Methods and systems for analyzing and planning the treatment of edema after hemorrhagic stroke

Abstract

The present invention discloses a method and system for analyzing edema after hemorrhagic stroke and for treatment planning. The method involves collecting and preprocessing edema volume data for each patient at different time points. Then, using a Gaussian fitting algorithm to construct a curve of the edema volume evolution over time and calculating the residual from the actual value, dividing the patients into subgroups using a deep learning clustering algorithm, fitting the edema volume for each subgroup at different time points and calculating the residual from the actual value. The change in edema volume is calculated, trends of change are identified, and the corresponding treatment plans are combined with the corresponding edema evolution trend and analyzed using an association rule algorithm to identify the treatment plan with the highest correlation. The present invention utilizes advanced algorithms to accurately analyze changes in edema volume after hemorrhagic stroke and improve the accuracy of patient prognosis prediction. Detailed analysis of different patient subgroups supports more personalized treatment plans and contributes to improved treatment efficacy.

Description

[Technical Field]

[0001] This invention relates to the technical field of clinical intelligence medicine, and more specifically to a method and system for analyzing edema and planning treatment after hemorrhagic stroke. [Background technology]

[0002] Hemorrhagic stroke is a severe intracerebral hemorrhage disorder caused by the rupture of blood vessels within the brain parenchyma without trauma. Its acute onset and rapid progression often lead to a high mortality rate and severe neurological impairment. This disease not only poses significant health risks to the patient but also places a heavy economic burden on society and families. In particular, in the later stages of hemorrhagic stroke, hematoma expansion is considered a significant risk factor for poor prognosis, leading to increased intracranial pressure, deterioration of neurological function, and potentially even death.

[0003] Furthermore, edema around hematomas has recently attracted increasing clinical attention as an important marker of secondary injury after hemorrhagic stroke. Such edema can cause compression of brain tissue, affect neuronal function, lead to further brain tissue damage, and worsen neurological dysfunction. Therefore, dynamic monitoring and early prediction of edema around hematomas are crucial for improving patient survival and quality of life.

[0004] With the continuous advancements in medical imaging technology, non-invasive dynamic monitoring of brain damage after hemorrhagic stroke has become possible. Furthermore, the rapid development of artificial intelligence technology has revolutionized data processing and analysis in the medical field. By integrating multi-source data such as imaging features, clinical information, and treatment plans, it is possible to accurately predict the prognosis of patients after hemorrhagic stroke and to evaluate the effectiveness of individualized treatment. [Overview of the project] [Problems that the invention aims to solve]

[0005] This invention proposes a comprehensive analysis method based on a Gaussian fitting algorithm, a deep learning clustering algorithm, and a correlation rule algorithm. The method aims to accurately predict and evaluate the trends in edema volume changes in patients after hemorrhagic stroke, and the correlation between these changes and different treatment options. Using the Gaussian fitting algorithm, the invention can construct a curve representing the time-dependent changes in edema volume and calculate the residuals between the actual observed values ​​and the fitted curve. Utilizing the deep learning clustering algorithm, the invention further distinguishes between different patient subgroups and performs a dedicated edema volume trend analysis for each subgroup. Finally, using the correlation rule algorithm, the invention can identify the treatment method most highly correlated with the progression of edema, providing crucial support for clinical treatment decisions. [Means for solving the problem]

[0006] In view of the aforementioned existing problems, the present invention was proposed.

[0007] Accordingly, the technical problem that the present invention aims to solve is as follows: to effectively monitor and predict the expansion of hematoma and the progression of peripheral edema after hemorrhagic stroke, and to accurately evaluate the relationship between treatment methods and the trend of changes in the volume of edema in patients, thereby creating an individualized treatment plan.

[0008] To solve the above technical problems, the present invention proposes the following technical solution: A method for analyzing edema and planning treatment after hemorrhagic stroke, Collect and pre-process edema data for each patient at different time points. The process involves fitting pre-processed data using a Gaussian fitting algorithm to construct a curve showing the temporal progression of edema volume across the entire patient, and calculating the residual between the curve and the actual values. Using a deep learning clustering algorithm, patients are divided into multiple subgroups, a Gaussian function is fitted to the edema levels of patients in each subgroup at different time points, and the residuals between the fitted curves for each subclass and the actual values ​​for each subclass are calculated. This includes calculating the change in edema volume, identifying the trend of change, and combining it with the corresponding treatment plan. A correlation rule algorithm is then used to analyze the degree of association between the treatment method and the trend of edema progression, thereby identifying the treatment plan with the highest correlation to the trend of edema progression.

