A method and system for monitoring patient health data during hemodialysis
By monitoring blood pressure and heart rate data in real time, calculating the heart rate fluctuation value and dynamically adjusting the frequency, combining LSTM model and genetic algorithm optimization, the discontinuity and timeliness problems in traditional monitoring methods are solved, and the health data of patients during hemodialysis is realized The precise monitoring and personalized monitoring improve the accuracy of heart rate abnormality prediction.
Patent Information
- Application Number
- CN202411237187.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-05
- Publication Date
- 2024-11-05
- Estimated Expiration
- 2044-09-05
AI Technical Summary
There are discontinuity and timeliness in monitoring patient health data during traditional hemodialysis, and existing methods are difficult to respond to mutations in patients' health status in real time, and there is a lack of effective prediction models to accurately predict and adjust monitoring frequency.
By monitoring blood pressure and heart rate data in real time, calculate short-term and long-term heart rate fluctuation values, set adjustment thresholds based on long-term fluctuation values, dynamically adjust the analysis frequency of heart rate data, and optimize the adjustment coefficients using LSTM model and genetic algorithm to perform frequency adjustment and Heart rate prediction measures to achieve accurate prediction and treatment of heart rate abnormalities.
Improve the timeliness and continuity of monitoring, can respond more accurately to changes in patients' health status, provide more detailed and personalized monitoring, improve the accuracy of predicting heart rate abnormalities, warning of potential health risks in advance, and provide a more comprehensive clinical decision-making. data support.
Smart Images

Figure CN118749932B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of data monitoring, and in particular to a method and system for monitoring patient health data during hemodialysis. Background Art
[0002] Hemodialysis is a common treatment for chronic renal failure. It uses an artificial kidney to simulate the filtering function of a normal kidney, helping to remove waste and excess water from the blood. During hemodialysis, the patient's health status needs to be closely monitored to ensure the safety and effectiveness of the dialysis process. However, traditional monitoring methods have some limitations.
[0003] First, traditional monitoring methods usually rely on fixed monitoring frequencies, such as blood pressure and heart rate measurements, which may lead to discontinuity and timeliness issues in the monitoring data; second, when abnormal data is monitored, existing treatment measures may not be timely or accurate enough to quickly respond to sudden changes in the patient's health status; in addition, the analysis of long-term monitoring data often lacks effective predictive models, making it difficult to accurately predict patients' heart rate fluctuations and adjust the monitoring frequency.
[0004] Although existing studies have made some progress in improving monitoring efficiency and prediction accuracy, there is still a lack of a patient health data monitoring method during hemodialysis that can comprehensively consider blood pressure and heart rate data and adjust the monitoring frequency in real time. Summary of the Invention
[0005] The purpose of the present invention is to overcome the deficiencies of the prior art and provide a method and system for monitoring patient health data during hemodialysis.
[0006] In one aspect, the present invention provides a method for monitoring patient health data during hemodialysis, comprising:
[0007] At a first monitoring time point, blood pressure data is acquired, including systolic and diastolic pressures; and at the same time, it is determined whether the acquired blood pressure data is within a normal blood pressure range. If not, an alarm is issued and abnormal blood pressure treatment measures are implemented;
[0008] Acquire heart rate data at the current second monitoring time point; determine at a preset analysis time point whether the heart rate data acquired during the current analysis time period are all within a normal heart rate range; if not, issue an alarm and implement heart rate abnormality treatment measures; the time between each two analysis time points is an analysis time period, and an analysis time period includes several second monitoring time points;
[0009] For any analysis time point, a short-term heart rate data set within a current short-term time window is obtained, the short-term heart rate data set includes the heart rate data obtained within the current short-term time window, and a short-term heart rate fluctuation degree value of the short-term heart rate data set is calculated; simultaneously, a long-term heart rate data set within a current long-term time window is obtained, the long-term heart rate data set includes the heart rate data obtained within the current long-term time window, and a long-term heart rate fluctuation degree value of the long-term heart rate data set is calculated; and a first adjustment threshold and a second adjustment threshold are set based on the currently obtained long-term heart rate fluctuation degree value;
[0010] If the short-term heart rate fluctuation value is greater than the first adjustment threshold and less than the second adjustment threshold, a frequency adjustment measure is performed, that is, the length of the analysis period and the time interval between each two first monitoring time points are halved;
[0011] If the short-term heart rate fluctuation value is greater than the second adjustment threshold, then on the basis of executing frequency adjustment measures, a heart rate data prediction set is obtained, the heart rate data prediction set is used as the input of the heart rate prediction model, and a predicted heart rate data set is output; it is determined whether the predicted heart rate data sets are all within the normal heart rate range. If not, an alarm is issued and abnormal heart rate preprocessing measures are executed.
