A method for identifying unexpected outflow of anticoagulant for plasma platelet collection

Abstract

The present invention discloses a method for identifying unexpected outflow of anticoagulant for plasma platelet collection, specifically related to the field of detecting unexpected outflow of anticoagulant. By virtue of the characteristic of processing time series data of the LSTM model, the present invention captures the changing trends of the anticoagulant dosage and chemical parameters over time, so as to provide a more accurate prediction of flow rationality, and analyzes chemical characteristics through the Dense model to identify the risk of unexpected outflow. The combination of the above two models in the present invention can analyze data from multiple dimensions, timely detect signs of unexpected outflow, and give early warnings.

Description

Technical Field

[0001] The present invention relates to the field of detecting unexpected outflow of anticoagulants, and particularly to a method for identifying unexpected outflow of anticoagulants used in plasma platelet collection. Background Art

[0002] In the field of plasma platelet collection, precise control of the dosage of anticoagulants is crucial to ensure the safety and efficiency of the collection process. Existing technologies usually rely on electronic scales and the mechanical rotation pulses of anticoagulant pumps to measure and control the dosage of anticoagulants. These devices can provide basic dosage data and assist in decision-making through the analysis of the concentration and viscosity parameters of anticoagulants. The electronic scale measures the dosage of anticoagulants, providing a direct method to monitor the accuracy of each collection. At the same time, the anticoagulant pump achieves precise control of the dosage through the conversion of the number of rotation pulses and the unit pulse volume. In addition, the real-time monitoring of the parameters of anticoagulants provides important information for evaluating the status of anticoagulants.

[0003] Although existing technologies can achieve the measurement and control of the dosage of anticoagulants to a certain extent, there are obvious deficiencies in the identification and prediction of unexpected outflows. Especially in a dynamically changing collection environment, simply relying on physical measurements and simple analysis is difficult to accurately predict and timely respond to unexpected outflow events. In addition, existing technologies often lack in-depth analysis of historical data and time series changes and cannot make full use of this information to improve the accuracy and reliability of prediction. Summary of the Invention

[0004] The purpose of the present invention is to provide a method for identifying unexpected outflow of anticoagulants used in plasma platelet collection, which can quickly and accurately determine whether there is an unexpected outflow of anticoagulants during the collection of plasma or platelets by the plasma / platelet collection system, so as to help the plasma / platelet collection system identify the occurrence of unexpected outflow of anticoagulants and give an alarm, improving the health and safety of plasma donors.

[0005] The present invention is achieved through the following technical solutions:

[0006] A method for identifying unexpected outflow of anticoagulants used in plasma platelet collection, which includes the following steps:

[0007] S1. Collect the dosage of anticoagulants measured by the electronic scale at time t, the number of rotation pulses of the anticoagulant pump at time t, and the concentration and viscosity parameters of the anticoagulant at time t. Convert the pulse dosage according to the number of rotation pulses of the anticoagulant pump at time t, calculate the dosage of anticoagulants corresponding to the number of rotation pulses of the anticoagulant pump, and calculate the deviation amount based on the dosage of anticoagulants measured by the electronic scale at time t and the dosage of anticoagulants corresponding to the number of rotation pulses of the anticoagulant pump;

[0008] S2. Extract the anticoagulant dosage and deviation measured by the electronic scale at time t as physical features, construct the feature matrix of the physical model, and construct the LSTM physical prediction model based on the feature matrix of the physical model. The LSTM physical prediction model performs flow rationality prediction;

[0009] S3. Extract the concentration and viscosity of the anticoagulant at time t as features, construct the feature matrix of the concentration and viscosity model, and construct the Dense prediction model based on the feature matrix of the concentration and viscosity model. The Dense prediction model predicts the non-expected outflow prediction result;

[0010] S4. Calculate the losses of flow rationality, non-expected outflow, and cumulative non-expected outflow respectively, define the joint loss function, and construct and train the joint model through the defined joint loss function and gradient descent optimization algorithm;

[0011] S5. Input new data, and predict the flow rationality and expected outflow prediction results through the joint model.

