Intensive care oriented sepsis kidney injury risk dynamic assessment method and system

Abstract

The invention discloses an intensive care oriented sepsis kidney injury risk dynamic assessment method and system. Critical sepsis kidney injury features are formed by combining brain electrophysiological features, electroencephalogram deviation correction physiological parameters and monitoring condition vector features, and kidney injury risk data evaluated by medical staff and the preferable severe sepsis kidney injury features are collected. Constructing a severe sepsis kidney injury feature generation model according to the severe sepsis kidney injury features and the preferable severe sepsis kidney injury features, generating a preferable severe sepsis kidney injury feature set, and constructing a kidney injury risk dynamic evaluation model according to the preferable severe sepsis kidney injury feature set and kidney injury risk data. According to the method, the physiological parameters are corrected on the basis of the electroencephalogram features, and the optimal model is constructed by fusing the related parameters of the subject and the object, so that the feature data of high-risk assessment fitting degree is obtained, the constructed scene adaptability injury assessment model is higher in robustness, assessment is more objective, and the dynamic assessment precision of the kidney injury is improved.

Description

Technical Field

[0001] This invention belongs to the field of biomedical engineering technology, and in particular relates to a method and system for dynamic assessment of the risk of sepsis-induced kidney injury in intensive care. Background Technology

[0002] Sepsis-associated acute kidney injury (SA-AKI) is a common complication in intensive care unit (ICU) patients, and early diagnosis and intervention are crucial for improving prognosis. Current clinical practice mainly relies on single indicators such as serum creatinine (Scr) and urine output. However, serum creatinine concentration only significantly increases 48-72 hours after renal parenchymal injury occurs, leading to an early missed detection rate exceeding 40%. Furthermore, urine output is typically only measured hourly, failing to promptly reflect the patient's current condition and risk of kidney damage.

[0003] Existing technologies based on single-modal data (such as ultrasound images or biochemical indicators) have significant limitations: 1. Insufficient data richness: While traditional ultrasound imaging can assess renal blood perfusion, it cannot capture molecular-level damage markers (such as early elevations in urinary NGAL and KI M-1); models relying solely on biochemical indicators are limited by detection lag and cannot provide timely warnings of kidney damage. 2. Lack of organ interaction mechanisms: Brain marrow and kidney essence ascend to the brain through marrow transformation, influencing a person's spirit, consciousness, and cognitive activities. Current medical theories suggest a close connection between the brain and kidneys, and the control relationship between physiological parameters between the brain and kidneys has not been effectively quantified and modeled. Existing methods often assume independent organ function, neglecting key pathological mechanisms such as brain-renal syndrome. 3. Poor generalization with small samples: Supervised learning-based prediction models heavily rely on labeled data, but labeling data from ICU patients is costly and inconsistent, resulting in insufficient generalization ability of the model in real-world scenarios. 4. Static decision-making rigidity: Fixed thresholds recommended by clinical guidelines (such as MAP≥65mmHg) cannot adapt to changes in patients' real-time risk, and existing decision-making systems lack dynamic adjustment capabilities, which can easily lead to undertreatment or over-intervention.

[0004] Although artificial intelligence technology has been gradually applied to the medical field in recent years, the aforementioned problems have not yet been systematically solved. How to utilize cross-modal data modeling and leverage the dynamic interaction between the brain and kidney organs to achieve personalized dynamic kidney injury risk assessment has become a key technical bottleneck in improving the diagnosis and treatment of SA-AK I.

[0005] Therefore, the key questions are: how to mine electroencephalographic features to correct the collected physiological parameters, making the physiological parameters of sepsis patients real-time, and how to construct a generative model of features severely correlated with sepsis-related kidney damage, thereby increasing the number of dimensions of features severely correlated with sepsis-related kidney damage? Furthermore, how to utilize existing basic machine learning models to construct a dynamic assessment model for kidney damage risk, and how to construct a dynamic assessment model for kidney damage risk based on kidney injury risk data annotated by medical staff over a limited time period? The questions also include: how to obtain sepsis-related physiological parameters that more closely reflect real-time changes in patients, how to make the trained model more objective and realistic, improve model robustness, enhance the model's generalization ability in kidney injury risk assessment scenarios, and make kidney injury risk assessment more accurate so that medical staff can make medical plans in advance, provide early warning of acute kidney injury, and improve the accuracy of the model's dynamic assessment. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention proposes a method and system for dynamic assessment of the risk of septic kidney injury in intensive care.

