CN116999689A

CN116999689A - Flow determination method, training method, device and medium of flow detection model

Info

Publication number: CN116999689A
Application number: CN202310694582.7A
Authority: CN
Inventors: 王越
Original assignee: Xinqing Medical Suzhou Co ltd
Current assignee: Xinqing Medical Suzhou Co ltd
Priority date: 2023-06-13
Filing date: 2023-06-13
Publication date: 2023-11-07

Abstract

The application relates to a flow determination method, a training method of a flow detection model, equipment and a medium, belonging to the technical field of medical appliances, wherein the method comprises the following steps: acquiring perfusion data and blood pressure data corresponding to the catheter ventricular assist device and motor operation data corresponding to the catheter ventricular assist device; and carrying out flow detection processing according to the perfusion data, the blood pressure data and the motor operation data to obtain flow data corresponding to the catheter ventricular assist device. The technical scheme provided by the application greatly reduces the dependence of flow detection on the flow sensor, effectively solves the technical problem that the pumping flow in the heart through the catheter ventricular assist device is difficult to detect, reduces the complexity of flow detection, and improves the efficiency of flow detection.

Description

Flow determination method, training method, device and medium of flow detection model

Technical Field

The application relates to a flow determination method, a training method of a flow detection model, equipment and a medium, and belongs to the technical field of medical appliances.

Background

A transcatheter ventricular assist device is used to mechanically assist the patient. When the catheter ventricular assist device works, blood in the heart chamber can be pumped to the artery, so that the ventricular assist function is realized. During operation of the transcatheter ventricular assist device, it is often necessary to measure the current pumped blood flow to determine the operational state of the transcatheter ventricular assist device.

However, the pump head of the interventional pump in the transcatheter ventricular assist device is required to be percutaneously placed into the heart through the peripheral blood vessel, and thus the pump head is small in volume, does not have a sufficient space for providing a flow sensor, and is difficult to measure the blood flow.

Disclosure of Invention

The application provides a flow determination method, a training method of a flow detection model, equipment and a medium, which solve the technical problem that the flow of blood pumped by a catheter ventricular assist device is difficult to monitor. The application provides the following technical scheme:

in one aspect, a flow determination method is provided, the method being applied to a transcatheter ventricular assist device, the method comprising:

acquiring perfusion data and blood pressure data corresponding to the transcatheter ventricular assist device and motor operation data corresponding to the transcatheter ventricular assist device;

and carrying out flow detection processing according to the perfusion data, the blood pressure data and the motor operation data to obtain flow data corresponding to the transcatheter ventricular assist device.

Optionally, the perfusion data comprises at least one perfusion parameter data, the at least one perfusion parameter data comprises at least one of perfusion pressure data, perfusion flow rate data, and perfusion pump rotational speed data, the blood pressure data comprises at least one blood pressure parameter data, the at least one blood pressure parameter data comprises at least one of arterial pressure data, ventricular pressure data, and pumping differential pressure data;

Optionally, the performing flow detection according to the perfusion data, the blood pressure data and the motor operation data to obtain flow data corresponding to the transcatheter ventricular assist device includes:

and carrying out flow detection processing according to the perfusion pressure data, the arterial pressure data and the motor operation data to obtain the flow data.

Optionally, the motor operation data includes motor rotation speed data and motor current data, and the flow detection processing is performed according to the perfusion pressure data, the arterial pressure data and the motor operation data to obtain the flow data, including:

and carrying out flow detection processing according to the perfusion pressure data, the arterial pressure data, the motor rotating speed data and the motor current data to obtain the flow data.

Optionally, the transcatheter ventricular assist device comprises:

an interventional catheter pump comprising a catheter, a pump head connected to a distal end of the catheter, a coupling assembly connected to a proximal end of the catheter, the pump head being deliverable by the catheter to a desired location of the heart for a pumping operation;

a perfusion channel penetrating at least the catheter, an inlet of the perfusion channel being disposed on the coupling assembly, an outlet of the perfusion channel being disposed at the pump head; the perfusion fluid in the perfusion channel enters a desired position in the cardiovascular system through the outlet; the desired location in the cardiovascular system includes at least one of the aorta, pulmonary artery, left ventricle, right ventricle, left atrium, right atrium;

The perfusate pressure data characterizes a sum of perfusate pressures generated by the perfusate at a pressure acquisition location to the outlet.

inputting the perfusion data, the blood pressure data and the motor operation data into a preset flow detection model to perform flow detection processing, and outputting the flow data;

the flow detection model is a machine learning model obtained after training based on a plurality of groups of sample data, and each group of sample data comprises perfusion sample data, blood pressure sample data, motor operation sample data and sample flow data which correspond to each other.

Optionally, inputting the perfusion data, the blood pressure data and the motor operation data into a preset flow detection model for flow detection processing, and outputting the flow data, including:

inputting perfusion pressure data, arterial pressure data, motor rotating speed data and motor current data into a preset Gaussian model for flow detection processing, and outputting the flow data;

The perfusion data comprise perfusion pressure data, the blood pressure data comprise arterial pressure data, the motor operation data comprise motor rotating speed data and motor current data, the preset flow detection model comprises the preset Gaussian model, and the preset Gaussian model is a Gaussian model obtained after training based on multiple groups of sample data.

Optionally, before inputting the perfusion data, the blood pressure data and the motor operation data into a preset flow detection model to perform flow detection processing, the method further includes:

determining a first flow detection model corresponding to the transcatheter ventricular assist device based on a data type of the perfusion data and/or a data type of the blood pressure data; the data type of the perfusion data is used for indicating the type of perfusion parameter data included in the perfusion data; the data type of the blood pressure data is used for indicating the type of blood pressure parameter data included in the blood pressure data.

Correspondingly, the inputting the perfusion data, the blood pressure data and the motor operation data into a preset flow detection model for flow detection processing, and outputting the flow data comprises the following steps:

And inputting the perfusion data, the blood pressure data and the motor operation data into the first flow detection model, and outputting the flow data.

The preset flow detection model comprises flow detection models corresponding to different data types, the data type of the perfusion sample data in each group of sample data corresponding to the first flow detection model is consistent with the data type of the perfusion data currently acquired, and the data type of the blood pressure sample data is consistent with the data type of the perfusion data currently acquired.

acquiring a first acquisition position of the perfusion data and a second acquisition position of the blood pressure data;

determining a second flow detection model based on the first acquisition location and the second acquisition location;

the preset flow detection model comprises flow detection models corresponding to different acquisition positions, and each group of sample data corresponding to the second flow detection model comprises: perfusion sample data acquired based on the first acquisition location, blood pressure sample data acquired based on the second acquisition location, and motor operation sample data.

Correspondingly, inputting the perfusion data, the blood pressure data and the motor operation data into a preset flow detection model for flow detection processing, and outputting the flow data, wherein the flow detection processing comprises the following steps:

and inputting the perfusion data, the blood pressure data and the motor operation data into the second flow detection model to perform flow detection processing, and outputting the flow data.

In another aspect, a method for training a flow detection model is provided, the method comprising:

acquiring a sample data set acquired in the operation process of a transcatheter ventricular assist device, wherein each group of sample data in the sample data set comprises perfusion sample data, blood pressure sample data, motor operation sample data and corresponding flow measurement data corresponding to the transcatheter ventricular assist device;

model training is carried out on a preset machine learning model according to the perfusion sample data, the blood pressure sample data, the motor operation sample data and the flow measurement data to obtain a flow detection model;

the flow detection model is used for detecting flow data corresponding to the transcatheter ventricular assist device according to perfusion data, blood pressure data and motor operation data generated in the application process of the transcatheter ventricular assist device.

Optionally, the sample data set includes a plurality of sets of training data and a plurality of sets of test data, and model training is performed on a preset machine learning model according to the perfusion sample data, the blood pressure sample data, the motor operation sample data and the flow measurement data to obtain a flow detection model, including:

modeling processing is carried out based on perfusion sample data, blood pressure sample data, motor operation sample data and flow measurement data in the plurality of groups of training data, so as to generate a first machine learning model;

inputting perfusion sample data, blood pressure sample data and motor operation sample data in each set of test data into the first machine learning model for flow detection processing, and outputting flow detection data corresponding to each set of test data;

comparing the flow measurement data in each group of test data with the flow detection data corresponding to each group of test data, and determining loss information corresponding to the first machine learning model;

and adjusting model parameters of the first machine learning model based on the loss information to obtain the flow detection model.

Optionally, the first machine learning model includes a gaussian model, the comparing the flow measurement data in each set of test data with the flow detection data corresponding to each set of test data, and determining loss information corresponding to the first machine learning model includes:

Determining average flow data according to the flow measurement data in each set of sample data;

and comparing the average flow data, the flow detection data corresponding to each group of test data and the flow measurement data in each group of test data to obtain variance data and root mean square error data corresponding to the Gaussian model, wherein the loss information comprises the variance data and the root mean square error data.