[0009] A preferred embodiment of the method for analyzing edema and planning treatment after hemorrhagic stroke described in the present invention includes the following preprocessing steps: cleans edema volume data at different time points for each patient, corrects errors and abnormal values, fills in missing edema volume data with mean values ​​and performs linear interpolation, synchronizes the edema volume data based on the patient's examination time and onset time, and performs standardization on the edema volume data.

[0010] A preferred embodiment of the method for analyzing edema and treatment planning after hemorrhagic stroke described in the present invention is to fit the pre-processed data using the Gaussian fitting algorithm. The pre-processed edema volume data is sorted by patient and time point to form a structured dataset, where each row represents the edema volume at a specific time point for each patient. The formula for selecting the Gaussian function is as follows: JPEG0007843414000001.jpg1231(1) However, v represents the time from the onset of symptoms to the examination, a, b, and c are parameters of the Gaussian function curve for the patient's edema time after fitting, a represents the peak value of the curve, b represents the mean value of the curve, and c represents the width of the curve. For each patient's data, the parameters a, b, and c of the Gaussian model are estimated using the least squares method, and then iterative parameter optimization is performed using gradient descent. This includes applying a Gaussian function to the time-series data of each patient to generate a fitting curve for the change in edema volume over time.

[0011] A preferred embodiment of the method for analyzing edema after hemorrhagic stroke and planning treatment according to the present invention is to calculate the residual between the curve and the actual value. The residual is defined as the difference between the predicted value of the Gaussian fitting curve and the actual observed value, and the calculation formula is as follows: JPEG0007843414000002.jpg1127(2) However, R i represents the fitting residual at time i, and Y i This represents the actual value of the edema volume at time point i. JPEG0007843414000003.jpg85 represents the predicted value of the Gaussian fitting curve at the same time point, This includes calculating the residuals between the edema volume data for each patient at each time point and the corresponding points on the Gaussian fitting curve.

[0012] A preferred embodiment of the method for analyzing edema and treatment planning after hemorrhagic stroke described in the present invention is to divide patients into multiple subgroups using the deep learning clustering algorithm. For different patient personal information features, an autoencoder is used to extract features from patient data. In the encoding process, the original various feature data x are converted into low-dimensional latent variables h by the encoder, and the calculation formula is as follows: JPEG0007843414000004.jpg831(3) However, W1 and b1 are the weights and biases of the encoder, respectively, σ is the activation function, and m is the feature value in the feature encoder. During the decoding process, the latent variable h, after undergoing the encoding process, is decoded back to its original data dimension, resulting in the reconstructed data. The file JPEG0007843414000005.jpg76 was obtained, and the calculation formula is as follows: JPEG0007843414000006.jpg1042(4) However, W2 represents the weights of the feature decoder, b2 represents the bias of the feature decoder, and the parameters of the autoencoder are iteratively optimized by the loss function, and the loss function is defined to minimize the difference between the original data x and the reconstructed data JPEG0007843414000007.jpg76, and its formula is as follows: JPEG0007843414000008.jpg841(5) However, G represents the feature information dataset of each original patient, and G R represents the feature information dataset of each original patient reconstructed by the encoder, and the low-dimensional representation h obtained by training the autoencoder is used for K-Means clustering. Based on the preset number k of clusters, the algorithm divides the patients into k clusters, and each cluster is composed of patients with similar features. (C1, C2…C k ) is represented, and the minimum squared error E is represented as follows: JPEG0007843414000009.jpg1031(6) However, TIFF0007843414000010.tif45 represents the norm of the vector, l represents the low-dimensional representation of the patient information, and μ i represents the mean vector of cluster C i .