[0007] In a first aspect of the invention, a method for dynamic assessment of the risk of septic kidney injury in intensive care is provided, the method comprising:

[0008] The brain electrophysiological characteristics were collected and processed, and the collected sepsis physiological parameters were processed based on the brain electrophysiological characteristics to obtain the brain electrophysiological deviation correction physiological parameters. The monitoring condition vector characteristics in the intensive care unit were obtained, and the kidney injury risk data assessed by medical staff were obtained.

[0009] The brain electrophysiological characteristics, the brain electrophysiological deviation correction physiological parameters, and the monitoring condition vector characteristics are combined to form the characteristics of severe septic kidney injury, and the preferred characteristics of severe septic kidney injury assessed by medical staff are obtained.

[0010] Based on the aforementioned severe septic kidney injury characteristics and the preferred severe septic kidney injury characteristics, a severe septic kidney injury characteristic generation model is constructed, and a preferred severe septic kidney injury characteristic set is generated;

[0011] Based on the preferred feature set of severe septic kidney injury and the corresponding kidney injury risk data, a dynamic assessment model for kidney injury risk is constructed, and real-time kidney injury risk data is generated using the dynamic assessment model for kidney injury risk.

[0012] Furthermore, the aforementioned brain electrophysiological characteristics are acquired and calculated using an EEG acquisition device.

[0013] Furthermore, the physiological parameters of sepsis include user urine output, neutrophil gelatinase-associated lipotransferase content, blood pressure, and D-dimer content.

[0014] Furthermore, the brain electrophysiological deviation correction physiological parameters obtained by processing the sepsis physiological parameters based on the brain electrophysiological characteristics are calculated by dimensionless feature vector fusion of the user's urine volume, the content of neutrophil gelatinase-associated lipid transport protein, the blood pressure, and the D-dimer content.

[0015] Furthermore, the monitoring condition vector features include the air temperature, oxygen concentration, carbon dioxide concentration, and humidity in the monitoring room;

[0016] The characteristics of severe septic kidney injury are obtained by vertically combining and plucking zeros from the brain electrophysiological characteristics, the brain electroencephalogram deviation correction physiological parameters, and the monitoring condition vector characteristics.

[0017] Furthermore, the preferred characteristics of severe septic kidney injury assessed by medical staff are obtained by using feature evaluation coefficients.

[0018] Furthermore, a severe septic kidney injury feature generation model is constructed based on the features of the severe septic kidney injury and the preferred features of the severe septic kidney injury.

[0019] Furthermore, the dynamic assessment model for kidney injury risk is constructed using a Fisher criterion-based classifier improved based on monitoring condition vector features.

[0020] It also provides a dynamic risk assessment system for septic kidney injury in intensive care, which implements the aforementioned dynamic risk assessment method for septic kidney injury in intensive care, including a module for acquiring electroencephalographic features, a module for correcting physiological parameters, a module for acquiring monitoring condition vector features, a module for constructing a feature generation model for severe septic kidney injury, and a module for constructing a dynamic risk assessment model for kidney injury.

[0021] The brain electrophysiological feature acquisition module is used to collect and process brain electrophysiological features.

[0022] The physiological parameter correction module is connected to the electroencephalogram (EEG) physiological feature acquisition module, acquires the EEG physiological features, and processes the acquired sepsis physiological parameters based on the EEG physiological features to obtain EEG deviation correction physiological parameters.

[0023] The monitoring condition vector feature acquisition module acquires the monitoring condition vector features in the intensive care unit and acquires the kidney injury risk data assessed by medical staff.

[0024] The severe septic kidney injury feature generation model construction module: uses the brain electrophysiological features, the brain electrophysiological deviation correction physiological parameters and the monitoring condition vector features to form severe septic kidney injury features, and obtains the preferred severe septic kidney injury features assessed by medical staff;

[0025] Based on the aforementioned severe septic kidney injury characteristics and the preferred severe septic kidney injury characteristics, a severe septic kidney injury characteristic generation model is constructed, and a preferred severe septic kidney injury characteristic set is generated;

[0026] The kidney injury risk dynamic assessment model construction module: constructs a kidney injury risk dynamic assessment model based on the preferred severe septic kidney injury feature set and the corresponding kidney injury risk data, and uses the kidney injury risk dynamic assessment model to generate real-time kidney injury risk data.

