after sorting, the sorted liver stiffness is [4.2, 5.1, 6.3, 7.5, 8.2, 9.1, 25.6].

[0076] In the ranking of liver stiffness, the 25th percentile data is 5.1, and the 75th percentile data is 9.1. The difference between the two is 4.0, which means the liver stiffness analysis value is 4.

[0077] The lower limit was set by subtracting 1.5 times the liver stiffness analysis value from the 25th percentile data, resulting in a lower limit of -0.9. This was then adjusted to 0 because liver stiffness cannot be negative.

[0078] The upper limit was set by adding 1.5 times the liver stiffness analysis value to the 75th percentile data, resulting in an upper limit of 15.1. All data with liver stiffness greater than 15.1 in the sorted liver stiffness were corrected to 15.1.

[0079] Finally, the liver stiffness was corrected to [4.2, 5.1, 6.3, 7.5, 8.2, 9.1, 15.1]. Similarly, the abnormal values ​​of the liver fat attenuation parameter were modified in a similar way.

[0080] Z-score standardization is used to transform the BMI data into a distribution with a mean of 0 and a standard deviation of 1, ensuring fairness in the model's weighting of each feature. The mean of the BMI data is calculated ( ) and standard deviation ( Each data point is standardized and transformed.

[0081] For example, the average BMI data of a certain group =24.5, standard deviation =3.2, a user's BMI=28.3.

[0082] The standardized value = (28.3 - 24.5) / 3.2 = 1.1875, according to the mathematical formula: ,in The raw values ​​of BMI data. This is the group mean of the BMI data. The group standard deviation of this BMI data. These are the standardized values. After the above processing, the BMI parameter normalization yields the corrected BMI parameter.

[0083] Set up mapping rules, encode male mapping values ​​as 1 and female mapping values ​​as 0, and obtain the corresponding gender mapping values ​​after mapping the gender parameters.

[0084] The parameters for corrected liver stiffness, corrected liver fat attenuation, corrected BMI, and gender mapping generated above are combined according to a certain set data structure to generate preprocessed integrated data.

[0085] 105. Based on the preset tag set, perform tag marking matrix processing on the preprocessed integrated data to generate a tag marking matrix;

[0086] In this embodiment, the data structure [[liver stiffness data', 'liver fat attenuation parameter', 'age', 'gender', 'BMI', 'lifestyle']] is sequentially matched and analyzed within a range of pre-defined labels to obtain corresponding labels, such as [5, 9.5, 15, 0, 0.7, 36]. Each row of samples corresponds one-to-one with five values ​​and one label. These labels are then integrated into the pre-processed integrated data to generate a label matrix.

[0087] For details, please refer to Figure 3 , Figure 3 This is a specific embodiment of step 105 of the liver chronic disease screening method in this invention, which includes the following specific implementation methods:

[0088] 1051. Perform element-matrix processing on the preprocessed integrated data to obtain an integrated data matrix;

[0089] 1052. Based on rows or columns, perform range matching and marking between the integrated data matrix and a preset tag set to obtain the tag data for rows or columns;

[0090] 1053. Merge the label data of the row or column with the integrated data matrix to generate a label marking matrix.

[0091] In steps 1051-1053, the preprocessed integrated data may exist in dictionary form. The data is then matrixed according to the dictionary elements to obtain an integrated data matrix in list form.

[0092] Based on the data structure of the integrated data matrix, data is sorted in columns, and the corresponding column data is matched with a preset label set for range marking. Alternatively, data is arranged in rows, and the corresponding row data is matched with a preset label set for range marking, resulting in label data consisting of rows or columns. Following the one-to-one correspondence between label data and entity data in the integrated data matrix, rows or columns are added to the integrated data matrix as markers to add label data, generating a label matrix.

