Pump blood flow determination method, apparatus, device, storage medium, and program product
By combining pump motor operation information, catheter usage time, and hemodynamic parameters, and using a neural network model to calculate pump flow rate, the problems of inaccurate calculation and insufficient robustness in existing technologies are solved, achieving more accurate flow monitoring.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- FENGKAI MEDICAL INSTR (SHANGHAI) CO LTD
- Filing Date
- 2024-12-30
- Publication Date
- 2026-06-30
AI Technical Summary
In the existing technology, the blood flow monitoring of ventricular assist devices is not accurate enough and lacks robustness because it only calculates the current of the blood pump motor. Furthermore, it fails to fully consider the wear of the blood pump catheter and the hemodynamic parameters of the target object.
By acquiring the operating information of the blood pumping motor, the usage time of the blood pumping catheter, and the hemodynamic parameters of the target object, a comprehensive calculation is performed using a neural network model to establish a blood pumping flow determination model, thereby improving the accuracy and robustness of the calculation.
This improved the accuracy of blood flow calculation, reduced the impact of single parameter errors on calculation results, and enhanced the stability and precision of the model.
Smart Images
Figure CN122297898A_ABST
Abstract
Description
Technical Field
[0001] This application belongs to the technical field of medical devices, and particularly relates to a method, apparatus, equipment, storage medium, and program product for determining blood flow rate. Background Technology
[0002] Ventricular assist devices are designed to assist the heart in pumping blood and improve its pumping capacity. To more accurately assist the heart in pumping blood, the pumping flow rate of ventricular assist devices is usually monitored.
[0003] Currently, the pumping motor of ventricular assist devices is usually located outside the body. That is, the pumping motor is outside the body, and the power generated by the pumping motor is transmitted to the impeller located inside the heart through the drive shaft in the pumping catheter, thereby controlling the rotation of the impeller to pump out blood.
[0004] When the blood pump motor is located outside the body, the blood flow rate is usually calculated using the current of the blood pump motor. However, the applicant found through monitoring the blood flow rate that the blood flow rate determined solely by the current of the blood pump motor is inaccurate and lacks robustness. Summary of the Invention
[0005] This application provides a method, apparatus, device, storage medium, and program product for determining blood flow rate, which can comprehensively determine blood flow rate by considering factors such as the usage time of the blood pumping catheter and the hemodynamic parameters of the target object, so as to improve the accuracy and robustness of blood flow rate determination.
[0006] On one hand, embodiments of this application provide a method for determining pumping blood flow rate, applied to a ventricular assist device, the ventricular assist device including a pumping motor and a pumping catheter connected to the pumping motor, the method comprising:
[0007] The system acquires the first operating information of the blood pumping motor, the first usage time of the blood pumping catheter, and the first hemodynamic parameter information of the target object; wherein, the target object includes the object on which the blood pumping motor acts.
[0008] The first operating information, the first usage duration, and the first hemodynamic parameter information are input into the pump blood flow determination model to obtain the pump blood flow output by the pump blood flow determination model.
[0009] Optionally, before inputting the first operating information, the first usage duration, and the first hemodynamic parameter information into the pump blood flow determination model to obtain the pump blood flow output by the pump blood flow determination model, the method further includes:
[0010] Multiple model training samples are obtained, including the first historical operating information of the blood pumping motor, the first historical usage time of the blood pumping catheter, and the first historical hemodynamic parameter information of the target object and the corresponding first historical blood pumping flow rate.
[0011] The first historical operation information, the first historical usage duration, and the first historical hemodynamic parameter information are used as inputs to a preset neural network model, and the first historical pumping blood flow rate is used as the output of the preset neural network model. The preset neural network model is trained to obtain a pumping blood flow rate determination model.
[0012] Optionally, the step of training the preset neural network model using the first historical operating information, the first historical usage duration, and the first historical hemodynamic parameter information as inputs to the preset neural network model, and using the first historical pumping flow rate as the output of the preset neural network model, to obtain a pumping flow rate determination model, includes:
[0013] The first historical operation information, the first historical usage duration, and the first historical hemodynamic parameter information are input into the preset neural network model to obtain the first prediction result of the pump blood flow rate;
[0014] Based on the first prediction result and the first historical blood flow rate, determine the loss function of the preset neural network model;
[0015] If the loss function does not meet the preset convergence condition, the model parameters of the preset neural network model are adjusted according to the backpropagation algorithm, and the first historical running information, the first historical usage time and the first historical hemodynamic parameter information are input into the preset neural network model to obtain the first prediction result of pumping blood flow, until the loss function meets the preset convergence condition.
[0016] Optionally, the preset neural network model includes a factorization machine, a fully connected neural network, and a gradient-enhanced decision tree network model; each of the factorization machine, fully connected neural network, and gradient-enhanced decision tree network model includes an input layer, an output layer, and at least one hidden layer.
[0017] The step of inputting the first historical operating information, the first historical usage duration, and the first historical hemodynamic parameter information into the preset neural network model to obtain the first predicted result of pump blood flow includes:
[0018] The first historical operation information, the first historical usage duration, and the first historical hemodynamic parameter information are respectively transformed into vectors to obtain the first input vector corresponding to the first historical operation information, the second input vector corresponding to the first historical usage duration, and the third input vector corresponding to the first historical hemodynamic parameter information.
[0019] The factorization machine extracts features from the first input vector, the second input vector, and the third input vector to obtain the first cross feature vector;
[0020] The fully connected neural network extracts features from the first input vector, the second input vector, and the third input vector to obtain the second cross feature vector.
[0021] The gradient-enhanced decision tree network model extracts features from the first input vector, the second input vector, and the third input vector to obtain the third cross feature vector;
[0022] The first cross feature vector, the second cross feature vector, and the third cross feature vector are combined and concatenated to obtain the first concatenated cross feature vector.
[0023] According to the preset forward propagation algorithm, the activation function is calculated on the first concatenated cross feature vector through at least one hidden layer to obtain the first prediction result;
[0024] The first prediction result is output through the output layer.
[0025] Optionally, after the factorization machine extracts features from the first input vector, the second input vector, and the third input vector to obtain the first cross feature vector, the method further includes:
[0026] The first cross feature vector is used as the input of the fully connected neural network and / or the gradient boosting decision tree network model, so that the fully connected neural network and / or the gradient boosting decision tree network model can extract features from the first cross feature vector to obtain the fourth cross feature vector extracted by the fully connected neural network and the fifth cross feature vector extracted by the gradient boosting decision tree network model.