[0027] Furthermore, the aforementioned brain electrophysiological characteristics are acquired and calculated using an EEG acquisition device.

[0028] Therefore, the beneficial effects of this invention are: utilizing real-time EEG data within a specific acquisition frequency during brain EEG parameter acquisition to analyze and correct the acquired sepsis physiological parameters, thus ensuring the real-time nature of the sepsis patient's physiological parameters; constructing a model to generate optimized features of severe septic kidney injury, employing a unique feature evaluation coefficient to acquire optimized features of severe septic kidney injury, thereby increasing the number of dimensions of optimized features of severe septic kidney injury; and using an improved version of an existing basic machine learning model to obtain a dynamic assessment model of kidney damage risk, based on kidney damage risk data annotated by medical staff over a limited time period to construct the dynamic assessment model of kidney damage risk. The sepsis-related physiological parameters acquired by this invention are closer to real-time changes, resulting in a more objective and realistic dynamic assessment model of kidney damage risk, improved model robustness, enhanced generalization ability in kidney damage risk assessment scenarios, and more accurate kidney damage risk assessment. This allows medical staff to make timely medical decisions, provide early warnings of acute kidney injury, and improve the accuracy of the model's dynamic assessment.

[0029] Further embodiments and improvements of the present invention will be described in conjunction with the accompanying drawings and specific examples. Attached Figure Description

[0030] Figure 1 This is a flowchart of the dynamic risk assessment method for septic kidney injury in intensive care according to the present invention;

[0031] Figure 2 This is a schematic diagram of the dynamic risk assessment system for septic kidney injury in intensive care according to the present invention;

[0032] Figure 3 This is an example diagram illustrating the principle of the feature generation model for severe septic kidney injury used in the embodiments of this invention;

[0033] Figure 4 This is a schematic diagram of the EEG channels collected in the embodiments of the present invention;

[0034] Figure 5This is a schematic diagram of an electronic device structure for implementing the method of the present invention in an embodiment of the present invention. Detailed Implementation

[0035] The invention will now be further described in conjunction with the accompanying drawings and specific embodiments. The neural network model used in this invention is a GAN model, while the classifier model based on the Fischer criterion is an improved model based on the findings of this application.

[0036] like Figure 2 As shown, the system of the present invention belongs to health-related information systems and cloud platforms, and therefore belongs to the biomedical engineering industry.

[0037] In a first aspect of the invention, a method for dynamic assessment of the risk of septic kidney injury in intensive care is provided, the method comprising:

[0038] The brain electrophysiological characteristics were collected and processed, and the collected sepsis physiological parameters were processed based on the brain electrophysiological characteristics to obtain the brain electrophysiological deviation correction physiological parameters. The monitoring condition vector characteristics in the intensive care unit were obtained, and the kidney injury risk data assessed by medical staff were obtained.

[0039] The brain electrophysiological characteristics, the brain electrophysiological deviation correction physiological parameters, and the monitoring condition vector characteristics are combined to form the characteristics of severe septic kidney injury, and the preferred characteristics of severe septic kidney injury assessed by medical staff are obtained.

[0040] Based on the aforementioned severe septic kidney injury characteristics and the preferred severe septic kidney injury characteristics, a severe septic kidney injury characteristic generation model is constructed, and a preferred severe septic kidney injury characteristic set is generated;

[0041] Based on the preferred feature set of severe septic kidney injury and the corresponding kidney injury risk data, a dynamic assessment model for kidney injury risk is constructed, and real-time kidney injury risk data is generated using the dynamic assessment model for kidney injury risk.

