Intelligent nursing management system and method for critically ill patients

Abstract

The present invention discloses an intelligent nursing management system and method for critically ill patients, which relates to the field of medical health monitoring technology, including setting up multiple physiological signal acquisition channels, synchronously acquiring multi-dimensional physiological signals of patients, and preprocessing the acquired signals to form a standardized multimodal physiological data stream; inputting the preprocessed multimodal physiological data into a deep learning model based on LSTM and multi-head attention mechanism, combining with a dynamic weight adaptive algorithm, to realize intelligent fusion analysis of different physiological signals, and establish an individualized physiological system coupling model for patients; based on the analysis results of the deep learning model, the patient's condition risk assessment value is calculated in real time, and a specific warning level is output according to the set multi-level warning threshold. The present invention realizes all-round monitoring and early warning of the patient's condition by constructing a multi-level and multi-dimensional warning threshold system. It can significantly improve the quality of monitoring of critically ill patients and reduce the rate of missed judgment and misjudgment.

Description

Technical Field

[0001] The present invention relates to the field of medical health monitoring technology, and in particular to an intelligent nursing management system and method for critically ill patients. Background Art

[0002] With the development of modern medical technology and the continuous updating of clinical monitoring equipment, the means of monitoring the condition of critically ill patients have become increasingly diverse, generating a large amount of multi-source heterogeneous medical data. However, traditional disease monitoring methods mainly rely on fixed thresholds of single physiological indicators for early warning, which often makes it difficult to detect complex pathophysiological changes in a timely manner.

[0003] Existing early warning systems generally have problems such as insufficient early warning accuracy, high false alarm rate, and lack of personalization. Most of them only focus on the functional abnormalities of a specific organ or system, ignoring the interactions between organs and their dynamic evolution characteristics. At the same time, the early warning results lack sufficient explanations, which makes it difficult for medical staff to quickly understand the specific reasons for the triggering of the early warning, affecting the accuracy and timeliness of clinical decision-making. In addition, the threshold settings of the existing early warning system are relatively fixed, and they fail to be dynamically adjusted according to the individual characteristics of the patient and the treatment process. There is no effective online learning mechanism to continuously optimize the early warning model. These limitations make it difficult for the early warning system to meet the complex and changeable monitoring needs of critically ill patients in actual applications. Especially in the treatment of critically ill patients with multiple organ dysfunction, the limitations of the existing early warning system are more obvious, and it is difficult to provide comprehensive and accurate early warning information for medical staff.

[0004] Therefore, there is an urgent need to develop an intelligent early warning system that can comprehensively consider multi-dimensional information, has dynamic optimization capabilities, and can provide clear explanations, so as to improve the quality of monitoring and treatment of critically ill patients. Summary of the invention

[0005] In view of the problems existing in the existing intelligent nursing management methods for critically ill patients, the present invention is proposed.

[0006] The present invention provides the following technical solutions:

[0007] In the first aspect, the present invention provides an intelligent nursing management method for critically ill patients, which includes: setting up multiple physiological signal acquisition channels, synchronously acquiring multi-dimensional physiological signals of patients, and preprocessing the acquired signals, including signal filtering, denoising and time alignment operations, to form a standardized multimodal physiological data stream; inputting the preprocessed multimodal physiological data into a deep learning model based on LSTM and multi-head attention mechanism, combining with a dynamic weight adaptive algorithm, to realize intelligent fusion analysis of different physiological signals, and establish an individualized physiological system coupling model for patients; based on the analysis results of the deep learning model, calculating the patient's disease risk assessment value in real time, and outputting a specific warning level according to the set multi-level warning threshold, and continuously optimizing the model parameters through an online learning mechanism to improve the accuracy of the warning.

[0008] As a preferred solution of the intelligent nursing management method for critically ill patients described in the present invention, wherein: the multiple physiological signal acquisition channels include a cardiovascular system signal acquisition channel, a respiratory system signal acquisition channel, a body temperature monitoring channel, a nervous system monitoring channel, a biochemical parameter monitoring channel and a body movement monitoring channel; the cardiovascular system signal acquisition channel acquires electrocardiogram signals and blood pressure signals; the respiratory system signal acquisition channel acquires respiratory waveform signals and blood oxygen saturation signals; the body temperature monitoring channel acquires body temperature data; the nervous system monitoring channel acquires electroencephalogram data and electromyogram data; the biochemical parameter monitoring channel monitors blood sugar data; and the body movement monitoring channel detects body position data.

