Critical medicine patient monitoring system

Abstract

The invention relates to the technical field of medical information, in particular to an intensive care medical patient monitoring system which comprises a dynamic threshold value calculation module used for setting an initial threshold value including a daytime mode blood pressure fluctuation tolerance value based on physical sign data and forcibly triggering an early warning signal when the blood pressure does not reach the initial threshold value; the night sensing and anti-interference module is used for generating a sleep depth parameter through a weighting algorithm according to the physical sign data and the environment data, synchronously transmitting the sleep depth parameter and the night specific physical sign data to the dynamic threshold value calculation module, and if the sleep depth parameter received by the dynamic threshold value calculation module continuously exceeds a preset threshold value, sending the sleep depth parameter to the night sensing and anti-interference module; if yes, the blood pressure fluctuation tolerance is improved to 1.2 times of the daytime mode, and organ perfusion critical parameters are obtained. By integrating the functions of multi-source data real-time processing, dynamic threshold self-adaptive adjustment, night physiological rhythm perception and interference resistance, false alarm intelligent correction and the like, the system is used for realizing complete-cycle accurate monitoring and timely early warning of critical patients.

Description

Technical Field

[0001] The present invention relates to the field of medical information technology, and in particular to a critical care patient monitoring system. Background Art

[0002] Critical care patients often experience unstable vital signs and complex, ever-changing conditions, and their organ functions may deteriorate rapidly at any time. Real-time monitoring of critical care patients is crucial. On the one hand, it allows for continuous and dynamic monitoring of changes in patients' vital signs, such as heart rate, blood pressure, and blood oxygen saturation, enabling timely detection of potential physiological abnormalities. On the other hand, it provides a deep understanding of the patient's organ function and metabolic status, providing a key basis for doctors to formulate precise treatment plans. Through effective monitoring, doctors can take intervention measures in the early stages of disease progression, improve treatment effectiveness, reduce mortality and disability rates, and significantly improve patient prognosis. Therefore, critical care patient monitoring systems play an indispensable role in modern healthcare.

[0003] Most existing critical care patient monitoring systems trigger alarms based on fixed thresholds. This approach fails to fully account for individual patient differences and the dynamic evolution of their condition. Taking blood pressure monitoring as an example, the system typically sets a uniform low blood pressure threshold for early warning. However, each patient's baseline blood pressure fluctuates differently. For patients with normally high baseline blood pressure, when their blood pressure drops below the general threshold but remains within their tolerable range, the system will issue an alarm, which is clearly a false alarm. This false alarm not only interferes with medical staff's judgment but can also cause them to become numb to genuine abnormalities, reducing their sensitivity to alarms and thus compromising patient safety.

[0004] Furthermore, while patients' physiological states are relatively stable at night, potential changes can be difficult to detect in a timely manner. Conventional monitoring systems have significant deficiencies in the accuracy and specificity of their warnings during the night. For one thing, the dim ambient light and fewer distracting factors at night can lead to deviations in some monitoring algorithms designed based on daily activities. For example, the system may misjudge changes in the patient's heart rate caused by slight movements, issuing unnecessary alerts. Furthermore, the system is unable to dynamically adjust monitoring and warning strategies based on the specific physiological rhythms of the night, such as the impact of sleep cycles on natural fluctuations in blood pressure and heart rate. This can mask important changes in the patient's condition at night, and by the time medical staff discover them, the patient may have already entered a state of decompensation. At this point, emergency treatment is often passive and has a relatively low success rate.

[0005] In summary, the existing critical care patient monitoring system has many defects in threshold setting and nighttime monitoring, which makes it difficult to meet the clinical needs for accurate monitoring and timely intervention of critically ill patients. Therefore, the development of a more advanced and complete critical care patient monitoring system has important practical significance. Summary of the Invention

[0006] To solve the above problems, the present invention provides a critical care patient monitoring system, which integrates functions such as real-time processing of multi-source data, adaptive adjustment of dynamic thresholds, nighttime physiological rhythm perception and anti-interference, and intelligent correction of false alarms to achieve accurate monitoring and timely warning of critically ill patients throughout the entire cycle.

[0007] In order to achieve the above-mentioned object, the technical solution of the present invention is as follows: A critical care patient monitoring system comprising:

[0008] The device unit is used to monitor the patient's vital sign data and collect environmental data in the ward, the vital sign data including daytime basic vital sign data and nighttime specific vital sign data; the device unit signal is connected to a driving mechanism that automatically injects pressor drugs into the patient by receiving injection instructions;

[0009] The acquisition module is used to standardize and transform the vital sign data and environmental data to obtain a time series vital sign data set and a time series environmental data set, both of which are marked with the patient ID and timestamp; it is also used to obtain the patient's physical examination data and ultrasonic cardiac output monitoring data;

[0010] The data processing module is used to process the time series vital sign data set and the time series environmental data set. The processing is divided into short-term prediction processing and long-term prediction processing. Among them, the short-term prediction processing is used to predict the visual attenuation trend of the patient's cardiopulmonary function in the first time period in the future. The long-term prediction processing is used to generate the risk index of organ failure in the second time period in the future by combining the physical examination data;

[0011] A dynamic threshold calculation module is used to set an initial threshold value including a tolerance value for daytime blood pressure fluctuations based on vital sign data. If the blood pressure does not reach the initial threshold value, an early warning signal is forcibly triggered.