[0013] As a preferred form of the method for analyzing edema and treatment plan after hemorrhagic stroke described in the present invention, calculating the residuals between the fitted curves of each subclass and the actual values of each subclass includes performing Gaussian fitting on the patient data within each subgroup divided by the deep learning clustering algorithm, and calculating the residuals between the Gaussian fitting curves of each subgroup and the actual edema volume data of each patient within the subgroup respectively.

[0014] A preferred form of the method for analyzing edema and treatment planning after hemorrhagic stroke described in the present invention is to analyze the degree of association between treatment methods and the trend of edema progression using the correlation rule algorithm. Extracting the change in edema volume and corresponding treatment plan information from the preprocessed dataset, Define item sets X and Y, where X represents a specific treatment plan and Y represents a specific trend in changes in edema volume. The support calculation is performed, and the support(X) of the data item set X is the ratio of the number of transactions containing X in D to the total number of transactions in D, and the formula is shown as follows: JPEG0007843414000011.jpg1033(7) However, D represents the progression of edema, x represents the specific treatment plan selected, count(x) represents the number of specific treatment plans selected, and count(D) represents the number of cases in which the progression of edema was suppressed. The degree of support for the correlation rule X=>Y is the item set The support level is equal to that of JPEG0007843414000012.jpg410, and the formula is shown as follows: JPEG0007843414000013.jpg1171(8) However, X represents a specific treatment plan that has been selected, and Y represents another specific treatment plan that has been selected. JPEG0007843414000014.jpg617 represents the number of two treatment plans selected, Calculate the confidence level of X=>Y, and use the following formula: JPEG0007843414000015.jpg1254(9) If a specific treatment plan is included, evaluate the rules for the possibility of another treatment plan emerging, however, JPEG0007843414000016.jpg721 represents the level of support where both treatment methods exist, and Support(x) represents the level of support where both of the former treatment methods exist, This includes iteratively generating sets of frequently occurring items and corresponding correlation rules using the Apriori algorithm, and identifying the rule with the highest support and confidence level from the generated correlation rules.

[0015] A system for analyzing edema and planning treatment after hemorrhagic stroke, It comprises a data preprocessing module, a Gaussian fitting module, a patient subgroup classification module, and a treatment planning-related analysis module. The aforementioned data preprocessing module preprocesses the edema volume data for each patient at different time points, The Gaussian fitting module uses a Gaussian fitting algorithm to fit the preprocessed data, constructs a curve showing the temporal progression of edema volume for the entire patient, and calculates the residual between the curve and the actual values. The patient subgroup classification module divides patients into multiple subgroups using a deep learning clustering algorithm, performs Gaussian function fitting on the edema levels of patients in each subgroup at different time points, calculates the residuals between the fitted curves for each subclass and the actual values ​​for each subclass, The aforementioned treatment plan-related analysis module calculates the change in edema volume, identifies the trend of change, and, in combination with the corresponding treatment plan, analyzes the degree of correlation between the treatment method and the trend of edema progression using a correlation rule algorithm, thereby identifying the treatment plan with the highest degree of correlation with the trend of edema progression.

[0016] A computer device comprising memory and a processor, wherein a computer program is stored in the memory and the processor executes the computer program, thereby realizing the steps of the above method.

[0017] A computer-readable storage medium, wherein a computer program is stored in the computer-readable storage medium. When the aforementioned computer program is executed by the processor, the steps of the above method are realized. [Effects of the Invention]

[0018] The beneficial effects of this invention are as follows: By utilizing advanced algorithms to accurately analyze volume changes in edema after hemorrhagic stroke, the invention improves the accuracy of predicting patient prognosis. By analyzing different patient subgroups in detail, the invention supports more personalized treatment methods and helps improve treatment efficacy. By identifying and accurately predicting the progression of hematoma and edema at an early stage, it helps to take effective measures in a timely manner and improve patient prognosis and quality of life. [Brief explanation of the drawing]