Optionally, the flow measurement data is determined by the amount of change in weight of the liquid per unit time.

Optionally, the perfusion sample data in the sample data set includes perfusion pressure sample data corresponding to the transcatheter ventricular assist device, the blood pressure sample data in the sample data set includes arterial pressure sample data corresponding to the transcatheter ventricular assist device, and the motor operation sample data in the sample data set includes motor speed sample data and motor current sample data corresponding to the transcatheter ventricular assist device;

correspondingly, the flow detection model is specifically used for detecting the flow data according to the perfusion pressure data, the arterial pressure data, the motor rotating speed data and the motor current data generated in the application process of the transcatheter ventricular assist device.

In another aspect, a flow determination device is provided for use with a transcatheter ventricular assist device, the device comprising:

the data acquisition module is used for acquiring perfusion data and blood pressure data corresponding to the transcatheter ventricular assist device and motor operation data corresponding to the transcatheter ventricular assist device;

and the flow determining module is used for carrying out flow detection processing according to the perfusion data, the blood pressure data and the motor operation data to obtain flow data corresponding to the transcatheter ventricular assist device.

In another aspect, a training apparatus for a flow detection model is provided, the apparatus comprising:

the data acquisition module is used for acquiring a sample data set acquired in the operation process of the catheter ventricular assist device, and each group of sample data in the sample data set comprises perfusion sample data, blood pressure sample data, motor operation sample data and corresponding flow measurement data corresponding to the catheter ventricular assist device;

the model training module is used for carrying out model training on a preset machine learning model according to the perfusion sample data, the blood pressure sample data, the motor operation sample data and the flow measurement data to obtain a flow detection model;

In another aspect, a transcatheter ventricular assist device is provided, the transcatheter ventricular assist device comprising: a computer device comprising a processor and a memory; the memory stores a program, and the program is loaded and executed by the processor to realize the flow determination method provided in the above aspect; or, the training method of the flow detection model provided by the aspect is realized.

In another aspect, a computer device is provided, the device comprising a processor and a memory; the memory stores a program, and the program is loaded and executed by the processor to realize the flow determination method provided in the above aspect; or, the training method of the flow detection model provided by the aspect is realized.

In another aspect, there is provided a computer-readable storage medium having stored therein a program for implementing the flow rate determination method provided in the above aspect when executed by a processor; or, the training method of the flow detection model provided by the aspect is realized.

In another aspect, a computer program product includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reading the computer instructions from a computer-readable storage medium, the processor executing the computer instructions such that the computer device executes to implement the flow determination method provided in the above aspect; or, the training method of the flow detection model provided by the aspect is realized.

According to the flow determination method provided by the embodiment of the application, the pumping flow of the catheter ventricular assist device can be detected according to the perfusion data, the blood pressure data and the motor operation data corresponding to the catheter ventricular assist device, so that the dependence of flow detection on a flow sensor is greatly reduced, the technical problem that the pumping flow of the catheter ventricular assist device in the heart is difficult to detect is effectively solved, the complexity of flow detection is reduced, and the efficiency of flow detection is improved.

In addition, since the perfusion state of the perfusion fluid, the pressure state of the blood, the operation state of the motor and the flow pumped through the catheter ventricular assist device are all related, the flow data is determined based on the perfusion pressure data representing the perfusion state of the perfusion fluid, the arterial pressure data representing the pressure state of the blood and the motor operation data representing the operation state of the motor, so that the accuracy of flow detection can be ensured.

In addition, by training the machine learning model as the flow detection model, the flow detection model can learn the data association relation among various parameters from the distribution of multiple groups of sample data, and the accuracy of the flow detection result can be ensured by detecting the flow data through the flow detection model.

The foregoing description is only an overview of the present application, and is intended to provide a better understanding of the present application, as it is embodied in the following description, with reference to the preferred embodiments of the present application and the accompanying drawings.

Drawings

FIG. 1 is a block diagram of a transcatheter ventricular assist device according to one embodiment of the present application;

FIG. 2 is a schematic illustration of the connection of a perfusion assembly to a working assembly according to one embodiment of the present application;

FIG. 3 is a flow chart of a flow determination method provided by one embodiment of the present application;

FIG. 4 is a flow chart of a method for training a flow detection model provided by one embodiment of the present application;

FIG. 5 is a block diagram of a flow determination device provided by one embodiment of the present application;

FIG. 6 is a block diagram of a training apparatus for a flow detection model provided by one embodiment of the present application;

FIG. 7 is a block diagram of a computer device provided in one embodiment of the application.

Detailed Description

The following describes in further detail the embodiments of the present application with reference to the drawings and examples. The following examples are illustrative of the application and are not intended to limit the scope of the application.

In the conventional flow rate determination method, flow rate data is generally detected by installing a flow rate sensor in the blood pump. However, for transcatheter ventricular assist devices, the volume of the pump head is typically relatively small because the blood pump in the device is an interventional catheter pump, and the pump head in an interventional catheter pump is required to be implanted in the body during use. In this case, it is difficult to implement a flow sensor installed in the pump head to detect flow data.

Based on the above, the application provides a flow determination method, which can perform flow detection processing according to the perfusion data, the blood pressure data and the motor operation data corresponding to the through-catheter ventricular assist device, so as to detect the flow data pumped through the through-catheter ventricular assist device, and can detect the flow without arranging a flow sensor in the through-catheter ventricular assist device.

In the application, before the description of the flow determination method, the application scene of the flow determination method is described.

The flow determination method provided by the application is applied to a transcatheter ventricular assist device, and the transcatheter ventricular assist device is used for carrying out mechanical circulation assistance on a patient. In one illustrative scenario, a transcatheter ventricular assist device may be used as left ventricular assist, and the device may be operated to pump blood in the left ventricle into the ascending aorta. In other possible and not explicitly excluded scenarios, the transcatheter ventricular assist device may also be used as a right ventricular assist, with the device operating to pump blood from the vein into the right ventricle. Alternatively, the transcatheter ventricular assist device may be adapted for use in pumping blood from the vena cava and/or the right atrium to the right ventricle, from the vena cava and/or the right atrium to the pulmonary artery and/or from the renal vein to the vena cava, and the application is not limited in its context of use. In the following embodiments, a transcatheter ventricular assist device is described for pumping blood in a ventricle into an aorta.

Referring to the block diagram of the transcatheter ventricular assist device 100 of one embodiment of the present application shown in fig. 1, the device 100 includes at least an interventional catheter pump 110.

The pump head of the interventional catheter pump 110 can be percutaneously placed into the heart through a peripheral blood vessel, for example, the pump head can be placed between the left ventricle and the ascending aorta, the blood inlet of the pump head can be placed into the left ventricle, and the blood outlet of the pump head can be placed into the ascending aorta, so that blood is pumped into the ascending aorta from the left ventricle, and a ventricular assist function is realized.

Alternatively, the interventional catheter pump 110 comprises a pump head 111, a coupler and a driver, the driver being detachably connected to the coupler, and the impeller in the pump head 111 being controllable to rotate via the coupler, the catheter after the driver is connected to the coupler.

The driver comprises a driving motor, an impeller is arranged in the pump head 111, and the impeller is in transmission connection with the driving motor so as to drive the impeller to rotate by controlling the motor to rotate, thereby realizing the driving of the pump head 111 to work. Pump head 111 may also include other components required for operation, such as: the pump head 111 further includes a pump housing to house the impeller, and the components included in the pump head 111 are not described here.

In the interventional catheter pump 110, the pump head 111 is in driving connection with a drive motor in the drive via a catheter.

The transcatheter ventricular assist device 100 further includes a perfusion assembly 120, the perfusion assembly 120 being configured to provide a perfusion fluid to the interventional catheter pump 110, which may be a physiological fluid required to maintain bodily functions, such as saline, dextrose solution, anticoagulant, or any combination thereof.

Optionally, the perfusion fluid is used to perfuse the interventional catheter pump through the catheter to exclude air from the catheter gap and avoid air from entering the blood vessel and heart through the catheter, thereby causing adverse physiological reactions. The perfusion fluid may flow through the perfusion channel and into a desired location in the cardiovascular system through an opening at the pump head. Optionally, the desired location in the cardiovascular system comprises at least one of the aorta, pulmonary artery, left ventricle, right ventricle, left atrium, right atrium.

In one example, as shown in fig. 2, the infusion assembly 120 includes an infusion fluid infusion port 121 through which infusion fluid is infused into an interventional catheter pump through the infusion fluid infusion port 121 and into the human body through the catheter 210. The perfusate injection interface 121 may be an inlet of a perfusate channel.