[0012] Preferably, the specific operations for calculating the short-term heart rate fluctuation value and the long-term heart rate fluctuation value are as follows:
[0013] Get the short-term heart rate dataset within the current short-term time window , i =1, 2, ..., n , n Indicates the number of heart rate data obtained in the current short-term time window; calculate and obtain short-term heart rate data set Average value , using the formula Calculate and obtain short-term heart rate data set The standard deviation A of the short-term heart rate fluctuation value; obtain the long-term heart rate data set within the current long-term time window , j = 1, 2, ..., m , m Indicates the number of heart rate data obtained in the current long-term time window; calculate and obtain the long-term heart rate data set Average value , using the formula Calculate and obtain long-term heart rate data sets The standard deviation B is the long-term heart rate fluctuation value.
[0014] Preferably, the specific operations of setting the first adjustment threshold and the second adjustment threshold based on the long-term heart rate fluctuation value are as follows:
[0015] Using the formula Calculate and obtain the first adjustment threshold , using the formula Calculate and obtain the second adjustment threshold , where k is the adjustment coefficient and is between (0, 1). The value of k is obtained by genetic algorithm selection.
[0016] Preferably, after executing the frequency adjustment measures, for any analysis time point, continue to obtain the short-term heart rate data set within the short-term time window, and calculate the short-term heart rate fluctuation degree value of the short-term heart rate data set; determine whether there are three consecutive analysis time points where the short-term heart rate fluctuation degree values obtained are all lower than the first adjustment threshold. If so, stop executing the frequency adjustment measures, that is, simultaneously restore the length of the analysis time period and the time interval between every two first monitoring time points; if not, continue to execute the frequency adjustment measures.
[0017] Preferably, the specific operation of using the genetic algorithm to select and obtain the value of k is as follows:
[0018] Step 1: Set the population size, randomly generate x candidate k values in the range of (0, 1), form the randomly generated candidate k values into the initial population, and set the maximum number of iterations;
[0019] Step 2: For any candidate k value, use the formula Calculate the fitness D of the candidate k value, where F Indicates the false alarm rate of applying the candidate k value for abnormal heart rate monitoring. U Indicates the missed reporting rate of heart rate abnormality monitoring using the candidate k value, and False alarm rate F and false negative rate U The weight coefficient of the calculated fitness is lower, which means the candidate k value performs better.
[0020] Step 3: Apply the roulette wheel selection method to select the corresponding number of candidate k values with the best fitness performance according to the preset selection ratio, and form the selected candidate k values into a selection individual set;
[0021] Step 4: Apply the preset crossover probability to perform a crossover operation on the candidate k values in the selected individual set to generate a corresponding number of new candidate k values, and form the obtained new candidate k values into a crossover individual set; randomly select candidate k values from the crossover individual set, and adjust the candidate k values in a small range according to the preset mutation probability to obtain a mutation individual set;
[0022] Step 5: Merge the selected individual set and the mutated individual set to obtain the merged individual set, calculate the fitness of all candidate k values in the merged individual set, and use the x candidate k values with the best fitness performance as the new generation population;
[0023] Step 6: Repeat steps 3 to 5 to iteratively update the population until the maximum number of iterations is reached. The population with the best fitness performance in the last generation is the optimal k value.
[0024] Preferably, the heart rate data prediction set is used as the input of the heart rate prediction model, and the specific operation of outputting the predicted heart rate data set is as follows:
[0025] Obtain a heart rate data prediction set, which includes the Q most recently obtained heart rate data sorted by time; set a sliding time window, and the window length of the sliding time window is Q data points; use the Q heart rate data in the heart rate data prediction set as the data sequence of the sliding time window; use the data sequence as the input of the heart rate prediction model, and output the predicted heart rate data of the next second monitoring time point; then remove the first heart rate data in the sliding time window, use the predicted heart rate data of the next second monitoring time point as the last data of the sliding time window, obtain a new data sequence, use the new data sequence as the input of the heart rate prediction model, and continue to output the predicted heart rate data of the next second monitoring time point, and so on, to obtain a predicted heart rate data set containing several predicted heart rate data.