[0027] The first cross feature vector, the fourth cross feature vector, and the fifth cross feature vector are combined and concatenated to obtain the second concatenated cross feature vector.
[0028] According to the preset forward propagation algorithm, the activation function is calculated on the second concatenated cross feature vector through at least one hidden layer to obtain the second prediction result;
[0029] The second prediction result is output through the output layer.
[0030] Optionally, before inputting the first operating information, the first usage duration, and the first hemodynamic parameter information into the pump blood flow determination model to obtain the pump blood flow output by the pump blood flow determination model, the method further includes:
[0031] The first operating information, the first usage duration, and the first hemodynamic parameter information are preprocessed to obtain the preprocessed second operating information, the second usage duration, and the second hemodynamic parameter information. The preprocessing includes normalization and / or standardization.
[0032] The step of inputting the first operating information, the first usage duration, and the first hemodynamic parameter information into the pump blood flow determination model to obtain the pump blood flow output by the pump blood flow determination model includes:
[0033] The second operating information, the second usage duration, and the second hemodynamic parameter information are input into the pump blood flow determination model to obtain the pump blood flow output by the pump blood flow determination model.
[0034] Optionally, after training the preset neural network model using the historical operating information, historical usage duration, and historical first hemodynamic parameter information as inputs and the historical pumping flow rate as the output of the preset neural network model to obtain a pumping flow rate determination model, the method further includes:
[0035] Obtain model test samples, which include the second historical operating information of the blood pumping motor, the second historical usage time of the blood pumping catheter, and the second historical hemodynamic parameter information of the target object and the corresponding second historical blood pumping flow rate.
[0036] The second historical operation information, the second historical usage duration, and the second historical hemodynamic parameter information are input into the pump blood flow determination model to obtain the third prediction result of the pump blood flow.
[0037] Based on the second historical blood flow rate and the third prediction result, the performance index of the blood flow rate determination model is determined;
[0038] If the performance indicators do not meet the preset conditions, the model parameters of the pump blood flow determination model are adjusted, and the second historical operation information, the second historical usage duration and the second historical hemodynamic parameter information are returned to the pump blood flow determination model to obtain the third prediction result of the pump blood flow, until the performance indicators meet the preset conditions.
[0039] On the other hand, embodiments of this application provide a blood flow determination device, the device comprising:
[0040] The acquisition module is used to acquire the first operating information of the blood pumping motor, the first usage time of the blood pumping catheter, and the first hemodynamic parameter information of the target object;
[0041] The input module is used to input the first operating information, the first usage duration, and the first hemodynamic parameter information into the pump blood flow determination model to obtain the pump blood flow output by the pump blood flow determination model.
[0042] In another aspect, embodiments of this application provide a ventricular assist device, the device comprising: a processor and a memory storing computer program instructions;
[0043] When the processor executes the computer program instructions, it implements the method for determining blood flow as described in the first aspect.
[0044] In another aspect, embodiments of this application provide a computer storage medium storing computer program instructions, which, when executed by a processor, implement the blood flow determination method as described in the first aspect.
[0045] In another aspect, embodiments of this application provide a computer program product in which instructions, when executed by the processor of a ventricular assist device, cause the ventricular assist device to perform the pumping flow determination method as described in the first aspect.
[0046] The pumping blood flow determination method, apparatus, device, storage medium, and program product of this application embodiment can input the acquired first operating information, first usage time, and first hemodynamic parameter information of the pumping motor into the pumping blood flow determination model. The pumping blood flow determination model performs mixed calculations on the first operating information, first usage time, and first hemodynamic parameter information to obtain a more accurate pumping blood flow. Compared with the prior art, which determines the pumping blood flow solely based on the pumping motor current, this application improves the accuracy of the pumping blood flow by using the pumping blood flow determination model to calculate the pumping blood flow using the first operating information, first usage time, and first hemodynamic parameter information. Moreover, determining the pumping blood flow from these three aspects—first operating information, first usage time, and first hemodynamic parameter information—reduces the possibility that any one of the acquired parameters might be incorrect, affecting the calculation of the pumping blood flow, thereby improving robustness. Attached Figure Description
[0047] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0048] Figure 1 This is a structural block diagram of a ventricular assist device provided in one embodiment of this application;
[0049] Figure 2This is a schematic diagram illustrating the connection relationship between the blood pumping catheter and the heart according to one embodiment of this application;
[0050] Figure 3 This is a flowchart illustrating the method for determining pump blood flow provided in the first embodiment of this application;
[0051] Figure 4 This is a flowchart illustrating the method for determining pump blood flow provided in the second embodiment of this application;
[0052] Figure 5 This is a flowchart illustrating the method for determining pump blood flow provided in the third embodiment of this application;
[0053] Figure 6 This is a schematic diagram of the structure of a preset neural network model provided in one embodiment of this application;
[0054] Figure 7 This is a flowchart illustrating the method for determining pump blood flow provided in the fourth embodiment of this application;
[0055] Figure 8 This is a flowchart illustrating the method for determining pump blood flow provided in the fifth embodiment of this application;
[0056] Figure 9 This is a schematic diagram of the structure of a blood flow determination device provided in one embodiment of this application;
[0057] Figure 10 This is a schematic diagram of the structure of a ventricular assist device provided in one embodiment of this application. Detailed Implementation
[0058] The features and exemplary embodiments of various aspects of this application will be described in detail below. To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only intended to explain this application and not to limit it. For those skilled in the art, this application can be implemented without some of these specific details. The following description of the embodiments is merely to provide a better understanding of this application by illustrating examples.
[0059] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes said element.
[0060] Before introducing the technical solutions of the embodiments of this application, we will first introduce the process by which the applicant discovered the technical problem:
[0061] Currently, when the pump motor of a ventricular assist device is running outside the body, the power generated by the pump motor is transmitted to the impeller located inside the heart through the drive shaft inside the pumping catheter, and then the impeller is controlled to rotate to pump blood out.
[0062] When the pumping motor is located outside the body, the pumping flow rate calculated solely from the pumping motor's current during actual use of the ventricular assist device is somewhat inaccurate and lacks robustness.
