Online monitoring system and method for patient hemodialysis process

Abstract

The present invention discloses an online monitoring system and method for a patient's hemodialysis process, which relates to the technical field of hemodialysis monitoring, including: obtaining multimodal monitoring data and preprocessing; using a deep learning model to fuse the preprocessed multimodal monitoring data into a comprehensive feature vector; generating a blood state evaluation matrix and a behavior state evaluation coefficient based on the fused comprehensive feature vector; constructing an infection risk prediction model based on the state value of blood components and the behavior state evaluation coefficient; optimizing the infection risk prediction model through adaptive learning; and generating a hemodialysis online monitoring report based on the infection risk prediction result. The present invention constructs an online hemodialysis monitoring model by fusing multimodal monitoring data and applying deep learning, effectively monitoring the infection risk of patients during dialysis, improving the prediction accuracy through comprehensive evaluation, helping medical staff to prevent infection in a timely manner, and improving the safety of patient treatment and the quality of life.

Description

Technical Field

[0001] The present invention relates to the technical field of hemodialysis monitoring, and in particular to an online monitoring system and method for a patient's hemodialysis process. Background Art

[0002] With the advancement of medical engineering and the continuous innovation of technology, the performance of hemodialysis equipment has been significantly improved, which not only improves the dialysis efficiency, but also improves the comfort of patients. Traditional hemodialysis monitoring mainly relies on manual observation and recording by medical staff. This method is limited by the workload and subjective judgment ability of medical staff, and it is difficult to achieve real-time and comprehensive monitoring of the patient's physiological state. With the development of sensing technology and artificial intelligence, multimodal data acquisition and analysis have become possible, which provides a new solution for online monitoring of the hemodialysis process.

[0003] Although current technological advances have greatly improved the monitoring capabilities during hemodialysis, there are still some challenges and shortcomings. First, most existing monitoring systems only focus on data from a single modality and lack the effective integration and utilization of multimodal data, which limits the system's understanding of the patient's comprehensive physiological state. Second, existing monitoring methods often focus on post-analysis rather than real-time warning and fail to achieve true online monitoring. Furthermore, the machine learning models used in existing technologies may not be able to fully utilize the characteristics of time series data. Finally, for anomaly detection and risk assessment, existing methods often rely on fixed thresholds. Such static thresholds may not accurately reflect individual differences and changes in disease progression, thereby affecting the accuracy and reliability of the monitoring system. Summary of the invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides an online monitoring system and method for a patient's hemodialysis process, which solves the problems of insufficient multimodal data fusion, lack of real-time warning, limited time series analysis capability and fixed abnormal detection threshold.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides an online monitoring method for a patient's hemodialysis process, which includes: acquiring multimodal monitoring data and performing preprocessing; using a deep learning model to fuse the preprocessed multimodal monitoring data into a comprehensive feature vector; generating a blood state evaluation matrix and a behavior state evaluation coefficient based on the fused comprehensive feature vector; constructing an infection risk prediction model based on the state values ​​of blood components and the behavior state evaluation coefficients; optimizing the infection risk prediction model through adaptive learning; and generating a hemodialysis online monitoring report based on the infection risk prediction results.

[0008] As a preferred solution of the method for online monitoring of the patient's hemodialysis process of the present invention, the step of obtaining multimodal monitoring data and preprocessing the data includes:

[0009] Regularly collect blood samples and analyze them through blood testing equipment to obtain blood test data;

[0010] By continuously monitoring physiological signals using wearable devices;

[0011] Obtain video data of patients through high-definition cameras in the dialysis room;

[0012] The blood test data were cleaned using mean interpolation method;

[0013] Butterworth filter is used to filter the noise of physiological signals;

[0014] Apply background subtraction algorithms to remove background noise from video data.

[0015] As a preferred solution of the method for online monitoring of the patient's hemodialysis process of the present invention, the method uses a deep learning model to fuse the pre-processed multimodal monitoring data into a comprehensive feature vector, including:

[0016] Arranging the preprocessed blood test data and physiological signal data into sequences in chronological order;

[0017] The sequence data is input into the LSTM model, and the feature vectors of the blood test data and the physiological signal data are extracted from the last layer of the LSTM model;

[0018] The preprocessed video data is input into the CNN model, and the model extracts the feature vector of the video frame through the forward propagation process;

[0019] The extracted feature vectors are fused through the LSTM model, and the expression is: ; Among them, f b The feature vector representing the blood test data, f p is the feature vector of physiological signal data, f v is the feature vector of the video data, F is the integrated feature vector after fusion, and ct represents the concatenation operation of the feature vectors.