[0012] The nighttime perception and anti-interference module is used to generate a sleep depth parameter based on vital sign data and environmental data using a weighted algorithm. The sleep depth parameter and nighttime-specific vital sign data are synchronously transmitted to the dynamic threshold calculation module. If the sleep depth parameter received by the dynamic threshold calculation module continuously exceeds the preset threshold, the blood pressure fluctuation tolerance is increased to 1.2 times that of the daytime mode to obtain the critical organ perfusion parameter;

[0013] It is also used to activate the anti-motion artifact algorithm based on environmental data. The anti-motion artifact algorithm uses Kalman filtering to eliminate SpO2 fluctuations caused by patients turning over at night. It is also used to automatically adjust the ultrasonic cardiac output monitoring frequency based on the organ perfusion critical parameters fed back by the dynamic threshold calculation module, so that the monitoring interval is negatively correlated with the patient's current cardiac load;

[0014] The early warning execution module is used to automatically wake up the driving mechanism to inject pressor drugs to the patient based on the risk index transmitted by the data processing module; it is also used to implement the transmission of alarm signals in different time periods based on the day and night differences in the frequency of ultrasonic cardiac output monitoring. The user determines whether the received early warning signal and alarm signal are false alarm signals. If so, the false alarm signal is transmitted to the false alarm correction module;

[0015] The false alarm correction module is used to perform continuous deep learning based on false alarm signals, analyze nighttime false alarm events through confusion matrices, and generate optimized parameters for the interference algorithm to optimize anti-motion artifacts and transmit them to the nighttime perception and anti-interference module. The nighttime perception and anti-interference module adjusts the Kalman filter parameters after receiving the optimized parameters of the interference algorithm; it is also used to transmit signals back to the data processing module to correct the metabolic risk weight based on the visual attenuation trend spectrum diagram when the blood gas analysis shows that the lactate value is greater than 4mmol / L but no alarm is triggered. The data processing module recalculates the risk index after receiving the corrected metabolic risk weight; it is also used to push the threshold compensation coefficient to the dynamic threshold calculation module. The dynamic threshold calculation module updates the initial threshold in real time after receiving the threshold compensation coefficient.

[0016] Further, the equipment unit includes a vital sign monitoring subunit and an environmental monitoring subunit;

[0017] The vital sign monitoring subunit is used to collect vital sign data based on the critical care equipment, and collect ECG, ABP, and SpO2 vital sign data by synchronizing the wearable bioelectrode array with the bedside monitor;

[0018] The environmental monitoring subunit is used to generate environmental data and transmit it to the acquisition module based on the light sensor and sound source acquisition equipment measuring the light in the ward and the sound pressure of the sound entering the ward.

[0019] Furthermore, the driving mechanism is used to dynamically adjust the injection flow rate of the driving mechanism using a PID algorithm according to the pressor injection instruction of the early warning execution module; it is also used to calculate the pressor dosage compensation coefficient through a nonlinear regression model based on ultrasonic cardiac output monitoring data, so that the injection rate is exponentially correlated with the patient's current vascular resistance.

[0020] Furthermore, the acquisition module includes a vital sign data standardization unit, an environmental data standardization unit, and a physical examination data integration unit;

[0021] The vital sign data normalization unit is used to generate a time series vital sign data set with patient ID and timestamp based on the vital sign data input by the equipment unit and align the time domain with the system time, and transmit it to the dynamic threshold calculation module;

[0022] The environmental data standardization unit is used to generate a time series environmental data set with spatial position coding based on the environmental data through Kalman filtering and noise reduction, and transmit it to the dynamic threshold calculation module;

[0023] The physical examination data integration unit is used to generate physical examination data based on the electronic medical records and patient physical examination reports in the hospital information system and transmit them to the data processing module.

[0024] Furthermore, the data processing module includes a short-term prediction unit and a long-term prediction unit;

[0025] The short-term prediction unit is used to generate a three-dimensional spectrum of the cardiopulmonary function decline trend in the first time period through modal data fusion based on the time series vital sign dataset and time series environmental dataset of the acquisition module, and transmit it to the early warning execution module. The modal data fusion processing method is to use the Daubechies 4 wavelet basis to perform a 5-layer decomposition of the ECG or ABP, extract the approximate coefficient after the γ-band electromyographic interference is suppressed, calculate the time offset δ between the ECG peak and the rising edge of the ABP waveform, and establish a delay compensation function:

[0026]

[0027] Where τ(t) represents the time delay compensation at time point t, which is used to dynamically adjust the time offset between modal signals to ensure the time alignment of physiological signals. N represents the number of heartbeat cycles contained in the sliding window, which is used to track the patient's heart rate changes in real time. α i represents the weight coefficient of the i-th heartbeat cycle, δ i (t) represents the time difference between the ECG peak and the rising edge of the ABP waveform in the i-th heart cycle;

[0028] The long-term prediction unit is used to generate an organ failure risk index using the Cox proportional hazard model based on the physical examination data, time series physical sign data set, and time series environmental data set collected by the acquisition module during the second time period, and transmit the index to the dynamic threshold calculation module. The survival analysis baseline risk calculation formula of the Cox proportional hazard model is:

[0029]

[0030] Among them, h(t|X) represents the conditional hazard function at time point t, h0(t) represents the baseline hazard function, and x i represents the i-th covariate of the individual, α iRepresents the regression coefficient, reflecting the covariate x i The intensity of contribution to risk, Indicates the relative impact of individual characteristics on risk.

[0031] Furthermore, the dynamic threshold calculation module includes an initial threshold unit, which is used to parse the historical data of the patient's baseline systolic blood pressure, diastolic blood pressure, and mean arterial pressure in the past 7 days in a quiet state based on the input time series vital sign data set to calculate the dynamic baseline and transmit it to the nighttime perception and anti-interference module:

[0032] Threshold 基础 =μ SBP / DBP ±1.5σ

[0033] Among them, μ SBP / DBP It represents the sliding average of the patient's systolic blood pressure (SBP) and diastolic blood pressure (DBP) in a resting state over the past 7 days. σ represents the dynamic standard deviation of blood pressure data. The coefficient of 1.5 defines the range of standard deviation multiples of the physiologically allowed fluctuation.

[0034] Based on the basic dynamic baseline, the organ perfusion demand parameters are superimposed to derive a personalized blood pressure threshold range calculation formula and transmit it to the early warning execution module:

[0035]

[0036] Among them, ScvO2 represents the central venous oxygen saturation sampled every five minutes by the acquisition module.