[0093] 106. According to the preset gradient boosting algorithm, the label matrix is ​​fitted with residuals to obtain extended decision tree data;

[0094] In this embodiment, the global probability calculation is first performed on the labels in the label matrix, and the global proportion of each sample label is generated as the predicted value F0(x) of the i-th sample in round 0. i For the predicted value F of the i-th sample in the m-th round. m (x) i ), which is converted into the estimated probability y of the i-th sample in the m-th round. i ^ =σ(F m (x) i Then, the true y of the i-th sample... i (Take 0 or 1), and finally obtain the residual r of the m-th round. i =y i -y i ^ residual r i A larger value indicates that the model underestimates or overestimates. Multiple rounds of data fitting are used to adjust the residuals r. i Sample i is split into a 3-layer regression tree using the squared error. Each split leaf stores the "mean residual of that leaf" as the initial residual value. The decision tree is generated iteratively, and finally the extended decision tree data is obtained.

[0095] For details, please refer to Figure 4 , Figure 4 This is a specific embodiment of the 106 steps of the liver chronic disease screening method in this invention, which includes the following specific implementation methods:

[0096] 1061. Globally scale the rows or columns of the corresponding tags in the tag marking matrix to obtain a global tag matrix;

[0097] 1062. Based on the corresponding tag elements of the global tag matrix, perform probability residual calculation on the rows or columns of the global tag matrix to obtain the probability residual;

[0098] 1063. Based on the preset CART regression tree, the rows or columns of the global label matrix are split to obtain extended decision tree data.

[0099] In steps 1061-1063, assuming there are 3 samples, label 1 is set to {1,0,1}, and label 2 is set to {0,0,1}. The global proportion of the three positive samples for label 1 is 2 / 3, and the global proportion of the three positive samples for label 2 is 1 / 3. Calculate the global proportions log(p / (1-p)) for label 1 and label 2 respectively, where p takes values ​​of 2 / 3 and 1 / 3, resulting in 0.6931 and -0.6931. Replace the original labels with the obtained global proportions for label 1 and label 2, so label 1 becomes {0.6931,0.6931,0.6931} and label 2 becomes {-0.6931,-0.6931,-0.6931}, obtaining the global label matrix. Then, set the global proportion of each sample label in the global label matrix to the predicted value F0(x) of the i-th sample in round 0. i ), for the predicted value F of the i-th sample in the m-th round m (x) i ), which is converted into the estimated probability y of the i-th sample in the m-th round. i ^ =σ(F m (x) i Then, the true y of the i-th sample... i (Take 0 or 1), and finally obtain the residual r of the m-th round. i =y i -y i ^ residual r i A larger value indicates that the model underestimates or overestimates. Multiple rounds of data fitting, based on a pre-set CART regression tree, are used to calculate the residuals r. i The sample i is split into a 3-layer regression tree using the squared error. Each leaf contains the mean residual of that leaf as the initial residual value r. i The process iteratively generates a decision tree, and the residuals are fitted using a CART regression tree. Each tree has a maximum of 3 layers, resulting in a maximum of 8 leaves. For samples falling into the same leaf, an optimal step size and learning rate γ are calculated, and F is updated. m (x) i )=F m-1 (x) i ) +0.1*γ*tree, where tree represents the sample weights equal to [r1, r2, ..., r i ]^T, output the new adjusted sample weights, save 100 trees, generate data for 3 data layers, each tree has 8 leaves (may be less), the total number of leaves is less than 800, and finally obtain the extended decision tree data.

[0100] Specifically, the step of globally scaling the rows or columns of the tag matrix to obtain the global tag matrix includes:

[0101] F0(x)=log(p / (1-p)),

[0102] Where F0(x) is the global proportion of a single label, and p is the positive sample probability proportion of row or column labels.