[0009] As a preferred solution of the intelligent nursing management method for critically ill patients described in the present invention, the preprocessing includes signal filtering, signal denoising and time alignment; the signal filtering includes ECG signal filtering and respiratory signal filtering. The ECG signal filtering uses a notch filter to eliminate power frequency interference. The transfer function is expressed as:

[0010] ;

[0011] in, is the transfer function of the system, is the power frequency, is the sampling frequency, is the filter pole radius, Indicates that the signal is delayed by one sampling period. Indicates that the signal is delayed by two sampling periods;

[0012] Baseline drift correction is used to process ECG signal filter drift. The baseline drift correction uses a high-pass Chebyshev filter, and the transfer function is expressed as:

[0013] ;

[0014] in, represents the transfer function, represents the passband ripple coefficient, represents the Chebyshev polynomial, represents the filter order, represents a complex variable, represents the imaginary unit, represents the cut-off frequency;

[0015] The Butterworth filter is used to eliminate power frequency interference in the respiratory signal filtering process. The square of the amplitude of the transfer function is expressed as:

[0016] ;

[0017] in, represents the magnitude squared response, represents the angular frequency, represents the cut-off frequency, Represents the filter order.

[0018] As a preferred solution of the intelligent nursing management method for critically ill patients described in the present invention, the signal denoising process is performed by wavelet threshold denoising, wherein the soft threshold function is expressed as:

[0019] ;

[0020] ;

[0021] in, is the signal to be processed; is the threshold, which is automatically selected using the VisuShrink criterion; is the noise standard deviation; is the signal length;

[0022] The time alignment process uses a dynamic time warping algorithm to find the best alignment path by accumulating the minimum distance. The formula is expressed as:

[0023] ;

[0024] in, Represents the local distance, which calculates the distance between the current two points. Indicates the selection of the minimum cumulative distance, where indicates the path from above, represents the path coming from the left, Indicates the path coming from the upper left.

[0025] As a preferred solution of the intelligent nursing management method for critically ill patients described in the present invention, based on the preprocessed multimodal physiological signal data, each signal contains a time series feature sequence, and the LSTM output of each input channel is calculated, which is expressed as:

[0026] ;

[0027] in, Indicates The output function of the LSTM unit is No. An input feature sequence, is the input weight matrix, is the circulant weight matrix, is the hidden state at the previous moment, is the bias vector, Indicates that the sigmoid function acts on the cell state;

[0028] Process the original time series data through LSTM and output the hidden state matrix. As the input of the multi-head attention mechanism, for each attention head , the calculation function is expressed as:

[0029] ;

[0030] in, For the The calculation function of the attention head, is the query matrix, is the transposed matrix of the Key matrix, is the Value matrix, is the scaling factor;

[0031] According to the calculation results of LSTM and multi-head attention mechanism, a deep learning model is constructed, which is expressed as:

[0032] ;

[0033] in, Represents the feature vector after fusion of multimodal physiological signals, Indicates The output function of the LSTM unit is For the The calculation function of the attention head, For the The dynamic weight coefficient of each feature, is the number of input features, is the number of attention heads, is the energy regulation coefficient, is the total energy of the system, is the spatial feature adjustment coefficient, is the total amount of spatial features.

[0034] As a preferred solution of the intelligent nursing management method for critically ill patients described in the present invention, the attention score based on the multi-head attention mechanism is extended to the physiological signal correlation by the similarity between the query and the key, the correlation relationship between the physiological signals is established, and the temporal dependency between the signals is analyzed, which is expressed as:

[0035] ;

[0036] in, For signal and The correlation coefficient between For the The signal is The value of the moment, For the The signal is The value of the moment, For signal The mean of For signal The mean of is the length of the time series;

[0037] On the basis of correlation, the time factor is introduced to calculate the coupling strength, which is expressed as:

[0038] ;

[0039] in, For signal and The coupling strength between is the absolute value of the correlation coefficient, is the time attenuation coefficient, For signal and The time delay between

[0040] By calculating the normal fluctuation range of each signal, identifying the typical signal interaction mode, analyzing the speed characteristics of the signal response, evaluating the amplitude of the system fluctuation, and integrating the system model to establish the patient's individualized physiological system coupling model, it is expressed as:

[0041] ;

[0042] in, The signal for prediction exist The value of the moment, is the associated signal, is the individualized parameter set, is a non-linear activation function.