[0042] Furthermore, the aforementioned brain electrophysiological characteristics are acquired and calculated using an EEG acquisition device, and the calculation formula is as follows:

[0043]

[0044] In the formula, B KhThe electrophysiological characteristics of the brain are represented by the dimensionless voltage amplitude. Due to varying degrees of contact between the wearable EEG acquisition device and the user's scalp, the amplitude of the acquired EEG voltage fluctuates. Therefore, a contact coefficient g is used to adjust the amplitude of the acquired EEG channel voltage. This application finds that the value of g ranges from 0.995 to 1.005, minimizing the impact on the acquisition device. Since the EEG acquisition frequency h is generally between 250Hz and 1000Hz, i.e., 250 data points are acquired per second, or approximately once every 1 / 250th of a second, to ensure the real-time correction of physiological parameters in the electrophysiological characteristics of the brain in this application, the electrophysiological characteristics are represented by the average amplitude of the EEG channel of a preferred acquisition frequency. h is the sampling frequency of the wireless EEG acquisition device, n is the preferred EEG channel, and V... ij This represents the EEG voltage amplitude of the i-th EEG channel acquired in the j-th time per second. Existing relevant medical theories reveal a connection between brain marrow and kidney essence. Therefore, the preferred EEG channels are the brainstem region and the EEG channel region near the cerebral cortex. In this embodiment, the preferred EEG voltage amplitudes are those acquired from nine EEG channels: Cz, C1, C2, FCz, FC2, FC1, O1, Oz, and O2. The EEG channel numbers are 1 to 9, corresponding to the value of n.

[0045] This embodiment is not limited to these nine EEG channels, but also includes others such as... Figure 4 The EEG channels shown are used to improve the classification accuracy, prediction accuracy, and objectivity of subsequent model input features.

[0046] In previous background technologies, there were methods to determine the degree of sepsis using electroencephalography (EEG). However, these methods relied on the time-frequency characteristics of EEG to assess the severity of sepsis without considering its impact on septic kidney damage. In the current medical context, human physiological signals are transmitted through the brain, the command organ. Therefore, the current state of physiological parameters often does not match the actual situation. This application utilizes an EEG deviation coefficient generated from brain electrophysiological characteristics to correct the user's physiological parameters, making them more consistent with the user's actual situation and providing accurate relevant features for subsequent real-time dynamic assessment of kidney damage.

[0047] Furthermore, the physiological parameters of sepsis include user urine output, neutrophil gelatinase-associated lipotransferase content, blood pressure, and D-dimer content.

[0048] This example also includes serum creatinine, and the addition of serum creatinine is consistent with the subsequent processing of feature vectors.

[0049] Furthermore, the brain electrophysiological deviation correction physiological parameters obtained by processing the sepsis physiological parameters based on the brain electrophysiological characteristics are calculated by dimensionless feature vector fusion of the user's urine volume, the content of neutrophil gelatinase-associated lipid transport protein, the blood pressure, and the D-dimer content, and are obtained by correcting the brain electrophysiological characteristics collected at a certain sampling frequency when collecting the sepsis physiological parameters.

[0050]

[0051] In the formula, S P To correct physiological parameters for the corrected EEG bias, B Khl B is a brain electrophysiological characteristic value at a sampling frequency h when collecting the physiological parameters of sepsis. Khf To collect the brain electrophysiological characteristics of the physiological markers of sepsis at a sampling frequency h after ex vivo, |10 4 *(B Khl -B Khf The value is the absolute value after removing dimensions. Since the amplitude of EEG voltage is generally in microvolts, 10 is used. 4 Perform deviation coefficient correction processing, N G The content of the neutrophil gelatinase-associated lipid transport protein, D d The D-dimer content, U o B represents the user's urine output. p For the blood pressure mentioned above, those skilled in the art use a logarithmic function to remove data discrepancies using dimensionality correction. In this calculation formula, This represents the dimensionless absolute value of a logarithmic function.

[0052] In this embodiment, B Khl One numerical value is represented as 8.65 * 10. -6 B Khf It is 9.53*10 -6 If the blood pressure obtained from the test analysis is 65 mmHg, then B p If the value is 65 and the D-dimer content is 5.2 μg / mL, then D... d If the value is 5.2, and other physiological parameters are taken in the same manner, then S... P Finally, a definite value is obtained.