[0103] In this embodiment, assuming there are 3 samples, label 1 is set to {1,0,1} and label 2 is set to {0,0,1}. The global proportion of the three positive samples for label 1 is 2 / 3, and the global proportion of the three positive samples for label 2 is 1 / 3. The global proportions log(p / (1-p)) for label 1 and label 2 are calculated respectively, where p takes values ​​of 2 / 3 and 1 / 3, resulting in 0.6931 and -0.6931. These global proportions are then used to replace the original labels, resulting in label 1 becoming {0.6931,0.6931,0.6931} and label 2 becoming {-0.6931,-0.6931,-0.6931}, thus obtaining the global label matrix.

[0104] 107. According to the preset sampling algorithm, the extended decision tree data is subjected to discrete sampling processing to obtain sparse vectors;

[0105] In this embodiment, for each sample, 100 trees are assigned the "leaf number" to which it falls. Each tree has 8 leaves, and 100 trees have a maximum of 800 features (actually sparse; missing leaves result in no corresponding column). If the test set contains a leaf number not seen in the training set, it is directly mapped to a vector of all zeros. After discrete sampling of the decision tree, a sparse vector containing only non-zero values ​​is obtained. It should be noted that the sampling logic and method are obtained during training.

[0106] Specifically, step 107 includes the following specific implementation methods:

[0107] 1071. Based on the preset OneHotEncoder() function, discrete features are extracted from the extended decision tree data to obtain sparse vectors.

[0108] In step 1071, the OneHotEncoder() function in the sci-sklearn module can be used to perform discrete leaf sampling on the extended decision tree data. OneHotEncoder(handle_unknown='ignore') is used here to extract discrete features. It is set to the function to directly map leaf numbers that appear in the test set but not in the training set to an all-zero vector. Based on OneHot encoding, the sparse vector is extracted into memory for computation.

[0109] 108. According to the preset logistic regression algorithm, the sparse vector is classified and processed to obtain the classification result;

[0110] In this embodiment, sparse vectors are linearly scored and compressed using the sigmoid function. Samples are then assigned to their respective categories according to a pre-set threshold to obtain the classification results. The LogisticRegression model can be used in the scikit-learn module for classification training of sparse vectors. The penalty can be set to L2 penalty, where L2 = 1 / C, and the penalty coefficient C is set to 2.

[0111] 109. Based on the classification results, generate a disease screening report and send the disease screening report to the mobile terminal system.

[0112] In this embodiment, based on the classification results obtained from the chronic disease classification analysis trained above, a basic report is generated by replacing placeholders in the classification results, user information, and collected feature data according to the original template. A large model with a disease analysis knowledge base is then invoked to analyze the entire classification result and generate analysis text. This analysis text is added to the basic report to generate a disease screening report. Finally, the disease screening report is sent to the mobile system so that users can receive the report promptly and view it on the app.

[0113] Specifically, following step 109, the following specific implementation methods are also included:

[0114] 1091. Send the disease screening report to the liver detection system and receive the clinical feedback data returned by the liver detection system;

[0115] 1092. Based on the clinical feedback data, adjust the label set to obtain a new label set.

[0116] In steps 1091-1092, the report includes the detection indicators (LSM / CAP), risk score, grading results, and personalized recommendations (such as lifestyle adjustments, follow-up suggestions, and retesting intervals). Simultaneously, the system pushes abnormal risk alerts to users or healthcare professionals and sends the disease screening report to the liver detection system. Doctors or health management personnel can view the summarized screening data through the cloud platform and, through clinical observation, receive clinical feedback data for each disease screening report.

[0117] The cloud service system can modify the label set matching the collected feature data based on the returned clinical feedback data to obtain a new label set, so as to adjust the label set in a timely manner to improve the accuracy of risk warning and provide a basis for actual deployment and optimization.