[0043] As a preferred solution of the intelligent nursing management method for critically ill patients described in the present invention, the patient's condition risk assessment value is calculated based on the analysis results of the deep learning model, which is expressed as:

[0044] ;

[0045] ;

[0046] in, for The total risk value at the moment, For the The weight coefficient of each indicator is Risk contribution for a single indicator, is the overall stability index of the system. is the risk mapping function, For the predicted The signal is The value of the moment, is the individual baseline value, To allow fluctuation range, is a dynamic weight; according to the patient's condition risk assessment value, a multi-level warning threshold is set, including a first threshold, a second threshold, and a third threshold, and the first threshold < the second threshold < the third threshold, to warn the patient's single organ function. Specifically: if the overall risk value of a single organ is abnormal, and the overall risk value exceeds its normal upper limit but is lower than the first warning threshold, the warning level is set to a first-level warning, indicating that the organ function is slightly abnormal, and the fluctuation is beyond the normal range. The monitoring frequency needs to be strengthened and targeted intervention plans need to be prepared; if the organ risk value exceeds the first warning threshold but is lower than the second warning threshold, the warning level is set to a second-level warning, indicating that the organ function is moderately abnormal and needs timely intervention, the organ-specific intervention plan is initiated, and the status of other organs is closely monitored; if the organ risk value exceeds the second warning threshold, the warning level is set to a third-level warning, indicating that the organ function is seriously abnormal and there is a risk of functional failure, and the emergency plan needs to be immediately initiated.

[0047] In the second aspect, the present invention provides an intelligent nursing management system for critically ill patients, which includes: an acquisition module, which sets up multiple physiological signal acquisition channels, synchronously acquires the patient's multi-dimensional physiological signals, and preprocesses the acquired signals, including signal filtering, denoising and time alignment operations, to form a standardized multimodal physiological data stream; a construction module, which is used to input the preprocessed multimodal physiological data into a deep learning model based on LSTM and multi-head attention mechanism, combined with a dynamic weight adaptive algorithm, to achieve intelligent fusion analysis of different physiological signals, and establish an individualized physiological system coupling model for patients; an early warning module, which calculates the patient's disease risk assessment value in real time based on the analysis results of the deep learning model, and outputs a specific warning level according to the set multi-level warning threshold.

[0048] In a third aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the processor executes the computer program, it implements any step of the above-mentioned intelligent nursing management method for critically ill patients.

[0049] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, any step of the above-mentioned intelligent nursing management method for critically ill patients is implemented.

[0050] The beneficial effects of the present invention are as follows: the present invention realizes all-round monitoring and early warning of the patient's condition by constructing a multi-level and multi-dimensional early warning threshold system. A dynamic threshold adjustment mechanism is adopted to perform real-time optimization according to the individual characteristics of the patient, treatment feedback and system stability, thereby improving the accuracy and timeliness of the early warning. It can significantly improve the quality of monitoring of critically ill patients, reduce the rate of missed judgment and misjudgment, provide powerful auxiliary decision-making support for medical staff, and ultimately achieve the purpose of improving treatment effects and reducing mortality. Especially in the treatment of complex critical illnesses such as multiple organ dysfunction, the system's multi-dimensional early warning mechanism can help medical staff better grasp the development trend of the disease and achieve precise and individualized treatment intervention. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. Among them:

[0052] Figure 1 Flowchart of intelligent nursing management method for critically ill patients. DETAILED DESCRIPTION

[0053] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0054] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0055] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0056] Example 1: Reference Figure 1 , which is the first embodiment of the present invention, and provides a smart nursing management method for critically ill patients, including:

[0057] S1: Set up multiple physiological signal acquisition channels, synchronously acquire the patient's multi-dimensional physiological signals, and pre-process the acquired signals to form a standardized multi-modal physiological data stream.

[0058] Multiple physiological signal acquisition channels are set up, including a cardiovascular system signal acquisition channel, a respiratory system signal acquisition channel, a body temperature monitoring channel, a nervous system monitoring channel, a biochemical parameter monitoring channel, and a body movement monitoring channel.

[0059] The cardiovascular system signal acquisition channel collects ECG signals and blood pressure signals. The sampling frequency of ECG signals is 250Hz, the resolution is 16 bits, and 12-lead ECG signals are collected. The sampling frequency of blood pressure signals is 0.2Hz.