Optionally, the transcatheter ventricular assist device 100 further comprises a peristaltic pump for pumping the perfusate. Illustratively, the perfusate tubing may be clamped to the peristaltic pump, and the present embodiment does not limit the placement of the peristaltic pump.

In an embodiment of the present application, in order to realize the detection of the flow data of the interventional catheter pump, the transcatheter ventricular assist device may further include a first detection unit 130, a second detection unit 140, a third detection unit 150, and a control unit 160 connected to the first detection unit 130, the second detection unit 140, and the third detection unit 150, respectively.

A first detection unit 130 for detecting perfusion data corresponding to the transcatheter ventricular assist device;

a second detecting unit 140 for detecting blood pressure data corresponding to the transcatheter ventricular assist device;

a third detecting unit 150 for detecting motor operation data corresponding to the transcatheter ventricular assist device;

A control unit 160, configured to obtain perfusion data and blood pressure data corresponding to the catheter ventricular assist device, and motor operation data corresponding to the catheter ventricular assist device; and carrying out flow detection processing according to the perfusion data, the blood pressure data and the motor operation data to obtain flow data corresponding to the catheter ventricular assist device.

According to the technical scheme provided by the embodiment of the application, the pumping flow of the catheter ventricular assist device can be detected according to the perfusion data, the blood pressure data and the motor operation data corresponding to the catheter ventricular assist device, so that the dependence of flow detection on a flow sensor is greatly reduced, the technical problem that the pumping flow of the catheter ventricular assist device in the heart is difficult to detect is effectively solved, the complexity of flow detection is reduced, and the efficiency of flow detection is improved.

Optionally, the control unit 160 performs flow detection processing according to the perfusion data, the blood pressure data and the motor operation data to obtain flow data corresponding to the transcatheter ventricular assist device, including: and inputting the perfusion data, the blood pressure data and the motor operation data into a preset flow detection model to perform flow detection processing, and outputting flow data.

The flow detection model in the control unit 160 may be obtained based on training of the control unit 160, or may be configured to the control unit 160 after training in other devices is completed, which is not limited by the training environment of the flow detection model in this embodiment.

Taking the example of obtaining the flow detection model based on the training of the control unit 160, the control unit 160 is further configured to: acquiring a sample data set acquired in the operation process of the catheter ventricular assist device, wherein each group of sample data in the sample data set comprises perfusion sample data, blood pressure sample data, motor operation sample data and corresponding flow measurement data corresponding to the catheter ventricular assist device; and carrying out model training on a preset machine learning model according to the perfusion sample data, the blood pressure sample data, the motor operation sample data and the flow measurement data to obtain a flow detection model.

The flow detection model is used for detecting flow data corresponding to the transcatheter ventricular assist device according to perfusion data, blood pressure data and motor operation data generated in the process of applying the transcatheter ventricular assist device.

The sample data set may be collected by the first detection unit, the second detection unit and the third detection unit, or may be sent by another device, which is not limited by the manner in which the sample data set is obtained in this embodiment.

In actual implementation, if the training process of the flow detection model is implemented in other devices, the steps are the same as those of the training process, and this embodiment will not be described in detail herein.

Alternatively, the control unit 160 may be implemented as a separate device from the transcatheter ventricular assist device, and the implementation of the control unit 160 is not limited in this embodiment.

Illustratively, the control unit 160 includes at least a processor and a memory, with a program stored in the memory, the program being loaded and executed by the processor to implement the flow determination method of the present application; or, the training method of the flow detection model in the application is realized.

In actual practice, the transcatheter ventricular assist device may also include other components required for operation, such as: the power supply assembly, the display screen, etc., are not described in this embodiment for components included in the transcatheter ventricular assist device.

The following describes a flow determination method according to an embodiment of the present application, where the method is applied to the transcatheter ventricular assist device shown in fig. 1, and the execution subject of each step of the method may be a control unit in the transcatheter ventricular assist device, or may be other apparatus communicatively connected to the transcatheter ventricular assist device, such as: computers, tablet computers, etc., the present embodiment does not limit the device types of other devices. Fig. 3 is a flow chart of a flow determination method according to an embodiment of the present application, the method at least includes the following steps:

Step 301, acquiring perfusion data, blood pressure data and motor operation data corresponding to the catheter ventricular assist device.

The perfusion data is used to characterize the perfusion status of the perfusate in the perfusion assembly. In transcatheter ventricular assist devices, the perfusate enters the heart through the catheter interior space, and the perfusion flow channel (which is formed by the perfusion tubing and the catheter) communicates with the blood vessel or the heart interior, i.e., with the blood flow channel, i.e., there is a correlation between the flow of perfusate and the flow of blood. Thus, perfusion data may be acquired to determine flow data.

Blood pressure data is used to characterize the pressure state of blood in the body. There is a strong correlation between blood flow and blood pressure. Thus, blood pressure data at a preset location can be acquired when determining blood flow.

The motor operating data is used to characterize the corresponding operating energy efficiency of the motor in the transcatheter ventricular assist device. The type of fluid (e.g., viscosity) that is pumped through the catheter ventricular assist device affects the energy efficiency of the motor operation at a set rotational speed. At a set rotational speed, the amount of power consumed by the motor to pump different types of liquids is also not different. It can be seen that there is a correlation between the motor operation data and the type of fluid pumped by the blood pump, and there is a correlation between the flow rate and the type of fluid, and the motor is also a power output source for driving the impeller to rotate, so that the motor operation data can be obtained to determine the flow rate data.

Optionally, the perfusion data includes at least one perfusion parameter data. Optionally, the at least one perfusion parameter data includes, but is not limited to, at least one of:

1. perfusion pressure data. Optionally, the first detection unit 130 comprises a perfusion hydraulic pressure sensor. Alternatively, the perfusion hydraulic pressure sensor may be disposed at a designated position on the perfusion channel, and the present application does not limit the disposed position of the perfusion hydraulic pressure sensor.

In an exemplary embodiment, a transcatheter ventricular assist device comprises:

an interventional catheter pump includes a catheter, a pump head connected to a distal end of the catheter, a coupling assembly connected to a proximal end of the catheter, the pump head being deliverable by the catheter to a desired location of the heart for a pumping operation.

The pouring channel at least penetrates through the catheter, an inlet of the pouring channel is arranged on the coupling assembly, and an outlet of the pouring channel is arranged at the pump head; perfusate in the perfusion channel enters a desired location in the cardiovascular system through the outlet; the desired location in the cardiovascular system includes at least one of the aorta, pulmonary artery, left ventricle, right ventricle, left atrium, right atrium.

The perfusate pressure data characterizes the sum of perfusate pressures generated by the perfusate at the pressure collection site to the outlet. That is, the perfusion pressure data is the sum of all line pressures and outlet pressures between the position of the pressure sensor (i.e., the pressure acquisition position) to the outlet.

2. Perfusion flow data. Optionally, the first detection unit 130 comprises a perfusate flow sensor. Alternatively, the perfusion flow sensor may be disposed in a perfusion line of the perfusion liquid, and the present application is not limited to the disposition position of the flow sensor. In addition, the perfusate flow rate can also be determined from the volumetric changes of the perfusate.

3. Perfusion pump rotational speed data. The perfusion pump is used for driving perfusion liquid to flow, and the rotation speed data of the perfusion pump is used for representing the rotation speed of the perfusion pump. The infusion pump may be a peristaltic pump or other type of pump (e.g., an embolic pump, etc.). Optionally, the first detection unit 130 includes a rotational speed sensor corresponding to the perfusion pump, and the rotational speed sensor is used for detecting rotational speed data of the rotational speed pump (such as a peristaltic pump).

The blood pressure data may be detected by a blood pressure sensor fitting in the transcatheter ventricular assist device, may be transmitted by other means, or may be manually entered.

Optionally, the second detection unit 140 comprises a blood pressure sensor that collects blood pressure data. Optionally, the blood pressure data comprises at least one blood pressure parameter data including, but not limited to, at least one of:

1. Arterial pressure data. Optionally, arterial pressure data includes, but is not limited to, aortic pressure data, pulmonary arterial pressure data. Since the interventional catheter pump in the transcatheter ventricular assist device is operative to pump blood from the ventricle to the artery, such as from the left ventricle to the aorta and from the right ventricle to the pulmonary artery. Thus, the aortic or pulmonary pressure data and the pumped flow are correlated, and the arterial pressure data may be selected to determine the flow.

In one possible embodiment, an arterial pressure sensor is provided on the catheter for detecting arterial blood pressure. Arterial pressure sensors are introduced into the body following percutaneous intervention of the pump head, and arterial blood pressure, such as aortic pressure data or pulmonary pressure data, is measured after reaching a specified location.

In another possible embodiment, an arterial pressure sensor is provided in the extracorporeal circuit in communication with the artery, also for detecting arterial blood pressure. The embodiment of the application does not limit the setting position of the arterial pressure sensor.