[0012] Further, in step S1, perform pulse dosage conversion according to the rotation pulses of the anticoagulant pump at time t, and calculate the anticoagulant dosage corresponding to the rotation pulses of the anticoagulant pump. The specific calculation method is as follows:

[0013] ;

[0014] wherein, the represents the anticoagulant dosage, the represents the rotation pulses of the anticoagulant pump at time t, and the represents the unit pulse volume.

[0015] Further, in step S1, calculate the deviation according to the anticoagulant dosage measured by the electronic scale at time t and the anticoagulant dosage corresponding to the rotation pulses of the anticoagulant pump. The specific calculation method is as follows:

[0016] ;

[0017] wherein, the represents the deviation, the represents the anticoagulant dosage measured by the electronic scale at time t, and the represents the anticoagulant dosage.

[0018] Further, step S2 specifically includes the following sub-steps:

[0019] S201. Construct the physical feature matrix, including the anticoagulant dosage and deviation measured by the electronic scale:

[0020] ;

[0021] Among them, the represents the physical feature matrix, and the represents the amount of anticoagulant measured by the electronic scale at a certain moment, and the represents the deviation amount per unit time, and the represents the time point, that is, the total number of time points in the time series data;

[0022] S202. Process the feature matrix through a one-dimensional convolutional layer to obtain the convolutional layer output: , where the represents the convolutional layer output at time point t, the represents the bias term of the convolutional layer, the represents, and the represents the width of the convolutional kernel, the represents the physical feature input at time point t-k, and the represents the weight in the convolutional kernel;

[0023] S203. Process the convolutional layer output through an LSTM layer to obtain the LSTM layer output;

[0024] S204. Use a fully connected layer to process the LSTM layer output to predict the flow rationality.

[0025] Furthermore, the specific content of step S204 is: , where the represents the predicted value of the flow rationality at time point t, the represents the weight matrix of the fully connected layer, the represents the LSTM layer output at time point t, and the represents the bias term of the fully connected layer.

[0026] Furthermore, step S3 specifically includes the following sub-steps:

[0027] S301. Construct a concentration and viscosity matrix, including anticoagulant concentration and viscosity:

[0028] ;

[0029] Among them, the represents the concentration and viscosity matrix, the represents the anticoagulant concentration at a certain moment, the represents the anticoagulant viscosity at a certain moment, and the represents the time point;

[0030] S302. Process the concentration and viscosity matrix through a fully connected layer to obtain the concentration and viscosity processing output;

[0031] S303. Output the unexpected outflow risk prediction through the fully connected layer.

[0032] Further, the step S303 is specifically as follows: , where the represents the unexpected outflow prediction result at time point t, and the represents the weight matrix of the fully connected layer for the unexpected outflow prediction, and the represents the output after the concentration and viscosity at time point t pass through the ReLU activation function, and the represents the bias term of the fully connected layer for the unexpected outflow prediction.

[0033] Further, in the step S4, calculating the losses for flow rationality, unexpected outflow, and cumulative unexpected outflow specifically includes:

[0034] Calculating the loss for flow rationality: , where the represents the mean squared error loss of flow rationality, the represents the predicted flow rationality at time point t, and the represents the actual flow rationality at time point t;

[0035] Calculating the loss for unexpected outflow: , where the represents the mean squared error loss of unexpected outflow, the represents the model-predicted unexpected outflow at time point t, and the represents the actual unexpected outflow at time point t;

[0036] Calculating the loss for cumulative unexpected outflow: , where the represents the mean squared error loss of cumulative unexpected outflow, the represents the actual unexpected outflow at time point t, and the represents the weight parameter of the regularization term.

[0037] Further, in the step S4, defining the joint loss function is specifically defined by weighted averaging of the flow rationality loss, unexpected outflow loss, and cumulative unexpected outflow loss: , where the represents the joint loss function, the represents the weight for balancing the flow rationality loss, and the represents the weight for balancing the cumulative unexpected outflow loss.

[0038] Further, in step S4, the gradient descent optimization algorithm specifically updates the model parameters through gradient descent:

[0039] ；

[0040] ;

[0041] Among them, the represents the weight matrix in the model, the represents the bias term in the model, the represents the learning rate, and the represents the loss function partial derivative of the weight, and the represents the loss function partial derivative of the bias.