[0037] Furthermore, the dynamic threshold calculation module also includes a day and night switching unit, which is used to obtain the patient's sleep depth parameters, cycles and body movement index based on the input time series vital sign data set, and generate a sleep stage mark. When the patient enters deep sleep, the blood pressure fluctuation tolerance is increased to 1.2 times that of the daytime under the conditions that the body movement index is less than 5 and the deep sleep / total sleep time is greater than 30%. If the above conditions are not met, the nighttime threshold remains the same as the daytime threshold; the patient's dynamic threshold is combined with the time period to obtain the organ perfusion critical parameter in the nighttime state, and it is transmitted to the early warning execution module to update the personalized blood pressure threshold range of the initial threshold unit, and is transmitted to the nighttime perception and anti-interference module to dynamically adjust the ultrasonic cardiac output monitoring frequency, and optimize the SpO2 threshold of the anti-motion artifact algorithm based on the organ perfusion critical parameter.

[0038] Furthermore, the nighttime perception and anti-interference module includes an anti-motion artifact unit and a monitoring frequency adjustment unit;

[0039] The anti-motion artifact unit is used to analyze the ambient light intensity and noise sound pressure in the patient's ward based on the time series environmental data set. When the average brightness in the past 5 minutes is less than 10 lux and the sound pressure is less than 40 decibels, the night mode signal is activated and transmitted to the dynamic threshold calculation module;

[0040] When night mode is activated, SpO2 is obtained from the time series environmental data set, and the state vector of the patient's body movement at the current moment is calculated based on the Kalman filter algorithm:

[0041] x k =Ax k-1 +Bμ k-1 +ω k

[0042] Among them, A represents the patient's state transfer matrix, which describes the evolution of state variables over time, x k-1 represents the state vector of the previous moment k-1, B represents the control input matrix, μ k-1 Represents the external control signal, ω k Represents natural fluctuations in physiological parameters and model errors;

[0043] Obtain the SpO2 observation vector based on the patient's state vector during body movement:

[0044] z k =Hx k +v k

[0045] Where H represents the measurement matrix that maps the patient's state vector to the observation vector when the patient moves, v k Represents the noise including photoelectric signal noise and motion artifact interference;

[0046] Adjust the Kalman gain:

[0047]

[0048] R=R0·(1+0.2·body kinetic energy)

[0049] Among them, K k Represents the Kalman gain matrix. The larger the gain, the more trustworthy the current SpO2 observation value under motion artifacts. R represents the real-time volume change matrix of the patient. R0 represents the standard deviation of SpO2 measurement when the patient is stationary. Body motion energy represents the mean of the three-axis acceleration when the patient is moving. The adjusted and corrected SpO2 change value is output, and the body motion coupling coefficient is updated at the same time to reflect the interference intensity of the current motion on the signal.

[0050] The monitoring frequency adjustment unit dynamically adjusts the ultrasound monitoring interval through the PID control algorithm according to the critical parameters of the organ perfusion of the dynamic threshold calculation module, generates the equipment frequency switching instruction and transmits it to the early warning execution module, the night perception and anti-interference module, and the false alarm correction module. After receiving the instruction, the early warning execution module synchronously adjusts the trigger logic of the time period alarm signal. After receiving the instruction, the night perception and anti-interference module dynamically adjusts the filtering strength through the monitoring frequency. After receiving the instruction, the false alarm correction module performs closed-loop optimization on the PID parameters.

[0051] Furthermore, the early warning execution module includes an alarm strategy unit and a device linkage unit;

[0052] The alarm strategy unit is used to select the alarm method based on the risk level signal of the dynamic threshold through the fuzzy logic decision tree, generate graded warning signals and alarm signals, and notify medical staff;

[0053] The device linkage unit is used to control the injection pump through the IEEE 11073 protocol based on the acute deterioration warning of the risk index, generate automatic infusion instructions for pressor drugs, and transmit them to the drive mechanism to execute the injection.

[0054] Furthermore, the false alarm correction module includes a false alarm analysis unit and a threshold compensation unit;

[0055] The false alarm analysis unit is used to calculate the feature contribution through the confusion matrix based on the manual confirmation records of the warning log, generate the anti-interference algorithm optimization parameters, and transmit them to the night perception and anti-interference module. The night perception and anti-interference module uses the anti-interference algorithm optimization parameters to adjust the noise covariance matrix of the Kalman filter and dynamically correct the weighting coefficient of the sleep depth parameter;

[0056] The threshold compensation unit is used to correct the metabolic weight through the gradient descent method according to the abnormal lactate level in the blood gas data without alarm, generate a dynamic threshold compensation coefficient and transmit it to the dynamic threshold calculation module. The dynamic threshold calculation module updates the metabolic risk weight in the organ perfusion critical parameters according to the compensation coefficient and adjusts the baseline of the blood pressure tolerance threshold.

[0057] The above scheme has the following beneficial effects:

[0058] 1. This solution uses equipment units to collect vital sign data and environmental data from critically ill patients. The collection module performs standardized conversions and transmits the data to the data processing module. The data processing module can process this data in real time, integrate the data interfaces of various monitoring devices, and use machine learning algorithms to analyze data in real time. For example, the short-term prediction unit in the data processing module generates a three-dimensional spectrum diagram of the trend of cardiopulmonary function decline in the next 30 minutes through modal data fusion, and the long-term prediction unit generates an organ failure risk index through the Cox proportional hazard model, realizing real-time analysis of data and identification of potential risks. The early warning execution module automatically wakes up the equipment unit to inject pressor drugs to the patient based on the risk index transmitted by the dynamic threshold calculation module. It can also issue different alarms during the day and at night, send early warning information to doctors, and promptly notify medical staff of abnormal patient conditions so that appropriate measures can be taken.

[0059] 2. In this solution, the dynamic threshold calculation module can dynamically adjust warning thresholds based on the patient's specific condition. For example, the initial threshold unit uses the input time series vital sign data set to parse the patient's baseline systolic blood pressure, diastolic blood pressure, and mean arterial pressure historical data from the previous seven days of resting state to calculate the dynamic baseline. This is then overlaid with organ perfusion demand parameters to derive a personalized blood pressure threshold range. For different patients, such as those with chronic hypertension, a more appropriate low blood pressure warning threshold can be set based on their usual blood pressure levels and tolerance range.