[0053] Because real-time EEG data is collected when monitoring users' sepsis physiological parameters, and the specific parameters are obtained after a certain period of testing, there are some post-in vitro parameter changes. However, the brain, as the command-issuing organ, can effectively correct for these post-in vitro parameter changes due to its changes over a certain period. Blood pressure, being monitored continuously in real-time, is not corrected for. Since oliguria is an early marker of kidney damage, and persistent oliguria indicates a risk of renal tubular necrosis, lower urine output indicates a greater risk of kidney damage. Blood pressure below 65 mmHg can lead to insufficient renal perfusion. Higher D-dimer levels indicate a greater risk of kidney damage, and neutrophil gelatinase-associated lipotransferase levels are also positively correlated with the risk of kidney damage.

[0054] Furthermore, the monitoring condition vector features include the air temperature, oxygen concentration, carbon dioxide concentration, and humidity in the monitoring room;

[0055] The characteristics of severe septic kidney injury are obtained by vertically combining and plucking zeros from the brain electrophysiological characteristics, the brain electroencephalogram deviation correction physiological parameters, and the monitoring condition vector characteristics.

[0056] Since the dimensions of each feature vector are different when the feature combination is formed, the vertical vector concatenation generally adopts vector supplementation to obtain a multi-dimensional concatenated vector, which facilitates the subsequent model processing of related feature vectors.

[0057] In one embodiment of the present invention, the monitoring condition vector feature includes the air temperature W in the monitoring room. P Oxygen concentration O P Carbon dioxide concentration (CO) P and humidity H P The above vector features have all been dimensionless.

[0058] The feature vector representing the characteristics of severe septic kidney injury is as follows:

[0059]

[0060] Using vector padding to obtain multi-dimensional concatenated vectors facilitates subsequent model processing of relevant feature vectors. This application found that padding with zero vectors makes the model training generation more in line with the needs and the model has high robustness. However, when padding with one vector features, the model classification results show obvious confusion and high generalization characteristics. Therefore, zero padding is used to supplement vector values ​​to facilitate the corresponding processing of feature vectors in subsequent model training.

[0061] Furthermore, the preferred characteristics of severe septic kidney injury assessed by medical staff are obtained using feature evaluation coefficients, whereby the feature evaluation coefficients are:

[0062]

[0063] In the formula, P s B is the characteristic evaluation coefficient. KhM The maximum value of brain electrophysiological characteristics for training on the characteristics of severe septic kidney injury, B KhN Minimum values ​​of brain electrophysiological characteristics in training sessions on the characteristics of severe septic kidney injury, S PM The maximum value of the physiological parameter for correcting EEG deviation in the training set, S PN The minimum value of physiological parameters for correcting EEG deviation in the training set.

[0064] In this embodiment, the feature vector values ​​from the characteristics of severe septic kidney injury are used to select the optimal features for severe septic kidney injury. Since the electroencephalographic features and EEG deviation correction physiological parameters have the greatest impact on the subsequent kidney injury assessment of the model, based on the feature evaluation coefficient, the selection of features not exceeding P is used. s The characteristics of severe septic kidney injury are preferred.

[0065] In one embodiment of this application, if P s If the value is 0.445, then the severe septic kidney injury features with a value not exceeding 0.445 are selected as the preferred severe septic kidney injury features and form the preferred severe septic kidney injury feature set.

[0066] Furthermore, the severe septic kidney injury feature generation model is constructed based on the aforementioned severe septic kidney injury features and the preferred severe septic kidney injury features. The activation function adopted by the severe septic kidney injury feature generation model is:

[0067]

[0068] In the formula, f(X) is the activation function value, and X is the feature value of the input severe septic kidney injury features linearly transformed by weights and biases.

[0069] A neural network consists of an input layer, hidden layers, and an output layer. The input layer is responsible for receiving external information, such as images, sounds, or text; the hidden layers are the key to processing and transforming this information; and the output layer provides the final prediction or decision based on the processing results of the hidden layers.

[0070] In a neural network, each neuron is connected to neurons in adjacent layers, and these connections are represented by weights. When information is passed from the input layer to the hidden layer, each neuron performs calculations based on the received information and weights, and then passes the result to the next neuron. This process is like a relay race of information, with each layer processing and refining the information until the final output layer provides the answer.

[0071] Backpropagation is the core of neural network training. During training, the network adjusts the weights of neurons in each layer based on the error between the actual output and the expected output using the backpropagation algorithm. This adjustment process is like a trial-and-error process; the network continuously tries different weight combinations until it finds the set that minimizes the error.