[0042] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0043] By using the characteristics of the LSTM model to process time series data, the present invention captures the changing trends of anticoagulant dosage, concentration, and viscosity parameters over time, thereby providing a more accurate prediction of flow rationality. Through the Dense model, the concentration and viscosity are analyzed to identify the risk of unexpected outflow. The combination of the above two models in the present invention can analyze data from multiple dimensions, timely detect signs of unexpected outflow, and give early warnings. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, form a part of this application, and do not limit the embodiments of the present invention. In the drawings:

[0045] Figure 1 is a method flow chart of a method for identifying unexpected outflow of anticoagulant for plasma platelet collection according to the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0046] The technical solution of the present invention will be further described in detail below with reference to the drawings, but the protection scope of the present invention is not limited to the following.

[0047] In order to make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention, that is, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings herein can be arranged and designed in various different configurations.

[0048] Accordingly, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative efforts fall within the scope of protection of the present invention. It should be noted that relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations.

[0049] Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, such that a process, method, article or device comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "comprising one..." does not exclude the presence of additional identical elements in the process, method, article or device comprising the element.

[0050] The features and performance of the present invention will be further described in detail below in conjunction with embodiments.

[0051] A method for identifying unexpected outflow of anticoagulant for plasma platelet collection, as Figure 1 , includes the following steps:

[0052] S1. Collect the anticoagulant dosage measured by the weighing scale at time t, the rotation pulses of the anticoagulant pump at time t, and the concentration and viscosity parameters of the anticoagulant at time t. Perform pulse dosage conversion according to the rotation pulses of the anticoagulant pump at time t, calculate the anticoagulant dosage corresponding to the rotation pulses of the anticoagulant pump, and calculate the deviation amount based on the anticoagulant dosage measured by the weighing scale at time t and the anticoagulant dosage corresponding to the rotation pulses of the anticoagulant pump.

[0053] S2. Extract the anticoagulant dosage measured by the weighing scale at time t and the deviation amount as physical features, construct the feature matrix of the physical model, construct the LSTM physical prediction model according to the feature matrix of the physical model, and the LSTM physical prediction model performs flow rationality prediction.

[0054] S3. Extract the concentration and viscosity parameters of the anticoagulant at time t as the concentration and viscosity, construct the feature matrix of the concentration and viscosity model, construct the Dense prediction model according to the feature matrix of the concentration and viscosity model, and the Dense prediction model predicts the prediction result of unexpected outflow.

[0055] S4. Calculate the losses for flow rationality, unexpected outflow, and cumulative unexpected outflow respectively, define a joint loss function, and construct and train a joint model through the defined joint loss function and gradient descent optimization algorithm;

[0056] S5. Input new data and predict the flow rationality and expected outflow prediction results through the joint model.

[0057] Specifically, the system collects the anticoagulant dosage measured by the electronic scale and the rotation pulses of the anticoagulant pump at each time point t, and simultaneously records the concentration and viscosity parameters of the anticoagulant. Through the known unit pulse volume, the rotation pulses are converted into anticoagulant dosage, and the deviation between the measured value of the electronic scale and the dosage calculated by the pump is calculated. The deviation is used to evaluate the accuracy of the anticoagulant pump. Secondly, the system extracts physical features, including the anticoagulant dosage measured by the electronic scale and the calculated deviation, constructs a feature matrix, and uses a one-dimensional convolutional layer to process the feature matrix to extract local features in the time series data; the LSTM network processes the output of the convolutional layer to capture long-term dependencies in the data and perform flow rationality prediction; further, the system extracts the concentration and viscosity. Exemplarily, in this embodiment, the concentration and viscosity of the anticoagulant are selected to construct a concentration and viscosity matrix. The concentration and viscosity matrix is processed through a fully connected layer to learn the relationship between the concentration and viscosity characteristics and the unexpected outflow, and the unexpected outflow risk prediction is output through another fully connected layer; the system calculates the mean square error loss of flow rationality, the mean square error loss of unexpected outflow, and the cumulative unexpected outflow loss respectively. The above loss functions evaluate the difference between the model prediction value and the actual value, define a joint loss function, balance the impact on model training by weighted average of the loss values, and then use the gradient descent optimization algorithm to update the weights and biases according to the gradient of the loss function with respect to the model parameters to minimize the joint loss function; when the model training is completed and its accuracy is verified, new data is input into the trained joint model, and the flow rationality and expected outflow prediction results are predicted through the joint model.