[0060] 3. In this solution, the day and night switching unit improves the tolerance for blood pressure fluctuations when the patient enters deep sleep based on the patient's sleep depth parameters, cycle and body movement index, and combines the patient's dynamic threshold with the time period to obtain the critical parameters of nocturnal organ perfusion. This reflects the dynamic adjustment of the warning threshold according to the patient's disease stage (such as the night sleep stage), meeting the need to dynamically adjust the warning threshold according to the patient's individual condition.

[0061] 4. This solution uses the night-specific monitoring function of the night perception and anti-interference module. When the anti-motion artifact unit is activated in night mode, it eliminates SpO2 fluctuations caused by the patient turning over at night based on the Kalman filter algorithm, optimizes the night monitoring algorithm, reduces interference from factors such as light and activity, improves data accuracy, and uses a blood oxygen sensor with better performance in low-light environments to ensure the accuracy of nighttime blood oxygen monitoring.

[0062] 5. This solution uses the day and night switching unit of the dynamic threshold calculation module to adjust the tolerance for blood pressure fluctuations according to the patient's sleep conditions at night, that is, to adjust the warning threshold; the monitoring frequency adjustment unit dynamically adjusts the ultrasound monitoring interval through the PID control algorithm based on the organ perfusion parameters of the dynamic threshold calculation module, and adjusts the warning threshold and monitoring frequency according to the characteristics of the human body's physiological rhythm at night, which can better adapt to the patient's physiological state at night and improve the monitoring effect.

[0063] Additional aspects and advantages of the present invention will be set forth in part in the description which follows and, in part, will be obvious from the description which follows, or may be learned by practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 Schematic diagram of the system framework of an embodiment of the critical care patient monitoring system of the present invention. DETAILED DESCRIPTION

[0065] The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0066] In the description of the present invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings and are intended solely to facilitate and simplify the description of the present invention. They are not intended to indicate or imply that the devices or components referred to must have, be constructed, or operate in a specific orientation, and therefore should not be construed as limitations on the present invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0067] In the description of the present invention, it should be noted that, unless otherwise expressly specified or limited, the terms "mounted," "connected," and "connected" should be understood in a broad sense. For example, they may refer to fixed, detachable, or integral connections; mechanical or electrical connections; direct or indirect connections through an intermediate medium; and internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on the specific circumstances.

[0068] The following is further described in detail through specific implementation methods:

[0069] Example 1:

[0070] As attached Figure 1 Shown: A critical care patient monitoring system, comprising:

[0071] The device unit is used to monitor the patient's vital sign data and collect environmental data in the ward, the vital sign data including daytime basic vital sign data and nighttime specific vital sign data; the device unit signal is connected to a driving mechanism that automatically injects pressor drugs into the patient by receiving injection instructions;

[0072] The equipment unit includes a vital signs monitoring subunit and an environmental monitoring subunit;

[0073] The vital sign monitoring subunit is used to collect vital sign data based on the critical care equipment, and collect ECG, ABP, and SpO2 vital sign data by synchronizing the wearable bioelectrode array with the bedside monitor;

[0074] The environmental monitoring subunit is used to generate environmental data and transmit it to the acquisition module based on the light sensor and sound source acquisition equipment measuring the light in the ward and the sound pressure of the sound entering the ward.

[0075] The driving mechanism is used to dynamically adjust the injection flow rate of the driving mechanism using a PID algorithm according to the pressor injection instruction of the early warning execution module; it is also used to calculate the pressor dosage compensation coefficient based on the ultrasonic cardiac output monitoring data through a nonlinear regression model, so that the injection rate is exponentially related to the patient's current vascular resistance.

[0076] The acquisition module is used to standardize and transform the vital sign data and environmental data to obtain a time series vital sign data set and a time series environmental data set, both of which are marked with the patient ID and timestamp; it is also used to obtain the patient's physical examination data and ultrasonic cardiac output monitoring data;

[0077] The acquisition module includes a physical sign data standardization unit, an environmental data standardization unit, and a physical examination data integration unit;

[0078] The vital sign data normalization unit is used to generate a time series vital sign data set with patient ID and timestamp based on the vital sign data input by the equipment unit and align the time domain with the system time, and transmit it to the dynamic threshold calculation module;

[0079] The environmental data standardization unit is used to generate a time series environmental data set with spatial position coding based on the environmental data through Kalman filtering and noise reduction, and transmit it to the dynamic threshold calculation module;

[0080] The physical examination data integration unit is used to generate physical examination data based on the electronic medical records and patient physical examination reports in the hospital information system and transmit them to the data processing module.

[0081] The data processing module is used to process the time series vital sign data set and the time series environmental data set. The processing is divided into short-term prediction processing and long-term prediction processing. Among them, the short-term prediction processing is used to predict the visual attenuation trend of the patient's cardiopulmonary function in the first time period in the future. The long-term prediction processing is used to generate the risk index of organ failure in the second time period in the future by combining the physical examination data;

[0082] The data processing module includes a short-term prediction unit and a long-term prediction unit;

[0083] The short-term prediction unit is used to generate a three-dimensional spectrogram of the cardiopulmonary function decline trend in the first time period through modal data fusion based on the time series vital sign dataset and time series environmental dataset of the acquisition module. Specifically, the first time period is limited to the next 30 minutes, and the spectrum is transmitted to the early warning execution module. The modal data fusion processing method is to use the Daubechies 4 wavelet basis to perform a 5-layer decomposition of the ECG or ABP, extract the approximate coefficient after the γ-band electromyographic interference is suppressed, calculate the time offset δ between the ECG peak and the rising edge of the ABP waveform, and establish a time delay compensation function:

[0084]

[0085] Where τ(t) represents the time delay compensation at time point t, which is used to dynamically adjust the time offset between modal signals to ensure the time alignment of physiological signals. N represents the number of heartbeat cycles contained in the sliding window, which is used to track the patient's heart rate changes in real time. α i represents the weight coefficient of the i-th heartbeat cycle, δ i (t) represents the time difference between the ECG peak and the rising edge of the ABP waveform in the i-th heart cycle;

[0086] The long-term prediction unit is configured to generate an organ failure risk index using the Cox proportional hazard model based on the physical examination data, time series physical sign data set, and time series environmental data set collected by the acquisition module during a second time period, specifically within the first 7 days, and transmit the index to the dynamic threshold calculation module. The survival analysis baseline risk calculation formula of the Cox proportional hazard model is:

[0087]

[0088] Among them, h(t|X) represents the conditional hazard function at time point t, h0(t) represents the baseline hazard function, and x i represents the i-th covariate of the individual, α i Represents the regression coefficient, reflecting the covariate x i The intensity of contribution to risk, Indicates the relative impact of individual characteristics on risk.