[0019] To more clearly explain the technical solutions of the embodiments of the present invention, the drawings necessary for describing the embodiments will be briefly described below. Clearly, the drawings in the following description are only a few embodiments of the present invention, and those skilled in the art can obtain other drawings based on these without requiring any creative effort. Here, [Figure 1] This is an overall flowchart of the method for analyzing edema after hemorrhagic stroke and planning treatment, provided by the first embodiment of the present invention. [Figure 2] This is a flowchart of the Apriori algorithm, a correlation rule for analyzing edema after hemorrhagic stroke and planning treatment, provided in the first embodiment of the present invention. [Figure 3] This is a fitting diagram of the total patient edema volume versus time for the method of analyzing edema and planning treatment after hemorrhagic stroke provided in the second embodiment of the present invention. [Figure 4] This is a curve diagram showing the temporal changes in edema volume in patients of subclass 0 of the method for analyzing edema and planning treatment after hemorrhagic stroke provided in the second embodiment of the present invention. [Figure 5] This is a curve diagram showing the temporal changes in edema volume in patients of subclass 1 of the method for analyzing edema and planning treatment after hemorrhagic stroke provided in a second embodiment of the present invention. [Figure 6] This is a curve diagram showing the temporal changes in edema volume in patients of subclass 2 of the method for analyzing edema and planning treatment after hemorrhagic stroke provided in the second embodiment of the present invention. [Figure 7] This is a curve diagram showing the temporal changes in edema volume in patients of subclass 3 of the method for analyzing edema and planning treatment after hemorrhagic stroke provided in the second embodiment of the present invention. [Modes for carrying out the invention]

[0020] To make the above-mentioned objectives, features, and advantages of the present invention clearer and easier to understand, specific embodiments of the present invention will be described in detail below with reference to the drawings of the specification. The embodiments described herein are only some, and not all, embodiments of the present invention. All other embodiments obtained by those skilled in the art without requiring creative effort based on the embodiments of the present invention should all fall within the scope of protection of the present invention.

[0021] Example 1 Referring to Figures 1 and 2, which represent one embodiment of the present invention, a method for analyzing edema after hemorrhagic stroke and for analyzing treatment planning is provided, and this method is In S1, preprocessing is performed on the edema volume data for each patient at different time points.

[0022] The system cleanses edema data from different time points for each patient, corrects errors and outliers, fills in missing edema data with mean values ​​and performs linear interpolation, synchronizes the edema data based on the patient's examination time and onset time, and then performs standardization on the edema data.

[0023] In S2, the pre-processed data is fitted using a Gaussian fitting algorithm to construct a curve showing the temporal progression of edema volume for the entire patient, and the residuals between the curve and the actual values ​​are calculated.

[0024] The pre-processed edema volume data is sorted by patient and time point to form a structured dataset, where each row represents the edema volume at a specific time point for each patient. The formula for selecting the Gaussian function is as follows: JPEG0007843414000017.jpg1231(1) However, x represents the time from the onset of symptoms to the examination, and a, b, and c are parameters of the Gaussian function, representing the peak value, mean value, and width of the descriptive curve, respectively.

[0025] For each patient's data, the parameters a, b, and c of the Gaussian model were estimated using the least squares method. A Gaussian function is applied to the time-series data of each patient to generate a fitted curve for the change in edema volume over time.

[0026] The residual is defined as the difference between the predicted value of the Gaussian fitting curve and the actual observed value, and the calculation formula is as follows: JPEG0007843414000018.jpg1127(2) However, R i This represents the residual of the i-th data point, Y i This represents the actual observed value, JPEG0007843414000019.jpg75 represents the predicted value of the Gaussian fitting curve at the same point in time. For each patient's edema volume data at each time point, the residuals between the data and the corresponding points on the Gaussian fitting curve are calculated.

[0027] In S3, a deep learning clustering algorithm is used to divide patients into multiple subgroups. Gaussian function fitting is performed on the edema levels of patients in each subgroup at different time points, and the residuals between the fitted curves for each subclass and the actual values ​​for each subclass are calculated.