2. Ventricular pressure data. Optionally, the ventricular pressure data includes, but is not limited to, left ventricular pressure data and right ventricular pressure data. For reasons similar to arterial pressure data, left or right ventricular pressure data correlates to the pumped flow, and therefore the above ventricular pressure data may be selected to determine flow. The ventricular pressure sensor is similar to the arterial pressure sensor and is provided in a different location than the arterial pressure sensor, and the embodiment of the present application is not limited thereto.

3. Pumping differential pressure data between arterial pressure data and ventricular pressure data. The pumping differential pressure data represents the pressure difference between the arterial pressure and the ventricular pressure, namely the differential pressure corresponding to the interventional catheter pump.

Optionally, the motor operating data includes, but is not limited to: motor speed data and motor current data. The motor speed data characterizes a motor speed of the drive motor in the catheter ventricular assist device and the motor current data characterizes a motor current of the drive motor in the catheter ventricular assist device. Alternatively, the drive motor is a brushless dc motor. The motor current data and the motor rotating speed data can have a time sequence corresponding relation, and the motor current data can reflect the power consumption of the motor under the corresponding motor rotating speed data, so that the running energy efficiency of the motor under the current liquid type is reflected.

In the embodiment of the application, the acquired perfusion data, blood pressure data and motor operation data correspond to each other, and based on the perfusion data, the blood pressure data and the motor operation data are acquired synchronously, or the maximum acquisition time interval between each data in a group of data is smaller than a threshold value.

And step 302, performing flow detection processing according to the perfusion data, the blood pressure data and the motor operation data to obtain flow data corresponding to the catheter ventricular assist device.

The above-mentioned perfusion data, blood pressure data and motor operation data have an association relationship with the flow rate, respectively, and in a possible implementation manner, the corresponding flow rate data is determined according to a data quantization relationship between the perfusion data, the blood pressure data and the motor operation data.

In another possible embodiment, the association relationship between the above several data may be learned based on a machine learning model, so as to determine the flow rate.

Optionally, the step 302 includes: and inputting the perfusion data, the blood pressure data and the motor operation data into a preset flow detection model to perform flow detection processing, and outputting flow data.

The flow detection model is a machine learning model obtained after training based on a plurality of groups of sample data, and each group of sample data comprises perfusion sample data, blood pressure sample data, motor operation sample data and sample flow data which correspond to each other. The specific training process of the flow detection model is described in the following embodiments, which are not described herein.

Since the motor operation data can represent the operation energy efficiency of the motor, and the motor operation energy efficiency is influenced by the liquid type of the pumped liquid, such as viscosity, the flow rate is directly related to the motor operation energy efficiency, and the correlation factor is complex, besides the change of the liquid type (such as viscosity) can cause the flow rate change, the operation state of the motor can also cause the flow rate change.

In particular, for motors, which pump liquids of different liquid types at a set rotational speed, the electrical energy consumed by the motor may also be different. For example, the motor consumes more power to pump a liquid with a higher viscosity than the motor consumes to pump a liquid with a lower viscosity. Thus, motor operation data may be used to determine flow data.

Moreover, in the case of a defined blood flow path, and where the blood flow path is in communication with the perfusion fluid flow path, the type of fluid and the pressure of the blood simultaneously affect the magnitude of the pumping flow of the blood pump, and the perfusion fluid flow is also affected by the change in blood flow. Therefore, through inputting the motor operation data, the perfusion data and the blood pressure data which are related to the liquid type into the trained flow detection model at the same time, the current flow data can be determined according to the association relationship among the learned flow, the motor operation data, the perfusion data and the blood pressure data which are extracted by the flow detection model during training under the condition that a flow sensor is not used, the dependence of flow detection on the flow sensor is greatly reduced, the flow detection complexity is reduced, and the flow detection efficiency is improved.

In addition, because the flow detection model is directly trained based on the perfusion sample data, the blood pressure sample data, the motor operation sample data and the flow measurement data, the flow detection model can directly and effectively learn the association relationship among the flow, the motor operation data, the perfusion data and the blood pressure data, namely the flow detection model is independent of the determination of the liquid type (such as the determination of the liquid viscosity). Compared with the manner that the viscosity needs to be calculated and the flow is calculated again in the Bernoulli equation, in the technical scheme provided by the embodiment of the application, the manner that the flow detection is performed through the flow detection model does not depend on the liquid type, such as the liquid viscosity, and particularly, it can be understood that no liquid type information is required to be input into the flow detection model, and no intermediate step of calculating or determining the liquid type, such as the intermediate step of calculating the viscosity, exists in the whole flow detection processing process.

Besides the motor operation data and the blood pressure data, the special perfusion data of the auxiliary device through the catheter ventricle is introduced, and the flow detection is carried out from more dimensions, so that the accuracy of the flow detection of the catheter pump can be ensured, the specific blood parameters are not required to be determined actually, the calculation steps are effectively reduced, the calculation complexity of detecting the flow data is reduced, the detection speed is effectively improved, and the device is more suitable for a control host with limited calculation force in the auxiliary device through the catheter ventricle.

In some embodiments, the sample dataset is not subject to constraints on the liquid type. For example, the sample data set does not set a viscosity range constraint condition on the liquid viscosity, and the liquid viscosity corresponding to each set of sample data in the sample data set is not limited by the preset viscosity range. Since there are data samples corresponding to various liquid types in the sample data set, the flow detection model trained based on these data is a flow detection model common to each liquid type and is an independent single model.

In one possible implementation, the flow detection model may be a trained neural network model. When the neural network model is trained, the flow measurement data in the sample data can be used as the supervision information of the neural network model, so that the accuracy of the output flow of the trained neural network model is restrained, the neural network model learns the characteristic association between the input data (perfusion sample data, blood pressure sample data and motor operation sample data) and the output data (flow measurement data), and the corresponding flow data is accurately predicted when new data is input.

In another possible embodiment, the flow detection model may be a trained gaussian model. And determining a mean vector and a covariance matrix corresponding to each group of sample data according to the perfusion sample data, the blood pressure sample data, the motor operation sample data and the corresponding flow measurement data in each group of sample data during Gaussian model training. The covariance matrix can effectively represent the association relation among various variables, so that when the model is applied, the Gaussian model can effectively predict flow data corresponding to the new input data according to the historical data distribution of the sample data for the input of the new data. Moreover, compared with a neural network model with huge parameter quantity, the Gaussian model has relatively small calculation quantity.

Illustratively, to improve the accuracy of flow detection, the data type of the perfusion sample data is consistent with the data type of the perfusion data; the data type of the blood pressure sample data is consistent with the data type of the blood pressure data; the motor operation sample data is consistent with the data type of the motor operation data.

In summary, according to the flow determination method provided by the embodiment, the pumping flow of the catheter ventricular assist device can be detected according to the perfusion data, the blood pressure data and the motor operation data corresponding to the catheter ventricular assist device, so that the dependence of flow detection on a flow sensor is greatly reduced, the technical problem that the pumping flow of the catheter ventricular assist device in the heart is difficult to detect is effectively solved, the complexity of flow detection is reduced, and the efficiency of flow detection is improved.

Moreover, since the perfusion state of the perfusion fluid, the pressure state of the blood and the running state of the motor can all influence or reflect the pumping flow of the interventional catheter pump, the accuracy of flow detection can be ensured by determining the flow data based on the perfusion data representing the liquid state of the perfusion fluid, the blood pressure data representing the pressure state of the blood and the running data of the motor representing the running state of the motor.

In addition, the flow data is detected through a flow detection model, the flow detection model is a machine learning model which is obtained based on sample data training, and the machine learning model can learn the association relation between more accurate data, so that the accuracy of a flow detection result can be ensured.

Optionally, based on the above embodiment, the perfusion data includes at least one perfusion parameter data and the blood pressure data includes at least one blood pressure parameter data, and the explanation is given below taking an example that the at least one perfusion parameter data includes perfusion pressure data and the at least one blood pressure parameter data includes arterial pressure data.

In this case, in step 302, flow detection processing is performed according to the perfusion data, the blood pressure data, and the motor operation data, so as to obtain flow data corresponding to the transcatheter ventricular assist device, including: and carrying out flow detection processing according to the perfusion pressure data, the arterial pressure data and the motor operation data to obtain flow data.

In one possible implementation, the catheter of the transcatheter ventricular assist device is not provided with a pressure sensor, so that ventricular pressure data is difficult to acquire, but arterial pressure data can be acquired through in vitro detection, in this case, flow detection can be performed only according to arterial pressure data, perfusion pressure data and motor operation data which are easy to detect, and flow of the interventional catheter pump can be accurately output, so that the difficulty of flow detection is reduced.