[0089] A dynamic threshold calculation module is used to set an initial threshold value including a tolerance value for daytime blood pressure fluctuations based on vital sign data. If the blood pressure does not reach the initial threshold value, an early warning signal is forcibly triggered.

[0090] The dynamic threshold calculation module includes an initial threshold unit, which is used to parse the patient's baseline systolic blood pressure, diastolic blood pressure, and mean arterial pressure in the past 7 days of resting state based on the input time series vital sign data set to calculate the dynamic baseline and transmit it to the nighttime perception and anti-interference module:

[0091] Threshold 基础 =μ SBP / DBP ±1.5σ

[0092] Among them, μ SBP / DBP It represents the sliding average of the patient's systolic blood pressure (SBP) and diastolic blood pressure (DBP) in a resting state over the past 7 days. σ represents the dynamic standard deviation of blood pressure data. The coefficient of 1.5 defines the range of standard deviation multiples of the physiologically allowed fluctuation.

[0093] Based on the basic dynamic baseline, the organ perfusion demand parameters are superimposed to derive a personalized blood pressure threshold range calculation formula and transmit it to the early warning execution module:

[0094]

[0095] Among them, ScvO2 represents the central venous oxygen saturation sampled every five minutes by the acquisition module;

[0096] It also includes a day and night switching unit, which is used to obtain the patient's sleep depth parameters, cycles and body movement index based on the input time series vital sign data set, and generate sleep stage marks. When the patient enters deep sleep, the blood pressure fluctuation tolerance is increased to 1.2 times that of the daytime under the conditions that the body movement index is less than 5 and the deep sleep / total sleep time is greater than 30%. If the above conditions are not met, the nighttime threshold remains the same as the daytime threshold; the patient's dynamic threshold is combined with the time period to obtain the organ perfusion critical parameter in the nighttime state, and it is transmitted to the early warning execution module to update the personalized blood pressure threshold range of the initial threshold unit, and is transmitted to the nighttime perception and anti-interference module to dynamically adjust the ultrasonic cardiac output monitoring frequency, and optimize the SpO2 threshold of the anti-motion artifact algorithm based on the organ perfusion critical parameter.

[0097] The nighttime perception and anti-interference module is used to generate a sleep depth parameter based on vital sign data and environmental data using a weighted algorithm. The sleep depth parameter and nighttime-specific vital sign data are synchronously transmitted to the dynamic threshold calculation module. If the sleep depth parameter received by the dynamic threshold calculation module continuously exceeds the preset threshold, the blood pressure fluctuation tolerance is increased to 1.2 times that of the daytime mode to obtain the critical organ perfusion parameter;

[0098] It is also used to activate the anti-motion artifact algorithm based on environmental data. The anti-motion artifact algorithm uses Kalman filtering to eliminate SpO2 fluctuations caused by patients turning over at night. It is also used to automatically adjust the ultrasonic cardiac output monitoring frequency based on the organ perfusion critical parameters fed back by the dynamic threshold calculation module, so that the monitoring interval is negatively correlated with the patient's current cardiac load;

[0099] The nighttime perception and anti-interference module includes an anti-motion artifact unit and a monitoring frequency adjustment unit;

[0100] The anti-motion artifact unit is used to analyze the ambient light intensity and noise sound pressure in the patient's ward based on the time series environmental data set. When the average brightness in the past 5 minutes is less than 10 lux and the sound pressure is less than 40 decibels, the night mode signal is activated and transmitted to the dynamic threshold calculation module;

[0101] When night mode is activated, SpO2 is obtained from the time series environmental data set, and the state vector of the patient's body movement at the current moment is calculated based on the Kalman filter algorithm:

[0102] x k =Ax k-1 +Bμ k-1 +ω k

[0103] Among them, A represents the patient's state transfer matrix, which describes the evolution of state variables over time, x k-1 represents the state vector of the previous moment k-1, B represents the control input matrix, μ k-1 Represents the external control signal, ω k Represents natural fluctuations in physiological parameters and model errors;

[0104] Obtain the SpO2 observation vector based on the patient's state vector during body movement:

[0105] z k =Hx k +v k

[0106] Where H represents the measurement matrix that maps the patient's state vector to the observation vector when the patient moves, v k Represents the noise including photoelectric signal noise and motion artifact interference;

[0107] Adjust the Kalman gain:

[0108]

[0109] R=R0·(1+0.2·body kinetic energy)

[0110] Among them, K kRepresents the Kalman gain matrix. The larger the gain, the more trustworthy the current SpO2 observation value under motion artifacts. R represents the real-time volume change matrix of the patient. R0 represents the standard deviation of SpO2 measurement when the patient is stationary. Body motion energy represents the mean of the three-axis acceleration when the patient is moving. The adjusted and corrected SpO2 change value is output, and the body motion coupling coefficient is updated at the same time to reflect the interference intensity of the current motion on the signal.

[0111] The monitoring frequency adjustment unit dynamically adjusts the ultrasound monitoring interval through the PID control algorithm according to the critical parameters of the organ perfusion of the dynamic threshold calculation module, generates the equipment frequency switching instruction and transmits it to the early warning execution module, the night perception and anti-interference module, and the false alarm correction module. After receiving the instruction, the early warning execution module synchronously adjusts the trigger logic of the time period alarm signal. After receiving the instruction, the night perception and anti-interference module dynamically adjusts the filtering strength through the monitoring frequency. After receiving the instruction, the false alarm correction module performs closed-loop optimization on the PID parameters.