[0058] Further, in step S1, the pulse dosage conversion is performed according to the rotation pulses of the anticoagulant pump at time t, and the anticoagulant dosage corresponding to the rotation pulses of the anticoagulant pump is calculated. The specific calculation method is as follows:

[0059] ;

[0060] wherein, the represents the anticoagulant dosage, the represents the rotation pulses of the anticoagulant pump at time t, and the represents the unit pulse volume.

[0061] Further, in the step S1, the deviation amount is calculated according to the anticoagulant dosage measured by the electronic scale at time t and the anticoagulant dosage corresponding to the rotation pulse number of the anticoagulant pump. The specific calculation method is as follows:

[0062] ;

[0063] wherein, the represents the deviation amount, the represents the anticoagulant dosage measured by the electronic scale at time t, and the represents the anticoagulant dosage.

[0064] Further, the step S2 specifically includes the following sub-steps:

[0065] S201. Construct a physical feature matrix, including the anticoagulant dosage measured by the electronic scale and the deviation amount:

[0066] ;

[0067] wherein, the represents the physical feature matrix, the represents the anticoagulant dosage measured by the electronic scale at a certain moment, the represents the deviation amount per unit time, and the represents the time point, that is, the total number of time points in the time series data;

[0068] S202. Process the feature matrix through a one-dimensional convolutional layer to obtain the output of the convolutional layer: , wherein, the represents the output of the convolutional layer at time point t, the represents the bias term of the convolutional layer, the represents, the represents the width of the convolutional kernel, the represents the physical feature input at time point t-k, and the represents the weight in the convolutional kernel;

[0069] S203. Process the output of the convolutional layer through an LSTM layer to obtain the output of the LSTM layer;

[0070] S204. Use a fully connected layer to process the output of the LSTM layer to predict the flow rationality.

[0071] Specifically, a physical feature matrix is constructed, which integrates the anticoagulant dosage measured by the electronic scale at each time point and the calculated deviation amount. The physical feature matrix is a two-dimensional array, where each row represents a feature vector at a time point, including the dosage and the deviation amount, and the columns correspond to different features. The physical feature matrix is processed by a one-dimensional convolutional layer to extract local features in the time series data. The convolution operation filters through a sliding window feature matrix; the LSTM layer receives the output of the convolutional layer and processes it to capture long-term dependencies in the time series data, and a fully connected layer is used to process the output of the LSTM layer to predict the flow rationality. Among them, the fully connected layer is a set of linear equations that convert the high-dimensional output of the LSTM layer into the final prediction result.

[0072] Further, the step S204 is specifically as follows: , where the represents the predicted value of the flow rationality at time point t, the represents the weight matrix of the fully connected layer, the represents the output of the LSTM layer at time point t, and the represents the bias term of the fully connected layer.

[0073] Further, the step S3 specifically includes the following sub-steps:

[0074] S301. Construct a concentration and viscosity matrix, including anticoagulant concentration and viscosity:

[0075] ;

[0076] where the represents the concentration and viscosity matrix, the represents the anticoagulant concentration at a certain moment, the represents the anticoagulant viscosity at a certain moment, and the represents the time point;

[0077] S302. Process the concentration and viscosity matrix through a fully connected layer to obtain a concentration and viscosity processing output;

[0078] S303. Predict the risk of unexpected outflow through the output of the fully connected layer.

[0079] Specifically, the construction of the concentration and viscosity matrix is similar to that of the physical characteristics matrix, but focuses on concentration and viscosity parameters, such as the concentration and viscosity of the anticoagulant. These parameters affect the flow characteristics and reactivity of the anticoagulant. The fully connected layer processes the concentration and viscosity matrix to learn the impact of concentration and viscosity parameters on the unintended outflow. The ReLU activation function is used to introduce non-linearity, enabling the model to learn and simulate more complex behaviors. The second fully connected layer converts the processed output of concentration and viscosity into a prediction of the unintended outflow, predicting potential unintended events by learning the relationship between concentration and viscosity parameters and the outflow risk.