The arterial pressure data may be aortic pressure (left ventricular assist), pulmonary arterial pressure data (right ventricular assist), or arterial pressure data at other locations. The embodiment of the present application is not limited thereto.

Wherein the motor operating data includes, but is not limited to, motor speed data and motor current data. Under the condition that the motor operation data comprises motor rotation speed data and motor current data, carrying out flow detection processing according to perfusion pressure data, arterial pressure data and motor operation data to obtain flow data, wherein the flow data comprises the following steps: and carrying out flow detection processing according to the perfusion pressure data, the arterial pressure data, the motor rotating speed data and the motor current data to obtain flow data.

The motor current data and the motor rotating speed data can have a time sequence corresponding relation, and the motor current data can reflect the power consumption of the motor under the corresponding motor rotating speed data, so that the running energy efficiency of the motor under the current liquid type is reflected. Therefore, specific motor current data and motor rotation speed data are used as motor operation data, the operation energy efficiency of the current driving motor can be effectively represented, and the accuracy of flow detection is ensured.

Correspondingly, if the flow detection processing mode is processing based on a preset flow detection model, the perfusion data, the blood pressure data and the motor operation data are input into the preset flow detection model to be subjected to flow detection processing, and the output flow data comprise: and inputting the perfusion pressure data, the arterial pressure data, the motor rotating speed data and the motor current data into a preset Gaussian model for flow detection processing, and outputting flow data.

Optionally, the preset flow detection model includes a preset gaussian model, where the preset gaussian model is a gaussian model obtained after training based on a plurality of sets of sample data, and each set of sample data includes perfusate pressure sample data, arterial pressure sample data, motor rotation speed sample data, and motor current sample data.

For the Gaussian model, the mean vector and the covariance matrix corresponding to each group of sample data are determined according to the perfusion pressure sample data, the arterial pressure sample data, the motor rotating speed sample data, the motor current sample data and the flow measurement data in each group of sample data during training. The covariance matrix can effectively represent the association relation among various variables, so that when the model is applied, the Gaussian model can effectively predict flow data corresponding to the new input data according to the historical data distribution of the sample data for the input of the new data. Compared with a neural network model with huge parameter quantity, the Gaussian model has relatively small calculated quantity and higher accuracy, and each parameter data in the training sample sets also accords with Gaussian distribution, so that the accuracy of flow detection is effectively improved.

In other embodiments, the preset flow detection model may also include other mathematical models such as a neural network model, and the type of the flow detection model is not limited in this embodiment. Since the sample data of the gaussian model is subject to normal distribution and has statistical significance, the present embodiment is described by taking the case that the preset flow detection model includes a preset gaussian model as an example.

In other embodiments, the motor operating data may also include motor current data, but not motor speed data; at this time, the sample data also includes motor current sample data, but does not include motor rotation speed sample data. Alternatively, the motor operating data may include motor speed data, but not motor current data; at this time, the sample data includes motor rotation speed sample data, but does not include motor current sample data, and the implementation of the motor operation data is not limited in this embodiment.

Optionally, the at least one perfusion parameter data may include perfusion fluid flow data and/or perfusion pump rotation speed data, in addition to perfusion fluid pressure data, where the flow detection processing needs to be combined with the perfusion fluid flow data and/or perfusion pump rotation speed data.

Optionally, the at least one blood pressure parameter data may include ventricular pressure data and/or pressure difference data, in addition to arterial pressure data, where the flow detection process is performed in combination with ventricular pressure data and/or pressure difference data.

Since the perfusion pressure data and the arterial pressure data have a larger influence on the pumping flow of the interventional catheter pump, in this embodiment, the accuracy of flow detection can be ensured by setting at least one perfusion parameter data including the perfusion pressure data and at least one blood pressure parameter data including the arterial pressure data.

In other embodiments, the at least one perfusion parameter data may include perfusion flow data and/or perfusion pump rotational speed data, but not perfusion pressure data; and/or the at least one blood pressure parameter data may include ventricular pressure data and/or pressure differential data, and not arterial pressure data, the present embodiment is not limited to implementations of perfusion parameter data and blood pressure parameter data.

Alternatively, based on the above embodiment, since the perfusion data may be a combination of at least two perfusion parameter data, and the flow detection models corresponding to different combinations need to be trained using the sample data of the corresponding combinations, there may be different flow detection models corresponding to different combinations. Similarly, there may be differences in the flow detection model corresponding to different combinations of at least two blood pressure parameter data of the blood pressure data. Based on this, before step 302, further includes:

Such as: the data types of the currently acquired perfusion data are as follows: the perfusion data comprises perfusion pressure data and perfusion flow data, and the data types of the blood pressure data are as follows: the blood pressure data includes arterial pressure data and ventricular pressure data, and the perfusion sample data in the sample data corresponding to the first flow detection model includes perfusion pressure sample data and perfusion flow sample data, and the blood pressure sample data in the sample data includes arterial pressure sample data and ventricular pressure sample data.

And, for example: the data types of the currently acquired perfusion data are as follows: the perfusion data comprises perfusion pressure data, and the data types of the blood pressure data are as follows: the blood pressure data includes arterial pressure data, and the perfusion sample data in the sample data corresponding to the first flow detection model includes perfusion pressure sample data, and the blood pressure sample data in the sample data includes arterial pressure sample data.

Alternatively, the data type may be received through a human-machine interaction interface; alternatively, the data amounts of the perfusion data and the blood pressure data of different data types may be different, and at this time, the data type may be determined based on the data amounts, and the method of acquiring the data type is not limited in this embodiment.

In this embodiment, by setting different flow detection models for perfusion data of different data types and/or blood pressure data of different data types, it may be determined that the adaptive first flow detection model performs flow detection processing based on the currently acquired data type, so that accuracy of flow detection may be improved.

Alternatively, based on the above embodiment, since there may be a plurality of acquisition positions for each perfusion parameter data and a plurality of acquisition positions for each blood pressure parameter data, the data acquired at different acquisition positions may represent the same data, but the specific values acquired correspondingly may be different. Based on this, the data collected at different collection locations may correspond to different flow detection models.

At this time, before step 302, further includes: acquiring a first acquisition position of the perfusion data and a second acquisition position of the blood pressure data; a second flow detection model is determined based on the first acquisition location and the second acquisition location.

In this embodiment, by setting flow detection models corresponding to different acquisition positions, an adapted second flow detection model can be determined to perform flow detection processing based on the first acquisition position of the currently acquired perfusion data and the second acquisition position of the blood pressure data, so that accuracy of flow detection can be improved.

Based on the above embodiment, the preset flow detection model is obtained by model training a preset machine learning model through a sample data set. A specific training procedure is described below.

FIG. 4 is a flowchart of a method for training a flow detection model according to an embodiment of the present application, the method at least includes the following steps:

step 401, a sample dataset acquired during operation of a transcatheter ventricular assist device is acquired.

Optionally, each set of sample data in the sample data set includes perfusion sample data, blood pressure sample data, motor operation sample data, and corresponding flow measurement data corresponding to the transcatheter ventricular assist device.

The perfusion sample data is used to characterize the perfusion status of the perfusate in the perfusion assembly at the time the sample data set was acquired. In transcatheter ventricular assist devices, the perfusate enters the heart through the catheter interior space, and the perfusion flow channel (which is formed by the perfusion tubing and the catheter) communicates with the blood vessel or the heart interior, i.e., with the blood flow channel, i.e., there is a correlation between the flow of perfusate and the flow of blood. Thus, determining flow data based on perfusion data may be achieved by learning a correlation between perfusion sample data and blood flow.

Optionally, the perfusate sample data includes, but is not limited to: at least one of the perfusate pressure data, the perfusate flow rate data and the perfusate pump rotation speed data, and the relevant description of each type of perfusate sample data is detailed in the description of the perfusate data, and the embodiment is not repeated here.

The blood pressure sample data is used to characterize the pressure state of blood passing through the operating environment of the catheter ventricular assist device at the time the sample data set is acquired. There is a strong correlation between blood flow and blood pressure. Thus, determining flow data based on blood pressure sample data may be achieved by learning a strong correlation between blood pressure sample data and blood flow.

Optionally, blood pressure sample data includes, but is not limited to: at least one of arterial pressure sample data, ventricular pressure sample data, and pumping pressure difference sample data between the arterial pressure sample data and the ventricular pressure sample data, and a description of each type of blood pressure sample data is detailed in the description of the blood pressure data, and the embodiment is not repeated here.