[0028] Feature extraction is performed on patient data using an autoencoder. During the encoding process, the original data x is transformed into a low-dimensional latent variable h by the encoder. The calculation formula is as follows: JPEG0007843414000020.jpg841(3) However, W1 and b1 are the weights and biases of the encoder, respectively, and σ is the activation function. In the process of decoding, the latent variable h is decoded into the original data dimension to obtain the reconstructed data to obtain JPEG0007843414000021.jpg76, and the calculation formula is as follows JPEG0007843414000022.jpg740(4) where W2 and b2 are the weights and biases of the decoder, and the parameters of the autoencoder are optimized to minimize the difference between the original data x and the reconstructed data to obtain JPEG0007843414000023.jpg76, and the formula is as follows JPEG0007843414000024.jpg841(5) The low-dimensional representation h obtained by training the autoencoder is used for K-Means clustering. Based on the preset number of clusters k, the algorithm divides the patients into k clusters, and each cluster is composed of patients with similar characteristics. (C1, C2…C k ) is represented as, and the minimum squared error E is represented as follows JPEG0007843414000025.jpg1031(6) where μ i is the mean vector of cluster c i , and is represented by the following formula JPEG0007843414000026.jpg1353(10) For the patient data within each subgroup divided by the deep learning clustering algorithm, Gaussian fitting is performed, and for the Gaussian fitting curve of each subgroup, the residual between the actual edema amount data of each patient within the subgroup is calculated.

[0029] The deep learning clustering algorithm (Autoencoder K-Means Clustering) combines the feature learning and K-means clustering capabilities of an autoencoder. First, it trains data using an autoencoder to obtain a low-dimensional encoded representation H. Next, it uses the K-Means clustering algorithm to cluster H into K clusters, improving the K-means clustering performance through autoencoder feature learning.

[0030] In S4, the change in edema volume is calculated, the trend of change is identified, and this is combined with the corresponding treatment plan. A correlation rule algorithm is then used to analyze the degree of association between the treatment method and the trend of edema progression, and the treatment plan with the highest degree of association with the trend of edema progression is identified.

[0031] From the preprocessed dataset, the change in edema volume and corresponding treatment plan information are extracted, and item sets X and Y are defined, where X represents a specific treatment plan and Y represents a specific trend in the change in edema volume. The support calculation is performed, and the support(X) of data item set X is the ratio of the number of transactions containing X in D to the total number of transactions in D, and the formula is shown as follows: JPEG0007843414000027.jpg1033(7) However, D represents the progression of edema, x represents the selected specific treatment plan, count(x) represents the number of selected specific treatment plans, and count(D) is the number of cases in which the trend of edema progression was suppressed. The support for the correlation rule X=>Y is the item set. The support level is equal to that of JPEG0007843414000028.jpg410, as shown by the following formula: JPEG0007843414000029.jpg1171(8) However, X represents a specific treatment plan that has been selected, and Y represents another specific treatment plan that has been selected. JPEG0007843414000030.jpg617 represents the number of the two selected treatment plans, and the confidence level of X=>Y is calculated using the following formula: JPEG0007843414000031.jpg1254(9) If a specific treatment plan is included, evaluate the rules for the possibility of another treatment plan emerging, however, JPEG0007843414000032.jpg721 represents the level of support where both treatment methods exist, while Support(X) represents the level of support where all company-wide treatment methods exist.

[0032] The Apriori algorithm is used to iteratively generate sets of frequently occurring items and their corresponding correlation rules, and from these generated correlation rules, the rule with the highest support and confidence level is identified.

[0033] As shown in Figure 2, the Apriori algorithm is an iterative method. The k-th most frequent item set L is used. k Connect using (k+1)th candidate set C k+1 Generate the (k+1)th most frequent item set L, and then perform a pruning operation. k+1 Obtain the data and scan the database multiple times, repeating the above operation until no more frequent item sets can be found. After mining of frequent item sets is complete, one k-th frequent item set L k For this, 2(2k-1) meaningful rules (where the preceding and succeeding items are not blank) can be generated, but none of them satisfy the condition. k A correlation rule "s→(1-s)" that satisfies the requirements is generated only if the confidence level of the non-blank subset S is greater than the minimum confidence level.