[0072] The core task of generative artificial intelligence is to generate new data that conforms to certain rules or patterns. Neural networks, with their powerful learning and expressive capabilities, play a crucial role in GAI.

[0073] Generative Adversarial Networks (GANs) are a typical example of a neural network-based GAN model. A GAN consists of two networks: a generator and a discriminator, which compete against and learn from each other. The generator is responsible for generating fake data, while the discriminator is responsible for judging the authenticity of the data. During training, the generator continuously attempts to generate more realistic images to deceive the discriminator, while the discriminator continuously improves its judgment to identify the generator's deceptions. This adversarial learning process allows the generator to gradually learn to generate high-quality data. The generator itself is constructed from a neural network, which requires the use of activation functions. The specific steps involved are readily available to those skilled in the art and will not be elaborated upon here.

[0074] Furthermore, the dynamic assessment model for kidney injury risk is constructed using a Fisher criterion-based classifier improved based on monitoring condition vector features:

[0075]

[0076] In the formula, F(A) represents the output kidney injury risk data, and W... T Let A be the normal vector perpendicular to the hyperplane, and W be the input preferred feature set of severe septic kidney injury. PA For the real-time temperature feature of the monitoring condition vector feature in the feature set of the preferred severe septic kidney injury feature, W PS For the temperature characteristics of the vector features of optimal monitoring conditions, O PA For the real-time oxygen concentration feature of the monitoring condition vector feature in the feature set of the preferred severe septic kidney injury feature, O PS Oxygen concentration characteristics as vector features of optimal monitoring conditions

[0077] In this embodiment, since the monitoring features have a certain impact on the classification of damage risk data, this application found that using the real-time status of the monitoring features and the optimal status during monitoring to correct the temperature and oxygen content makes the classification effect more accurate.

[0078] In this embodiment, the kidney injury risk data is output based on the value of F(A). Generally, according to those skilled in the art, based on the value of F(A) and the corresponding data of the actual training model, the data is divided into the following categories: F(A) less than -1, the output kidney injury risk data is low; F(A) less than 0 and greater than or equal to -1, the output kidney injury risk data is medium; F(A) equal to 0, the output kidney injury risk data is no risk of kidney injury; and F(A) greater than 0, the output kidney injury risk data is high.

[0079] It also provides a dynamic risk assessment system for septic kidney injury in intensive care, which implements the aforementioned dynamic risk assessment method for septic kidney injury in intensive care, including a module for acquiring electroencephalographic features, a module for correcting physiological parameters, a module for acquiring monitoring condition vector features, a module for constructing a feature generation model for severe septic kidney injury, and a module for constructing a dynamic risk assessment model for kidney injury.

[0080] The brain electrophysiological feature acquisition module is used to collect and process brain electrophysiological features.

[0081] The physiological parameter correction module is connected to the electroencephalogram (EEG) physiological feature acquisition module, acquires the EEG physiological features, and processes the acquired sepsis physiological parameters based on the EEG physiological features to obtain EEG deviation correction physiological parameters.

[0082] The monitoring condition vector feature acquisition module acquires the monitoring condition vector features in the intensive care unit and acquires the kidney injury risk data assessed by medical staff.

[0083] The severe septic kidney injury feature generation model construction module: uses the brain electrophysiological features, the brain electrophysiological deviation correction physiological parameters and the monitoring condition vector features to form severe septic kidney injury features, and obtains the preferred severe septic kidney injury features assessed by medical staff;

[0084] Based on the aforementioned severe septic kidney injury characteristics and the preferred severe septic kidney injury characteristics, a severe septic kidney injury characteristic generation model is constructed, and a preferred severe septic kidney injury characteristic set is generated;

[0085] The kidney injury risk dynamic assessment model construction module: constructs a kidney injury risk dynamic assessment model based on the preferred severe septic kidney injury feature set and the corresponding kidney injury risk data, and uses the kidney injury risk dynamic assessment model to generate real-time kidney injury risk data.