[0080] Further, the specific content of step S303 is as follows: , where the represents the unintended result at time point t, and the represents the weight matrix of the fully connected layer for the prediction of unintended outflow, and the represents the output of the concentration and viscosity at time point t after passing through the ReLU activation function, and the represents the bias term of the fully connected layer for the prediction of unintended outflow.

[0081] Further, in step S4, the loss calculation for flow rationality, unintended outflow, and cumulative unintended outflow specifically includes:

[0082] Loss calculation for flow rationality: , where the represents the mean squared error loss of flow rationality, and the represents the predicted flow rationality at time point t, and the represents the actual flow rationality at time point t;

[0083] Loss calculation for unintended outflow: , where the represents the mean squared error loss of unintended outflow, and the represents the model-predicted unintended outflow at time point t, and the represents the actual unintended outflow at time point t;

[0084] Loss calculation for cumulative unintended outflow: , where the represents the mean squared error loss of cumulative unintended outflow, and the represents the actual unintended outflow at time point t, and the represents the weight parameter of the regularization term.

[0085] Specifically, the mean squared error loss and The difference between the flow rationality predicted by the quantization model and the actual measurement value, and the cumulative unexpected outflow loss To evaluate the long-term accuracy of the model prediction, it should be noted that in plasma platelet collection, even a small amount of unexpected outflow may have little impact in the short term, but the long-term accumulation may have a significant impact on the safety and efficiency of the entire collection process. Therefore, through the regularization term of the cumulative unexpected outflow loss function, additional penalties are imposed on the cases where the predicted outflow volume is non-zero. In addition, for the flow rationality, it represents the rationality of the flow under specific conditions (such as specific anticoagulant dosage, time point, etc.). This can be a continuous value representing the degree of flow rationality, or a binary value representing whether the flow is reasonable. For the above actual flow rationality and actual unexpected outflow values, the flow rationality is obtained by cross-verifying the anticoagulant dosage measured by the electronic scale and the output of the anticoagulant pump, while the actual unexpected outflow value is obtained from the unexpected outflow data collected by the system, which is mainly determined according to the deviation value between the expected anticoagulant dosage and the actual anticoagulant dosage. Exemplarily, when the deviation value is within the set threshold range, it is determined as an expected outflow, and when the deviation value is not within the set threshold range, it is determined as an unexpected outflow.

[0086] Furthermore, in step S4, the defined combined loss function is specifically defined as the weighted average of the flow rationality loss, unexpected outflow loss, and cumulative unexpected outflow loss: , where the represents the combined loss function, the represents the weight for balancing the flow rationality loss, and the represents the weight for balancing the cumulative unexpected outflow loss. Specifically, the purpose of the combined loss function is to provide a unified optimization objective, enabling the model to simultaneously learn features related to flow rationality and unexpected outflow risk prediction during training, coordinating the optimization processes of the two sub-tasks, balancing the impact on model parameter updates, and improving the overall performance and generalization ability of the model.

[0087] Furthermore, in step S4, the gradient descent optimization algorithm is specifically to update the model parameters through gradient descent:

[0088] ；

[0089] ;

[0090] where the represents the weight matrix in the model, the represents the bias term in the model, the represents the learning rate, and the represents the loss function Partial derivative with respect to the weight, where the represents the loss function Partial derivative with respect to the bias. Specifically, the parameters updated by the gradient descent optimization algorithm are used for the output of the integration layer of the joint model, that is: .

[0091] Furthermore, as a preferred implementation manner of the above embodiment, a specific implementation scheme for the non-linear integration of the above joint model is proposed. The process is as follows: For the output non-expected outflow risk result and flow rationality prediction result, non-linear activation functions are used for non-linear integration: , where the represents the output of the joint model, the represents the activation function, the and are respectively determined through gradient descent optimization and determined according to the non-expected outflow risk result and the flow rationality prediction model. The represents the output non-expected outflow risk result, the represents the output flow rationality prediction result, and the represents the bias term of the joint model. And the output of the two subtasks is used as features to train the joint model. During training, the models of each subtask are trained independently.

[0092] Furthermore, as a preferred implementation scheme of this embodiment, a method for outputting results through decision rules is proposed. Exemplarily:

[0093] ;

[0094] where the represents a set threshold, which is used to determine the magnitude of the difference between the two outputs, so as to determine which output to select as the final output.