[0034] The above embodiment further includes an analysis system for edema analysis and treatment planning after hemorrhagic stroke, specifically, The data preprocessing module preprocesses the edema volume data for each patient at different time points. The Gaussian fitting module uses a Gaussian fitting algorithm to fit the pre-processed data to construct a curve showing the temporal progression of edema volume for the entire patient, and calculates the residual between the curve and the actual values. The patient subgroup classification module uses a deep learning clustering algorithm to divide patients into multiple subgroups, performs Gaussian function fitting on the edema levels of patients in each subgroup at different time points, and calculates the residuals between the fitted curves for each subclass and the actual values ​​for each subclass. The treatment plan-related analysis module calculates the change in edema volume, identifies trends in change, and combines them with corresponding treatment plans. Using a correlation rule algorithm, it analyzes the degree of correlation between treatment methods and the trend of edema progression, identifying the treatment plan with the highest correlation to the trend of edema progression.

[0035] The computer device may be a server. The computer device includes a processor, memory, an input / output interface (I / O), and a communication interface. The processor, memory, and input / output interface are connected via a system bus, and the communication interface is connected to the system bus via the input / output interface. The processor of the computer device is used to provide computational and control capabilities. The memory of the computer device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system, computer programs, and a database. The internal memory provides an environment for executing the operating system and computer programs in the non-volatile storage medium. The database of the computer device is used to store data cluster data for a power monitoring system. The input / output interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with external terminals via a network connection. When the computer program is executed by the processor, it realizes a method for analyzing edema and planning treatment after hemorrhagic stroke.

[0036] As those skilled in the art will understand, all or part of the processes in the above embodiments can be implemented by instructing the relevant hardware with a computer program, and such computer program can be stored in a non-volatile computer-readable storage medium, and when the computer program is executed, it can include the processes of each embodiment of the above embodiments. Here, any reference to memory, database or other medium used in each embodiment provided by this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an explanation rather than an limitation, RAM may take various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The databases relating to each embodiment provided in this application may include at least one of relational databases and non-relational databases. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors relating to each embodiment provided in this application may include, but are not limited to, general-purpose processors, central processors, graphics processors, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc.

[0037] Example 2 Figures 3 to 7 illustrate one embodiment of the present invention, providing a method for analyzing edema and planning treatment after hemorrhagic stroke. To verify the beneficial effects of the present invention, scientific evidence is provided through calculations of economic effects and simulations / comparative experiments.

[0038] Figure 3 shows the curve of the time course of edema in the entire patient population using the Gaussian function fitting method of the present invention, and the curves obtained by dividing 100 patients into four subclasses using a deep learning clustering algorithm and fitting each subclass with the Gaussian function.

[0039] In real-world scenarios, randomness, independence, and countability exist in the time from symptom onset to hospitalization and the first test. The number of subsequent tests and test times for each patient also demonstrate strong independence and randomness in their statistical data. Therefore, for large patient datasets, data points where no tests were performed in subsequent data are cleared. With a total of 450 data points across all patients, the results of fitting the curve using a Gaussian Fitting Algorithm are shown in Figure 3 below.

[0040] As can be seen from the analysis by referring to Figure 3, many of the points in time when the patient's edema volume is large are concentrated within xx hours after simultaneous onset, the data shows a tendency to decrease over time, there is a clear peak, and it can be confirmed that the fitted data (curve) has remarkable accuracy. Fitting the curve with a Gaussian Fitting Algorithm function clearly shows that it maintains the distribution characteristics of the data and the superiority of the smoothness of the fitted curve well. By calculating the residuals between the actual values ​​and the corresponding points in time on the fitted curve, the serial number of the point in time corresponding to the residual is found, and the residual values ​​corresponding to all serial numbers for each patient are matched. Then, the mean squared error is calculated for the residual values ​​of all serial numbers corresponding to the patient, and finally the residual value between the patient's actual value and the fitted curve of the change in edema over time for the entire patient is obtained.

[0041] Similarly, to further investigate individual differences in the temporal progression patterns of edema volume in patients, we added information related to each patient's personal history, medical history, and onset. Using a deep learning algorithm, we performed clustering and clustered 100 patients into four subclasses (subclass 0, subclass 1, subclass 2, and subclass 3). As shown in Figure 4, we fitted the curves using a Gaussian Fitting Algorithm function.

