Blood purification equipment data monitoring method and system

Abstract

The present invention relates to the field of blood purification, and specifically to a method and system for monitoring data of blood purification equipment. The blood purification equipment data monitoring system includes a patient vital sign data acquisition module, an operation data monitoring module and a fault prediction module. The present invention acquires the vital sign data of the patient and executes corresponding fault prediction schemes for the vital signs corresponding to different patients; this personalized method can more accurately reflect the impact of individual differences of patients on the operating status of the equipment, thereby improving the accuracy and timeliness of fault prediction; this targeted monitoring and prediction strategy helps to identify potential problems in advance, reduce equipment failure rate, and improve the continuity and safety of treatment.

Description

Technical Field

[0001] The present invention relates to the field of blood purification, and in particular to a method and system for monitoring data of blood purification equipment. Background Art

[0002] Blood purification equipment is an indispensable medical device for the treatment of renal failure and other blood-related diseases. These devices help maintain the patient's vital signs by removing waste and excess water from the blood. Blood purification equipment may encounter a variety of faults during operation, such as filter blockage, pump failure, abnormal conductivity, etc. If these faults are not discovered and handled in time, they may affect the treatment effect and even pose a threat to the patient's health. However, the reliability of blood purification equipment is largely affected by individual differences in patients, especially blood characteristics. Differences in blood viscosity and composition in patients may cause certain parts of the equipment to wear faster or increase the risk of blockage. For example, high-viscosity blood may increase the probability of filter blockage, and abnormal concentrations of specific components in the blood may affect the conductivity of the dialysate, thereby affecting the dialysis efficiency.

[0003] Existing blood purification equipment monitoring methods often ignore these individual differences and use uniform monitoring standards and failure thresholds, which can lead to inaccurate monitoring and failure to timely predict and respond to equipment failures caused by different patient blood characteristics. Summary of the invention

[0004] The present invention provides a blood purification equipment data monitoring method and system to solve the technical problem of inaccurate blood purification equipment monitoring in the above-mentioned background technology.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0006] In a first aspect, the present invention provides a method for monitoring data of a blood purification device, comprising:

[0007] Obtain patient vital signs data and equipment operation data at the current monitoring time point. Patient vital signs data include blood pressure, heart rate and blood viscosity; equipment operation data include filter element pressure difference, filter element flow, pump output flow, pump pressure, dialysate conductivity, dialysate flow, system overall pressure and equipment temperature;

[0008] The patient vital sign data obtained within the current preset time window of duration T are combined into a short-term vital sign data set, the standard deviation of the patient vital sign data in the current short-term vital sign data set is calculated, and the patient vital sign fluctuation level is obtained; based on the current patient vital sign fluctuation level, the normal operating range of the equipment at the current monitoring time point is set; it is determined whether the equipment operation data obtained at the current monitoring time point is within the preset normal operating range, and if so, the fault prediction plan is executed; if not, an alarm is issued and alarm response measures are executed.

[0009] Preferably, based on the current fluctuation level of the patient's vital signs, the normal operating range of the device at the current monitoring time point is set to include:

[0010] Set the normal operating range and also set the standard deviation threshold and the maximum standard deviation ;

[0011] If the standard deviation of the patient's vital signs data in the currently acquired short-term vital signs data set is Below or equal to the standard deviation threshold , then the normal operating range of the equipment at the current monitoring time point is the regular operating range;

[0012] Set the increment range to , if the standard deviation of the patient's vital signs data in the currently acquired short-term vital signs data set is Above standard deviation threshold , then use the formula Calculate the gain rate , and the increase rate Applied to the normal operating range, obtain the normal operating range of the equipment at the current monitoring time point.

[0013] Preferably, executing the fault prediction scheme includes:

[0014] An update cycle of 10T is set. At the beginning of any update cycle, the patient vital sign data acquired in the previous update cycle are combined into a long-term vital sign data set; the average value of the patient vital sign data in the acquired long-term vital sign data set is calculated to acquire the overall patient vital sign data in the current update cycle;

[0015] Set the blood pressure threshold, heart rate threshold and blood viscosity threshold, and compare the overall blood pressure value, heart rate value and blood viscosity value in the overall vital sign data of the patient with the blood pressure threshold, heart rate threshold and blood viscosity threshold; if the overall blood pressure value of the current patient is higher than the blood pressure threshold, perform fault prediction operations on the filter element and the pump; if the overall heart rate value of the current patient is higher than the heart rate threshold, perform fault prediction operations on the pump and the flow control system; if the overall blood viscosity value of the current patient is higher than the blood viscosity threshold, perform fault prediction operations on the filter element, conductivity measurement module and flow control system;