[0060] The pressure range is -50~300mmHg.

[0061] The respiratory system signal acquisition channel collects respiratory waveform signals and blood oxygen saturation signals. The sampling frequency of the respiratory waveform is 60Hz, the resolution is 12 bits, and the chest and abdominal respiratory movement waveforms are collected; the sampling frequency of blood oxygen saturation is 1Hz, the resolution is 8 bits, and its waveform data is collected.

[0062] The body temperature monitoring channel collects body temperature data with a sampling frequency of 0.2Hz and an accuracy of 0.1℃, and collects core body temperature data.

[0063] The nervous system monitoring channel collects EEG data and EMG data. The sampling frequency of the EEG data is 250Hz, the resolution is 16 bits, and the noise level is less than 0.5μVrms; the sampling frequency of the EMG data is 1000Hz, the bandwidth is 200Hz, and the common mode rejection ratio is greater than 100dB.

[0064] The biochemical parameter monitoring channel monitors blood glucose data with a sampling frequency of 0.2 Hz and a measurement range of 2.2~22.2 mmol / L.

[0065] The body motion monitoring channel monitors body position data with a sampling frequency of 50 Hz and a resolution of 14 bits.

[0066] The data acquired by all physiological signal acquisition channels need to have a unified clock source and synchronized trigger signals.

[0067] The collected data are preprocessed, including signal filtering, signal denoising and time alignment.

[0068] Signal filtering includes ECG signal filtering and respiratory signal filtering. ECG signal filtering uses a notch filter to eliminate power frequency interference. The transfer function is expressed as:

[0069] ;

[0070] in, is the transfer function of the system, is the power frequency, is the sampling frequency, is the filter pole radius, Indicates that the signal is delayed by one sampling period. Indicates that the signal is delayed by two sampling periods.

[0071] Baseline drift correction is used to process ECG signal filter drift. Baseline drift correction uses a high-pass Chebyshev filter, and the transfer function is expressed as:

[0072] ;

[0073] in, represents the transfer function, represents the passband ripple coefficient, represents the Chebyshev polynomial, represents the filter order, represents a complex variable, represents the imaginary unit, Indicates the cutoff frequency.

[0074] The Butterworth filter is used to eliminate power frequency interference in the respiratory signal filtering process. The square of the amplitude of the transfer function is expressed as:

[0075] ;

[0076] in, represents the magnitude squared response, represents the angular frequency, represents the cut-off frequency, Represents the filter order.

[0077] Signal denoising is performed by wavelet threshold denoising, where the soft threshold function is expressed as:

[0078] ;

[0079] ;

[0080] in, is the signal to be processed; is the threshold, which is automatically selected using the VisuShrink criterion; is the noise standard deviation; is the signal length.

[0081] The denoising process includes wavelet decomposition, threshold processing and wavelet reconstruction;

[0082] In the wavelet decomposition stage, the wavelet basis is selected, the number of decomposition layers is determined, and the wavelet coefficients are obtained; in the threshold processing stage, the noise standard deviation is estimated , calculate the threshold , soft threshold processing is performed on the wavelet coefficients; in the wavelet reconstruction stage, the processed coefficients are used to perform inverse wavelet transform to obtain the denoised signal.

[0083] The time alignment process uses a dynamic time warping algorithm to find the best alignment path by accumulating the minimum distance. The formula is expressed as:

[0084] ;

[0085] in, Represents the local distance, which calculates the distance between the current two points. Indicates the selection of the minimum cumulative distance, where indicates the path from above, represents the path coming from the left, Indicates the path coming from the upper left.

[0086] S2: The preprocessed multimodal physiological data is input into a deep learning model based on LSTM and multi-head attention mechanism, combined with a dynamic weight adaptive algorithm to achieve intelligent fusion analysis of different physiological signals and establish a patient-specific physiological system coupling model.

[0087] Collect preprocessed multimodal physiological signal data, each signal contains a time series feature sequence, and calculate the LSTM output of each input channel, expressed as:

[0088] ;

[0089] in, Indicates The output function of the LSTM unit is No. An input feature sequence, is the input weight matrix, is the circulant weight matrix, is the hidden state at the previous moment, is the bias vector, Indicates that the sigmoid function acts on the cell state.