[0042] As can be seen from Figures 4 to 7, the edema volume of patients in different subclasses shows a significant change within 1000 hours after the onset of symptoms, and the change in edema volume stabilizes after 1000 hours. By calculating the residuals between the actual values ​​of each subclass and their fitting curves and analyzing the data, it can be seen that the residuals between the actual data of each patient in each subclass obtained based on the deep learning clustering algorithm and the curve fitted with the Gaussian (Guess) function, as well as the residuals with the curve of the change in edema volume over time for the entire patient who was fitted, have decreased significantly. As can be seen from this, the subclass group obtained by the deep learning clustering algorithm, and the curve further fitted using the Gaussian (Guess) function, match the laws of change in the actual data, are closer to the actual situation, and have important implications for the analysis of changes in patients' conditions in the field of remedy in the future.

[0043] The above embodiments are merely for illustrating the technical solutions of the present invention and are not limiting. While the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and that any such modifications or substitutions should be included within the scope of the claims of the present invention.

Claims

1. A method for analyzing edema and planning treatment after hemorrhagic stroke, performed by a computer, Collect and pre-process edema data for each patient at different time points. The process involves fitting pre-processed data using a Gaussian fitting algorithm to construct a curve showing the temporal progression of edema volume across the entire patient, and calculating the residual between the curve and the actual values. Using a deep learning clustering algorithm, patients are divided into multiple subgroups, a Gaussian function is fitted to the edema levels of patients in each subgroup at different time points, and the residuals between the fitted curves for each subclass and the actual values ​​for each subclass are calculated. A method for analyzing edema and treatment planning after hemorrhagic stroke, characterized by comprising: calculating the change in edema volume, identifying the trend of change, combining it with a corresponding treatment plan, analyzing the degree of association between the treatment method and the trend of edema progression using a correlation rule algorithm, and identifying the treatment plan that has the highest degree of association with the trend of edema progression.

2. The method for analyzing edema and planning treatment after hemorrhagic stroke according to claim 1, characterized in that the preprocessing includes cleansing edema volume data at different time points for each patient, correcting errors and outliers, performing mean filling and linear interpolation on missing edema volume data, time synchronization of the edema volume data based on the patient's examination time and onset time, and standardization processing of the edema volume data.

3. Fitting the preprocessed data using the Gaussian fitting algorithm described above is The pre-processed edema volume data is sorted by patient and time point to form a structured dataset, where each row represents the edema volume at a specific time point for each patient. The formula for selecting the Gaussian function is as follows: However, v represents the time from the onset of symptoms to the examination, a, b, and c are parameters of the Gaussian function curve with respect to the patient's edema time after fitting, a represents the peak value of the curve, b represents the mean value of the curve, and c represents the width of the curve. For each patient's data, the parameters a, b, and c of the Gaussian model are estimated using the least squares method, and then iterative parameter optimization is performed using gradient descent. The method for analyzing edema after hemorrhagic stroke and planning treatment, as described in claim 2, further comprising applying a Gaussian function to the time-series data of each patient to generate a fitting curve for the temporal change in edema volume.

4. Calculating the residual between the aforementioned curve and the actual value is, The residual is defined as the difference between the predicted value of the Gaussian fitting curve and the actual observed value, and the calculation formula is as follows: However, R ｉ represents the fitting residual at time i, Y ｉ This represents the actual value of the edema volume at time i, This represents the predicted value of the Gaussian fitting curve at the same point in time, The method for analyzing edema after hemorrhagic stroke and planning treatment, as described in claim 3, further comprising calculating the residuals between the edema volume data for each patient at each time point and the corresponding points on a Gaussian fitting curve.