[0063] Through extensive experiments, the applicant discovered that not only does the current of the pump motor affect the pumping flow rate, but also, because the power is transmitted to the impeller via a drive shaft inside the pumping catheter, the degree of wear on the inner wall of the pumping catheter during power transmission interferes with the actual pumping flow rate. In other words, the more severe the wear on the pumping catheter, the lower the actual pumping flow rate under the same current. Since the pumping catheter is connected to the heart, its wear is difficult to quantify. During the use of ventricular assist devices, the wear of the pumping catheter is mainly related to the corrosion of the perfusion fluid. The applicant found that the corrosion of the perfusion fluid is related to the usage time of the ventricular assist device; therefore, the wear of the pumping catheter can be estimated by its usage time.
[0064] On the other hand, during the applicant's multiple experiments, it was found that the pumping blood flow rate is not only related to the wear and tear of the pumping catheter, but also to the hemodynamic parameters of the target object of the ventricular assist device. Since different target objects have different bodies, the resistance that blood encounters when flowing in the vascular system is different. Therefore, the difficulty for the heart to pump blood into the peripheral blood vessels is different. Correspondingly, the auxiliary power provided by the ventricular assist device to the heart is also different. Therefore, when calculating the pumping blood flow rate, it is also necessary to consider the hemodynamic parameters of the target object.
[0065] In summary, the applicant found that calculating the blood flow rate solely based on the current of the blood pump motor is not accurate enough and lacks robustness. After multiple experiments, it was discovered that the blood flow rate is also related to the usage time of the blood pump catheter and the hemodynamic parameters of the target object. Therefore, in order to calculate the blood flow rate more accurately, this application provides a method, apparatus, device, storage medium, and program product for determining the blood flow rate. The first operating information, first usage time, and first hemodynamic parameters of the pumping motor are input into the pumping flow determination model. The pumping flow determination model performs a mixed calculation on the first operating information, first usage time, and first hemodynamic parameters to obtain a more accurate pumping flow. Compared with the prior art, which determines the pumping flow solely based on the pumping motor current, this application improves the accuracy of the pumping flow by using the pumping flow determination model to calculate the pumping flow using the first operating information, first usage time, and first hemodynamic parameters. Furthermore, determining the pumping flow from these three aspects reduces the possibility that any error in any of the acquired parameters could affect the calculation of the pumping flow, thus improving robustness.
[0066] In this embodiment, the method for determining pump blood flow is applied to a ventricular assist device. Therefore, to facilitate understanding of the method for determining pump blood flow, the ventricular assist device will be introduced first:
[0067] Reference Figure 1 and Figure 2 Ventricular assist devices may include:
[0068] A blood pumping catheter assembly 1 is positioned between the heart 5 and the aorta 6 for blood flow;
[0069] The blood pump motor 2, connected to the blood pump conduit assembly 1, is used to provide blood pumping capability to the heart;
[0070] Control device 3, blood pump motor 2 is electrically connected to control device 3, and control device 3 is used to obtain the first operating information of blood pump motor 2.
[0071] In this embodiment, the control device 3 can also acquire the first hemodynamic parameter information of the target object input by the user, and can also communicate with other monitoring systems 5 to acquire the first hemodynamic parameter information of the target object.
[0072] The memory 4 is electrically connected to the control device 3 and is used to store the usage time of the blood pumping catheter, all first operating information during the operation of the blood pumping motor 2, and the first hemodynamic parameters of the target of the ventricular assist device.
[0073] As an example, the memory 4 can be integrated with the control device 3, or it can be a separate device.
[0074] Reference Figure 2 In some embodiments, taking the auxiliary pumping of blood to the left ventricle of the heart 5 as an example, the blood pumping catheter assembly 1 may include: an inlet tube 11 disposed in the left ventricle, one end of the inlet tube 11 being connected to an inhalation channel 12, and the end of the inhalation channel 12 away from the inlet tube 11 being connected to a transvalvular flexible tube 13, wherein the transvalvular flexible tube 13 passes through the aortic valve span 51 and is connected to an outflow channel 14 and an impeller (not shown in the figure), the end of the outflow channel 14 away from the transvalvular flexible tube 13 being connected to a catheter sheath 15, the catheter sheath 15 being disposed in the aorta 6, and the impeller being connected to the blood pumping motor 2 via a drive shaft (not shown in the figure).
[0075] In this embodiment, the control device 3 can acquire the first operating information of the blood pumping motor 2, the usage time of the blood pumping catheter stored in the memory 4, and the first hemodynamic parameters of the target object. Then, it determines the blood pumping flow rate by deploying a blood pumping flow rate determination model in the control device 3 to improve the accuracy and robustness of the blood pumping flow rate. The operation of the blood pumping motor is controlled by the blood pumping flow rate output by the blood pumping flow rate determination model to better assist the heart of the target object in pumping blood.
[0076] The method for determining the blood flow rate provided in the embodiments of this application is described below. The method for determining the blood flow rate is applied to the above-mentioned ventricular assist device.
[0077] Figure 3 A flowchart illustrating a method for determining pump blood flow rate according to an embodiment of this application is shown. Figure 3 As shown, the method for determining pump blood flow can include S301-S302:
[0078] S301, acquire the first operating information of the blood pump motor, the first usage time of the blood pump catheter, and the first hemodynamic parameter information of the target object.
[0079] In this embodiment, when determining the blood pumping flow rate, the control device can directly obtain the first operating information of the blood pumping motor, wherein the first operating information may include the operating current of the blood pumping motor, and the control device can obtain the first usage time of the blood pumping catheter through the memory.
[0080] In some embodiments, the control device can acquire first hemodynamic parameters of the target object with other monitoring systems. The first hemodynamic parameters may include peripheral vascular resistance parameters. It is worth noting that peripheral vascular resistance parameters can reflect the ease with which the heart pumps blood into peripheral blood vessels and are an important indicator for measuring the function of the cardiovascular system. They can be calculated using existing technologies. The control device can directly acquire peripheral vascular resistance parameters through other monitoring systems.
[0081] In this embodiment, the target object may include the object on which the blood pump motor operates, which may be a patient using a ventricular assist device.
[0082] S302, input the first operating information, the first usage duration and the first hemodynamic parameter information into the pump blood flow determination model to obtain the pump blood flow output by the pump blood flow determination model.