[0016] The fault prediction operations in the fault prediction scheme include:

[0017] Predict the fault of the filter element, and compose the filter element pressure difference and filter element flow obtained at the current monitoring time point and the most recent N-1 monitoring time points into a filter element fault prediction set in chronological order. Use the filter element fault prediction set as the input of the filter element fault prediction model, output the filter element pressure difference prediction value and filter element flow prediction value at the next monitoring time point, and judge whether the obtained filter element pressure difference prediction value and filter element flow prediction value are both within the normal operating range of the equipment. If so, no operation is performed; if not, perform fault pre-processing measures for the filter element;

[0018] Predict pump failures, organize the pump output flow and pump pressure obtained at the current monitoring time point and the most recent N-1 monitoring time points into a pump failure prediction set in chronological order, use the pump failure prediction set as the input of the pump failure prediction model, output the pump output flow prediction value and pump pressure prediction value at the next monitoring time point, and determine whether the obtained pump output flow prediction value and pump pressure prediction value are both within the normal operating range of the equipment. If so, no operation is performed; if not, perform fault pre-processing measures for the pump;

[0019] Perform fault prediction on the conductivity measurement module, organize the dialysis fluid conductivity obtained at the current monitoring time point and the most recent N-1 monitoring time points into a conductivity fault prediction set in chronological order, use the conductivity fault prediction set as the input of the conductivity fault prediction model, output the predicted value of the dialysis fluid conductivity at the next monitoring time point, and determine whether the obtained predicted value of the dialysis fluid conductivity is within the normal operating range of the equipment. If so, no operation is performed; if not, perform fault pre-processing measures for the conductivity measurement module;

[0020] Fault prediction is performed on the liquid flow control system. The dialysate flow obtained at the current monitoring time point and the most recent N-1 monitoring time points are organized into a flow fault prediction set in chronological order. The flow fault prediction set is used as the input of the flow fault prediction model. The dialysate flow prediction value at the next monitoring time point is output. It is determined whether the obtained dialysate flow prediction value is within the normal operating range of the equipment. If so, no operation is performed; if not, fault pre-processing measures are performed on the liquid flow control system.

[0021] Preferably, the filter element fault prediction model is established based on the LSTM model, including N filter element fault prediction LSTM units, 1 first fully connected layer and 1 first output layer, wherein the filter element fault prediction LSTM unit is used to extract the temporal characteristics of the filter element fault prediction set; the first fully connected layer is used to extract features from the outputs of all filter element fault prediction LSTM units; the first output layer is used to output the filter element pressure difference prediction value and the filter element flow prediction value at the next monitoring time point;

[0022] The training of the filter element failure prediction model includes:

[0023] A plurality of filter element failure training samples are obtained, each of which stores N+1 filter element pressure differences and filter element flow rates sorted by time, and the first N filter element pressure differences and filter element flow rates in the filter element failure training samples are used as the input of the filter element failure prediction model, and the last filter element pressure difference and filter element flow rate in the filter element failure training samples are used as the output of the filter element failure prediction model; all filter element failure training samples are divided into a filter element failure training set and a filter element failure test set, and the filter element failure training set is sent to the filter element failure prediction model with parameter initialization for training, and then the filter element failure test set is input to the filter element failure prediction model to obtain a first test result; a first training condition is set to determine whether the first test result meets the first training condition, and if so, the trained filter element failure prediction model is output; otherwise, the filter element failure prediction model is continuously trained through the filter element failure training set.

[0024] Preferably, the pump fault prediction model is established based on the LSTM model, including N pump fault prediction LSTM units, 1 second fully connected layer and 1 second output layer. The training of the pump fault prediction model includes:

[0025] A plurality of pump fault training samples are obtained, each of which stores N+1 pump output flow rates and pump pressures sorted by time, and the first N pump output flow rates and pump pressures in the pump fault training samples are used as the input of the pump fault prediction model, and the last pump output flow rate and pump pressure in the pump fault training samples are used as the output of the pump fault prediction model; all pump fault training samples are divided into a pump fault training set and a pump fault test set, and the pump fault training set is sent to the pump fault prediction model with parameter initialization for training, and then the pump fault test set is input to the pump fault prediction model to obtain a second test result; a second training condition is set to determine whether the second test result meets the second training condition, and if so, the trained pump fault prediction model is output; otherwise, the pump fault prediction model is continuously trained using the pump fault training set.