[0026] Preferably, the heart rate prediction model is established based on an LSTM model, including Q heart rate prediction LSTM units, 1 fully connected layer, and 1 output layer, wherein the heart rate prediction LSTM unit is used to extract the temporal features of the heart rate data in the sliding time window; the fully connected layer is used to extract features from the outputs of all heart rate prediction LSTM units; and the output layer is used to output the heart rate prediction data;
[0027] The specific operations for training the heart rate prediction model are as follows:
[0028] Obtain several training samples, each of which stores Q+1 heart rate data sorted by time, use the first Q heart rate data in the training sample as the input of the heart rate prediction model, and use the last heart rate data in the training sample as the output of the heart rate prediction model; divide all training samples into training sets and test sets, send the training set to the parameter-initialized heart rate prediction model for training, and then input the test set to the heart rate prediction model to obtain the test results; set training conditions, and judge whether the test results meet the training conditions. If the training conditions are met, output the trained heart rate prediction model; otherwise, continue to train the heart rate prediction model through the training set.
[0029] On the other hand, the present invention further provides a system for monitoring patient health data during hemodialysis, the system being applied to the method for monitoring patient health data during hemodialysis as described above, the system comprising:
[0030] A data acquisition module, configured to acquire blood pressure data at a first monitoring time point and heart rate data at a current second monitoring time point;
[0031] A heart rate fluctuation abnormality discrimination module is configured to calculate a short-term heart rate fluctuation degree value and a long-term heart rate fluctuation degree value, and set a first adjustment threshold and a second adjustment threshold value based on the current long-term heart rate fluctuation degree value; if the short-term heart rate fluctuation degree value is greater than the first adjustment threshold value and less than the second adjustment threshold value, then perform frequency adjustment measures; if the short-term heart rate fluctuation degree value is greater than the second adjustment threshold value, then, based on the execution of the frequency adjustment measures, use the acquired heart rate data prediction set as the input of the heart rate prediction model, and output a predicted heart rate data set;
[0032] The exception handling module is used to determine whether the blood pressure data obtained at the first detection time point is within the normal blood pressure range. If not, an alarm is issued and blood pressure abnormality handling measures are implemented; at a preset analysis time point, it is determined whether the heart rate data obtained within the current analysis time period are all within the normal heart rate range. If not, an alarm is issued and heart rate abnormality handling measures are implemented; and it is determined whether the predicted heart rate data set is all within the normal heart rate range. If not, an alarm is issued and heart rate abnormality pre-processing measures are implemented.
[0033] The present invention has the following advantages:
[0034] The present invention monitors the patient's blood pressure and heart rate data in real time, calculates short-term and long-term heart rate fluctuation values based on the acquired heart rate data, and calculates an adjustment threshold for judging whether the current patient's heart rate fluctuation has abnormalities based on the long-term heart rate fluctuation value, and dynamically adjusts the analysis frequency of the heart rate data according to the short-term heart rate fluctuation value; this adaptive adjustment mechanism can respond to changes in the patient's health status more accurately, while paying more attention to the current patient's physiological characteristics, rather than judging through fixed thresholds, thereby improving the timeliness and continuity of monitoring; compared with traditional fixed frequencies, the present invention can more flexibly respond to fluctuations in the patient's health status, thereby providing more detailed and personalized monitoring.