[0095] Furthermore, through the joint model proposed in the above embodiment, the final output of the model can be set according to the threshold and decision rules. According to the changes in the threshold and decision rules, the output priorities of the prediction results of the two task outputs can be set. Specifically, as in the above example content, when the difference between the two prediction result values is too small, the can be selected to be output first, that is has a higher priority.

[0096] Furthermore, as a preferred implementation manner of this embodiment, a non-expected outflow recognition system for anticoagulants used in plasma platelet collection is proposed. This system can be applied to a non-expected outflow recognition method for anticoagulants used in plasma platelet collection, including:

[0097] A data acquisition and calculation module is used to collect the anticoagulant dosage measured by the electronic scale at time t, the rotation pulse number of the anticoagulant pump at time t, and the concentration and viscosity parameters of the anticoagulant at time t. It converts the pulse dosage according to the rotation pulse number of the anticoagulant pump at time t, calculates the anticoagulant dosage corresponding to the rotation pulse number of the anticoagulant pump, and calculates the deviation amount based on the anticoagulant dosage measured by the electronic scale at time t and the anticoagulant dosage corresponding to the rotation pulse number of the anticoagulant pump.

[0098] A physical model construction module is used to extract the anticoagulant dosage and deviation amount measured by the electronic scale at time t as physical features, construct a feature matrix of the physical model, and construct an LSTM physical prediction model based on the feature matrix of the physical model. The LSTM physical prediction model performs flow rationality prediction.

[0099] A concentration and viscosity model construction module is used to extract the concentration and viscosity parameters of the anticoagulant at time t as features, construct a feature matrix of the concentration and viscosity model, and construct a Dense prediction model based on the feature matrix of the concentration and viscosity model. The Dense prediction model predicts the prediction result of unexpected outflow.

[0100] A model training and optimization module is used to calculate the losses of flow rationality, unexpected outflow, and cumulative unexpected outflow respectively, define a joint loss function, and construct and train a joint model through the defined joint loss function and gradient descent optimization algorithm.

[0101] A data prediction module is used to input new data and predict the flow rationality and unexpected outflow prediction results through the joint model.

[0102] Furthermore, the specific implementation principle process of the above embodiment is as follows:

[0103] The first step: The electronic scale self-check before sampling, checks the working state of the current electronic scale and obtains the zero value of the electronic scale, and stores it in the system.

[0104] The second step: The anticoagulant pump self-check before sampling, checks the working state of the current anticoagulant pump.

[0105] The third step: Pipeline pre-filling, uses anticoagulant to pre-fill the pipeline, and obtains the anticoagulant dosage of the electronic scale and the rotation pulse number of the anticoagulant pump.

[0106] The fourth step: Calculate the anticoagulant pump flow rate, divide the anticoagulant dosage of the electronic scale by the rotation pulse number of the anticoagulant pump to calculate the anticoagulant pump flow rate.

[0107] The fifth step: Anticoagulant pump flow rate rationality verification. After calculating the current anticoagulant pump flow rate data, the verification threshold range is [M0, M1]. If the verification is qualified, the process continues; if the verification is unqualified, an alarm is given.

[0108] Step 6: Preprocess the collected data, record the current weighing data of the electronic scale, and zero the rotation pulses of the anticoagulant pump;

[0109] Step 7: During the collection process, the electronic scale calculates the dosage of the anticoagulant, and calculates the dosage of the anticoagulant W0 according to the weighing reading of the electronic scale; Step 8: During the collection process, the anticoagulant pump calculates the dosage of the anticoagulant, and calculates the dosage of the anticoagulant W1 by multiplying the number of rotation pulses of the anticoagulant pump by the flow rate of the anticoagulant pump;

[0110] Step 9: Verify the unexpected outflow of the anticoagulant. The non-expected outflow value is calculated by subtracting the dosage of the anticoagulant W1 calculated by the anticoagulant pump from the dosage of the anticoagulant W0 calculated by the electronic scale, and the deviation amount [-B0, B0] is verified.