[0118] In this embodiment of the invention, a three-system architecture is established, comprising a liver detection system, a mobile terminal system, and a cloud service system. The liver detection system collects liver tissue stiffness values ​​and liver fat attenuation parameters, which, combined with basic information from the target individual's electronic medical record or health record, are sent to the mobile terminal system. After user confirmation on the mobile terminal system, the collected feature data is sent to the cloud service system. The cloud service system performs feature analysis and regression calculations on the collected feature data to generate high-precision disease screening results, which are then sent back to the mobile terminal system. This improves the timeliness and accuracy of early warning for chronic liver diseases, solving the technical problems of insufficient early warning timeliness and inaccurate analysis results for chronic liver diseases.

[0119] Figure 5 This is a schematic diagram of the structure of a liver chronic disease screening device 500 provided in an embodiment of the present invention. The liver chronic disease screening device 500 can vary considerably due to different configurations or performance. It may include one or more central processing units (CPUs) 510 (e.g., one or more processors) and a memory 520, and one or more storage media 530 (e.g., one or more mass storage devices) for storing application programs 533 or data 532. The memory 520 and storage media 530 can be temporary or persistent storage. The program stored in the storage media 530 may include one or more modules (not shown in the diagram), each module may include a series of instruction operations on the liver chronic disease screening device 500. Furthermore, the processor 510 may be configured to communicate with the storage media 530 and execute the series of instruction operations in the storage media 530 on the liver chronic disease screening device 500.

[0120] The liver chronic disease screening device 500 may also include one or more power supplies 540, one or more wired or wireless network interfaces 550, one or more input / output interfaces 560, and / or one or more operating systems 531, such as Windows Server, Mac OS X, Unix, Linux, Free BSD, etc. Those skilled in the art will understand that... Figure 5 The illustrated structure of the liver chronic disease screening device does not constitute a limitation on liver chronic disease screening devices, which may include more or fewer components than illustrated, or combine certain components, or have different component arrangements.

[0121] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium, wherein the computer-readable storage medium stores instructions that, when the instructions are executed on a computer, cause the computer to perform the steps of the liver chronic disease screening method.

[0122] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0123] Furthermore, although the operations are described in a specific order, this should be understood as requiring that such operations be performed in the specific order shown or in sequential order, or requiring that all illustrated operations be performed to achieve the desired result. In certain environments, multitasking and parallel processing may be advantageous. Similarly, although several specific implementation details are included in the above discussion, these should not be construed as limiting the scope of this disclosure. Certain features described in the context of individual embodiments may also be implemented in combination in a single implementation. Conversely, various features described in the context of a single implementation may also be implemented individually or in any suitable sub-combination in multiple implementations.

[0124] Although the subject matter has been described using language specific to structural features and / or methodological logic, it should be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or actions described above. Rather, the specific features and actions described above are merely illustrative examples of implementing the claims.

Claims

1. A method for screening chronic liver diseases, characterized in that, The liver chronic disease screening method is applied in a liver chronic disease screening system, which includes: a liver detection system, a mobile terminal system, and a cloud service system. The chronic disease screening method includes: The liver detection system collects liver stiffness data and liver fat attenuation parameters of the target object, and sends the liver stiffness data and liver fat attenuation parameters to the mobile terminal system; The mobile terminal system receives the liver stiffness data and the liver fat attenuation parameter, combines the liver stiffness data and the liver fat attenuation parameter with the basic information of the preset target object to generate collected feature data, and displays the collected feature data. When a confirmation command is received, the collected feature data is sent to the cloud service system. The cloud service system receives the collected feature data and preprocesses the collected feature data to obtain preprocessed integrated data; Based on a preset tag set, the preprocessed integrated data is subjected to tag matrix processing to generate a tag matrix; According to the preset gradient boosting algorithm, the label matrix is ​​fitted with residuals to obtain extended decision tree data; According to the preset sampling algorithm, the extended decision tree data is subjected to discrete sampling processing to obtain sparse vectors; The sparse vectors are classified according to a preset logistic regression algorithm to obtain the classification result. Based on the classification results, a disease screening report is generated, and the disease screening report is sent to the mobile terminal system; The step of performing tagging matrix processing on the preprocessed integrated data according to the preset tag set to generate a tagging matrix includes: The preprocessed integrated data is then subjected to element-matrix processing to obtain an integrated data matrix; Based on rows or columns, the integrated data matrix is ​​matched and labeled with a preset tag set to obtain the tag data of the rows or columns; The label data of the rows or columns are merged with the integrated data matrix to generate a label marking matrix; The step of performing residual processing on the label matrix according to a preset gradient boosting algorithm to obtain extended decision tree data includes: The rows or columns of the corresponding tags in the tag marking matrix are globally scaled to obtain a global tag matrix; Based on the corresponding tag elements of the global tag matrix, probability residuals are calculated on the rows or columns of the global tag matrix to obtain probability residuals; Based on a preset CART regression tree, the rows or columns of the global label matrix are split to obtain extended decision tree data.