[0112] The early warning execution module is used to automatically wake up the driving mechanism to inject pressor drugs to the patient based on the risk index transmitted by the data processing module; it is also used to implement the transmission of alarm signals in different time periods based on the day and night differences in the frequency of ultrasonic cardiac output monitoring. The user determines whether the received early warning signal and alarm signal are false alarm signals. If so, the false alarm signal is transmitted to the false alarm correction module;

[0113] The early warning execution module includes an alarm strategy unit and a device linkage unit;

[0114] The alarm strategy unit is used to select the alarm method based on the risk level signal of the dynamic threshold through the fuzzy logic decision tree, generate graded warning signals and alarm signals, and notify medical staff;

[0115] The device linkage unit is used to control the injection pump through the IEEE 11073 protocol based on the acute deterioration warning of the risk index, generate automatic infusion instructions for pressor drugs, and transmit them to the drive mechanism to execute the injection.

[0116] The false alarm correction module is used to conduct continuous deep learning based on false alarm signals, analyze nighttime false alarm events through confusion matrices, and generate optimized parameters for the interference algorithm to optimize anti-motion artifacts and transmit them to the nighttime perception and anti-interference module. The nighttime perception and anti-interference module adjusts the Kalman filter parameters after receiving the optimized parameters of the interference algorithm. It is also used to transmit signals back to the data processing module to correct the metabolic risk weight based on the visual attenuation trend spectrum diagram when the blood gas analysis shows that the lactate value is greater than 4mmol / L but no alarm is triggered. The data processing module recalculates the risk index after receiving the corrected metabolic risk weight. It is also used to push the threshold compensation coefficient to the dynamic threshold calculation module. The dynamic threshold calculation module updates the initial threshold in real time after receiving the threshold compensation coefficient.

[0117] The false alarm correction module includes a false alarm analysis unit and a threshold compensation unit;

[0118] The false alarm analysis unit is used to calculate the feature contribution through the confusion matrix based on the manual confirmation records of the warning log, generate the anti-interference algorithm optimization parameters, and transmit them to the night perception and anti-interference module. The night perception and anti-interference module uses the anti-interference algorithm optimization parameters to adjust the noise covariance matrix of the Kalman filter and dynamically correct the weighting coefficient of the sleep depth parameter;

[0119] The threshold compensation unit is used to correct the metabolic weight through the gradient descent method according to the abnormal lactate level in the blood gas data without alarm, generate a dynamic threshold compensation coefficient and transmit it to the dynamic threshold calculation module. The dynamic threshold calculation module updates the metabolic risk weight in the organ perfusion critical parameters according to the compensation coefficient and adjusts the baseline of the blood pressure tolerance threshold.

[0120] The specific implementation process is as follows: After the patient is admitted to the ICU, the vital signs such as ECG, ABP, SpO2 are continuously monitored through the vital sign monitoring subunit, and the light and sound pressure in the ward are monitored through the environmental monitoring subunit. When the patient sleeps at night, the night perception and anti-interference module detects that the light intensity is continuously <10lux and the sound pressure is <40dB, and activates the anti-motion artifact algorithm. At this time, if the patient turns over and causes the fingertip SpO2 probe to be compressed, the signal will drop sharply, such as the signal drops from 98% to 70%. The anti-motion artifact shadow unit identifies the artifact through Kalman filtering, dynamically adjusts the noise covariance matrix R, suppresses noise interference, and re-outputs the SpO2 value to 96% after correction. After receiving the corrected data, the dynamic threshold calculation module determines that there is no need to trigger an alarm to avoid false alarms.

[0121] When a patient's blood pressure drops suddenly at night due to septic shock, for example, from 120 mmHg to 80 mmHg, the vital sign monitoring subunit detects the abnormal ABP signal in real time and transmits it synchronously to the acquisition module via the bedside monitor. The dynamic threshold calculation module detects that the current SBP is 80 mmHg, which is lower than the dynamic baseline of 105 mmHg set by the personalized threshold lower limit, triggering an immediate alarm. Simultaneously, the day / night switching unit recognizes that the patient is in deep sleep (stage N3 accounts for 35%) and increases the threshold tolerance to 1.2 times that of daytime sleep (with the lower limit adjusted to 84 mmHg). However, if the current value remains below the threshold, the alarm strategy unit activates a Level 1 alarm with an audible and visual alarm and text message notification. The device linkage unit controls the syringe pump via the IEEE 11073 protocol to infuse norepinephrine at a rate of 0.1 μg / kg / min. If the medical staff confirms the event as a true positive, the system records the response time. If it is a false alarm (e.g., sensor detachment), the false alarm correction module updates the motion artifact detection parameters.

[0122] When the patient's blood lactate level rises from 1.5mmol / L to 4.8mmol / L, but the current vital signs are stable, the long-term prediction unit combines the physical examination data of abnormal liver and kidney function and the real-time lactate value, and calculates the 24-hour organ failure risk index to 85 points (high risk) through the Cox model. The false alarm correction module detects that the lactate abnormality does not trigger an alarm, and reversely corrects the metabolic risk weight (gradient descent method adjusts the weight by +0.2). The dynamic threshold calculation module synchronously updates the perfusion parameter threshold. The early warning execution module activates a pop-up prompt and a secondary alarm broadcast by the nursing station, prompting blood culture and antibiotic treatment.

[0123] Obviously, the above embodiments are merely examples for clarity of explanation and are not intended to limit the implementation methods. Those skilled in the art will readily appreciate that other variations or modifications based on the above descriptions are possible. It is not necessary and impossible to enumerate all implementation methods here. Obvious variations or modifications arising therefrom remain within the scope of protection of the present invention.