[0020] As a preferred solution of the method for online monitoring of the patient's hemodialysis process of the present invention, the blood state evaluation matrix and the behavior state evaluation coefficient are generated based on the fused comprehensive feature vector, including:

[0021] Define an n×m blood status evaluation matrix M to evaluate the changes in blood components. Each row of the matrix represents a specific blood component and the column represents a time point.

[0022] For each time point t, according to the feature vector f of the blood monitoring data in F b Calculate the state value of the corresponding blood component, the expression is: ; Among them, f bct The feature vector f represents the blood test data b The part related to blood component c at time point t, W1 is the weight matrix of the first layer, b1 is the bias vector of the first layer, W2 is the weight matrix of the second layer, b2 is the bias vector of the second layer, s ct Represents the state value of blood component c at time point t;

[0023] Use the feature vector f of the physiological signal data in the fused comprehensive feature vector F p and the feature vector f of the video data v , calculate the behavior state evaluation coefficient, the expression is: ; Among them, g p (f p ) represents the stability of the patient’s physiological characteristics at time point t, g v (f v ) represents the degree of change in the patient's behavioral and limb characteristics at time point t, B(t) is the behavioral state evaluation coefficient at time t, and w p is the physiological signal data feature vector f p The weight factor, w v is the video data feature vector f v The weight factor of .

[0024] As a preferred solution of the method for online monitoring of the patient's hemodialysis process of the present invention, the infection risk prediction model is constructed based on the state value of the blood component and the behavior state evaluation coefficient, which specifically includes:

[0025] Combining the state values ​​of blood components and the behavior state evaluation coefficient B(t), an infection risk evaluation model is constructed to evaluate the infection risk of patients at a given time point t. The expression is: ;in, represents the probability of infection risk at time t, n represents the number of blood components, and c is the blood component.

[0026] As a preferred solution of the method for online monitoring of the patient's hemodialysis process of the present invention, the optimization of the infection risk prediction model through adaptive learning includes:

[0027] Using the cross entropy loss function, for each time point t, the cross entropy loss between the predicted value P(t) and the actual label y(t) is calculated, and the expression is: ; Where y(t) is the actual label and L(t) is the difference measure between the model prediction value P(t) and the actual label y(t);

[0028] Integrate α and β to form a vector w = [α1, α2, …, α n ,β], where α n represents the weight of the nth blood component c, β is the weight of the behavior state evaluation function B(t);

[0029] Use gradient descent to update the weights, the expression is: ; Among them, w new is the updated weight vector, w old is the weight vector before update, η is the learning rate, ∇L is the gradient of the loss function L with respect to the weight vector w;

[0030] The optimized weight vector w new =[α * 1, α * 2, …, α * n , β * ].

[0031] As a preferred solution of the method for online monitoring of the patient's hemodialysis process of the present invention, the generating of the hemodialysis online monitoring report according to the infection risk prediction result includes:

[0032] The updated model parameters, the weight α of blood component c * and the weight β of the behavior evaluation function B(t) * Input the infection risk assessment model to predict the infection risk;

[0033] The predicted probability p based on the optimized infection risk prediction model * (t), generate hemodialysis online monitoring report.

[0034] In a second aspect, the present invention provides an online monitoring system for a patient's hemodialysis process, comprising: a data acquisition and preprocessing module, a multimodal data fusion module, a state evaluation module, an infection risk prediction module, an adaptive learning optimization module, and an online monitoring report generation module;

[0035] The data acquisition and preprocessing module is used to obtain multimodal monitoring data from different sensors and devices, and to preprocess the acquired data; the multimodal data fusion module is used to fuse the preprocessed multimodal monitoring data; the state evaluation module is used to generate a blood state evaluation matrix based on the comprehensive feature vector to evaluate the physiological indicator state of the patient's blood; the infection risk prediction module is used to construct an infection risk prediction model using the state values ​​of blood components and the behavioral state evaluation coefficients; the adaptive learning optimization module is used to continuously train and optimize the infection risk prediction model to improve the prediction accuracy; the online monitoring report generation module is used to generate corresponding online monitoring reports based on the infection risk probability.