[0035] The present invention adopts heart rate prediction technology based on the LSTM model, combined with an adjustment coefficient optimized by a genetic algorithm. When the short-term heart rate fluctuation value exceeds the preset threshold, the system will not only perform frequency adjustment measures, but also use the heart rate prediction model to predict future heart rate data to further determine whether there is an abnormal risk, thereby improving the accuracy of heart rate abnormality prediction. This method, which combines real-time monitoring and predictive analysis, can effectively warn of potential health risks in advance and provide more comprehensive data support for clinical decision-making. BRIEF DESCRIPTION OF THE DRAWINGS
[0036] Figure 1 This is a schematic diagram of the structure of a patient health data monitoring system during hemodialysis used in an embodiment of the present invention. DETAILED DESCRIPTION
[0037] In order to enable those 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 provides a method for monitoring patient health data during hemodialysis, comprising:
[0039] Blood pressure data is obtained at the first monitoring time point. Blood pressure is the pressure of blood on the blood vessel wall and is usually expressed by two values: systolic pressure and diastolic pressure. Systolic pressure is the pressure of blood on the blood vessel wall when the heart contracts, while diastolic pressure is the pressure when the heart relaxes. The normal range of blood pressure varies from individual to individual, but generally systolic pressure of 90-120 mmHg and diastolic pressure of 60-80 mmHg are considered normal. The normal range is generally set by professional medical staff. It is determined whether the obtained blood pressure data is within the normal blood pressure range. If not, an alarm is issued. This alarm can be an audible signal, a visual indication, or other means to notify medical staff to ensure that they can promptly notice the patient's abnormal blood pressure and implement blood pressure abnormality treatment measures, which include adjusting the parameters of the dialysis machine, giving the patient medication, or taking other necessary medical interventions to stabilize the patient's blood pressure and prevent potential complications.
[0040] At the current second monitoring time point, the heart rate data is obtained. The heart rate is the number of heartbeats per minute and is usually used to evaluate heart function and overall health status. At the preset analysis time point, it is determined whether the heart rate data obtained during the current analysis time period are all within the normal heart rate range. The normal heart rate is usually between 60-100 beats / minute, but this range can be adjusted by professional medical staff according to the patient's age, health status and individual differences. If not, an alarm will be issued and abnormal heart rate treatment measures will be implemented, including adjusting dialysis parameters, giving drug treatment, changing the patient's position or other appropriate medical interventions to restore the normal heart rate and prevent potential problems. Heart problems; the time between every two analysis time points is an analysis period, and an analysis period includes several second monitoring time points; since heart rate data is mostly acquired through electrocardiogram monitoring, the frequency of data acquisition is close to real-time, and the setting of analysis time points is to avoid making judgments every time data is acquired, which would put too much pressure on the system; blood pressure is usually monitored intermittently, that is, a cuff-type blood pressure monitor can be used to monitor every half an hour to an hour, so there is no need to set analysis time points for blood pressure monitoring, and the blood pressure can be immediately judged whether it is abnormal after obtaining the blood pressure data;
[0041] For any analysis time point, a short-term heart rate dataset within the current short-term time window is obtained. The short-term heart rate dataset includes heart rate data obtained within the current short-term time window. The short-term time window refers to a short period of time before the current analysis time point, such as five minutes or ten minutes. These data reflect the patient's heart rate changes in a short period of time. The short-term heart rate fluctuation degree value of the short-term heart rate dataset is calculated. The larger the fluctuation degree value, the more severe the heart rate fluctuation, which may indicate that the patient has heart problems or reactions to the dialysis process; at the same time, a long-term heart rate dataset within the current long-term time window is obtained. The long-term heart rate dataset includes heart rate data obtained within the current long-term time window. The long-term time window refers to a longer period of time before the current analysis time point, such as 1 to 2 hours. These data reflect the patient's heart rate fluctuation characteristics over a longer period of time. The long-term heart rate fluctuation degree value of the long-term heart rate dataset is calculated. This value helps to evaluate the overall stability of the patient's heart rate changes during the dialysis process; a first adjustment threshold and a second adjustment threshold are set based on the currently obtained long-term heart rate fluctuation degree value. These two thresholds are used to determine whether the short-term heart rate fluctuation is within an acceptable range;
[0042] If the short-term heart rate fluctuation value is greater than the first adjustment threshold and less than the second adjustment threshold, this indicates that the patient's heart rate changes exceed the lower limit of the normal range, and frequency adjustment measures are implemented, that is, the length of the analysis period and the time interval between each two first monitoring time points are halved at the same time. This can timely detect when the heart rate and blood pressure exceed the normal value and prevent the patient from being in danger. The reason for increasing the frequency of blood pressure monitoring when there are abnormal fluctuations in heart rate is that there is a certain correlation between heart rate and blood pressure. Heart rate and blood pressure jointly determine cardiac output (the amount of blood pumped by the heart per minute), which is calculated by multiplying the ejection volume (stroke volume) of the heart with each beat by the heart rate. Changes in heart rate can affect blood pressure. For example, an increased heart rate usually leads to an increase in systolic blood pressure, while diastolic blood pressure may not change much or may decrease slightly. Abnormal heart rate is often accompanied by abnormal blood pressure.