[0090] Process the original time series data through LSTM and output the hidden state matrix. As the input of the multi-head attention mechanism, for each attention head , the calculation function is expressed as:

[0091] ;

[0092] in, For the The calculation function of the attention head, is the query matrix, is the transposed matrix of the Key matrix, is the Value matrix, is the scaling factor.

[0093] The attention score is obtained by calculating the similarity through the dot product of the query and the key, and then divided by the square root of the dimension to scale it to avoid gradient disappearance. The score is converted into probability weights using softmax, and these weights are multiplied with the value matrix to obtain the weighted result, thereby obtaining a new representation of the current input. Dynamic feature selection and information fusion are achieved.

[0094] According to the calculation results of LSTM and multi-head attention mechanism, a deep learning model is constructed, which is expressed as:

[0095] ;

[0096] in, Represents the feature vector after fusion of multimodal physiological signals, Indicates The output function of the LSTM unit is For the The calculation function of the attention head, For the The dynamic weight coefficient of each feature, is the number of input features, is the number of attention heads, is the energy regulation coefficient, is the total energy of the system, is the spatial feature adjustment coefficient, is the total amount of spatial features.

[0097] The deep learning model captures temporal features through LSTM, uses a multi-head attention mechanism to select features, and achieves information fusion through a nonlinear combination of energy and spatial features. The fractional structure is used to balance temporal and spatial features, and energy constraints are introduced to achieve adaptive fusion of multimodal features.

[0098] Through the deep learning model, various physiological signals are used as input channels, each signal is equipped with an independent LSTM unit for feature extraction, and preliminary feature selection is achieved through the multi-head attention mechanism.

[0099] based on The output of Calculate signal importance and use the attention score Evaluate feature relevance and dynamically adjust weight coefficients considering signal quality, timing stability, and historical importance .

[0100] The attention score is calculated by calculating the similarity between the query and the key, which is extended to the correlation of physiological signals, establishing the correlation between physiological signals, and analyzing the temporal dependency between signals, which is expressed as:

[0101] ;

[0102] in, For signal and The correlation coefficient between For the The signal is The value of the moment, For the The signal is The value of the moment, For signal The mean of For signal The mean of is the length of the time series.

[0103] The attention mechanism is a short-term, dynamic similarity calculation, and the correlation is a global similarity assessment of the entire time series. It provides a stable correlation strength measure between signals and provides basic parameters for subsequent system coupling modeling.

[0104] On the basis of correlation, the time factor is introduced to calculate the coupling strength, which is expressed as:

[0105] ;

[0106] in, For signal and The coupling strength between is the absolute value of the correlation coefficient, is the time attenuation coefficient, For signal and The time delay between.

[0107] Based on all signal pairs Construct a coupling matrix with each physiological signal as a node. The value is used as the connection strength between nodes to form the initial network structure;

[0108] LSTM is used to capture the temporal characteristics of each signal, and the coupling weights are updated in combination with the attention mechanism, and the network structure is dynamically adjusted over time.

[0109] Calculate the normal fluctuation range of each signal, identify typical signal interaction patterns, analyze the speed characteristics of signal response, evaluate the amplitude of system fluctuations, integrate the system model to establish a patient-specific physiological system coupling model, expressed as:

[0110] ;

[0111] in, The signal for prediction exist The value of the moment, is the associated signal, is the individualized parameter set, is a non-linear activation function.

[0112] S3: Based on the analysis results of the deep learning model, the patient's disease risk assessment value is calculated in real time, and according to the set multi-level warning threshold, the specific warning level, warning reason and intervention suggestion are output in combination with the explainability module. At the same time, the model parameters are continuously optimized through the online learning mechanism to improve the accuracy of the warning.

[0113] According to the analysis results of the deep learning model, the patient's condition risk assessment value is calculated and expressed as:

[0114] ;

[0115] ;

[0116] in, for The total risk value at the moment, For the The weight coefficient of each indicator is Risk contribution for a single indicator, is the overall stability index of the system. is the risk mapping function, For the predicted The signal is The value of the moment, is the individual baseline value, To allow fluctuation range, is a dynamic weight.

[0117] According to the patient's condition risk assessment value, set a multi-level warning threshold, including the first threshold, the second threshold, and the third threshold, and the first threshold < the second threshold < the third threshold, to warn the patient's single organ function, specifically:

[0118] If the overall risk value of a single organ is abnormal, and the overall risk value exceeds its normal upper limit but is lower than the first warning threshold, the warning level will be set to level one, indicating that the organ function is slightly abnormal and the fluctuation is beyond the normal range. It is necessary to increase the monitoring frequency and prepare targeted intervention plans.