[0083] In some embodiments, first operating information, first usage duration, and first hemodynamic parameter information are input into the pump blood flow determination model. The pump blood flow determination model can perform mixed calculations internally based on the input first operating information, first usage duration, and first hemodynamic parameter information, and output the pump blood flow.
[0084] In this embodiment, the pump blood flow determination model can analyze the intrinsic relationship between the first operating information, the first usage time, and the first hemodynamic parameter information and the pump blood flow, so as to obtain a more accurate pump blood flow.
[0085] In this embodiment, the first operating information, first usage time, and first hemodynamic parameter information of the pumping motor are input into the pumping flow determination model. The pumping flow determination model performs a mixed calculation on the first operating information, first usage time, and first hemodynamic parameter information to obtain a more accurate pumping flow. Compared with the prior art, which determines the pumping flow solely based on the pumping motor current, this application improves the accuracy of the pumping flow by using the pumping flow determination model to calculate the pumping flow using the first operating information, first usage time, and first hemodynamic parameter information. Furthermore, determining the pumping flow from these three aspects reduces the possibility that any one of the acquired parameters might be incorrect, affecting the calculation of the pumping flow, thus improving robustness.
[0086] Reference Figure 4In some embodiments, the blood flow rate is determined by three aspects: the first operating information of the blood pumping motor, the first usage time of the blood pumping catheter, and the first hemodynamic parameters of the target object. However, after multiple experiments and data analysis by the applicant, the intrinsic relationship between the data can be automatically learned through a neural network model to obtain a more accurate blood flow rate. Therefore, in order to obtain a more accurate blood flow rate, a training process for the blood flow rate determination model can be included before S302, specifically including S401-S402:
[0087] S401, obtain multiple model training samples;
[0088] S402, the preset neural network model is trained by taking the first historical operation information, the first historical usage time and the first historical hemodynamic parameter information as inputs and the first historical pumping blood flow as output, so as to obtain the pumping blood flow determination model.
[0089] In some embodiments, in S401, as an example, the model training samples include the first historical operating information of the pumping motor, the first historical usage time of the pumping catheter, and the first historical hemodynamic parameter information of the target object and the corresponding first historical pumping flow rate.
[0090] In some embodiments, before training the blood flow determination model, in order to make the blood flow determination model more accurate in determining the blood flow, the first historical operating information of the blood pumping motor, the first historical usage time of the blood pumping catheter, the first historical hemodynamic information of the target object, and the corresponding first historical blood flow can be obtained through the memory. The first historical operating information, the first historical usage time, and the first historical hemodynamic information all correspond to the same first historical blood flow.
[0091] In some embodiments, in S402, a preset neural network model is iteratively trained using model training samples to obtain a pumping blood flow determination model. The trained pumping blood flow model can deeply analyze the potential relationship between the first historical operating information, the first usage time, and the first hemodynamic parameter information and the first historical pumping blood flow, so as to continuously train and optimize the preset neural network model, so that the trained preset neural network model can more accurately predict the pumping blood flow.
[0092] As an example, the preset neural network model can be a convolutional neural network model or a forward propagation neural network model. That is, the preset neural network model can be a basic neural network model that can learn autonomously, and there is no limitation here.
[0093] In this embodiment, by using the first historical operating information, the first historical usage duration, and the first historical hemodynamic parameters as inputs to a preset neural network model, and using the first historical pumping flow rate as the output of the preset neural network model, the preset neural network model is continuously trained. The preset neural network model automatically calculates the pumping flow rate based on the first historical operating information, the first historical usage duration, and the first historical hemodynamic parameters. Furthermore, it enables the pumping flow rate determination model to determine the pumping flow rate from different perspectives, making the calculated pumping flow rate more accurate and further improving the robustness of the pumping flow rate.
[0094] Reference Figure 5 In some other embodiments, S402 may specifically include:
[0095] S4021, Input the first historical operation information, the first historical usage time and the first historical hemodynamic parameter information into the preset neural network model to obtain the first prediction result of pump blood flow;
[0096] S4022, Based on the first prediction result and the first historical blood flow rate, determine the loss function of the preset neural network model;
[0097] S4023, if the loss function does not meet the preset convergence condition, adjust the model parameters of the preset neural network model according to the backpropagation algorithm, and return the first historical running information, the first historical usage time and the first historical hemodynamic parameter information to the preset neural network model to obtain the first prediction result of pumping blood flow until the loss function meets the preset convergence condition.
[0098] In some embodiments, when training a preset neural network model, the preset neural network model extracts features from the first historical operating information, the first historical usage time, and the first historical hemodynamic parameters based on its own learning ability, explores their intrinsic relationships, performs hybrid calculations to obtain the predicted pump blood flow, uses the predicted pump blood flow as the first prediction result, and then uses the first historical pump blood flow as the true value to determine the loss function of the preset neural network model.
[0099] If the loss function meets the preset convergence condition, stop training the preset neural network model.
[0100] In other embodiments, in S4023, the preset convergence condition may include at least one of the following: the trend of change of the loss function value gradually flattens and tends to stabilize, the parameter change rate is less than a preset threshold, and the preset maximum number of iterations is reached.
[0101] In other embodiments, if the loss function does not meet the preset convergence condition, the model parameters of the preset neural network model can be adjusted according to the backpropagation algorithm, that is, the weights and slice quality of the preset neural network model can be adjusted, and the weights can be updated by the gradient descent method to continuously optimize the preset neural network model and obtain the pump blood flow determination model.
[0102] Specifically, the weights can be updated according to the following formula (1):
[0103]
[0104] Among them, w ij Let η be the weight between neuron i and neuron j, η be the learning rate, and E be the loss function value.
[0105] In other embodiments, the preset neural network model may include a factorization machine, a fully connected neural network, and a gradient-boosting decision tree network model, wherein the factorization machine, the fully connected neural network, and the gradient-boosting decision tree network model all include an input layer, an output layer, and at least one hidden layer.
[0106] Reference Figure 6 As an example, the presupposed structure of a neural network is as follows: Figure 6 As shown, the input layer has 4 neuron computing nodes, there are two hidden layers, the hidden layer connected to the input layer has 5 neuron computing nodes, the hidden layer connected to the output layer has 3 neuron computing nodes, and the output layer has only one neuron computing node.