[0043] If the short-term heart rate fluctuation value is greater than the second adjustment threshold, this may indicate that the patient has a higher health risk. Then, on the basis of executing frequency adjustment measures, a heart rate data prediction set is obtained, and the heart rate data prediction set is used as the input of the heart rate prediction model to output a predicted heart rate data set; it is determined whether the predicted heart rate data sets are all within the normal heart rate range. If not, an alarm is issued and abnormal heart rate pretreatment measures are performed, which include adjusting the settings of the dialysis machine, giving the patient appropriate medical intervention, or other preventive measures.
[0044] The specific operations for calculating the short-term heart rate fluctuation value and the long-term heart rate fluctuation value are as follows:
[0045] Get the short-term heart rate dataset within the current short-term time window , i=1,2,…, n , n Indicates the number of heart rate data obtained in the current short-term time window; calculate and obtain short-term heart rate data set Average value , using the formula Calculate and obtain short-term heart rate data set The standard deviation A of the short-term heart rate fluctuation value; obtain the long-term heart rate data set within the current long-term time window , j = 1, 2, ..., m , m Indicates the number of heart rate data obtained in the current long-term time window; calculate and obtain the long-term heart rate data set Average value , using the formula Calculate and obtain long-term heart rate data sets The standard deviation B is the long-term heart rate fluctuation value.
[0046] The specific operations for setting the first adjustment threshold and the second adjustment threshold based on the long-term heart rate fluctuation value are as follows:
[0047] Using the formula Calculate and obtain the first adjustment threshold , using the formula Calculate and obtain the second adjustment threshold , where k is the adjustment coefficient and is between (0, 1). The value of k is obtained through genetic algorithm selection. The genetic algorithm simulates natural selection and genetic mechanisms. The algorithm can perform a global search in the solution space to find the optimal or approximately optimal solution. The algorithm automatically adjusts the k value without human intervention, reducing the workload and subjectivity of parameter adjustment.
[0048] After executing the frequency adjustment measures, for any analysis time point, continue to obtain the short-term heart rate data set within the short-term time window, and calculate the short-term heart rate fluctuation degree value of the short-term heart rate data set; determine whether there are three consecutive analysis time points where the short-term heart rate fluctuation degree values obtained are all lower than the first adjustment threshold. This judgment is to determine whether the heart rate has been stable within the normal range for a period of time. If so, this indicates that the patient's heart rate has stabilized and no more intensive monitoring is required, then stop executing the frequency adjustment measures, that is, simultaneously restore the length of the analysis time period and the time interval between every two first monitoring time points; if not, this means that the patient's heart rate may still have unstable factors, then continue to execute the frequency adjustment measures.
[0049] The specific operation of using genetic algorithm to select the value of k is as follows:
[0050] Step 1: Set the population size, randomly generate x candidate k values in the range of (0, 1), form the randomly generated candidate k values into the initial population, and set the maximum number of iterations;
[0051] Step 2: For any candidate k value, use the formula Calculate the fitness D of the candidate k value, where F Indicates the false alarm rate of applying the candidate k value for abnormal heart rate monitoring. U Indicates the missed reporting rate of heart rate abnormality monitoring using the candidate k value, and False alarm rate F and false negative rate U The weight coefficient of the calculated fitness is lower, which means the candidate k value performs better.
[0052] Step 3: Apply the roulette wheel selection method to select the corresponding number of candidate k values with the best fitness performance at a preset selection ratio, and form the selected candidate k values into a selection set; this step imitates the natural selection process and retains individuals with higher fitness for continued optimization in the next generation;
[0053] Step 4: Apply the preset crossover probability to perform a crossover operation on the candidate k values in the selected individual set to generate a corresponding number of new candidate k values. The obtained new candidate k values are combined into a crossover individual set. New genetic feature combinations are generated through crossover, which increases population diversity and improves the possibility of finding the global optimal solution. Randomly select candidate k values from the crossover individual set, and adjust the candidate k values in a small range according to the preset mutation probability to obtain a mutant individual set. The mutation is to simulate gene mutation, provide new genetic information for the algorithm, and prevent the algorithm from converging to the local optimal solution too early.