[0119] If the risk value of the organ exceeds the first warning threshold but is lower than the second warning threshold, the warning level is set to level 2, indicating that the organ function is moderately abnormal and requires timely intervention. The organ-specific intervention plan is initiated and the status of other organs is closely monitored.

[0120] If the risk value of the organ exceeds the second warning threshold, the warning level is set to level three, indicating that the organ function is seriously abnormal and there is a risk of functional failure, and the emergency plan needs to be activated immediately.

[0121] If multiple organ risk values ​​are abnormal at the same time, including:

[0122] If two organs are in a mild abnormality at the same time, the warning level will be set to level 2, indicating that multiple organ functions are slightly impaired and there is a risk of mutual influence. A comprehensive assessment is needed to formulate a multi-system intervention plan.

[0123] If at least one of the two organs reaches a moderate abnormality, the warning level will be set to level three, indicating that multiple organ functions are significantly abnormal and system stability is reduced. A multi-system intervention plan must be immediately initiated to prepare for emergency treatment.

[0124] If three or more organs are abnormal at the same time, the warning level will be set to level 4, indicating the risk of multiple organ failure and severe system instability. It is necessary to activate the multiple organ failure treatment plan and comprehensive intervention.

[0125] If the coupling strength between organs is abnormal, including:

[0126] If the coupling intensity increases significantly, the warning level will be raised by one level compared to the current level, indicating that the abnormal interaction between organs is increasing, there is a risk of chain reaction, and preventive intervention is needed for other organs that may be affected.

[0127] If the coupling strength is significantly weakened, the warning level will remain unchanged compared to the current level, indicating that the coordination between organs has decreased and the functional independence has increased, and the independent functional status of each organ needs to be evaluated.

[0128] If the rate of change of the risk value shows a slow upward trend, the warning level will remain unchanged compared to the current level, indicating that the condition is slowly worsening and the current intervention plan and preventive measures need to be strengthened.

[0129] If the rate of change of the risk value shows a rapid upward trend, the warning level will be raised by one level compared to the current level, indicating that the condition is rapidly deteriorating and requires emergency intervention, adjustment of the treatment plan, and strengthening of monitoring efforts.

[0130] If the rate of change of the risk value increases suddenly, the warning level will be directly raised to the highest level, indicating that the condition has suddenly worsened and is extremely dangerous. The emergency plan must be immediately activated and all-out treatment must be provided.

[0131] In summary, the present invention realizes all-round monitoring and early warning of the patient's condition by constructing a multi-level and multi-dimensional early warning threshold system. A dynamic threshold adjustment mechanism is adopted to perform real-time optimization according to the individual characteristics of the patient, treatment feedback and system stability, thereby improving the accuracy and timeliness of the early warning. It can significantly improve the quality of monitoring of critically ill patients, reduce the rate of missed judgment and misjudgment, provide powerful auxiliary decision-making support for medical staff, and ultimately achieve the purpose of improving treatment effects and reducing mortality. Especially in the treatment process of complex critical diseases such as multiple organ dysfunction, the system's multi-dimensional early warning mechanism can help medical staff better grasp the development trend of the disease and realize precise and individualized treatment intervention.

[0132] This embodiment further provides an intelligent nursing management system for critically ill patients, including:

[0133] The acquisition module sets up multiple physiological signal acquisition channels to synchronously acquire the patient's multi-dimensional physiological signals and pre-processes the acquired signals, including signal filtering, denoising and time alignment operations, to form a standardized multi-modal physiological data stream;

[0134] The construction module is used to input the pre-processed multimodal physiological data into the deep learning model based on LSTM and multi-head attention mechanism, and realize the intelligent fusion analysis of different physiological signals by extracting time domain features, frequency domain features and nonlinear features, combined with the dynamic weight adaptive algorithm, and establish a patient-specific physiological system coupling model;

[0135] In the early warning module, based on the analysis results of the deep learning model, the system calculates the patient's disease risk assessment value in real time, and outputs a specific warning level based on the set multi-level warning thresholds.

[0136] This embodiment also provides a computer device, which is suitable for the intelligent nursing management method for critically ill patients, including a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement the intelligent nursing management method for critically ill patients proposed in the above embodiment.