[0036] In a third aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein when the computer program is executed by the processor, any step of the method for online monitoring of a patient's hemodialysis process as described in the first aspect of the present invention is implemented.

[0037] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, any step of the method for online monitoring of a patient's hemodialysis process as described in the first aspect of the present invention is implemented.

[0038] The beneficial effects of the present invention are as follows: by acquiring and preprocessing multimodal monitoring data, fusing these data with a deep learning model, constructing a blood status evaluation matrix and a behavioral status evaluation coefficient, establishing an infection risk prediction model and continuously optimizing it, an online hemodialysis monitoring report is finally generated, which effectively monitors the infection risk of patients during the dialysis process, and improves the accuracy of risk prediction by comprehensively evaluating the blood status and behavioral status, thereby helping medical staff to take preventive measures in a timely manner, reducing the possibility of infection, and improving the treatment safety and quality of life of patients. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0040] Figure 1 This is a flow chart of the method for online monitoring of the patient's hemodialysis process in Example 1.

[0041] Figure 2 This is a block diagram of the online monitoring system for the patient's hemodialysis process in Example 1. DETAILED DESCRIPTION

[0042] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0043] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0044] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0045] Example 1, reference Figure 1 and Figure 2 , which is the first embodiment of the present invention, provides an online monitoring method for a patient's hemodialysis process, comprising the following steps:

[0046] S1: Obtain multimodal monitoring data and perform preprocessing.

[0047] Furthermore, blood samples are collected and analyzed regularly through blood testing equipment to obtain blood test data, which are arranged in chronological order and plotted with time as the horizontal axis and blood component concentration as the vertical axis. The curve can show the dynamic change trend of blood components during the dialysis process;

[0048] By continuously monitoring physiological signals using wearable devices;

[0049] Obtain video data of patients through high-definition cameras in the dialysis room;

[0050] The blood test data were cleaned using mean interpolation method;

[0051] Butterworth filter is used to filter the noise of physiological signals;

[0052] Apply background subtraction algorithms to remove background noise from video data.

[0053] It should also be noted that mean interpolation is a method for dealing with missing data, especially when there are fewer missing values ​​or the distribution is more random. The specific steps are as follows: First, for each column (feature), calculate the arithmetic mean of all non-missing values; then, replace each missing value in the column with the calculated mean. This method assumes that the missing value can be reasonably estimated by the mean of other observations in the same column.

[0054] Butterworth filter is a digital filter. For physiological signals such as electrocardiogram (ECG), electroencephalogram (EEG), etc., Butterworth filter can be used to remove unwanted frequency components, such as 50Hz / 60Hz power frequency noise or other high-frequency noise. The general steps of designing a low-pass Butterworth filter include determining the filter specifications (such as passband cutoff frequency, blocking cutoff frequency, maximum allowable attenuation of passband, and minimum allowable attenuation of stopband), and then using these specifications to design filter coefficients and apply them to physiological signal data to achieve effective filtering of noise, thereby obtaining clearer and more accurate physiological signals.

[0055] Background subtraction is a technique used in video analysis that aims to separate foreground objects (usually moving objects) from the background in a video stream. This process usually includes several key steps: first, a background model is built through a series of video frames; then, for each frame of video, the current frame is compared with the background model to detect and segment the foreground area; finally, the boundaries of the foreground object are refined through thresholding or other image processing techniques to remove background interference. This method is very useful in video surveillance and behavioral analysis because it helps the system focus on important activities in the video while reducing the impact of background noise.

[0056] S2: Use deep learning models to fuse the preprocessed multimodal monitoring data into a comprehensive feature vector.