[0054] Step 5: Merge the selected individual set and the mutated individual set to obtain the merged individual set, calculate the fitness of all candidate k values in the merged individual set, and use the x candidate k values with the best fitness performance as the new generation population;
[0055] Step 6: Repeat steps 3 to 5 to iteratively update the population until the maximum number of iterations is reached. The population with the best fitness performance in the last generation is the optimal k value.
[0056] The specific operations of using the predicted heart rate data set as the input of the heart rate prediction model and outputting the predicted heart rate data set are as follows:
[0057] Obtain a heart rate data prediction set, which includes the Q most recently obtained heart rate data sorted by time; set a sliding time window, and the window length of the sliding time window is Q data points; use the Q heart rate data in the heart rate data prediction set as the data sequence of the sliding time window; use the data sequence as the input of the heart rate prediction model, and output the predicted heart rate data of the next second monitoring time point; then remove the first heart rate data in the sliding time window, use the predicted heart rate data of the next second monitoring time point as the last data of the sliding time window, obtain a new data sequence, use the new data sequence as the input of the heart rate prediction model, and continue to output the predicted heart rate data of the next second monitoring time point, and so on, to obtain a predicted heart rate data set containing several predicted heart rate data.
[0058] The heart rate prediction model is built based on the LSTM model, which includes Q heart rate prediction LSTM units, one fully connected layer, and one output layer. The heart rate prediction LSTM unit is used to extract the temporal features of the heart rate data in the sliding time window; the fully connected layer is used to extract features from the outputs of all heart rate prediction LSTM units; and the output layer is used to output the heart rate prediction data.
[0059] The specific operations for training the heart rate prediction model are as follows:
[0060] Obtain several training samples, each of which stores Q+1 heart rate data sorted by time, use the first Q heart rate data in the training sample as the input of the heart rate prediction model, and use the last heart rate data in the training sample as the output of the heart rate prediction model; divide all training samples into training sets and test sets, send the training set to the parameter-initialized heart rate prediction model for training, and then input the test set to the heart rate prediction model to obtain the test results; set training conditions, and judge whether the test results meet the training conditions. If the training conditions are met, output the trained heart rate prediction model; otherwise, continue to train the heart rate prediction model through the training set.
[0061] Embodiment 2 of the present invention is a system for monitoring patient health data during hemodialysis, such as Figure 1 Shown, including:
[0062] A data acquisition module, configured to acquire blood pressure data at a first monitoring time point and heart rate data at a current second monitoring time point;
[0063] A heart rate fluctuation abnormality discrimination module is configured to calculate a short-term heart rate fluctuation degree value and a long-term heart rate fluctuation degree value, and set a first adjustment threshold and a second adjustment threshold value based on the current long-term heart rate fluctuation degree value; if the short-term heart rate fluctuation degree value is greater than the first adjustment threshold value and less than the second adjustment threshold value, then perform frequency adjustment measures; if the short-term heart rate fluctuation degree value is greater than the second adjustment threshold value, then, based on the execution of the frequency adjustment measures, use the acquired heart rate data prediction set as the input of the heart rate prediction model, and output a predicted heart rate data set;
[0064] The exception handling module is used to determine whether the blood pressure data obtained at the first detection time point is within the normal blood pressure range. If not, an alarm is issued and blood pressure abnormality handling measures are implemented; at a preset analysis time point, it is determined whether the heart rate data obtained within the current analysis time period are all within the normal heart rate range. If not, an alarm is issued and heart rate abnormality handling measures are implemented; and it is determined whether the predicted heart rate data set is all within the normal heart rate range. If not, an alarm is issued and heart rate abnormality pre-processing measures are implemented.
[0065] It should be understood that those skilled in the art may make improvements or modifications based on the above description, and all such improvements and modifications shall fall within the scope of protection of the appended claims. Any portion of this specification not described in detail is prior art known to those skilled in the art.