5. Dividing patients into multiple subgroups using the deep learning clustering algorithm described above is possible. For different patient personal information features, an autoencoder is used to extract features from patient data. In the encoding process, the original various feature data x are converted into low-dimensional latent variables h by the encoder, and the calculation formula is as follows: However, W １ and b １ However, these are the weights and biases of the encoder, σ is the activation function, and m is the feature value in the feature encoder, During the decoding process, the latent variable h, after undergoing the encoding process, is decoded back to its original data dimension, resulting in the reconstructed data. The result is obtained, and the calculation formula is as follows: However, W ２ The features represent the weights of the decoder, and b ２ This represents the bias of the feature decoder, The parameters of the autoencoder are iteratively optimized by the loss function, and the loss function is calculated based on the original data x and the reconstructed data. It is defined to minimize the difference between and , and its formula is as follows: However, G represents the original dataset of characteristic information for each patient, Ｒ This represents the original dataset of each patient's characteristic information reconstructed by the encoder, The low-dimensional representation h obtained by training the self-symbolizer is used for K-Means clustering. Based on the preset number k of clusters, the algorithm divides the patients into k clusters, and each cluster is composed of patients with similar characteristics, (C １ , C ２ … C ｋ ), and the minimized squared error E is represented as follows: however, Here, represents the norm of the vector, l represents the low-dimensional representation of patient information, and μ ｉ Cluster C ｉ The method for analyzing edema after hemorrhagic stroke and planning treatment according to claim 4, characterized in that it includes representing the mean vector of .

6. Calculating the residuals between the fitted curves for each subclass and the actual values ​​for each subclass is: The method for analyzing edema after hemorrhagic stroke and planning treatment according to claim 5, characterized in that it includes performing Gaussian fitting on patient data within each subgroup separated by a deep learning clustering algorithm, and calculating the residuals for each subgroup's Gaussian fitting curve with the actual edema amount data of each patient within the subgroup.

7. Analyzing the degree of association between treatment methods and the tendency for edema progression using the aforementioned correlation rule algorithm is possible. Extracting the change in edema volume and corresponding treatment plan information from the preprocessed dataset, We define item sets X and Y, where X represents a specific treatment plan and Y represents a specific trend in changes in edema volume. Support calculations are performed, and the support (X) of a data item set X is the ratio of the number of transactions containing X in D to the total number of transactions in D, and the formula is shown as follows: However, D represents the progression of edema, x represents the specific treatment plan selected, count(x) represents the number of specific treatment plans selected, and count(D) represents the number of cases in which the progression of edema was suppressed. The degree of support for the correlation rule X => Y is the item set It is equal to the degree of support, and the equation is shown as follows: However, X represents a specific treatment plan that has been selected, and Y represents another specific treatment plan that has been selected. This represents the number of two treatment plans selected, Calculate the confidence level of X => Y, and use the following formula: If a specific treatment plan is included, evaluate the rules for the possibility of another treatment plan emerging, however, The first option represents the degree of support where both treatment methods exist, and Support(x) represents the degree of support where both of the former treatment methods exist. The method for analyzing edema and planning treatment after hemorrhagic stroke according to claim 6, comprising: iteratively generating a set of frequently occurring items and corresponding correlation rules using the Apriori algorithm; and identifying the rule with the highest support and confidence level from the generated correlation rules.

8. An analysis system for edema analysis and treatment planning after hemorrhagic stroke using the method described in claim 1, It comprises a data preprocessing module, a Gaussian fitting module, a patient subgroup classification module, and a treatment planning-related analysis module. The aforementioned data preprocessing module preprocesses the edema volume data for each patient at different time points, The Gaussian fitting module uses a Gaussian fitting algorithm to fit the preprocessed data, constructs a curve showing the temporal progression of edema volume for the entire patient, and calculates the residual between the curve and the actual values. The patient subgroup classification module divides patients into multiple subgroups using a deep learning clustering algorithm, performs Gaussian function fitting on the edema levels of patients in each subgroup at different time points, calculates the residuals between the fitted curves for each subclass and the actual values ​​for each subclass, The aforementioned treatment plan-related analysis module is characterized by calculating the change in edema volume, identifying the trend of change, combining it with the corresponding treatment plan, analyzing the degree of correlation between the treatment method and the trend of edema progression using a correlation rule algorithm, and identifying the treatment plan that has the highest degree of correlation with the trend of edema progression.

9. A computer device, Including memory and processor, A computer program is stored in the aforementioned memory. A computer device characterized in that when the processor executes the computer program, the steps of the method according to claim 1 are realized.

10. A computer-readable storage medium, A computer program is stored in the aforementioned computer-readable storage medium. A computer-readable storage medium characterized in that, when the computer program is executed by a processor, the steps of the method according to claim 1 are realized.

Images