[0026] Preferably, the conductivity fault prediction model is established based on the LSTM model, including N conductivity fault prediction LSTM units, 1 third fully connected layer and 1 third output layer. The training of the conductivity fault prediction model includes:

[0027] A plurality of conductivity fault training samples are obtained, each of which stores N+1 dialysis fluid conductivities sorted by time, and the first N dialysis fluid conductivities in the conductivity fault training samples are used as inputs of the conductivity fault prediction model, and the last dialysis fluid conductivity in the conductivity fault training samples is used as outputs of the conductivity fault prediction model; all conductivity fault training samples are divided into a conductivity fault training set and a conductivity fault test set, and the conductivity fault training set is sent to the conductivity fault prediction model with parameter initialization for training, and then the conductivity fault test set is input to the conductivity fault prediction model to obtain a third test result; a third training condition is set, and it is determined whether the third test result meets the third training condition. If the third training condition is met, the trained conductivity fault prediction model is output; otherwise, the conductivity fault prediction model is continuously trained through the conductivity fault training set.

[0028] Preferably, the traffic fault prediction model is established based on the LSTM model, including N traffic fault prediction LSTM units, 1 fourth fully connected layer and 1 fourth output layer. The training of the traffic fault prediction model includes:

[0029] A plurality of flow fault training samples are obtained, each flow fault training sample stores N+1 dialysate flow rates sorted by time, the first N dialysate flow rates in the flow fault training sample are used as the input of the flow fault prediction model, and the last dialysate flow rate in the flow fault training sample is used as the output of the flow fault prediction model; all flow fault training samples are divided into a flow fault training set and a flow fault test set, the flow fault training set is sent to the parameter-initialized flow fault prediction model for training, and then the flow fault test set is input to the flow fault prediction model to obtain a fourth test result; a fourth training condition is set to determine whether the fourth test result meets the fourth training condition, and if so, the trained flow fault prediction model is output; otherwise, the flow fault prediction model is continuously trained using the flow fault training set.

[0030] In a second aspect, the present invention provides a blood purification equipment data monitoring system, comprising:

[0031] The patient vital sign data acquisition module is used to acquire the patient vital sign data, and form the patient vital sign data acquired within the time window of the current preset duration T into a short-term vital sign data set, calculate the standard deviation of the patient vital sign data in the current acquired short-term vital sign data set, and acquire the patient vital sign fluctuation level; based on the current patient vital sign fluctuation level, set the normal operating range of the equipment at the current monitoring time point;

[0032] The operation data monitoring module is used to obtain the equipment operation data at the current monitoring time point, determine whether the equipment operation data obtained at the current monitoring time point is within the preset normal operating range, and if so, execute the fault prediction plan; if not, issue an alarm and execute alarm response measures;

[0033] The fault prediction module is used to compare the overall vital sign data of the patient in the current update cycle with the blood pressure threshold, heart rate threshold and blood viscosity threshold. If the overall blood pressure value of the current patient is higher than the blood pressure threshold, fault prediction operations are performed on the filter element and the pump; if the overall heart rate value of the current patient is higher than the heart rate threshold, fault prediction operations are performed on the pump and the flow control system; if the overall blood viscosity value of the current patient is higher than the blood viscosity threshold, fault prediction operations are performed on the filter element, conductivity measurement module and flow control system.

[0034] The present invention has the following advantages:

[0035] The present invention acquires the patient's vital sign data and executes corresponding fault prediction schemes for different patient signs; this personalized method can more accurately reflect the impact of individual differences of patients on the operating status of the equipment, thereby improving the accuracy and timeliness of fault prediction; this targeted monitoring and prediction strategy helps to identify potential problems in advance, reduce the equipment failure rate, and improve the continuity and safety of treatment. In addition, by automatically executing corresponding fault pre-processing measures when it is predicted that the equipment operation data may exceed the normal range; this response mechanism can quickly respond to predicted faults, reduce treatment interruptions or delays caused by equipment problems, thereby ensuring the quality and safety of patient treatment; at the same time, it can dynamically adjust the normal operating range of the equipment according to the fluctuation level of the patient's vital signs, improve the flexibility and adaptability of monitoring, and ensure the accuracy of monitoring and the stable operation of the equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 This is a schematic diagram of the structure of a blood purification equipment data monitoring system used in an embodiment of the present invention. DETAILED DESCRIPTION

[0037] In order to enable persons skilled in the art to better understand the technical solutions in the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0038] Embodiment 1 of the present invention is a method for monitoring data of a blood purification device, comprising:

[0039] Obtain patient vital signs data and equipment operation data at the current monitoring time point. Patient vital signs data include blood pressure, heart rate and blood viscosity; blood viscosity refers to the viscosity of blood, which is an important factor affecting blood flow characteristics; the level of blood viscosity will affect the flow rate of blood in blood vessels, thereby affecting the filtration efficiency and dialysis effect of blood purification equipment; equipment operation data includes filter element pressure difference, filter element flow, pump output flow, pump pressure, dialysate conductivity, dialysate flow, system overall pressure and equipment temperature;