[0057] Furthermore, the preprocessed blood test data and physiological signal data are arranged into a sequence in chronological order;

[0058] The sequence data is input into the LSTM model, and the feature vectors of the blood test data and the physiological signal data are extracted from the last layer of the LSTM model;

[0059] The preprocessed video data is input into the CNN model, and the model extracts the feature vector of the video frame through the forward propagation process;

[0060] The extracted feature vectors are fused through the LSTM model, and the expression is: ; Among them, f b The feature vector representing the blood test data, f p is the feature vector of physiological signal data, f v is the feature vector of the video data, F is the integrated feature vector after fusion, and ct represents the concatenation operation of the feature vectors.

[0061] It should also be noted that LSTM is used to extract the features of blood test data and physiological signal data because: blood test data and physiological signal data usually have temporal continuity and correlation, and LSTM can capture this long-term dependency to better extract features. LSTM can effectively process data points in different time periods, which is very important for understanding the changes in the patient's physiological state over time.

[0062] CNN is used to extract features from video frames because: video frames are two-dimensional images composed of pixels, and the convolution layer in CNN can automatically learn local features in the image, such as edges, textures, etc., through the convolution kernel. Using CNN, high-level visual features can be extracted from video frames, which is very useful for identifying patient behavior patterns.

[0063] The extracted feature vectors are fused through the LSTM model because: LSTM can not only process sequence data, but also process sequences of different lengths, and it can retain important information in time series. Using LSTM in the fusion stage can effectively integrate information from different sources (blood test data, physiological signal data, and behavioral video data), thereby capturing the potential correlation between multimodal data, which is very critical for subsequent risk prediction.

[0064] S3: Generate the blood state evaluation matrix and behavior state evaluation coefficient based on the fused comprehensive feature vector.

[0065] Furthermore, an n×m blood status evaluation matrix M is defined to evaluate the changes in blood components, such as hemoglobin, blood potassium and other key indicators. Each row of the matrix represents a specific blood component, and the column represents the time point t. The expression is: ; Where n is the number of blood components and m is the number of time points monitored;

[0066] For each time point t, according to the feature vector f of the blood monitoring data in F b Calculate the state value of the corresponding blood component, the expression is: ; Among them, f bct The feature vector f represents the blood test data b The part related to blood component c at time point t, W1 is the weight matrix of the first layer, b1 is the bias vector of the first layer, W2 is the weight matrix of the second layer, b2 is the bias vector of the second layer, s ct Represents the state value of blood component c at time point t;

[0067] Use the feature vector f of the physiological signal data in the fused comprehensive feature vector F p and the feature vector f of the video data v, calculate the behavior state evaluation coefficient, the expression is: ; Among them, g p (f p ) represents the stability of the patient’s physiological characteristics at time point t, g v (f v ) represents the degree of change in the patient's behavioral and limb characteristics at time point t, B(t) is the behavioral state evaluation coefficient at time t, and w p is the physiological signal data feature vector f p The weight factor, w v is the video data feature vector f v The weight factor of

[0068] It should be noted that The expression is: ; where d is the dimension of the feature vector, is the weight assigned to the ith physiological signal feature vector, is the physiological signal feature vector The i-th element of ;

[0069] The expression is: ;in, is the weight assigned to the i-th video data feature vector, It is the feature of video data. The i-th element of the quantity;

[0070] It should also be noted that the behavioral status evaluation coefficient B(t) is used to evaluate the patient's behavioral status during dialysis. This coefficient should reflect whether the patient's behavior is normal;

[0071] Although CNN can process a single video frame well, it is not good at processing the temporal relationship between video frames. Video data is essentially a series of frames arranged in chronological order. There are temporal dependencies between these frames. The Transformer model can capture these temporal dependencies through its attention mechanism, which is a capability that CNN does not have.

[0072] S4: Construct an infection risk prediction model based on the blood status evaluation matrix and behavioral status evaluation coefficient.

[0073] Furthermore, by combining the blood state evaluation moment M and the behavior state evaluation coefficient B(t), an infection risk evaluation model is constructed to evaluate the infection risk of patients at a given time point t. The expression is: ;in, represents the probability of infection risk at time t, n represents the number of blood components, and c is the blood component.

[0074] It should also be noted that the weight α c It can be determined based on historical data through machine learning methods, such as training using methods such as support vector machine (SVM) or random forest (RF).

[0075] S5: Optimizing infection risk prediction model through adaptive learning.