Claims

1. A critical care patient monitoring system, characterized in that: include: Equipment unit, used to monitor the patient's vital signs data and collect environmental data in the ward, the vital signs data including daytime basic vital signs data and nighttime specific vital signs data; The device unit is signal-connected to a drive mechanism that automatically injects pressor drugs into the patient upon receiving an injection command; The acquisition module is used to standardize and transform the vital sign data and environmental data to obtain a time series vital sign data set and a time series environmental data set, both of which are marked with the patient ID and timestamp; it is also used to obtain the patient's physical examination data and ultrasonic cardiac output monitoring data; The data processing module is used to process the time series vital sign data set and the time series environmental data set. The processing is divided into short-term prediction processing and long-term prediction processing. Among them, the short-term prediction processing is used to predict the visual attenuation trend of the patient's cardiopulmonary function in the first time period in the future. The long-term prediction processing is used to generate the risk index of organ failure in the second time period in the future by combining the physical examination data; A dynamic threshold calculation module is used to set an initial threshold value including a tolerance value for daytime blood pressure fluctuations based on vital sign data. If the blood pressure does not reach the initial threshold value, an early warning signal is forcibly triggered. The nighttime perception and anti-interference module is used to generate a sleep depth parameter based on vital sign data and environmental data using a weighted algorithm. The sleep depth parameter and nighttime-specific vital sign data are synchronously transmitted to the dynamic threshold calculation module. If the sleep depth parameter received by the dynamic threshold calculation module continuously exceeds the preset threshold, the blood pressure fluctuation tolerance is increased to 1.2 times that of the daytime mode to obtain the critical organ perfusion parameter; It is also used to activate the anti-motion artifact algorithm based on environmental data. The anti-motion artifact algorithm uses Kalman filtering to eliminate SpO2 fluctuations caused by patients turning over at night. It is also used to automatically adjust the ultrasonic cardiac output monitoring frequency based on the organ perfusion critical parameters fed back by the dynamic threshold calculation module, so that the monitoring interval is negatively correlated with the patient's current cardiac load; The early warning execution module is used to automatically wake up the driving mechanism to inject pressor drugs to the patient based on the risk index transmitted by the data processing module; it is also used to implement the transmission of alarm signals in different time periods based on the day and night differences in the frequency of ultrasonic cardiac output monitoring. The user determines whether the received early warning signal and alarm signal are false alarm signals. If so, the false alarm signal is transmitted to the false alarm correction module; The false alarm correction module is used to perform continuous deep learning based on false alarm signals, analyze nighttime false alarm events through confusion matrices, and generate optimized parameters for the interference algorithm to optimize anti-motion artifacts and transmit them to the nighttime perception and anti-interference module. The nighttime perception and anti-interference module adjusts the Kalman filter parameters after receiving the optimized parameters of the interference algorithm; it is also used to transmit signals back to the data processing module to correct the metabolic risk weight based on the visual attenuation trend spectrum diagram when the blood gas analysis shows that the lactate value is greater than 4mmol / L but no alarm is triggered. The data processing module recalculates the risk index after receiving the corrected metabolic risk weight; it is also used to push the threshold compensation coefficient to the dynamic threshold calculation module. The dynamic threshold calculation module updates the initial threshold in real time after receiving the threshold compensation coefficient.

2. The critical care patient monitoring system according to claim 1, characterized in that: The equipment unit includes a vital signs monitoring subunit and an environmental monitoring subunit; The vital sign monitoring subunit is used to collect vital sign data based on the critical care equipment, and collect ECG, ABP, and SpO2 vital sign data by synchronizing the wearable bioelectrode array with the bedside monitor; The environmental monitoring subunit is used to generate environmental data and transmit it to the acquisition module based on the light sensor and sound source acquisition equipment measuring the light in the ward and the sound pressure of the sound entering the ward.

3. The critical care patient monitoring system according to claim 2, characterized in that: The driving mechanism is used to dynamically adjust the injection flow rate of the driving mechanism using a PID algorithm according to the pressor injection instruction of the early warning execution module; it is also used to calculate the pressor dosage compensation coefficient based on the ultrasonic cardiac output monitoring data through a nonlinear regression model, so that the injection rate is exponentially related to the patient's current vascular resistance.

4. The critical care patient monitoring system according to claim 3, characterized in that: The acquisition module includes a physical sign data standardization unit, an environmental data standardization unit, and a physical examination data integration unit; The vital sign data normalization unit is used to generate a time series vital sign data set with patient ID and timestamp based on the vital sign data input by the equipment unit and align the time domain with the system time, and transmit it to the dynamic threshold calculation module; The environmental data standardization unit is used to generate a time series environmental data set with spatial position coding based on the environmental data through Kalman filtering and noise reduction, and transmit it to the dynamic threshold calculation module; The physical examination data integration unit is used to generate physical examination data based on the electronic medical records and patient physical examination reports in the hospital information system and transmit them to the data processing module.

5. The critical care patient monitoring system according to claim 4, characterized in that: The data processing module includes a short-term prediction unit and a long-term prediction unit; The short-term prediction unit is used to generate a three-dimensional spectrum of the cardiopulmonary function decline trend in the first time period through modal data fusion based on the time series vital sign dataset and time series environmental dataset of the acquisition module, and transmit it to the early warning execution module. The modal data fusion processing method is to use the Daubechies 4 wavelet basis to perform a 5-layer decomposition of the ECG or ABP, extract the approximate coefficient after the γ-band electromyographic interference is suppressed, calculate the time offset δ between the ECG peak and the rising edge of the ABP waveform, and establish a delay compensation function: Where τ(t) represents the time delay compensation at time point t, which is used to dynamically adjust the time offset between modal signals to ensure the time alignment of physiological signals. N represents the number of heartbeat cycles contained in the sliding window, which is used to track the patient's heart rate changes in real time. α i represents the weight coefficient of the i-th heartbeat cycle, δ i (t) represents the time difference between the ECG peak and the rising edge of the ABP waveform in the i-th heart cycle; The long-term prediction unit is used to generate an organ failure risk index using the Cox proportional hazard model based on the physical examination data, time series physical sign data set, and time series environmental data set collected by the acquisition module during the second time period, and transmit the index to the dynamic threshold calculation module. The survival analysis baseline risk calculation formula of the Cox proportional hazard model is: Among them, h(t|X) represents the conditional hazard function at time point t, h0(t) represents the baseline hazard function, and x i represents the i-th covariate of the individual, α i Represents the regression coefficient, reflecting the covariate x i The intensity of contribution to risk, Indicates the relative impact of individual characteristics on risk.