[0040] The patient vital sign data obtained within the current preset time window of duration T are combined into a short-term vital sign data set, the standard deviation of the patient vital sign data in the current short-term vital sign data set is calculated, and the patient vital sign fluctuation level is obtained; based on the current patient vital sign fluctuation level, the normal operating range of the equipment at the current monitoring time point is set; it is determined whether the equipment operation data obtained at the current monitoring time point is within the preset normal operating range, and if so, the fault prediction plan is executed; if not, an alarm is issued and alarm response measures are executed.

[0041] Based on the current fluctuation level of the patient's vital signs, set the specific operation of the normal operating range of the equipment at the current monitoring time point:

[0042] Set the normal operating range, which is the baseline state of the equipment operation, defines the basic parameter range that the equipment should follow under normal working conditions, and sets the standard deviation threshold and the maximum standard deviation ; Standard deviation threshold Used to determine whether the fluctuation of patient vital sign data is within an acceptable range; the highest standard deviation in historical patient data can be selected as the maximum standard deviation ;

[0043] If the standard deviation of the patient's vital signs data in the currently acquired short-term vital signs data set is Below or equal to the standard deviation threshold , it means that the fluctuation of the patient's vital signs is within the normal range, and the equipment can operate within the normal operating range. The normal operating range of the equipment at the current monitoring time point is the normal operating range;

[0044] Set the increment range to , if the standard deviation of the patient's vital signs data in the currently acquired short-term vital signs data set is Above standard deviation threshold , then use the formula Calculate the gain rate , and the increase rate Applied to the routine operating range, obtain the normal operating range of the equipment at the current monitoring time point to adapt to changes in patient vital signs and ensure that the equipment can still operate safely and effectively under the new level of vital sign fluctuations.

[0045] The reason for performing this step is that during the operation of the blood purification equipment, fluctuations in the patient's vital signs will indeed affect the equipment's various operating data (such as filter element pressure difference, filter element flow, pump output flow, etc.), so it is necessary to adjust the alarm threshold to adapt to the vital sign status of different patients.

[0046] Execute the specific operations of the fault prediction solution, including:

[0047] An update cycle of 10T is set. At the beginning of any update cycle, the patient vital sign data acquired in the previous update cycle are combined into a long-term vital sign data set; the average value of the patient vital sign data in the acquired long-term vital sign data set is calculated to acquire the overall patient vital sign data in the current update cycle;

[0048] The blood pressure threshold, heart rate threshold and blood viscosity threshold are set, and the overall blood pressure value, heart rate value and blood viscosity value in the patient's overall physical sign data are compared with the blood pressure threshold, heart rate threshold and blood viscosity threshold.

[0049] If the current patient's overall blood pressure is higher than the blood pressure threshold, the filter element and pump will be subjected to fault prediction operations. This is because increased blood pressure will directly lead to increased filter element input pressure, increase the burden on the filter element, and easily lead to increased filter element pressure difference, increasing the probability of filter element failure; at the same time, high blood pressure will increase the workload of the pump, because the pump needs to overcome higher pressure to maintain a stable flow output; long-term high blood pressure may cause increased wear of the pump and increase the failure rate of the pump.

[0050] If the current patient's overall heart rate is higher than the heart rate threshold, the pump and flow control system will be fault predicted. Because large heart rate fluctuations may cause unstable blood flow, the pump needs to constantly adjust the flow output to adapt to blood flow changes; this frequent adjustment will increase the mechanical wear of the pump and the load of the control system, thereby increasing the failure rate of the pump; blood flow fluctuations caused by heart rate changes may affect the accuracy of the flow control system, increase the probability of control errors, and may cause unstable flow, affecting the stable operation of the entire system.

[0051] If the current patient's overall blood viscosity value is higher than the blood viscosity threshold, fault prediction operations are performed on the filter element, conductivity measurement module and flow control system. This is because higher blood viscosity means higher concentrations of red blood cells, platelets and other components in the blood, which may increase the flow resistance of blood when passing through the filter element, thereby accelerating the clogging of the filter element; high-viscosity blood may require a larger pressure difference to maintain the same flow rate, which will increase the risk of wear and damage to the filter element; this predictive maintenance helps reduce the occurrence of emergencies, improve the reliability of the equipment and the effectiveness of treatment; at the same time, high-viscosity blood may require longer dialysis time or higher dialysis pressure to achieve the same purification effect; this requirement may affect the flow rate and pressure of the dialysate, and then affect the stability of the dialysate conductivity. The conductivity directly affects the concentration and distribution of ions in the dialysate, which in turn affects the exchange efficiency of substances during the dialysis process. Improper conductivity may cause complications in patients, such as electrolyte disorders, posing a safety hazard to patients;