Claims
1. A method for monitoring patient health data during hemodialysis, characterized in that: include: At the first monitoring time point, blood pressure data is obtained, the blood pressure data including systolic pressure and diastolic pressure; at the same time, it is determined whether the obtained blood pressure data is within the normal blood pressure range, if not, an alarm is issued, and abnormal blood pressure treatment measures are executed; Acquire heart rate data at the current second monitoring time point; determine at a preset analysis time point whether the heart rate data acquired in the current analysis time period are all within the normal heart rate range, if not, issue an alarm and execute heart rate abnormality processing measures; the time between every two analysis time points is an analysis time period, and one analysis time period includes several second monitoring time points; For any analysis time point, a short-term heart rate data set within the current short-term time window is obtained, the short-term heart rate data set includes the heart rate data obtained within the current short-term time window, and a short-term heart rate fluctuation degree value of the short-term heart rate data set is calculated; at the same time, a long-term heart rate data set within the current long-term time window is obtained, the long-term heart rate data set includes the heart rate data obtained within the current long-term time window, and a long-term heart rate fluctuation degree value of the long-term heart rate data set is calculated; a first adjustment threshold and a second adjustment threshold are set based on the currently obtained long-term heart rate fluctuation degree value; If the short-term heart rate fluctuation value is greater than the first adjustment threshold and less than the second adjustment threshold, a frequency adjustment measure is performed, that is, the length of the analysis time period and the time interval between every two first monitoring time points are halved at the same time; If the short-term heart rate fluctuation value is greater than the second adjustment threshold, then on the basis of executing the frequency adjustment measures, a heart rate data prediction set is obtained, the heart rate data prediction set is used as the input of the heart rate prediction model, and a predicted heart rate data set is output; it is determined whether the predicted heart rate data sets are all within the normal heart rate range, if not, an alarm is issued, and abnormal heart rate preprocessing measures are executed; The specific operations for calculating the short-term heart rate fluctuation degree value and the long-term heart rate fluctuation degree value are as follows: Get the short-term heart rate dataset within the current short-term time window , i =1, 2, …, n , n Indicates the number of heart rate data obtained in the current short-term time window; calculate and obtain short-term heart rate data set The average , using the formula Calculate and obtain short-term heart rate data set The standard deviation A of the short-term heart rate fluctuation value; obtain the long-term heart rate data set in the current long-term time window , j =1, 2, …, m , m Indicates the number of heart rate data obtained in the current long-term time window; calculate and obtain the long-term heart rate data set The average , using the formula Calculate and obtain long-term heart rate data set The standard deviation B is the value of long-term heart rate fluctuation; The specific operation of setting the first adjustment threshold and the second adjustment threshold based on the long-term heart rate fluctuation value is as follows: Using the formula Calculate and obtain the first adjustment threshold Using the formula Calculate and obtain the second adjustment threshold , where k is the adjustment coefficient and is between (0, 1), and the value of k is obtained by genetic algorithm selection; After executing the frequency adjustment measure, for any analysis time point, continue to obtain the short-term heart rate data set within the short-term time window, and calculate the short-term heart rate fluctuation degree value of the short-term heart rate data set; determine whether there are three consecutive analysis time points where the short-term heart rate fluctuation degree values obtained are all lower than the first adjustment threshold value, if so, stop executing the frequency adjustment measure, that is, simultaneously restore the length of the analysis time period and the time interval between every two first monitoring time points; if not, continue to execute the frequency adjustment measure; The specific operation of using genetic algorithm to select the value of k is as follows: Step 1: Set the population size, randomly generate x candidate k values in the range of (0, 1), form the randomly generated candidate k values into the initial population, and set the maximum number of iterations; Step 2: For any candidate k value, use the formula Calculate the fitness D of the candidate k value, where F Indicates the false alarm rate of using the candidate k value for abnormal heart rate monitoring. U It indicates the underreporting rate of heart rate abnormality monitoring using the candidate k value. and False alarm rate F and false negative rate U The weight coefficient of ; the lower the calculated fitness, the better the performance of the candidate k value; Step 3: Apply the roulette wheel selection method to select the corresponding number of candidate k values with the best fitness performance according to the preset selection ratio, and form the selected candidate k values into a selection individual set; Step 4: Apply the preset crossover probability to perform a crossover operation on the candidate k values in the selected individual set to generate a corresponding number of new candidate k values, and form the obtained new candidate k values into a crossover individual set; randomly select candidate k values in the crossover individual set, and adjust the candidate k values in a small range according to the preset mutation probability to obtain a mutation individual set; Step 5: Merge the selected individual set and the mutated individual set to obtain the merged individual set, calculate the fitness of all candidate k values in the merged individual set, and take the x candidate k values with the best fitness performance as the new generation population; Step 6: Repeat steps 3 to 5 to iteratively update the population until the maximum number of iterations is reached. The population with the best fitness performance in the last generation is the optimal k value.