[0107] Reference Figure 7 To ensure that the pumping flow rate determined by the trained model is more accurate in outputting the pumping flow rate, S4021 may specifically include:
[0108] S40211, Perform vector transformation on the first historical operation information, the first historical usage duration and the first historical hemodynamic parameter information respectively to obtain the first input vector corresponding to the first historical operation information, the second input vector corresponding to the first historical usage duration and the third input vector corresponding to the first historical hemodynamic parameter information;
[0109] S40212, the factorization machine extracts features from the first input vector, the second input vector and the third input vector to obtain the first cross feature vector;
[0110] S40213, The fully connected neural network extracts features from the first input vector, the second input vector and the third input vector to obtain the second cross feature vector;
[0111] S40214, the gradient boosting decision tree network model extracts features from the first input vector, the second input vector and the third input vector to obtain the third cross feature vector;
[0112] S40215, The first cross feature vector, the second cross feature vector and the third cross feature vector are mixed and concatenated to obtain the first concatenated cross feature vector;
[0113] S40216, according to the preset forward propagation algorithm, the activation function is calculated on the first concatenated cross feature vector through at least one hidden layer to obtain the first prediction result;
[0114] S40217 outputs the first prediction result through the output layer.
[0115] In some embodiments, in S40211, when the model training samples are input into the preset neural network model, in order to facilitate the training of the preset neural network model, the first historical running information, the first historical usage duration, and the first historical hemodynamic parameter information can be vectorized to obtain the first input vector, the second input vector, and the third input vector. Then, the first input vector, the second input vector, and the third input vector are input into the input layer of the preset neural network to facilitate the training of the preset neural network model in the form of vectors.
[0116] In some embodiments, in S40212, the factor decomposition machine can analyze the implicit relationships of the data of the first input vector, the second input vector, and the third input vector, which can preserve low-order features and enhance the interaction of high-order features, so that the extracted first cross feature vector can reflect the cross characteristics between the first historical running information, the first usage duration, and the first hemodynamic parameters, and automatically learn the relationship between the first historical running information, the first usage duration, and the first hemodynamic parameters, so that the pumping flow rate can be determined from a single input after the pumping flow rate determination model is trained.
[0117] It is worth noting that the first cross feature vector extracted by the factorization machine is an explicit cross feature of first and second order.
[0118] In some embodiments, in S40213, the fully connected neural network model can extract implicit higher-order cross features of the first input vector, the second input vector, and the third input vector. These implicit higher-order cross features can also reflect the implicit relationships between the data, resulting in a second cross feature vector. The second cross feature vector reflects the implicit higher-order cross features between the first input vector, the second input vector, and the third input vector.
[0119] In some other embodiments, in S40214, the gradient boosting decision tree network model can extract explicit high-order cross features between the first input vector, the second input vector, and the third input vector, namely the third cross feature vector, wherein the third cross feature vector can reflect the explicit high-order cross features between the first input vector, the second input vector, and the third input vector.
[0120] In some embodiments, in S40215, in order to enable the trained pump blood flow determination model to analyze the cross relationship between the first input vector, the second input vector and the third input vector, the first cross feature vector, the second cross feature vector and the third cross feature vector can be mixed and concatenated to obtain the first concatenated cross feature vector, and the first concatenated cross feature vector is used as the output of the input layer.
[0121] It is worth noting that when concatenating the first, second, and third cross feature vectors, any two cross feature vectors can be combined, or the three cross feature vectors can be concatenated in different quantities and orders; no restrictions are imposed here.
[0122] In this embodiment, the first cross feature vector, the second cross feature vector, and the third cross feature vector are combined and concatenated. This allows the blood flow determination model to comprehensively consider the relationship between these three cross feature vectors during the training process, so as to obtain a more accurate blood flow rate.
[0123] In some embodiments, in S40216, the activation function of the first concatenated cross feature vector can be calculated using a preset forward propagation algorithm. Specifically, the output of each neuron's computation node can be calculated according to the following formula (2):
[0124]
[0125] Where y is the output of the neuron's computation node, w i x represents the weights between neuron i and that neuron. i Let b be the input of neuron i to neuron i, b be the bias of neuron i, and f() be the activation function.
[0126] It is worth noting that the output of the first neuron computing node is used as the input of the next neuron computing node, thus calculating each output layer by layer to finally obtain the first prediction result.
[0127] In some embodiments, in S40217, since the output layer has only one neuron computing node, the predicted pump blood flow calculated by the neuron computing node can be determined as the first prediction result.
[0128] In this embodiment, a fully connected neural network is used to extract implicit high-order cross features from the first, second, and third input vectors. A factorization machine is used to extract explicit first- and second-order cross features from the first, second, and third input vectors. A gradient-enhanced decision tree network model is used to extract explicit high-order cross features from the first, second, and third input vectors. This helps the blood flow determination model to fully learn the low- and high-order feature information hidden behind the data, such as the operating information of the blood pump motor, the usage time of the blood pump catheter, and the hemodynamic parameters of the target object. This improves the blood flow determination model's ability to represent nonlinear classification scenarios, thereby enabling more accurate calculation of blood flow.
[0129] Reference Figure 8 In some other embodiments, to accelerate the training process of the blood flow determination model, after S40212, the method may further include:
[0130] S801, the first cross feature vector is used as the input of the fully connected neural network and / or gradient boosting decision tree network model, so that the fully connected neural network and / or gradient boosting decision tree network model can extract features from the first cross feature vector to obtain the fourth cross feature vector extracted by the fully connected neural network and the fifth cross feature vector extracted by the gradient boosting decision tree network model.
[0131] S802, the first cross feature vector, the fourth cross feature vector and the fifth cross feature vector are mixed and concatenated to obtain the second concatenated cross feature vector;
[0132] S803, according to the preset forward propagation algorithm, the activation function is calculated on the second concatenated cross feature vector through at least one hidden layer to obtain the second prediction result;
[0133] S804 outputs the second prediction result through the output layer.
[0134] In this embodiment, after the factorization machine extracts the first and second order explicit cross features, i.e. the first cross feature vector, the first cross feature vector can be used as the embedding vector of the input layer of the pump blood flow. Then, the first input vector, the second input vector, the third input vector and the first cross feature vector are respectively input into the fully connected neural network and the gradient boosting decision tree network model to obtain the fourth cross feature vector and the fifth cross feature vector.
[0135] In this embodiment, by inputting the first cross feature vector into the fully connected neural network and the gradient boosting decision tree network respectively, the fully connected neural network and the gradient boosting decision tree network can learn to write the output results of the factorization machine more quickly, thereby accelerating the training speed of the blood flow determination model.