[0052] The fault prediction operations in the fault prediction scheme specifically include:

[0053] Predict the fault of the filter element, and compose the filter element pressure difference and filter element flow obtained at the current monitoring time point and the most recent N-1 monitoring time points into a filter element fault prediction set in chronological order. Use the filter element fault prediction set as the input of the filter element fault prediction model, output the filter element pressure difference prediction value and filter element flow prediction value at the next monitoring time point, and judge whether the obtained filter element pressure difference prediction value and filter element flow prediction value are both within the normal operating range of the equipment. If so, no operation is performed; if not, perform fault pre-processing measures for the filter element;

[0054] Predict pump failures, organize the pump output flow and pump pressure obtained at the current monitoring time point and the most recent N-1 monitoring time points into a pump failure prediction set in chronological order, use the pump failure prediction set as the input of the pump failure prediction model, output the pump output flow prediction value and pump pressure prediction value at the next monitoring time point, and determine whether the obtained pump output flow prediction value and pump pressure prediction value are both within the normal operating range of the equipment. If so, no operation is performed; if not, perform fault pre-processing measures for the pump;

[0055] Perform fault prediction on the conductivity measurement module, organize the dialysis fluid conductivity obtained at the current monitoring time point and the most recent N-1 monitoring time points into a conductivity fault prediction set in chronological order, use the conductivity fault prediction set as the input of the conductivity fault prediction model, output the predicted value of the dialysis fluid conductivity at the next monitoring time point, and determine whether the obtained predicted value of the dialysis fluid conductivity is within the normal operating range of the equipment. If so, no operation is performed; if not, perform fault pre-processing measures for the conductivity measurement module;

[0056] Fault prediction is performed on the liquid flow control system. The dialysate flow obtained at the current monitoring time point and the most recent N-1 monitoring time points are organized into a flow fault prediction set in chronological order. The flow fault prediction set is used as the input of the flow fault prediction model. The dialysate flow prediction value at the next monitoring time point is output. It is determined whether the obtained dialysate flow prediction value is within the normal operating range of the equipment. If so, no operation is performed; if not, fault pre-processing measures are performed on the liquid flow control system.

[0057] The filter element fault prediction model is established based on the LSTM model, including N filter element fault prediction LSTM units, 1 first fully connected layer and 1 first output layer. The filter element fault prediction LSTM unit is used to extract the temporal characteristics of the filter element fault prediction set; the first fully connected layer is used to extract features from the outputs of all filter element fault prediction LSTM units; the first output layer is used to output the filter element pressure difference prediction value and the filter element flow prediction value at the next monitoring time point;

[0058] For the training of filter element failure prediction model, the specific operations are as follows:

[0059] A plurality of filter element failure training samples are obtained, each of which stores N+1 filter element pressure differences and filter element flow rates sorted by time, and the first N filter element pressure differences and filter element flow rates in the filter element failure training samples are used as the input of the filter element failure prediction model, and the last filter element pressure difference and filter element flow rate in the filter element failure training samples are used as the output of the filter element failure prediction model; all filter element failure training samples are divided into a filter element failure training set and a filter element failure test set, and the filter element failure training set is sent to the filter element failure prediction model with parameter initialization for training, and then the filter element failure test set is input to the filter element failure prediction model to obtain a first test result; a first training condition is set to determine whether the first test result meets the first training condition, and if so, the trained filter element failure prediction model is output; otherwise, the filter element failure prediction model is continuously trained through the filter element failure training set.

[0060] The pump fault prediction model is established based on the LSTM model, including N pump fault prediction LSTM units, 1 second fully connected layer and 1 second output layer. The specific operations for the training of the pump fault prediction model are as follows:

[0061] A plurality of pump fault training samples are obtained, each of which stores N+1 pump output flow rates and pump pressures sorted by time, and the first N pump output flow rates and pump pressures in the pump fault training samples are used as the input of the pump fault prediction model, and the last pump output flow rate and pump pressure in the pump fault training samples are used as the output of the pump fault prediction model; all pump fault training samples are divided into a pump fault training set and a pump fault test set, and the pump fault training set is sent to the pump fault prediction model with parameter initialization for training, and then the pump fault test set is input to the pump fault prediction model to obtain a second test result; a second training condition is set to determine whether the second test result meets the second training condition, and if so, the trained pump fault prediction model is output; otherwise, the pump fault prediction model is continuously trained using the pump fault training set.