[0076] Furthermore, using the cross entropy loss function, for each time point t, the cross entropy loss between the predicted value P(t) and the actual label y(t) is calculated, and the expression is: ; Where y(t) is the actual label and L(t) is the difference measure between the model prediction value P(t) and the actual label y(t);

[0077] Integrate α and β to form a vector w = [α1, α2, …, α n ,β], where α n represents the weight of the nth blood component c, β is the weight of the behavior state evaluation function B(t);

[0078] Use gradient descent to update the weights, the expression is: ; Among them, w new is the updated weight vector, w old is the weight vector before update, η is the learning rate, ∇L is the gradient of the loss function L with respect to the weight vector w;

[0079] The optimized weight vector w new =[α * 1, α * 2, …, α * n , β * ];

[0080] It should also be noted that the components of the gradient vector ∇L are: ;

[0081] The reason for choosing "using the cross entropy loss function" is that it is very suitable for classification tasks, especially in binary or multi-classification problems. The cross entropy loss function can intuitively measure the difference between the probability distribution predicted by the model and the true label, and provides good gradient information to promote the rapid convergence of the model. It can effectively guide the update of model parameters during the optimization process, thereby improving the prediction accuracy.

[0082] S6: Generate a hemodialysis online monitoring report based on the infection risk prediction results.

[0083] Furthermore, the updated model parameters, the weight α of blood component c * and the weight β of the behavior evaluation function B(t) *Input the infection risk assessment model to optimize it and make infection risk prediction;

[0084] The predicted probability p based on the optimized infection risk prediction model * (t), generate hemodialysis online monitoring report;

[0085] It should also be noted that during the optimization and prediction process, it should be ensured that the model can accurately reflect the changes in the patient's blood components and their relationship with the risk of infection. In addition, it is also necessary to consider the impact of the patient's behavioral habits and other external factors on the prediction results to ensure that the prediction model is not only accurate but also has a high practical application value. By continuously adjusting and optimizing the model parameters, the accuracy of infection risk prediction can be improved, thereby providing more reliable support for online monitoring of hemodialysis patients.

[0086] The present embodiment also provides an online monitoring system for a patient's hemodialysis process, comprising: a data acquisition and preprocessing module, a multimodal data fusion module, a state evaluation module, an infection risk prediction module, an adaptive learning optimization module, and an online monitoring report generation module; the data acquisition and preprocessing module is used to acquire multimodal monitoring data from different sensors and devices, and preprocess the acquired data; the multimodal data fusion module is used to perform fusion processing on the preprocessed multimodal monitoring data; the state evaluation module is used to generate a blood state evaluation matrix based on a comprehensive feature vector to evaluate the physiological indicator state of the patient's blood; the infection risk prediction module is used to construct an infection risk prediction model using the state values ​​of blood components and the behavioral state evaluation coefficients; the adaptive learning optimization module is used to continuously train and optimize the infection risk prediction model to improve the prediction accuracy; the online monitoring report generation module is used to generate a corresponding online monitoring report based on the infection risk probability.

[0087] This embodiment also provides a computer device, which is suitable for the case of an online monitoring method for a patient's hemodialysis process, and includes: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the online monitoring method for a patient's hemodialysis process as proposed in the above embodiment.

[0088] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covered on the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0089] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the method for online monitoring of a patient's hemodialysis process as proposed in the above embodiment is implemented; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, referred to as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, referred to as EEPROM), erasable programmable read-only memory (Erasable Programmable Read Only Memory, referred to as EPROM), programmable read-only memory (Programmable Red-Only Memory, referred to as PROM), read-only memory (Read-Only Memory, referred to as ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0090] In summary, the present invention ensures the integrity and accuracy of information through the acquisition and preprocessing of multimodal data, fuses these data using a deep learning model, extracts a comprehensive feature vector, and provides a quantitative basis for the evaluation of blood status and behavioral status. An infection risk prediction model is constructed based on these evaluations, and the model parameters are continuously optimized through adaptive learning to improve the accuracy of the prediction. The online monitoring report generated according to the prediction results provides a real-time and accurate reference for clinical decision-making, effectively improving the safety and effectiveness of hemodialysis treatment, and has significant clinical significance and practical value in improving the quality of monitoring of dialysis patients, reducing the risk of infection, and optimizing the allocation of clinical resources.