[0111] On the basis of the above steps, this embodiment also adds deep learning technology and performs joint prediction on concentration and viscosity, making the prediction result more accurate. Specifically:

[0112] The physical model construction module specifically includes:

[0113] The physical feature construction unit is used to construct a physical feature matrix, including the dosage of the anticoagulant measured by the electronic scale and the deviation amount:

[0114] ;

[0115] Among them, the represents the physical feature matrix, the represents the dosage of the anticoagulant measured by the electronic scale at a certain moment, the represents the deviation amount per unit time, and the represents the time point, that is, the total number of time points in the time series data;

[0116] The model convolution unit is used to process the feature matrix through a one-dimensional convolution layer to obtain the output of the convolution layer: , where the represents the output of the convolution layer at time point t, the represents the bias term of the convolution layer, the represents, the width of the convolution kernel is represented, and the represents the physical feature input at time point t-k, and the represents the weight in the convolution kernel;

[0117] The LSTM output unit is used to process the output of the convolution layer through the LSTM layer to obtain the output of the LSTM layer;

[0118] The first prediction output unit uses a fully connected layer to process the output of the LSTM layer to predict the flow rationality.

[0119] Further, the concentration and viscosity model construction module specifically includes:

[0120] Concentration and viscosity building blocks for constructing a concentration and viscosity matrix, including anticoagulant concentration and viscosity:

[0121] ;

[0122] Among them, the represents the concentration and viscosity matrix, and the represents the anticoagulant concentration at a certain moment, and the represents the anticoagulant viscosity at a certain moment, and the represents the time point;

[0123] A model processing unit for processing the concentration and viscosity matrix through a fully connected layer to obtain a concentration and viscosity processing output;

[0124] A second prediction output unit for outputting an unexpected outflow risk prediction through a fully connected layer.

[0125] The specific embodiments described above further elaborate on the purpose, technical solutions, and beneficial effects of the present invention. It should be understood that the above are only specific embodiments of the present invention and are not used to limit the protection scope of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included in the protection scope of the present invention.

Claims

1. An identification method for unexpected outflow of anticoagulant used in plasma platelet collection, characterized in that, Including the following steps: S1. Collect the anticoagulant dosage measured by the electronic scale at time t, the rotation pulse number of the anticoagulant pump at time t, and the concentration and viscosity parameters of the anticoagulant at time t. Perform pulse dosage conversion according to the rotation pulse number of the anticoagulant pump at time t, calculate the anticoagulant dosage corresponding to the rotation pulse number of the anticoagulant pump, and calculate the deviation amount according to the anticoagulant dosage measured by the electronic scale at time t and the anticoagulant dosage corresponding to the rotation pulse number of the anticoagulant pump; S2. Extract the anticoagulant dosage measured by the electronic scale at time t and the deviation amount as physical features, construct the feature matrix of the physical model, and construct the LSTM physical prediction model according to the feature matrix of the physical model. The LSTM physical prediction model performs flow rationality prediction; S3. Extract the concentration and viscosity of the anticoagulant at time t as features, construct the feature matrix of the concentration and viscosity model, and construct the Dense prediction model according to the feature matrix of the concentration and viscosity model. The Dense prediction model predicts the non-expected outflow prediction result; S4. Calculate the losses for flow rationality, non-expected outflow, and cumulative non-expected outflow respectively, and define the joint loss function. Construct and train the joint model through the defined joint loss function and the gradient descent optimization algorithm; S5. Input new data, and predict the flow rationality and the expected outflow prediction result through the joint model.

2. The method for identifying unexpected outflow of anticoagulant for plasma platelet collection according to claim 1, wherein In step S1, perform pulse dosage conversion according to the rotation pulse number of the anticoagulant pump at time t, and calculate the anticoagulant dosage corresponding to the rotation pulse number of the anticoagulant pump. The specific calculation method is as follows: ； Among them, the represents the anticoagulant dosage, and the represents the rotation pulse number of the anticoagulant pump at time t, and the represents the unit pulse volume.

3. A method for identifying unexpected outflow of anticoagulant for plasma platelet collection according to claim 1, characterized in that, In step S1, calculate the deviation amount according to the anticoagulant dosage measured by the electronic scale at time t and the anticoagulant dosage corresponding to the rotation pulse number of the anticoagulant pump. The specific calculation method is as follows: ； Among them, the represents the deviation amount, the represents the anticoagulant dosage measured by the electronic scale at time t, and the represents the anticoagulant dosage.