[0086] Furthermore, the aforementioned brain electrophysiological characteristics are acquired and calculated using an EEG acquisition device, and the calculation formula is as follows:

[0087]

[0088] In the formula, B Kh The electrophysiological characteristics of the brain are represented by the dimensionless voltage amplitude. Due to varying degrees of contact between the wearable EEG acquisition device and the user's scalp, the amplitude of the acquired EEG voltage fluctuates. Therefore, a contact coefficient g is used to adjust the amplitude of the acquired EEG channel voltage. This application finds that the value of g ranges from 0.995 to 1.005, minimizing the impact on the acquisition device. Since the EEG acquisition frequency h is generally between 250Hz and 1000Hz, i.e., 250 data points are acquired per second, or approximately once every 1 / 250th of a second, to ensure the real-time correction of physiological parameters in the electrophysiological characteristics of the brain in this application, the electrophysiological characteristics are represented by the average amplitude of the EEG channel of a preferred acquisition frequency. h is the sampling frequency of the wireless EEG acquisition device, n is the preferred EEG channel, and V... ij The amplitude of the brain voltage in the i-th brain channel is represented by the j-th acquisition in one second. Existing relevant medical theories reveal that there is a connection between brain marrow and kidney essence. Therefore, the preferred brain channels are the brainstem region and the brain channel region near the cerebral cortex. In this embodiment, the preferred brain voltage amplitudes are the brain voltage amplitudes acquired by nine brain channels: Cz, C1, C2, FCz, FC2, FC1, O1, Oz, and O2.

[0089] In previous background technologies, there were methods to determine the degree of sepsis using electroencephalography (EEG). However, these methods relied on the time-frequency characteristics of EEG to assess the severity of sepsis without considering its impact on septic kidney damage. In the current medical context, human physiological signals are transmitted through the brain, the command organ. Therefore, the current state of physiological parameters often does not match the actual situation. This application utilizes an EEG deviation coefficient generated from brain electrophysiological characteristics to correct the user's physiological parameters, making them more consistent with the user's actual situation and providing accurate relevant features for subsequent real-time dynamic assessment of kidney damage.

[0090] The beneficial effects of this invention are as follows: It utilizes real-time EEG data within a specific acquisition frequency during brain electroencephalography (EEG) parameter collection to analyze and correct the acquired physiological parameters of sepsis patients, ensuring the real-time nature of these parameters. It constructs a model for generating optimized features of severe septic kidney injury (SKI), employing a unique feature evaluation coefficient to acquire these optimized features, thereby increasing the number of dimensions for optimized SKI features. Furthermore, it utilizes existing basic machine learning models to improve upon the obtained dynamic assessment model for kidney damage risk, constructing this model based on a limited amount of time-labeled kidney damage risk data collected by medical personnel. The physiological parameters related to sepsis obtained by this invention more closely reflect real-time changes, resulting in a more objective and realistic dynamic assessment model for kidney damage risk. This improves model robustness, enhances the model's generalization ability in kidney damage risk assessment scenarios, and makes kidney damage risk assessment more accurate. This allows medical personnel to make timely medical decisions, provide early warnings of acute kidney injury, and improve the accuracy of the model's dynamic assessment.

[0091] Of course, it is understood that each embodiment of the present invention can achieve one of the effects individually, and the combination of multiple embodiments of the present invention can achieve all the above effects. However, it is not required that each embodiment of the present invention achieve all the above advantages and effects, because each embodiment of the present invention can constitute a separate technical solution and make one or more contributions to the prior art. Each embodiment of the present invention is not affected, and the solution can still be implemented even if one embodiment is deleted.

[0092] For any module structures not specifically defined in this invention, the existing technical descriptions shall prevail. The prior art mentioned in the foregoing background and specific embodiments sections can be considered part of this invention and used to understand the meaning of certain technical features or parameters.

Claims

1. A method for dynamic assessment of septic kidney injury risk in intensive care, characterized in that, The method includes: The brain electrophysiological characteristics were collected and processed, and the collected sepsis physiological parameters were processed based on the brain electrophysiological characteristics to obtain the brain electrophysiological deviation correction physiological parameters. The monitoring condition vector characteristics in the intensive care unit were obtained, and the kidney injury risk data assessed by medical staff were obtained. The brain electrophysiological characteristics, the brain electrophysiological deviation correction physiological parameters, and the monitoring condition vector characteristics are combined to form the characteristics of severe septic kidney injury, and the preferred characteristics of severe septic kidney injury assessed by medical staff are obtained. Based on the aforementioned severe septic kidney injury characteristics and the preferred severe septic kidney injury characteristics, a severe septic kidney injury characteristic generation model is constructed, and a preferred severe septic kidney injury characteristic set is generated; Based on the preferred feature set of severe septic kidney injury and the corresponding kidney injury risk data, a dynamic assessment model for kidney injury risk is constructed, and real-time kidney injury risk data is generated using the dynamic assessment model for kidney injury risk.