[0136] In this set of embodiments, the implementation process of S802-S804 is the same as that of S40215-S40217, and will not be described in detail here.
[0137] In some other embodiments, prior to S301, the method may further include:
[0138] The first operating information, the first usage duration, and the first hemodynamic parameter information are preprocessed to obtain the preprocessed second operating information, the second usage duration, and the second hemodynamic parameter information.
[0139] In this embodiment, S301 may specifically include:
[0140] The second operating information, the second usage duration, and the second hemodynamic parameter information are input into the pump blood flow determination model to obtain the pump blood flow output by the pump blood flow determination model.
[0141] As an example, preprocessing may include normalization and / or standardization; by normalizing and / or standardizing the first operating information, the first usage duration, and the first hemodynamic parameter information, the influence of different feature data units and orders of magnitude can be eliminated, accelerating the optimization speed of the pump flow determination model and improving the stability and accuracy of the model.
[0142] In some other embodiments, in order to ensure the stability of the blood flow determination model, it is also necessary to test the performance of the blood flow determination model. Therefore, after S402, the method may further include:
[0143] Obtain model test samples, which include the second historical operating information of the blood pumping motor, the second historical usage time of the blood pumping catheter, and the second historical hemodynamic parameters of the target object and the corresponding second historical blood pumping flow rate.
[0144] The second historical operation information, the second historical usage duration, and the second historical hemodynamic parameter information are input into the pump blood flow determination model to obtain the third prediction result of the pump blood flow.
[0145] Based on the second historical blood flow rate and the third prediction results, the performance indicators of the blood flow rate determination model are determined.
[0146] If the performance indicators do not meet the preset conditions, adjust the model parameters of the pump blood flow determination model, and return to input the second historical operation information, the second historical usage time and the second historical hemodynamic parameter information into the pump blood flow determination model to obtain the third prediction result of the pump blood flow, until the performance indicators meet the preset conditions.
[0147] In this embodiment, the test method for determining the pump blood flow rate model and the above-mentioned Figure 5 The training method shown is the same, the difference being that the input data is the model test sample, which will not be elaborated here.
[0148] It is also worth noting that the model test samples were preprocessed, and the model test samples can be used to test the generalization ability of the trained blood flow determination model to unseen data, and to verify the performance of the blood flow determination model in real-world scenarios.
[0149] Based on the method for determining blood flow rate provided in the above embodiments, this application also provides a specific implementation of the device for determining blood flow rate.
[0150] Please refer to the following examples.
[0151] See Figure 9 The blood flow determination device 900 provided in this application embodiment may include:
[0152] The acquisition module 901 is used to acquire the first operating information of the blood pumping motor, the first usage time of the blood pumping catheter, and the first hemodynamic parameter information of the target object; wherein, the target object includes the object on which the blood pumping motor acts.
[0153] The input module 902 is used to input the first operating information, the first usage duration and the first hemodynamic parameter information into the pump blood flow determination model to obtain the pump blood flow output by the pump blood flow determination model.
[0154] As an optional implementation, the input module 902 may include:
[0155] The acquisition unit is used to acquire multiple model training samples, including the first historical operating information of the blood pumping motor, the first historical usage time of the blood pumping catheter, and the first historical hemodynamic parameter information of the target object and the corresponding first historical blood pumping flow rate.
[0156] The training unit is used to train the preset neural network model by taking the first historical operation information, the first historical usage time and the first historical hemodynamic parameters as inputs and the first historical pumping flow rate as the output of the preset neural network model, so as to obtain the pumping flow rate determination model.
[0157] As an optional implementation, the training unit is also specifically used for:
[0158] The first historical operation information, the first historical usage time, and the first historical hemodynamic parameters are input into the preset neural network model to obtain the first prediction result of the pump blood flow.
[0159] Based on the first prediction result and the first historical blood flow rate, determine the loss function of the preset neural network model;
[0160] If the loss function does not meet the preset convergence condition, the model parameters of the preset neural network model are adjusted according to the backpropagation algorithm, and the first historical running information, the first historical usage time and the first historical hemodynamic parameter information are input into the preset neural network model to obtain the first prediction result of the pumping blood flow until the loss function meets the preset convergence condition.
[0161] The preset neural network models include factorization machines, fully connected neural networks, and gradient-boosting decision tree network models; each of the factorization machine, fully connected neural network, and gradient-boosting decision tree network models includes an input layer, an output layer, and at least one hidden layer; as an optional implementation, the training unit can also be used for:
[0162] Vector transformations are performed on the first historical operation information, the first historical usage duration, and the first historical hemodynamic parameter information to obtain the first input vector corresponding to the first historical operation information, the second input vector corresponding to the first historical usage duration, and the third input vector corresponding to the first historical hemodynamic parameter information.
[0163] The factorization machine extracts features from the first input vector, the second input vector, and the third input vector to obtain the first cross feature vector;
[0164] The fully connected neural network extracts features from the first input vector, the second input vector, and the third input vector to obtain the second cross feature vector;
[0165] The gradient-enhanced decision tree network model extracts features from the first input vector, the second input vector, and the third input vector to obtain the third cross feature vector.
[0166] The first cross feature vector, the second cross feature vector, and the third cross feature vector are combined and concatenated to obtain the first concatenated cross feature vector.
[0167] According to the preset forward propagation algorithm, the activation function is calculated on the first concatenated cross feature vector through at least one hidden layer to obtain the first prediction result;
[0168] The first prediction result is output through the output layer.
[0169] As an optional implementation, the training unit can also be used for:
[0170] The first cross feature vector is used as input to the fully connected neural network and / or gradient boosting decision tree network model, so that the fully connected neural network and / or gradient boosting decision tree network model can extract features from the first cross feature vector to obtain the fourth cross feature vector extracted by the fully connected neural network and the fifth cross feature vector extracted by the gradient boosting decision tree network model.
[0171] The first, fourth, and fifth cross feature vectors are combined and concatenated to obtain the second concatenated cross feature vector.
[0172] The second prediction result is obtained by performing activation function calculation on the second concatenated cross feature vector through at least one hidden layer according to the preset forward propagation algorithm.
[0173] The second prediction result is output through the output layer.