The motor operation sample data is used to characterize the corresponding operation energy efficiency of the motor in the transcatheter ventricular assist device at the time the sample data set was acquired. The type of fluid (e.g., viscosity) that is pumped through the catheter ventricular assist device affects the energy efficiency of the motor operation at a set rotational speed. At a set rotational speed, the amount of power consumed by the motor to pump different types of liquids is also not different. It can be seen that there is a correlation between the motor operation sample data and the type of fluid pumped by the blood pump, and there is a correlation between the flow rate and the type of fluid, and the motor is also a power output source for driving the impeller to rotate, so that the flow rate data can be determined based on the motor operation sample data by learning the correlation between the motor operation sample data and the blood flow rate.

Optionally, the motor operation sample data includes, but is not limited to, at least one of motor rotation speed sample data and motor current sample data, and a description of each type of motor operation sample data is detailed in the description of the motor operation data, and this embodiment will not be repeated here.

The perfusate sample data can be u collected by the first detection unit, the blood pressure sample data can be collected by the second detection unit, and the motor operation sample data can be collected by the third detection unit; alternatively, at least one of the perfusate sample data, the blood pressure sample data and the motor operation sample data may be collected by other devices, and the data collection manner in the sample data set is not limited in this embodiment.

The operating environment of the transcatheter ventricular assist device may be a real ventricle or may be a simulated environment obtained by simulating a ventricle when acquiring a sample data set, and the embodiment does not limit the acquisition environment of the sample data set.

The perfusion sample data, the blood pressure sample data, the motor operation sample data and the flow measurement data in each set of sample data correspond to each other. Schematically, the perfusion sample data, the blood pressure sample data, the motor operation sample data and the flow measurement data in each group of sample data are synchronously acquired; alternatively, the maximum acquisition time interval between each item of data in the set of sample data is less than a threshold.

In one possible embodiment, the flow measurement data may be detected by a flow sensor.

In another possible embodiment, the flow measurement data is determined by the amount of change in weight of the liquid per unit of time. For example, the flow measurement data may be obtained by measuring a change in the weight of the liquid using a balance. Therefore, the influence of the measurement error of the flow sensor on the accuracy of the flow measurement data can be eliminated, the detection accuracy of the flow detection model trained by using the flow measurement data can get rid of the dependence on the accuracy of the flow sensor, and the model accuracy can exceed the measurement accuracy of the flow sensor, so that the accuracy of determining the flow data is improved.

In one example, the transcatheter ventricular assist device includes an adjustment device for adjusting a pumping flow or rotational speed of the interventional catheter pump, and accordingly, acquiring a sample dataset acquired during operation of the transcatheter ventricular assist device includes:

controlling the adjusting device to work in a first state, and acquiring at least one group of sample data in the first state; adjusting the working state of the adjusting device from a first state to a second state, and acquiring at least one group of sample data in the second state; the pumping speed corresponding to the second state is larger than that corresponding to the first state, or the pumping speed corresponding to the second state is smaller than that of the first state; determining whether the adjustment times of the working state reach preset times; under the condition that the preset times are not reached, taking the second state as the first state, and executing the steps of adjusting the working state of the adjusting device from the first state to the second state again and acquiring at least one group of sample data in the second state; stopping until the adjustment times reach the preset times or the adjustment times reach the maximum value or the minimum value, and obtaining a sample data set.

At this time, when sample data is collected, the adjusting device is operated to collect the sample data at different pumping speeds, so that the richness of the sample data set can be ensured, and the performance of the model obtained by training is improved.

In other embodiments, the working state of the adjusting device may also be set randomly when collecting the sample data, and the embodiment does not limit the collection mode of the sample data.

Optionally, acquiring a sample dataset acquired during operation of the transcatheter ventricular assist device comprises: sample data of different types of transcatheter ventricular assist devices are acquired to obtain a sample data set.

At this time, when sample data is collected, by collecting sample data of different types of transcatheter ventricular assist devices, a flow detection model of a corresponding type can be obtained by training sample data of each type of transcatheter ventricular assist device, and the accuracy of flow detection can be improved.

In some embodiments, the sample dataset is not subject to constraints on the liquid type. For example, the sample data set does not set a viscosity range constraint condition on the liquid viscosity, and the liquid viscosity corresponding to each set of sample data in the sample data set is not limited by the preset viscosity range. Since there are data samples corresponding to each liquid type in the sample data set, the flow rate detection model trained based on these data is a flow rate detection model common to each liquid type and is an independent single model. Step 402, model training is performed on a preset machine learning model according to the perfusion sample data, the blood pressure sample data, the motor operation sample data and the flow measurement data, so as to obtain a flow detection model.

In one example, the sample data u includes a plurality of sets of training data and a plurality of sets of test data, and model training is performed on a preset machine learning model according to the perfusion sample data, the blood pressure sample data, the motor operation sample data and the flow measurement data to obtain a flow detection model, including: modeling processing is carried out based on perfusion sample data, blood pressure sample data, motor operation sample data and flow measurement data in a plurality of groups of training data, and a first machine learning model is generated; inputting the perfusion sample data, the blood pressure sample data and the motor operation sample data in each group of test data into a first machine learning model for flow detection processing, and outputting flow detection data corresponding to each group of test data; comparing flow measurement data in each group of test data with flow detection data corresponding to each group of test data, and determining loss information corresponding to the first machine learning model; and adjusting model parameters of the first machine learning model based on the loss information to obtain a flow detection model.

Optionally, the first machine learning model comprises a gaussian model. In one possible embodiment, the first machine learning model is a gaussian model modeled from a first sample dataset; correspondingly, the flow detection model is a Gaussian model with adjusted parameters. Because the sample data of the Gaussian model obeys normal distribution and has statistical significance, the correlation among the parameters can be learned, and therefore, the Gaussian model can be selected for flow detection.

Optionally, comparing the flow measurement data in each set of test data with the flow detection data corresponding to each set of test data to determine loss information corresponding to the first machine learning model, including: determining average flow data according to the flow measurement data in each set of sample data; and comparing the average flow data, the flow detection data corresponding to each group of test data and the flow measurement data in each group of test data to obtain variance data and root mean square error data corresponding to the Gaussian model. At this time, the loss information includes variance data and root mean square error data.

Variance data r ² The calculation formula of (2) can be expressed by the following formula:

wherein n represents the total number of groups of samples in the test data; i represents the i-th set of test data, i being an integer from 1 to n. Representing flow measurement data in the i-th set of test data; />Average flow data representing flow detection data corresponding to the n sets of test data; x is X _i And the flow detection data corresponding to the i-th group of test data is represented.

The calculation formula of the root mean square error data RMSE can be expressed by the following formula:

wherein n represents the total number of groups of samples in the test data; i represents the i-th set of test data, i being an integer from 1 to n.Representing flow measurement data in the i-th set of test data; x is X _i And the flow detection data corresponding to the i-th group of test data is represented.

Alternatively, in the case where the variance and the root mean square error are used together to determine the loss information, the variance and the root mean square error also have corresponding weight parameters, and at this time, a weighted sum of the variance and the root mean square error is determined to obtain the loss information. Alternatively, the sum of the variance and the root mean square error may be determined as the loss information, and the present embodiment does not limit the calculation method of the loss information.

In the Gaussian model training process, the Gaussian model corresponds to a plurality of groups of model parameters, loss data corresponding to each group of parameters are respectively determined, so that the target model parameter with the highest flow detection accuracy (namely, the loss data is the smallest) is determined according to the loss data corresponding to each group of model parameters, and the Gaussian model is configured according to the target model parameter, so that the flow detection model is obtained.

In another possible implementation, the machine learning model is a neural network model. The foregoing adjusting, based on the loss information, model parameters of the first machine learning model to obtain a flow detection model includes: determining whether a loss value represented by the loss information is larger than a loss threshold value and whether the adjustment times of the model parameters are smaller than or equal to a times threshold value; if the loss value is larger than the loss threshold value or the adjustment frequency is smaller than the frequency threshold value, the model parameters of the first machine learning model are adjusted, pump load sample data and motor operation sample data in each group of test data are triggered to be input into the first machine learning model for flow detection processing, and flow detection data corresponding to each group of test data are output; comparing the flow measurement data in each group of test data with the flow detection data corresponding to each group of test data, and determining loss information corresponding to the first machine learning model; and if the loss value is smaller than or equal to the loss threshold value or the adjustment frequency is larger than or equal to the frequency threshold value, outputting the adjusted model parameters to obtain the flow detection model.

In this embodiment, the flow detection model is obtained by training in advance by using the sample data set, so that the flow data of the catheter ventricular assist device is determined by using the flow detection model, the dependence of flow detection on the flow sensor is greatly reduced, the technical problem that the pumping flow of the catheter ventricular assist device in the heart is difficult to detect is effectively solved, the complexity of flow detection is reduced, and the efficiency of flow detection is improved.