2. The method for monitoring patient health data during hemodialysis according to claim 1, characterized in that: The specific operations of taking the predicted heart rate data set as the input of the heart rate prediction model and outputting the predicted heart rate data set are as follows: Obtain a heart rate data prediction set, which includes Q heart rate data recently obtained and sorted by time; set a sliding time window, and the window length of the sliding time window is Q data points; Use the Q heart rate data of the heart rate data prediction set as the data sequence of the sliding time window; use the data sequence as the input of the heart rate prediction model, and output the predicted heart rate data at the next second monitoring time point; Then, the first heart rate data in the sliding time window is removed, and the predicted heart rate data at the next second monitoring time point is used as the last data in the sliding time window to obtain a new data sequence, and the new data sequence is used as the input of the heart rate prediction model to continue to output the predicted heart rate data at the next second monitoring time point, and so on, to obtain a predicted heart rate data set containing several predicted heart rate data.
3. The method for monitoring patient health data during hemodialysis according to claim 2, characterized in that: The heart rate prediction model is built based on the LSTM model, including Q heart rate prediction LSTM units, 1 fully connected layer and 1 output layer. The heart rate prediction LSTM unit is used to extract the temporal characteristics of the heart rate data in the sliding time window; the fully connected layer is used to extract features from the outputs of all heart rate prediction LSTM units; the output layer is used to output the heart rate prediction data; The specific operations for training the heart rate prediction model are as follows: Obtain several training samples, each of which stores Q+1 heart rate data sorted by time, using the first Q heart rate data in the training sample as the input of the heart rate prediction model, and using the last heart rate data in the training sample as the output of the heart rate prediction model; All training samples are divided into a training set and a test set. The training set is sent to the heart rate prediction model with initialized parameters for training. Then the test set is input to the heart rate prediction model to obtain the test results. Set training conditions and determine whether the test results meet the training conditions. If the training conditions are met, output the trained heart rate prediction model. Otherwise, continue to train the heart rate prediction model using the training set.
4. A patient health data monitoring system during hemodialysis, characterized in that: The system is applied to the method for monitoring patient health data during hemodialysis according to any one of claims 1 to 3, and the system comprises: A data acquisition module, used to acquire blood pressure data at a first monitoring time point and acquire heart rate data at a current second monitoring time point; The heart rate fluctuation abnormality discrimination module is used to calculate the short-term heart rate fluctuation degree value and the long-term heart rate fluctuation degree value, and set the first adjustment threshold and the second adjustment threshold based on the current long-term heart rate fluctuation degree value; if the short-term heart rate fluctuation degree value is greater than the first adjustment threshold and less than the second adjustment threshold, the frequency adjustment measure is executed; if the short-term heart rate fluctuation degree value is greater than the second adjustment threshold, on the basis of executing the frequency adjustment measure, the acquired heart rate data prediction set is used as the input of the heart rate prediction model, and the predicted heart rate data set is output; The abnormality handling module is used to determine whether the blood pressure data obtained at the first detection time point is within the normal range of blood pressure. If not, an alarm is issued and blood pressure abnormality handling measures are executed; at a preset analysis time point, determine whether the heart rate data obtained in the current analysis time period are all within the normal range of heart rate. If not, an alarm is issued and heart rate abnormality handling measures are executed; determine whether the predicted heart rate data sets are all within the normal range of heart rate. If not, an alarm is issued and heart rate abnormality preprocessing measures are executed.
Citation Information
Patent Citations
Blood pressure monitoring system and method based on ECG signal and PPG signal
CN113749630A
System and method for automatically detecting vital signs for handheld PDA (Personal Digital Assistant)
CN117850601A
Intelligent health monitoring method and system
CN117954135A
Big-data-driven health status assessment method for middle-aged and elderly patients
CN118098545A
System and method for optimization of a database for the training and testing of prediction algorithms
US20060230006A1