[0062] The conductivity fault prediction model is established based on the LSTM model, including N conductivity fault prediction LSTM units, 1 third fully connected layer and 1 third output layer. The specific operations for the training of the conductivity fault prediction model are as follows:

[0063] A plurality of conductivity fault training samples are obtained, each of which stores N+1 dialysis fluid conductivities sorted by time, and the first N dialysis fluid conductivities in the conductivity fault training samples are used as inputs of the conductivity fault prediction model, and the last dialysis fluid conductivity in the conductivity fault training samples is used as outputs of the conductivity fault prediction model; all conductivity fault training samples are divided into a conductivity fault training set and a conductivity fault test set, and the conductivity fault training set is sent to the conductivity fault prediction model with parameter initialization for training, and then the conductivity fault test set is input to the conductivity fault prediction model to obtain a third test result; a third training condition is set, and it is determined whether the third test result meets the third training condition. If the third training condition is met, the trained conductivity fault prediction model is output; otherwise, the conductivity fault prediction model is continuously trained through the conductivity fault training set.

[0064] The traffic fault prediction model is built based on the LSTM model, including N traffic fault prediction LSTM units, 1 fourth fully connected layer and 1 fourth output layer. The specific operations for the training of the traffic fault prediction model are as follows:

[0065] A plurality of flow fault training samples are obtained, each flow fault training sample stores N+1 dialysate flow rates sorted by time, the first N dialysate flow rates in the flow fault training sample are used as the input of the flow fault prediction model, and the last dialysate flow rate in the flow fault training sample is used as the output of the flow fault prediction model; all flow fault training samples are divided into a flow fault training set and a flow fault test set, the flow fault training set is sent to the parameter-initialized flow fault prediction model for training, and then the flow fault test set is input to the flow fault prediction model to obtain a fourth test result; a fourth training condition is set to determine whether the fourth test result meets the fourth training condition, and if so, the trained flow fault prediction model is output; otherwise, the flow fault prediction model is continuously trained using the flow fault training set.

[0066] Embodiment 2 of the present invention is a blood purification equipment data monitoring system, such as Figure 1 As shown, including:

[0067] The patient vital sign data acquisition module is used to acquire the patient vital sign data, and form the patient vital sign data acquired within the time window of the current preset duration T into a short-term vital sign data set, calculate the standard deviation of the patient vital sign data in the current acquired short-term vital sign data set, and acquire the patient vital sign fluctuation level; based on the current patient vital sign fluctuation level, set the normal operating range of the equipment at the current monitoring time point;

[0068] The operation data monitoring module is used to obtain the equipment operation data at the current monitoring time point, determine whether the equipment operation data obtained at the current monitoring time point is within the preset normal operating range, and if so, execute the fault prediction plan; if not, issue an alarm and execute alarm response measures;

[0069] The fault prediction module is used to compare the overall vital sign data of the patient in the current update cycle with the blood pressure threshold, heart rate threshold and blood viscosity threshold. If the overall blood pressure value of the current patient is higher than the blood pressure threshold, fault prediction operations are performed on the filter element and the pump; if the overall heart rate value of the current patient is higher than the heart rate threshold, fault prediction operations are performed on the pump and the flow control system; if the overall blood viscosity value of the current patient is higher than the blood viscosity threshold, fault prediction operations are performed on the filter element, conductivity measurement module and flow control system.

[0070] It should be understood that those skilled in the art can make improvements or changes based on the above description, and all such improvements and changes should fall within the scope of protection of the appended claims of the present invention. Parts not described in detail in this specification belong to the prior art known to those skilled in the art.