Moreover, because the perfusion state of the perfusion liquid, the pressure state of the blood and the running state of the motor can influence or reflect the pumping flow of the interventional catheter pump, the flow detection model is obtained by training based on the perfusion sample data representing the liquid state of the perfusion liquid, the blood pressure sample data representing the pressure state of the blood and the running sample data of the motor representing the running state of the motor, and the accuracy of the flow detection model in determining the flow data can be ensured.

In addition, the flow data is detected through a flow detection model, the flow detection model is a machine learning model which is obtained based on sample data training, and the machine learning model can learn the relatively accurate corresponding relation between the data, so that the accuracy of a flow detection result can be ensured.

In addition, by realizing the machine learning model as a Gaussian model, the flow detection result conforming to normal distribution can be obtained, and the accuracy of flow detection is improved; in addition, compared with a neural network model with huge parameter quantity, the Gaussian model has relatively small calculated quantity, so that the calculation efficiency of flow detection can be improved, and the calculation resources are saved.

Based on the above embodiments, if the perfusion data includes at least one perfusion parameter data, the at least one perfusion parameter data includes perfusion pressure data; the blood pressure data comprises at least one blood pressure parameter data comprising arterial pressure data, and the perfusion sample data in the sample data set comprises perfusion pressure sample data corresponding to the transcatheter ventricular assist device, and the blood pressure sample data in the sample data set comprises arterial pressure sample data corresponding to the transcatheter ventricular assist device.

Correspondingly, model training is carried out on a preset machine learning model according to perfusion sample data, blood pressure sample data, motor operation sample data and flow measurement data to obtain a flow detection model, and the method comprises the following steps:

and carrying out model training on a preset machine learning model according to the perfusion pressure sample data, the arterial pressure sample data, the motor operation sample data and the flow measurement data to obtain a flow detection model.

In one possible embodiment, the catheter of the transcatheter ventricular assist device is not provided with a pressure sensor, the ventricular pressure data is difficult to acquire, and the arterial pressure sample data can be acquired in vitro, in which case the flow detection model can be trained only according to the arterial pressure sample data, the perfusion pressure sample data and the motor operation sample data which are easy to detect, and the flow of the interventional catheter pump can be accurately output, so that the difficulty of flow detection is reduced.

The arterial pressure sample data is the same as the data type of the arterial pressure data, and the arterial pressure sample data may be aortic pressure (when the left ventricle is assisted), pulmonary arterial pressure data (when the right ventricle is assisted), or arterial pressure sample data of other positions. The embodiment of the present application is not limited thereto.

In the above embodiment, the motor operation data includes at least one of motor rotation speed data and motor current data. Under the condition that the motor operation data comprise motor rotation speed data and motor current data, the motor operation sample data in the sample data set comprise motor rotation speed sample data and motor current sample data corresponding to the catheter ventricular assist device;

under the condition, model training is carried out on a preset machine learning model according to perfusion sample data, blood pressure sample data, motor operation sample data and flow measurement data to obtain a flow detection model, and the method comprises the following steps: and carrying out model training on a preset machine learning model according to the perfusion pressure sample data, the arterial pressure sample data, the motor rotating speed sample data, the motor current sample data and the flow measurement data to obtain a flow detection model.

Accordingly, the flow detection model is specifically configured to detect flow data based on perfusion pressure sample data, arterial pressure sample data, motor speed data, and motor current data generated during application of the transcatheter ventricular assist device.

The motor current sample data and the motor rotating speed sample data can have a time sequence corresponding relation, and the motor current sample data can reflect the power consumption of the motor under the corresponding motor rotating speed sample data, so that the operation energy efficiency of the motor under the current liquid type is reflected. Therefore, specific motor current sample data and motor rotating speed sample data are used as motor operation data, the operation energy efficiency of the current driving motor can be effectively represented, and the accuracy of flow data determination of a flow detection model obtained through training is guaranteed.

Because there is a correlation between the perfusion pressure data and the arterial pressure data and the pumping flow of the interventional catheter pump, in this embodiment, the accuracy of the flow detection model in detecting the flow data can be ensured by training the flow detection model based on the perfusion pressure sample data and the arterial pressure sample data.

Optionally, the at least one perfusion parameter data may further comprise perfusion fluid flow rate data and/or perfusion pump rotational speed data, etc. in addition to perfusion fluid pressure data, where the perfusion sample data further comprises perfusion fluid flow rate sample data and/or perfusion pump rotational speed sample data.

Optionally, the at least one blood pressure parameter data may include ventricular pressure data and/or pressure difference data, in addition to arterial pressure data, and the blood pressure sample data may include ventricular pressure sample data and/or pressure difference sample data.

In other embodiments, the at least one perfusion parameter data may include perfusion flow data and/or perfusion pump rotational speed data, but not perfusion pressure data; at this time, the perfusion sample data includes perfusion fluid flow sample data and/or perfusion pump rotational speed sample data, and does not include perfusion fluid pressure sample data. And/or, the at least one blood pressure parameter data may include ventricular pressure data and/or differential pressure data, and not arterial pressure data; at this time, the perfusion sample data includes ventricular pressure sample data and/or differential pressure sample data, but does not include arterial pressure sample data, and the implementation of the perfusion sample data and the blood pressure sample data is not limited in this embodiment.

Alternatively, based on the above embodiment, since the perfusion data may be a combination of at least two perfusion parameter data, the blood pressure data may be a combination of at least two blood pressure parameter data, the data types formed by different combinations may be different, and the flow detection models corresponding to different data types may be different. Therefore, the flow detection model corresponding to different data types needs to be trained.

At this point, acquiring a sample dataset acquired during operation of the transcatheter ventricular assist device, comprising: perfusion sample data of different data types and blood pressure sample data of different data types are acquired.

Correspondingly, model training is carried out on a preset machine learning model according to perfusion sample data, blood pressure sample data, motor operation sample data and flow measurement data to obtain a flow detection model, and the method comprises the following steps: model training is carried out on a preset machine learning model according to perfusion sample data with the same data type, blood pressure sample data with the same data type, motor operation sample data and flow measurement data, and a flow detection model corresponding to the data type of the perfusion sample data and the data type of the blood pressure sample data is obtained.

In this embodiment, by training the corresponding flow detection model based on the perfusion sample data of each data type and/or the blood pressure sample data of each data type, the catheter ventricular assist device may determine that the adapted first flow detection model performs flow detection based on the currently acquired data type, so that accuracy of flow detection may be improved.

Alternatively, based on the above embodiment, the data collected at different collection positions may correspond to different flow detection models, and thus, the flow detection models corresponding to the different collection positions need to be trained.

At this point, acquiring a sample dataset acquired during operation of the transcatheter ventricular assist device, comprising: and acquiring perfusion sample data acquired at different first acquisition positions and blood pressure sample data acquired at different second acquisition positions.

and carrying out model training on a preset machine learning model according to the perfusion sample data acquired at the same first acquisition position, the blood pressure sample data acquired at the same second acquisition position, the motor operation sample data and the flow measurement data to obtain a flow detection model corresponding to the first acquisition position and the second acquisition position.

In this embodiment, the corresponding flow detection model is obtained by training according to the perfusion sample data acquired at each first acquisition position and the blood pressure sample data acquired at each second acquisition position, so that the catheter ventricular assist device can determine the adaptive second flow detection model based on the first acquisition position of the currently acquired perfusion data and the second acquisition position of the blood pressure data to perform flow detection processing, thereby improving the accuracy of flow detection.

Fig. 5 is a block diagram of a flow determination device provided in one embodiment of the present application as applied to a transcatheter ventricular assist device. The device at least comprises the following modules: a data acquisition module 510 and a flow determination module 520.

A data acquisition module 510, configured to acquire perfusion data and blood pressure data corresponding to the transcatheter ventricular assist device, and motor operation data corresponding to the transcatheter ventricular assist device;

the flow determining module 520 is configured to perform flow detection according to the perfusion data, the blood pressure data, and the motor operation data, so as to obtain flow data corresponding to the transcatheter ventricular assist device.

Optionally, the perfusion data comprises at least one perfusion parameter data, the at least one perfusion parameter data comprises at least one of perfusion pressure data, perfusion flow rate data, and perfusion pump rotational speed data, the blood pressure data comprises at least one blood pressure parameter data, and the at least one blood pressure parameter data comprises at least one of arterial pressure data, ventricular pressure data, and pumping differential pressure data.

Optionally, the flow determining module 520 is configured to:

Optionally, the motor operation data includes motor rotation speed data and motor current data, and the flow determining module 520 is configured to:

Optionally, the flow determination module 520 includes a determination sub-module 521.

The determination submodule 521 is configured to input the perfusion data, the blood pressure data, and the motor operation data into a preset flow detection model to perform flow detection processing, and output the flow data;

Optionally, the determining submodule 521 is configured to:

For relevant details reference is made to the method embodiments described above.