[0174] As an optional implementation, the input module 902 can also be used for:
[0175] The first operational information, the first usage duration, and the first hemodynamic parameter information are preprocessed to obtain the preprocessed second operational information, the second usage duration, and the second hemodynamic parameter information. The preprocessing includes normalization and / or standardization.
[0176] The second operating information, the second usage duration, and the second hemodynamic parameter information are input into the pump blood flow determination model to obtain the pump blood flow output by the pump blood flow determination model.
[0177] As an optional implementation, the input module 902 can also be used for:
[0178] Obtain model test samples, which include the second historical operating information of the blood pumping motor, the second historical usage time of the blood pumping catheter, and the second historical hemodynamic parameters of the target object and the corresponding second historical blood pumping flow rate.
[0179] The second historical operation information, the second historical usage duration, and the second historical hemodynamic parameter information are input into the pump blood flow determination model to obtain the third prediction result of the pump blood flow.
[0180] Based on the second historical blood flow rate and the third prediction results, the performance indicators of the blood flow rate determination model are determined.
[0181] If the performance indicators do not meet the preset conditions, adjust the model parameters of the pump blood flow determination model, and return to input the second historical operation information, the second historical usage time and the second historical hemodynamic parameter information into the pump blood flow determination model to obtain the third prediction result of the pump blood flow, until the performance indicators meet the preset conditions.
[0182] Figure 10 A schematic diagram of the hardware structure of the ventricular assist device provided in an embodiment of this application is shown.
[0183] The ventricular assist device may include a processor 1001 and a memory 1002 storing computer program instructions.
[0184] Specifically, the processor 1001 may include a central processing unit (CPU), an application specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of this application.
[0185] Memory 1002 may include mass storage for data or instructions. For example, and not limitingly, memory 1002 may include a hard disk drive (HDD), floppy disk drive, flash memory, optical disk, magneto-optical disk, magnetic tape, or Universal Serial Bus (USB) drive, or a combination of two or more of these. In one instance, memory 1002 may include removable or non-removable (or fixed) media, or memory 1002 may be non-volatile solid-state memory. Memory 1002 may be internal or external to the integrated gateway disaster recovery device.
[0186] In one instance, memory 1002 may be read-only memory (ROM). In one instance, the ROM may be a mask-programmed ROM, a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), an electrically rewritable ROM (EAROM), or flash memory, or a combination of two or more of these.
[0187] The memory 1002 may include read-only memory (ROM), random access memory (RAM), disk storage media device, optical storage media device, flash memory device, electrical, optical, or other physical / tangible memory storage device. Therefore, typically, the memory includes one or more tangible (non-transitory) computer-readable storage media (e.g., memory devices) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it is operable to perform the operations described with reference to the blood flow determination method according to the first aspect of this disclosure.
[0188] The processor 1001 reads and executes computer program instructions stored in the memory 1002 to achieve... Figure 3 A method for determining pump blood flow rate in the illustrated embodiment.
[0189] In one example, the ventricular assist device may also include a communication interface 1003 and a bus 1004. Wherein, as... Figure 10As shown, the processor 1001, memory 1002, and communication interface 1003 are connected through bus 1004 and complete communication with each other.
[0190] The communication interface 1003 is mainly used to realize communication between various modules, devices, units and / or equipment in the embodiments of this application.
[0191] Bus 1004 includes hardware, software, or both, that couples components of a ventricular assist device together. For example, and not limitingly, the bus may include an Accelerated Graphics Port (AGP) or other graphics bus, an Extended Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a Hyper Transport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an Infinite Bandwidth Interconnect, a Low Pin Count (LPC) bus, a memory bus, a Microchannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local (VLB) bus, or other suitable buses, or combinations of two or more of these. Where appropriate, bus 1004 may include one or more buses. Although specific buses are described and illustrated in embodiments of this application, any suitable bus or interconnect is contemplated herein.
[0192] This ventricular assist device can execute the pumping flow determination method in the embodiments of this application, thereby achieving a combination of Figures 3-9 The method and apparatus for determining pump blood flow are described.
[0193] Furthermore, in conjunction with the blood flow determination method in the above embodiments, this application embodiment can provide a computer storage medium for implementation. This computer storage medium stores computer program instructions; when these computer program instructions are executed by a processor, they implement any of the blood flow determination methods in the above embodiments.
[0194] In an optional embodiment, in conjunction with the blood flow determination method in the above embodiments, this application embodiment can provide a computer program product to implement it. The instructions in the computer program product are executed by the processor of the ventricular assist device, enabling the ventricular assist device to implement any of the blood flow determination methods in the above embodiments.
[0195] It should be clarified that this application is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of this application is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of this application.
[0196] The functional blocks shown in the above block diagram can be implemented as hardware, software, firmware, or a combination thereof. When implemented in hardware, they can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this application are programs or code segments used to perform the required tasks. Programs or code segments can be stored on a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried on a carrier wave. "Machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, fiber optic media, radio frequency (RF) links, etc. Code segments can be downloaded via computer networks such as the Internet, intranets, etc.
[0197] It should also be noted that the exemplary embodiments mentioned in this application describe methods or systems based on a series of steps or apparatus. However, this application is not limited to the order of the above steps; that is, the steps can be performed in the order mentioned in the embodiments, or in a different order, or several steps can be performed simultaneously.
[0198] The aspects of this disclosure have been described above with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It should be understood that each block in the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that these instructions, executable via the processor of the computer or other programmable data processing apparatus, enable the implementation of the functions / actions specified in one or more blocks of the flowchart illustrations and / or block diagrams. Such a processor can be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor, or a field-programmable logic circuit. It is also understood that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can also be implemented by special-purpose hardware performing the specified functions or actions, or can be implemented by a combination of special-purpose hardware and computer instructions.
[0199] The above are merely specific embodiments of this application. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, modules, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. It should be understood that the protection scope of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the protection scope of this application.
Claims
1. A method for determining pump blood flow rate, characterized in that, The method, applied to a ventricular assist device, the ventricular assist device comprising a pumping motor and a pumping catheter connected to the pumping motor, includes: The system acquires the first operating information of the blood pumping motor, the first usage time of the blood pumping catheter, and the first hemodynamic parameter information of the target object; wherein, the target object includes the object on which the blood pumping motor acts. The first operating information, the first usage duration, and the first hemodynamic parameter information are input into the pump blood flow determination model to obtain the pump blood flow output by the pump blood flow determination model.