2. The method for screening chronic liver diseases according to claim 1, characterized in that, The step of globally scaling the rows or columns of the corresponding tags in the tag marking matrix to obtain the global tag matrix includes: F0(x)=log(p / (1-p)), Where F0(x) is the global proportion of a single label, and p is the positive sample probability proportion of row or column labels.

3. The method for screening chronic liver diseases according to claim 1, characterized in that, The step of performing discrete sampling processing on the extended decision tree data according to a preset sampling algorithm to obtain sparse vectors includes: Based on the preset OneHotEncoder() function, discrete features will be extracted from the extended decision tree data to obtain sparse vectors.

4. The method for screening chronic liver diseases according to claim 1, characterized in that, The collected feature data includes: liver stiffness, liver fat attenuation parameters, BMI parameters, and gender parameters. The step of preprocessing the collected feature data to obtain preprocessed integrated data includes: The liver stiffness and liver fat attenuation parameters are sorted from smallest to largest to obtain sorted liver stiffness and liver fat sorting parameters. Calculate the difference between the 75th percentile and 25th percentile data of the sorted liver stiffness and liver fat sorting parameters to obtain the liver stiffness analysis value and liver fat analysis value. Based on the liver stiffness analysis values, the sorted liver stiffness is truncated by upper and lower limits to obtain the corrected liver stiffness. Based on the liver fat analysis values, the upper and lower limits of the liver fat ranking parameters are truncated to obtain the corrected liver fat attenuation parameters. The BMI parameters are normalized according to the preset Z-score algorithm to obtain the corrected BMI parameters; According to the preset mapping rules, the gender parameter is mapped to obtain a gender mapping value; The modified liver stiffness, the modified liver fat attenuation parameter, the modified BMI parameter, and the gender mapping value are combined to obtain preprocessed integrated data.

5. The method for screening chronic liver diseases according to claim 1, characterized in that, Following the step of displaying the acquired feature data, the method further includes: When an error feedback instruction is received, the feedback input instruction is received and error description data is generated. The abnormality description data and the collected feature data are sent to the liver detection system in order to adjust the abnormal data in the collected feature data.

6. The method for screening chronic liver diseases according to claim 1, characterized in that, After the steps of generating a disease screening report based on the classification results and sending the disease screening report to the mobile terminal system, the method further includes: The disease screening report is sent to the liver detection system, and the clinical feedback data returned by the liver detection system is received. Based on the clinical feedback data, the label set is adjusted to obtain a new label set.

7. A liver chronic disease screening device, characterized in that, The liver chronic disease screening device includes: a memory and at least one processor, wherein the memory stores instructions, and the memory and the at least one processor are interconnected via a circuit; The at least one processor invokes the instructions in the memory to cause the liver chronic disease screening device to perform the liver chronic disease screening method as described in any one of claims 1-6.

8. A computer-readable storage medium storing a computer program thereon, characterized in that, When the computer program is executed by a processor, it implements the liver chronic disease screening method as described in any one of claims 1-6.

Images