It should be noted that: in the flow determining device provided in the above embodiment, only the division of the above functional modules is used for illustration, and in practical application, the above functional allocation may be performed by different functional modules according to needs, that is, the internal structure of the flow determining device is divided into different functional modules to perform all or part of the functions described above. In addition, the flow determining device provided in the foregoing embodiments and the flow determining method embodiment belong to the same concept, and specific implementation processes of the flow determining device are detailed in the method embodiment, which is not described herein again.

FIG. 6 is a block diagram of a training apparatus for a flow detection model provided in one embodiment of the application. The device at least comprises the following modules: a data acquisition module 610 and a model training module 620.

A data acquisition module 610, configured to acquire a sample data set acquired during operation of the transcatheter ventricular assist device, where each set of sample data in the sample data set includes perfusion sample data, blood pressure sample data, motor operation sample data, and corresponding flow measurement data corresponding to the transcatheter ventricular assist device;

the model training module 620 is configured to perform model training on a preset machine learning model according to the perfusion sample data, the blood pressure sample data, the motor operation sample data and the flow measurement data, so as to obtain a flow detection model;

Optionally, the sample data set includes a plurality of sets of training data and a plurality of sets of test data, and the model training module 620 includes: a model generation sub-module 621, a flow detection sub-module 622, a loss calculation sub-module 623, and a parameter adjustment sub-module 624.

A model generating sub-module 621, configured to perform modeling processing based on the perfusion sample data, the blood pressure sample data, the motor operation sample data, and the flow measurement data in the plurality of sets of training data, and generate a first machine learning model;

the flow detection sub-module 622 is configured to input the perfusion sample data, the blood pressure sample data, and the motor operation sample data in each set of test data into the first machine learning model to perform flow detection processing, and output flow detection data corresponding to each set of test data;

a loss calculation sub-module 623, configured to compare the flow measurement data in each set of test data with the flow detection data corresponding to each set of test data, and determine loss information corresponding to the first machine learning model;

and a parameter adjustment sub-module 624, configured to adjust model parameters of the first machine learning model based on the loss information, so as to obtain the flow detection model.

Optionally, the first machine learning model comprises a gaussian model, and the loss calculation submodule 623 is configured to:

It should be noted that: in the training device for a flow detection model provided in the foregoing embodiment, only the division of the functional modules is used for illustration when the training of the flow detection model is performed, and in practical application, the functional allocation may be performed by different functional modules according to needs, that is, the internal structure of the training device for a flow detection model is divided into different functional modules, so as to complete all or part of the functions described above. In addition, the training device of the flow detection model provided in the above embodiment and the training method embodiment of the flow detection model belong to the same concept, and the specific implementation process is detailed in the method embodiment, which is not described herein again.

FIG. 7 is a block diagram of a computer device provided in one embodiment of the application. The computer device comprises at least the control unit in the transcatheter ventricular assist device shown in fig. 1, but in other embodiments may be other devices communicatively connected to the transcatheter ventricular assist device, such as: computer devices with processing capabilities such as computers, tablet computers, cell phones, and the like. The computer device comprises at least a processor 701 and a memory 702.

The processor 701 may include one or more processing cores, such as: 4 core processors, 8 core processors, etc. The processor 701 may be implemented in at least one hardware form of DSP (Digital Signal Processing ), FPGA (Field-Programmable Gate Array, field programmable gate array), PLA (Programmable Logic Array ). The processor 701 may also include a main processor, which is a processor for processing data in an awake state, also referred to as a CPU (Central Processing Unit ); a coprocessor is a low-power processor for processing data in a standby state. In some embodiments, the processor 701 may integrate a GPU (Graphics Processing Unit, image processor) for rendering and drawing of content required to be displayed by the display screen. In some embodiments, the processor 701 may also include an AI (Artificial Intelligence ) processor for processing computing operations related to machine learning.

Memory 702 may include one or more computer-readable storage media, which may be non-transitory. The memory 702 may also include high-speed random access memory, as well as non-volatile memory, such as one or more magnetic disk storage devices, flash memory storage devices. In some embodiments, a non-transitory computer readable storage medium in memory 702 is used to store at least one instruction for execution by processor 701 to implement the flow determination method or the training method of the flow detection model provided by the method embodiments of the present application.

In some embodiments, the computer device may further optionally include: a peripheral interface and at least one peripheral. The processor 701, the memory 702, and the peripheral interfaces may be connected by buses or signal lines. The individual peripheral devices may be connected to the peripheral device interface via buses, signal lines or circuit boards. Illustratively, peripheral devices include, but are not limited to: radio frequency circuitry, touch display screens, audio circuitry, and power supplies, among others.

Of course, the computer device may also include fewer or more components, as the present embodiment is not limited in this regard.

Optionally, the present application further provides a computer readable storage medium, where a program is stored, where the program is loaded and executed by a processor to implement the method for determining a flow rate or the training method for a flow rate detection model according to the above method embodiment.

Optionally, the present application further provides a computer product, where the computer product includes a computer readable storage medium, where a program is stored in the computer readable storage medium, and the program is loaded and executed by a processor to implement the method for determining a flow rate or the training method for a flow rate detection model according to the above method embodiment.

The technical features of the above-described embodiments may be arbitrarily combined, and all possible combinations of the technical features in the above-described embodiments are not described for brevity of description, however, as long as there is no contradiction between the combinations of the technical features, they should be considered as the scope of the description.

The above examples illustrate only a few embodiments of the application, which are described in detail and are not to be construed as limiting the scope of the application. It should be noted that it will be apparent to those skilled in the art that several variations and modifications can be made without departing from the spirit of the application, which are all within the scope of the application. Accordingly, the scope of protection of the present application is to be determined by the appended claims.

Claims

1. A method of flow determination, the method being applied to a transcatheter ventricular assist device, the method comprising:

2. The method of claim 1, wherein the perfusion data comprises at least one perfusion parameter data comprising at least one of perfusion fluid pressure data, perfusion fluid flow data, and perfusion pump rotational speed data, the blood pressure data comprising at least one blood pressure parameter data comprising at least one of arterial pressure data, ventricular pressure data, and pumping differential pressure data.

3. The method according to claim 2, wherein the performing flow detection processing according to the perfusion data, the blood pressure data, and the motor operation data to obtain flow data corresponding to the transcatheter ventricular assist device includes:

4. The method of claim 2, wherein the motor operation data includes motor rotation speed data and motor current data, and wherein the performing a flow detection process according to the perfusion pressure data, the arterial pressure data, and the motor operation data to obtain the flow data includes:

5. The method of any one of claims 2 to 4, wherein the transcatheter ventricular assist device comprises:

6. The method of claim 1, wherein the performing flow detection processing according to the perfusion data, the blood pressure data, and the motor operation data to obtain flow data corresponding to the transcatheter ventricular assist device comprises:

7. The method according to claim 6, wherein inputting the perfusion data, the blood pressure data, and the motor operation data into a preset flow detection model for flow detection processing, and outputting the flow data, comprises:

8. A method of training a flow detection model, the method comprising:

9. The method according to claim 8, wherein the sample data u includes a plurality of sets of training data and a plurality of sets of test data, and the model training the preset machine learning model according to the perfusion sample data, the blood pressure sample data, the motor operation sample data and the flow measurement data to obtain a flow detection model includes:

10. The method of claim 9, wherein the first machine learning model comprises a gaussian model, wherein comparing the flow measurement data in each set of test data with the flow detection data corresponding to each set of test data determines loss information corresponding to the first machine learning model, comprising:

11. A method according to any one of claims 8 to 10, wherein the flow measurement data is determined by the amount of change in weight of liquid per unit time.

12. A flow determination device for use with a transcatheter ventricular assist device, the device comprising:

13. A training device for a flow detection model, the device comprising:

the data acquisition module is used for acquiring sample data u acquired in the operation process of the catheter ventricular assist device, and each group of sample data in the sample data set comprises perfusion sample data, blood pressure sample data, motor operation sample data and corresponding flow measurement data corresponding to the catheter ventricular assist device;

14. A transcatheter ventricular assist device, the transcatheter ventricular assist device comprising a computer device comprising a processor and a memory; the memory has stored therein a program that is loaded and executed by the processor to implement the flow rate determination method according to any one of claims 1 to 6; alternatively, a training method of the flow detection model according to any one of claims 7 to 10 is implemented.

15. A computer device, the device comprising a processor and a memory; the memory has stored therein a program that is loaded and executed by the processor to implement the flow rate determination method according to any one of claims 1 to 6; alternatively, a training method of the flow detection model according to any one of claims 7 to 10 is implemented.

16. A computer-readable storage medium, characterized in that the storage medium has stored therein a program for realizing the flow rate determination method according to any one of claims 1 to 6 when loaded and executed by a processor; alternatively, a training method of the flow detection model according to any one of claims 7 to 10 is implemented.