Claims

1. A blood purification equipment data monitoring method, characterized in that: include: Obtain patient vital signs data and equipment operation data at the current monitoring time point. Patient vital signs data include blood pressure, heart rate and blood viscosity; equipment operation data include filter element pressure difference, filter element flow, pump output flow, pump pressure, dialysate conductivity, dialysate flow, system overall pressure and equipment temperature; The patient vital sign data acquired within the time window of the current preset duration T are combined into a short-term vital sign data set, the standard deviation of the patient vital sign data in the current short-term vital sign data set is calculated, and the patient vital sign fluctuation level is obtained; based on the current patient vital sign fluctuation level, the normal operating range of the equipment at the current monitoring time point is set; it is determined whether the equipment operating data acquired at the current monitoring time point is within the preset normal operating range, and if so, the fault prediction plan is executed; If not, an alarm is sounded and alarm response measures are executed; Execution of the fault prediction solution includes: An update cycle of 10T is set. At the beginning of any update cycle, the patient vital sign data acquired in the previous update cycle are combined into a long-term vital sign data set; the average value of the patient vital sign data in the acquired long-term vital sign data set is calculated to acquire the overall patient vital sign data in the current update cycle; Set the blood pressure threshold, heart rate threshold and blood viscosity threshold, and compare the overall blood pressure value, heart rate value and blood viscosity value in the overall vital sign data of the patient with the blood pressure threshold, heart rate threshold and blood viscosity threshold; if the overall blood pressure value of the current patient is higher than the blood pressure threshold, perform fault prediction operations on the filter element and the pump; if the overall heart rate value of the current patient is higher than the heart rate threshold, perform fault prediction operations on the pump and the flow control system; if the overall blood viscosity value of the current patient is higher than the blood viscosity threshold, perform fault prediction operations on the filter element, conductivity measurement module and flow control system; The fault prediction operations in the fault prediction scheme include: Predict the fault of the filter element, and compose the filter element pressure difference and filter element flow obtained at the current monitoring time point and the most recent N-1 monitoring time points into a filter element fault prediction set in chronological order. Use the filter element fault prediction set as the input of the filter element fault prediction model, output the filter element pressure difference prediction value and filter element flow prediction value at the next monitoring time point, and judge whether the obtained filter element pressure difference prediction value and filter element flow prediction value are both within the normal operating range of the equipment. If so, no operation is performed; if not, perform fault pre-processing measures for the filter element; Predict pump failures, organize the pump output flow and pump pressure obtained at the current monitoring time point and the most recent N-1 monitoring time points into a pump failure prediction set in chronological order, use the pump failure prediction set as the input of the pump failure prediction model, output the pump output flow prediction value and pump pressure prediction value at the next monitoring time point, and determine whether the obtained pump output flow prediction value and pump pressure prediction value are both within the normal operating range of the equipment. If so, no operation is performed; if not, perform fault pre-processing measures for the pump; Perform fault prediction on the conductivity measurement module, organize the dialysis fluid conductivity obtained at the current monitoring time point and the most recent N-1 monitoring time points into a conductivity fault prediction set in chronological order, use the conductivity fault prediction set as the input of the conductivity fault prediction model, output the predicted value of the dialysis fluid conductivity at the next monitoring time point, and determine whether the obtained predicted value of the dialysis fluid conductivity is within the normal operating range of the equipment. If so, no operation is performed; if not, perform fault pre-processing measures for the conductivity measurement module; Fault prediction is performed on the liquid flow control system. The dialysate flow obtained at the current monitoring time point and the most recent N-1 monitoring time points are organized into a flow fault prediction set in chronological order. The flow fault prediction set is used as the input of the flow fault prediction model. The dialysate flow prediction value at the next monitoring time point is output. It is determined whether the obtained dialysate flow prediction value is within the normal operating range of the equipment. If so, no operation is performed; if not, fault pre-processing measures are performed on the liquid flow control system.

2. The blood purification equipment data monitoring method according to claim 1, characterized in that: Based on the current fluctuation level of the patient's vital signs, the normal operating range of the device at the current monitoring time point is set to include: Set the normal operating range and also set the standard deviation threshold and the maximum standard deviation ; If the standard deviation of the patient's vital signs data in the currently acquired short-term vital signs data set is Below or equal to the standard deviation threshold , then the normal operating range of the equipment at the current monitoring time point is the regular operating range; Set the increment range to , if the standard deviation of the patient's vital signs data in the currently acquired short-term vital signs data set is Above standard deviation threshold , then use the formula Calculate the gain rate , and the increase rate Applied to the normal operating range, obtain the normal operating range of the equipment at the current monitoring time point.

3. The blood purification equipment data monitoring method according to claim 1, characterized in that: The filter element fault prediction model is established based on the LSTM model, including N filter element fault prediction LSTM units, 1 first fully connected layer and 1 first output layer. The filter element fault prediction LSTM unit is used to extract the temporal characteristics of the filter element fault prediction set; the first fully connected layer is used to extract features from the outputs of all filter element fault prediction LSTM units; the first output layer is used to output the filter element pressure difference prediction value and the filter element flow prediction value at the next monitoring time point; The training of the filter element failure prediction model includes: A plurality of filter element failure training samples are obtained, each of which stores N+1 filter element pressure differences and filter element flow rates sorted by time, and the first N filter element pressure differences and filter element flow rates in the filter element failure training samples are used as the input of the filter element failure prediction model, and the last filter element pressure difference and filter element flow rate in the filter element failure training samples are used as the output of the filter element failure prediction model; all filter element failure training samples are divided into a filter element failure training set and a filter element failure test set, and the filter element failure training set is sent to the filter element failure prediction model with parameter initialization for training, and then the filter element failure test set is input to the filter element failure prediction model to obtain a first test result; a first training condition is set to determine whether the first test result meets the first training condition, and if so, the trained filter element failure prediction model is output; otherwise, the filter element failure prediction model is continuously trained through the filter element failure training set.