2. The method for dynamic assessment of septic kidney injury risk in intensive care as described in claim 1, characterized in that: The brain electrophysiological characteristics are obtained through acquisition and calculation using electroencephalogram (EEG) acquisition equipment.

3. The method for dynamic assessment of septic kidney injury risk in intensive care as described in claim 1 or 2, characterized in that: The physiological parameters of sepsis include user urine output, neutrophil gelatinase-associated lipotransferase content, blood pressure, and D-dimer content.

4. The method for dynamic assessment of septic kidney injury risk in intensive care as described in claim 3, characterized in that: The electroencephalogram (EEG) deviation correction physiological parameters obtained by processing the sepsis physiological parameters based on the brain's electrophysiological characteristics are calculated by dimensionless feature vector fusion using the user's urine volume, neutrophil gelatinase-associated lipid transport protein content, blood pressure, and D-dimer content.

5. The method for dynamic assessment of septic kidney injury risk in intensive care as described in claim 4, characterized in that: The monitoring condition vector features include the air temperature, oxygen concentration, carbon dioxide concentration, and humidity in the monitoring room; The characteristics of severe septic kidney injury are obtained by vertically combining and plucking zeros from the brain electrophysiological characteristics, the brain electroencephalogram deviation correction physiological parameters, and the monitoring condition vector characteristics.

6. The method for dynamic assessment of septic kidney injury risk in intensive care as described in claim 1 or 2, characterized in that: The preferred features of severe septic kidney injury assessed by medical staff are obtained by using feature evaluation coefficients.

7. The method for dynamic assessment of septic kidney injury risk in intensive care as described in claim 6, characterized in that: The severe septic kidney injury feature generation model is constructed based on the features of the severe septic kidney injury and the preferred features of the severe septic kidney injury.

8. The method for dynamic assessment of septic kidney injury risk in intensive care as described in claim 7, characterized in that: The dynamic assessment model for kidney injury risk is constructed using a Fisher criterion-based classifier improved from monitoring condition vector features.

9. A dynamic risk assessment system for septic kidney injury in intensive care, the system implementing the method described in any one of claims 1-8, comprising an electroencephalogram (EEG) feature acquisition module, a physiological parameter correction module, a monitoring condition vector feature acquisition module, a severe septic kidney injury feature generation model construction module, and a kidney injury risk dynamic assessment model construction module, characterized in that: The brain electrophysiological feature acquisition module is used to collect and process brain electrophysiological features. The physiological parameter correction module is connected to the electroencephalogram (EEG) physiological feature acquisition module, acquires the EEG physiological features, and processes the acquired sepsis physiological parameters based on the EEG physiological features to obtain EEG deviation correction physiological parameters. The monitoring condition vector feature acquisition module acquires the monitoring condition vector features in the intensive care unit and acquires the kidney injury risk data assessed by medical staff. The severe septic kidney injury feature generation model construction module: uses the brain electrophysiological features, the brain electrophysiological deviation correction physiological parameters and the monitoring condition vector features to form severe septic kidney injury features, and obtains the preferred severe septic kidney injury features assessed by medical staff; Based on the aforementioned severe septic kidney injury characteristics and the preferred severe septic kidney injury characteristics, a severe septic kidney injury characteristic generation model is constructed, and a preferred severe septic kidney injury characteristic set is generated; The kidney injury risk dynamic assessment model construction module: constructs a kidney injury risk dynamic assessment model based on the preferred severe septic kidney injury feature set and the corresponding kidney injury risk data, and uses the kidney injury risk dynamic assessment model to generate real-time kidney injury risk data.

10. The dynamic risk assessment system for septic kidney injury in intensive care as described in claim 9, characterized in that: The brain electrophysiological characteristics are obtained through acquisition and calculation using electroencephalogram (EEG) acquisition equipment.

Images