6. The critical care patient monitoring system according to claim 5, characterized in that: The dynamic threshold calculation module includes an initial threshold unit, which is used to parse the patient's baseline systolic blood pressure, diastolic blood pressure, and mean arterial pressure in the past 7 days of resting state based on the input time series vital sign data set to calculate the dynamic baseline and transmit it to the nighttime perception and anti-interference module: Threshold 基础 =μ SBP / DBP ±1.5σ Among them, μ SBP / DBP It represents the sliding average of the patient's systolic blood pressure (SBP) and diastolic blood pressure (DBP) in a resting state over the past 7 days. σ represents the dynamic standard deviation of blood pressure data. The coefficient of 1.5 defines the range of standard deviation multiples of the physiologically allowed fluctuation. Based on the basic dynamic baseline, the organ perfusion demand parameters are superimposed to derive a personalized blood pressure threshold range calculation formula and transmit it to the early warning execution module: Among them, ScvO2 represents the central venous oxygen saturation sampled every five minutes by the acquisition module.

7. The critical care patient monitoring system according to claim 6, characterized in that: The dynamic threshold calculation module also includes a day and night switching unit, which is used to obtain the patient's sleep depth parameters, cycles and body movement index based on the input time series vital sign data set, and generate a sleep stage mark. When the patient enters deep sleep, the blood pressure fluctuation tolerance is increased to 1.2 times that of the daytime under the conditions that the body movement index is less than 5 and the deep sleep / total sleep time is greater than 30%. If the above conditions are not met, the nighttime threshold remains the same as the daytime threshold; the patient's dynamic threshold is combined with the time period to obtain the organ perfusion critical parameter in the nighttime state, and it is transmitted to the early warning execution module to update the personalized blood pressure threshold range of the initial threshold unit, and is transmitted to the nighttime perception and anti-interference module to dynamically adjust the ultrasonic cardiac output monitoring frequency, and optimize the SpO2 threshold of the anti-motion artifact algorithm based on the organ perfusion critical parameter.

8. The critical care patient monitoring system according to claim 7, characterized in that: The nighttime perception and anti-interference module includes an anti-motion artifact unit and a monitoring frequency adjustment unit; The anti-motion artifact unit is used to analyze the ambient light intensity and noise sound pressure in the patient's ward based on the time series environmental data set. When the average brightness in the past 5 minutes is less than 10 lux and the sound pressure is less than 40 decibels, the night mode signal is activated and transmitted to the dynamic threshold calculation module; When night mode is activated, SpO2 is obtained from the time series environmental data set, and the state vector of the patient's body movement at the current moment is calculated based on the Kalman filter algorithm: x k =Ax k-1 +Bμ k-1 +oh k Among them, A represents the patient's state transfer matrix, which describes the evolution of state variables over time, x k-1 represents the state vector of the previous moment k-1, B represents the control input matrix, μ k-1 Represents the external control signal, ω k Represents natural fluctuations in physiological parameters and model errors; Obtain the SpO2 observation vector based on the patient's state vector during body movement: z k =Hx k +v k Where H represents the measurement matrix that maps the patient's state vector to the observation vector when the patient moves, v k Represents the noise including photoelectric signal noise and motion artifact interference; Adjust the Kalman gain: R=R0·(1+0.2·body kinetic energy) Among them, K k Represents the Kalman gain matrix. The larger the gain, the more trustworthy the current SpO2 observation value under motion artifacts. R represents the real-time volume change matrix of the patient. R0 represents the standard deviation of SpO2 measurement when the patient is stationary. Body motion energy represents the mean of the three-axis acceleration when the patient is moving. The adjusted and corrected SpO2 change value is output, and the body motion coupling coefficient is updated at the same time to reflect the interference intensity of the current motion on the signal. The monitoring frequency adjustment unit dynamically adjusts the ultrasound monitoring interval through the PID control algorithm according to the critical parameters of the organ perfusion of the dynamic threshold calculation module, generates the equipment frequency switching instruction and transmits it to the early warning execution module, the night perception and anti-interference module, and the false alarm correction module. After receiving the instruction, the early warning execution module synchronously adjusts the trigger logic of the time period alarm signal. After receiving the instruction, the night perception and anti-interference module dynamically adjusts the filtering strength through the monitoring frequency. After receiving the instruction, the false alarm correction module performs closed-loop optimization on the PID parameters.

9. The critical care patient monitoring system according to claim 8, characterized in that: The early warning execution module includes an alarm strategy unit and a device linkage unit; The alarm strategy unit is used to select the alarm method based on the risk level signal of the dynamic threshold through the fuzzy logic decision tree, generate graded warning signals and alarm signals, and notify medical staff; The device linkage unit is used to control the injection pump through the IEEE 11073 protocol based on the acute deterioration warning of the risk index, generate automatic infusion instructions for pressor drugs, and transmit them to the drive mechanism to execute the injection.

10. The critical care patient monitoring system according to claim 9, characterized in that: The false alarm correction module includes a false alarm analysis unit and a threshold compensation unit; The false alarm analysis unit is used to calculate the feature contribution through the confusion matrix based on the manual confirmation records of the warning log, generate the anti-interference algorithm optimization parameters, and transmit them to the night perception and anti-interference module. The night perception and anti-interference module uses the anti-interference algorithm optimization parameters to adjust the noise covariance matrix of the Kalman filter and dynamically correct the weighting coefficient of the sleep depth parameter; The threshold compensation unit is used to correct the metabolic weight through the gradient descent method according to the abnormal lactate level in the blood gas data without alarm, generate a dynamic threshold compensation coefficient and transmit it to the dynamic threshold calculation module. The dynamic threshold calculation module updates the metabolic risk weight in the organ perfusion critical parameters according to the compensation coefficient and adjusts the baseline of the blood pressure tolerance threshold.

Images