4. The method for identifying unexpected outflow of anticoagulant for plasma platelet collection according to claim 1, characterized in that, The specific steps of step S2 include the following sub-steps: S201. Construct the physical feature matrix, including the anticoagulant dosage measured by the electronic scale and the deviation amount: ； Among them, the represents a physical characteristic matrix, and the represents the amount of anticoagulant measured by an electronic scale at a certain moment. The represents the deviation amount at a certain moment, and the represents a time point, that is, the total number of time points in time series data; S202. Process the feature matrix through a one-dimensional convolutional layer to obtain the convolutional layer output: , where the represents the convolutional layer output at time point t, the represents the bias term of the convolutional layer, the represents the width of the convolutional kernel, the represents the physical feature input at time point t-k, and the represents the weight in the convolutional kernel; S203. Process the output of the convolutional layer through the LSTM layer to obtain the output of the LSTM layer; S204. Use the fully connected layer to process the output of the LSTM layer to predict the flow rationality.

5. The method for identifying unexpected outflow of anticoagulant for plasma platelet collection according to claim 4, characterized in that, The specific content of step S204 is as follows: , where the represents the predicted value of the flow rationality at time point t, and the represents the weight matrix of the fully connected layer, and the represents the output of the LSTM layer at time point t, and the represents the bias term of the fully connected layer.

6. The method for identifying unexpected outflow of anticoagulant for plasma platelet collection according to claim 1, characterized in that, The specific steps of step S3 include the following sub-steps: S301. Construct the concentration and viscosity matrix, including the anticoagulant concentration and viscosity: ； Among them, the represents the concentration and viscosity matrix, the represents the anticoagulant concentration at a certain moment, the represents the anticoagulant viscosity at a certain moment, and the represents the time point; S302. Process the concentration and viscosity matrix through the fully connected layer to obtain the processed output of the concentration and viscosity; S303. Output the non-expected outflow risk prediction through the fully connected layer.

7. A method for identifying unexpected outflow of anticoagulant for plasma platelet collection according to claim 6, characterized in that, The specific content of step S303 is as follows: , where the represents the unexpected outflow prediction result at time point t, and the represents the weight matrix of the fully connected layer for the unexpected outflow prediction, and the represents the output after the concentration and viscosity at time point t pass through the ReLU activation function, and the represents the bias term of the fully connected layer for the unexpected outflow prediction.

8. The method for identifying unexpected outflow of anticoagulant for plasma platelet collection according to claim 1, characterized in that, In step S4, the specific calculation of the losses for flow rationality, non-expected outflow, and cumulative non-expected outflow includes: Calculate the loss of flow rationality: , where the represents the mean square error loss of flow rationality, and the represents the predicted flow rationality at time point t, and the represents the actual flow rationality at time point t; Calculate the loss for unexpected outflows: , where the represents the mean square error loss of unexpected outflows, and the represents the model's predicted unexpected outflows at time point t, and the represents the actual unexpected outflows at time point t; Calculate the loss for the cumulative unexpected outflow: , where the represents the mean squared error loss of the cumulative unexpected outflow, and the represents the actual unexpected outflow at time point t, and the represents the weight parameter of the regularization term.

9. The method for identifying unexpected outflow of anticoagulant for plasma platelet collection according to claim 8, characterized in that, In the step S4, the combined loss function is specifically defined as a weighted average of the flow rationality loss, the unexpected outflow loss, and the cumulative unexpected outflow loss: , where the represents the combined loss function, and the represents the weight for balancing the flow rationality loss, and the represents the weight for balancing the cumulative unexpected outflow loss.

10. The method for identifying unexpected outflow of anticoagulant for plasma platelet collection according to claim 1, characterized in that, In step S4, the gradient descent optimization algorithm is specifically to update the model parameters through gradient descent: ； ； Among them, the represents the weight matrix in the model, the represents the bias term in the model, the represents the learning rate, the represents the loss function is the partial derivative of the loss function with respect to the weights, the represents the loss function is the partial derivative of the loss function with respect to the bias.

Images