[0091] Example 2, referring to Table 1, is the second example of the present invention. To further verify the advancement of the present invention, experimental simulation data of an online monitoring method for a patient's hemodialysis process are provided.

[0092] Two groups were set up: the experimental group was monitored and analyzed using the method of the present invention, and the control group was monitored and analyzed using the existing technology;

[0093] First, blood samples are collected regularly using high-precision blood testing equipment, and physiological signals and video data are collected using wearable devices and high-definition cameras. The mean interpolation method is used to process missing values ​​in the blood test data, the Butterworth filter is used to remove noise in the physiological signals, and the background subtraction algorithm is used to separate foreground objects from the video data.

[0094] Secondly, the preprocessed data is input into the deep learning model. The LSTM model processes the blood test data and physiological signal data to extract time series features. The CNN model processes the video data to extract visual features. These feature vectors are fused through the LSTM model to form a comprehensive feature vector.

[0095] Then, based on feature fusion, the blood status evaluation matrix and behavior status evaluation coefficient were constructed, and the infection risk prediction model was further constructed. Through adaptive learning, the model parameters were continuously optimized using the cross entropy loss function to improve the accuracy of the prediction.

[0096] Finally, the hemodialysis online monitoring report was generated based on the optimized model, and the risk threshold θ was defined. P , and by comparing the predicted probability p*(t) with θ P To assess the patient's health status.

[0097] The details are shown in Table 1:

[0098] Table 1 Experimental record table

[0099] Test subjects Time Blood component concentration (mg / dL) Behavior status evaluation coefficient Prediction probability of infection risk Patient 1 in the experimental group T1 110 0.5 0.05 Patient 1 in the experimental group T2 112 0.7 0.10 Patient 2 in the experimental group T1 98 0.3 0.20 Patient 2 in the experimental group T2 100 0.6 0.15 Control group patient 1 T1 105 0.4 0.30 Control group patient 1 T2 108 0.5 0.25 Control group patient 2 T1 101 0.2 0.35 Control group patient 2 T2 104 0.4 0.40

[0100] Through the collection and analysis of the above table data, it can be clearly seen that the present invention has advantages in predicting the infection risk of hemodialysis patients. For example, in the data of the experimental group of patients, the change of the behavior state evaluation coefficient over time leads to an increase in the infection risk prediction probability. This shows that the present invention can monitor and evaluate the risk status of patients in real time, providing strong support for clinical decision-making.

[0101] Through the online monitoring method of the patient's hemodialysis process of the present invention, not only can the patient's blood composition changes and physiological signals be captured and analyzed in real time, but also multimodal data can be integrated through a deep learning model to provide a more accurate behavior and infection risk assessment. In particular, through the prediction model that is continuously optimized through adaptive learning, the present invention can predict and manage health risks that may occur during the dialysis process with unprecedented accuracy.