4. The blood purification equipment data monitoring method according to claim 3, characterized in that: The pump fault prediction model is established based on the LSTM model, including N pump fault prediction LSTM units, 1 second fully connected layer and 1 second output layer. The training of the pump fault prediction model includes: A plurality of pump fault training samples are obtained, each of which stores N+1 pump output flow rates and pump pressures sorted by time, and the first N pump output flow rates and pump pressures in the pump fault training samples are used as the input of the pump fault prediction model, and the last pump output flow rate and pump pressure in the pump fault training samples are used as the output of the pump fault prediction model; all pump fault training samples are divided into a pump fault training set and a pump fault test set, and the pump fault training set is sent to the pump fault prediction model with parameter initialization for training, and then the pump fault test set is input to the pump fault prediction model to obtain a second test result; a second training condition is set to determine whether the second test result meets the second training condition, and if so, the trained pump fault prediction model is output; otherwise, the pump fault prediction model is continuously trained using the pump fault training set.

5. The blood purification equipment data monitoring method according to claim 4, characterized in that: The conductivity fault prediction model is established based on the LSTM model, including N conductivity fault prediction LSTM units, 1 third fully connected layer and 1 third output layer. The training of the conductivity fault prediction model includes: A plurality of conductivity fault training samples are obtained, each of which stores N+1 dialysis fluid conductivities sorted by time, and the first N dialysis fluid conductivities in the conductivity fault training samples are used as inputs of the conductivity fault prediction model, and the last dialysis fluid conductivity in the conductivity fault training samples is used as outputs of the conductivity fault prediction model; all conductivity fault training samples are divided into a conductivity fault training set and a conductivity fault test set, and the conductivity fault training set is sent to the conductivity fault prediction model with parameter initialization for training, and then the conductivity fault test set is input to the conductivity fault prediction model to obtain a third test result; a third training condition is set, and it is determined whether the third test result meets the third training condition. If the third training condition is met, the trained conductivity fault prediction model is output; otherwise, the conductivity fault prediction model is continuously trained through the conductivity fault training set.

6. The blood purification equipment data monitoring method according to claim 5, characterized in that: The traffic fault prediction model is built based on the LSTM model, including N traffic fault prediction LSTM units, one fourth fully connected layer, and one fourth output layer. The training of the traffic fault prediction model includes: A plurality of flow fault training samples are obtained, each flow fault training sample stores N+1 dialysate flow rates sorted by time, the first N dialysate flow rates in the flow fault training sample are used as the input of the flow fault prediction model, and the last dialysate flow rate in the flow fault training sample is used as the output of the flow fault prediction model; all flow fault training samples are divided into a flow fault training set and a flow fault test set, the flow fault training set is sent to the parameter-initialized flow fault prediction model for training, and then the flow fault test set is input to the flow fault prediction model to obtain a fourth test result; a fourth training condition is set to determine whether the fourth test result meets the fourth training condition, and if so, the trained flow fault prediction model is output; otherwise, the flow fault prediction model is continuously trained using the flow fault training set.

7. A blood purification equipment data monitoring system, characterized in that: The system is applied to the blood purification equipment data monitoring method according to any one of claims 1 to 6, comprising: The patient vital sign data acquisition module is used to acquire the patient vital sign data, and form the patient vital sign data acquired within the time window of the current preset duration T into a short-term vital sign data set, calculate the standard deviation of the patient vital sign data in the current acquired short-term vital sign data set, and acquire the patient vital sign fluctuation level; based on the current patient vital sign fluctuation level, set the normal operating range of the equipment at the current monitoring time point; The operation data monitoring module is used to obtain the equipment operation data at the current monitoring time point, determine whether the equipment operation data obtained at the current monitoring time point is within the preset normal operating range, and if so, execute the fault prediction plan; if not, issue an alarm and execute alarm response measures; The fault prediction module is used to compare the overall vital sign data of the patient in the current update cycle with the blood pressure threshold, heart rate threshold and blood viscosity threshold. If the overall blood pressure value of the current patient is higher than the blood pressure threshold, fault prediction operations are performed on the filter element and the pump; if the overall heart rate value of the current patient is higher than the heart rate threshold, fault prediction operations are performed on the pump and the flow control system; if the overall blood viscosity value of the current patient is higher than the blood viscosity threshold, fault prediction operations are performed on the filter element, conductivity measurement module and flow control system.

Images