2. The method according to claim 1, characterized in that, Before inputting the first operating information, the first usage duration, and the first hemodynamic parameter information into the pump blood flow determination model to obtain the pump blood flow output by the pump blood flow determination model, the method further includes: Multiple model training samples are obtained, including the first historical operating information of the blood pumping motor, the first historical usage time of the blood pumping catheter, and the first historical hemodynamic parameter information of the target object and the corresponding first historical blood pumping flow rate. The first historical operation information, the first historical usage duration, and the first historical hemodynamic parameter information are used as inputs to a preset neural network model, and the first historical pumping blood flow rate is used as the output of the preset neural network model. The preset neural network model is trained to obtain a pumping blood flow rate determination model.
3. The method according to claim 2, characterized in that, The step of training the preset neural network model by using the first historical operating information, the first historical usage duration, and the first historical hemodynamic parameters as inputs and the first historical pumping flow rate as the output of the preset neural network model to obtain a pumping flow rate determination model includes: The first historical operation information, the first historical usage duration, and the first historical hemodynamic parameter information are input into the preset neural network model to obtain the first prediction result of the pump blood flow rate; Based on the first prediction result and the first historical blood flow rate, determine the loss function of the preset neural network model; If the loss function does not meet the preset convergence condition, the model parameters of the preset neural network model are adjusted according to the backpropagation algorithm, and the first historical running information, the first historical usage time and the first historical hemodynamic parameter information are input into the preset neural network model to obtain the first prediction result of pumping blood flow, until the loss function meets the preset convergence condition.
4. The method according to claim 3, characterized in that, The preset neural network model includes a factorization machine, a fully connected neural network, and a gradient-enhanced decision tree network model; each of the factorization machine, fully connected neural network, and gradient-enhanced decision tree network model includes an input layer, an output layer, and at least one hidden layer. The step of inputting the first historical operating information, the first historical usage duration, and the first historical hemodynamic parameter information into the preset neural network model to obtain the first predicted result of pump blood flow includes: The first historical operation information, the first historical usage duration, and the first historical hemodynamic parameter information are respectively transformed into vectors to obtain the first input vector corresponding to the first historical operation information, the second input vector corresponding to the first historical usage duration, and the third input vector corresponding to the first historical hemodynamic parameter information. The factorization machine extracts features from the first input vector, the second input vector, and the third input vector to obtain the first cross feature vector; The fully connected neural network extracts features from the first input vector, the second input vector, and the third input vector to obtain the second cross feature vector. The gradient-enhanced decision tree network model extracts features from the first input vector, the second input vector, and the third input vector to obtain the third cross feature vector; The first cross feature vector, the second cross feature vector, and the third cross feature vector are combined and concatenated to obtain the first concatenated cross feature vector. According to the preset forward propagation algorithm, the activation function is calculated on the first concatenated cross feature vector through at least one hidden layer to obtain the first prediction result; The first prediction result is output through the output layer.
5. The method according to claim 4, characterized in that, After the factorization machine extracts features from the first input vector, the second input vector, and the third input vector to obtain the first cross feature vector, the method further includes: The first cross feature vector is used as the input of the fully connected neural network and / or the gradient boosting decision tree network model, so that the fully connected neural network and / or the gradient boosting decision tree network model can extract features from the first cross feature vector to obtain the fourth cross feature vector extracted by the fully connected neural network and the fifth cross feature vector extracted by the gradient boosting decision tree network model. The first cross feature vector, the fourth cross feature vector, and the fifth cross feature vector are combined and concatenated to obtain the second concatenated cross feature vector. According to the preset forward propagation algorithm, the activation function is calculated on the second concatenated cross feature vector through at least one hidden layer to obtain the second prediction result; The second prediction result is output through the output layer.
6. The method according to claim 1, characterized in that, Before inputting the first operating information, the first usage duration, and the first hemodynamic parameter information into the pump blood flow determination model to obtain the pump blood flow output by the pump blood flow determination model, the method further includes: The first operating information, the first usage duration, and the first hemodynamic parameter information are preprocessed to obtain the preprocessed second operating information, the second usage duration, and the second hemodynamic parameter information. The preprocessing includes normalization and / or standardization. The step of inputting the first operating information, the first usage duration, and the first hemodynamic parameter information into the pump blood flow determination model to obtain the pump blood flow output by the pump blood flow determination model includes: The second operating information, the second usage duration, and the second hemodynamic parameter information are input into the pump blood flow determination model to obtain the pump blood flow output by the pump blood flow determination model.
7. The method according to claim 2, characterized in that, After training the preset neural network model using the historical operating information, historical usage duration, and historical first hemodynamic parameter information as inputs and the historical pumping flow rate as the output to obtain a pumping flow rate determination model, the method further includes: Obtain model test samples, which include the second historical operating information of the blood pumping motor, the second historical usage time of the blood pumping catheter, and the second historical hemodynamic parameter information of the target object and the corresponding second historical blood pumping flow rate. The second historical operation information, the second historical usage duration, and the second historical hemodynamic parameter information are input into the pump blood flow determination model to obtain the third prediction result of the pump blood flow. Based on the second historical blood flow rate and the third prediction result, the performance index of the blood flow rate determination model is determined; If the performance indicators do not meet the preset conditions, the model parameters of the pump blood flow determination model are adjusted, and the second historical operation information, the second historical usage duration and the second historical hemodynamic parameter information are returned to the pump blood flow determination model to obtain the third prediction result of the pump blood flow, until the performance indicators meet the preset conditions.
8. A device for determining pump blood flow rate, characterized in that, The device includes: The acquisition module is used to acquire the first operating information of the blood pumping motor, the first usage time of the blood pumping catheter, and the first hemodynamic parameter information of the target object; The input module is used to input the first operating information, the first usage duration and the first hemodynamic parameter information into the pump blood flow determination model to obtain the pump blood flow output by the pump blood flow determination model.
9. A ventricular assist device, characterized in that, The device includes: a processor and a memory storing computer program instructions; When the processor executes the computer program instructions, it implements the method for determining blood flow as described in any one of claims 1-7.
10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer program instructions, which, when executed by a processor, implement the method for determining blood flow rate as described in any one of claims 1-7.
11. A computer program product, characterized in that, When the instructions in the computer program product are executed by the processor of the ventricular assist device, the ventricular assist device performs the pumping flow determination method as described in any one of claims 1-7.