[0102] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for online monitoring of a patient's hemodialysis process, characterized in that: include: Acquire multimodal monitoring data and perform preprocessing; Use deep learning models to fuse preprocessed multimodal monitoring data into comprehensive feature vectors; Based on the fused comprehensive feature vector, a blood state evaluation matrix and a behavior state evaluation coefficient are generated; Construct an infection risk prediction model based on the status values ​​of blood components and behavioral status evaluation coefficients; Optimizing infection risk prediction models through adaptive learning; Generate hemodialysis online monitoring report based on infection risk prediction results; The obtaining of multimodal monitoring data and preprocessing thereof include: Regularly collect blood samples and analyze them through blood testing equipment to obtain blood test data; By continuously monitoring physiological signals using wearable devices; Obtain video data of patients through high-definition cameras in the dialysis room; The blood test data were cleaned using mean interpolation method; Butterworth filter is used to filter the noise of physiological signals; Apply background subtraction algorithms to remove background interference from video data; The deep learning model is used to fuse the pre-processed multimodal monitoring data into a comprehensive feature vector, including: Arranging the preprocessed blood test data and physiological signal data into sequences in chronological order; The sequence data is input into the LSTM model, and the feature vectors of the blood test data and the physiological signal data are extracted from the last layer of the LSTM model; The preprocessed video data is input into the CNN model, and the model extracts the feature vector of the video frame through the forward propagation process; The extracted feature vectors are fused through the LSTM model, and the expression is: Among them, f b The feature vector representing the blood test data, f p is the feature vector of physiological signal data, f v is the feature vector of the video data, F is the integrated feature vector after fusion, and ct represents the concatenation operation of the feature vectors; The method of generating a blood state evaluation matrix and a behavior state evaluation coefficient based on the fused comprehensive feature vector includes: Define an n×m blood status evaluation matrix M to evaluate the changes in blood components. Each row of the matrix represents a specific blood component and the column represents the time point t. For each time point t, according to the feature vector f of the blood monitoring data in F b Calculate the state value of the corresponding blood component, the expression is: Among them, f bct The feature vector f represents the blood test data b The part related to blood component c at time point t, W1 is the weight matrix of the first layer, b1 is the bias vector of the first layer, W2 is the weight matrix of the second layer, b2 is the bias vector of the second layer, s ct Represents the state value of blood component c at time point t; Use the feature vector f of the physiological signal data in the fused comprehensive feature vector F p and the feature vector f of the video data v , calculate the behavior state evaluation coefficient, the expression is: Among them, g p (f p ) represents the stability of the patient’s physiological characteristics at time point t, g v (f v ) represents the degree of change in the patient's behavioral and limb characteristics at time point t, B(t) is the behavioral state evaluation coefficient at time t, and w p is the physiological signal data feature vector f p The weight factor, w v is the video data feature vector f v The weight factor of .

2. The method for online monitoring of a patient's hemodialysis process as claimed in claim 1, characterized in that: The infection risk prediction model is constructed based on the state value of blood components and the behavior state evaluation coefficient, including: Combining the state values ​​of blood components and the behavior state evaluation coefficient B(t), an infection risk evaluation model is constructed to evaluate the infection risk of patients at a given time point t. The expression is: in, represents the probability of infection risk at time t, n represents the number of blood components, and c is the blood component.

3. The method for online monitoring of a patient's hemodialysis process as claimed in claim 2, characterized in that: The method of optimizing the infection risk prediction model through adaptive learning includes: Using the cross entropy loss function, for each time point t, the cross entropy loss between the predicted value P(t) and the actual label y(t) is calculated, and the expression is: Where y(t) is the actual label and L(t) is the difference measure between the model prediction value P(t) and the actual label y(t); Integrate α and β to form a vector w = [α1, α2, …, α n ,β], where α n represents the weight of the nth blood component c, β is the weight of the behavior state evaluation function B(t); Use gradient descent to update the weights, the expression is: Among them, w new is the updated weight vector, w old is the weight vector before update, η is the learning rate, ∇L is the gradient of the loss function L with respect to the weight vector w; The optimized weight vector w new =[α * 1, α * 2, …, α * n , β * ].

4. The method for online monitoring of a patient's hemodialysis process as claimed in claim 3, characterized in that: The hemodialysis online monitoring report is generated according to the infection risk prediction result, including: The updated model parameters, the weight α of blood component c * and the weight β of the behavior evaluation function B(t) * Input the infection risk assessment model to predict the infection risk; The predicted probability p based on the optimized infection risk prediction model * (t), generate hemodialysis online monitoring report.

5. An online monitoring system for a patient's hemodialysis process, based on the online monitoring method for a patient's hemodialysis process according to any one of claims 1 to 4, characterized in that: include: Data acquisition and preprocessing module, multimodal data fusion module, status evaluation module, infection risk prediction module, adaptive learning optimization module, online monitoring report generation module; The data acquisition and preprocessing module is used to obtain multimodal monitoring data from different sensors and devices and preprocess the acquired data; The multimodal data fusion module is used to fuse the pre-processed multimodal monitoring data; The state evaluation module is used to generate a blood state evaluation matrix based on the comprehensive feature vector to evaluate the physiological index state of the patient's blood; The infection risk prediction module is used to construct an infection risk prediction model using the state values ​​of blood components and the behavioral state evaluation coefficients; The adaptive learning optimization module is used to continuously train and optimize the infection risk prediction model to improve prediction accuracy; The online monitoring report generation module is used to generate corresponding online monitoring reports according to the infection risk probability.

Images