[0137] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0138] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the method for intelligent nursing management of critically ill patients proposed in the above embodiment is implemented.

[0139] The storage medium proposed in this embodiment and the data storage method proposed in the above embodiment belong to the same inventive concept. The technical details not fully described in this embodiment can be found in the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0140] Example 2: This example provides an intelligent nursing management method for critically ill patients. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0141] In order to verify the effectiveness of the early warning system of the present invention, data from 300 critically ill patients in the intensive care unit of a tertiary hospital from January 2023 to December 2023 were selected for comparative experiments. The patients were randomly divided into an experimental group (150 cases) and a control group (150 cases). The experimental group was monitored using the early warning system of the present invention, and the control group used the traditional single threshold early warning method. There was no statistical difference between the two groups of patients in baseline characteristics such as age, gender, and disease severity (P>0.05). The experiment lasted 30 days, and the results are shown in Table 1:

[0142] Table 1 Comparative analysis table

[0143]

[0144] The results of a randomized controlled trial of 300 critically ill patients showed that the early warning system of the present invention was significantly superior to the traditional single-threshold early warning method in multiple key indicators. The early warning accuracy of the experimental group reached 92.4%, an increase of 13.8 percentage points over the control group, while achieving lower missed and false alarm rates. It is particularly noteworthy that the system can issue an early warning signal 6.8±1.2 hours in advance, about 3.6 hours earlier than the traditional method, which wins valuable time for timely intervention by medical staff.

[0145] In terms of clinical effects, the patient group using this early warning system showed better prognostic indicators, the timely adjustment rate of the intervention plan increased to 95.6%, and the incidence of patient complications was significantly reduced to 15.3%. These data fully confirm that the present invention has significant clinical application value in improving early warning accuracy, improving patient prognosis, and optimizing the use of medical resources, and the improvement of all evaluation indicators is statistically significant (P<0.001).

[0146] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. Intelligent nursing management method for critically ill patients, characterized by: include: Set up multiple physiological signal acquisition channels to synchronously acquire multi-dimensional physiological signals of patients, and pre-process the acquired signals, including signal filtering, denoising and time alignment operations, to form a standardized multi-modal physiological data stream; The pre-processed multimodal physiological data is input into the deep learning model based on LSTM and multi-head attention mechanism, combined with the dynamic weight adaptive algorithm to realize the intelligent fusion analysis of different physiological signals and establish a patient-specific physiological system coupling model; Based on the analysis results of the deep learning model, the patient's disease risk assessment value is calculated in real time, and the specific warning level is output according to the set multi-level warning threshold. At the same time, the model parameters are continuously optimized through the online learning mechanism to improve the accuracy of the warning; The multiple physiological signal acquisition channels include a cardiovascular system signal acquisition channel, a respiratory system signal acquisition channel, a body temperature monitoring channel, a nervous system monitoring channel, a biochemical parameter monitoring channel, and a body movement monitoring channel; The cardiovascular system signal acquisition channel acquires electrocardiogram signals and blood pressure signals; The respiratory system signal acquisition channel acquires respiratory waveform signals and blood oxygen saturation signals; The body temperature monitoring channel collects body temperature data; The nervous system monitoring channel collects electroencephalogram data and electromyogram data; The biochemical parameter monitoring channel monitors blood sugar data; The body movement monitoring channel detects body position data; The preprocessing includes signal filtering, signal denoising and time alignment; The signal filtering process includes ECG signal filtering and respiratory signal filtering. The ECG signal filtering process uses a notch filter to eliminate power frequency interference. The transfer function is expressed as: ； in, is the transfer function of the system, is the power frequency, is the sampling frequency, is the filter pole radius, Indicates that the signal is delayed by one sampling period. Indicates that the signal is delayed by two sampling periods; Baseline drift correction is used to process ECG signal filter drift. The baseline drift correction uses a high-pass Chebyshev filter, and the transfer function is expressed as: ； in, represents the transfer function, represents the passband ripple coefficient, represents the Chebyshev polynomial, represents the filter order, represents a complex variable, represents the imaginary unit, represents the cut-off frequency; The Butterworth filter is used to eliminate power frequency interference in the respiratory signal filtering process. The square of the amplitude of the transfer function is expressed as: ； in, represents the magnitude squared response, represents the angular frequency, represents the cut-off frequency, represents the filter order; The signal denoising process is performed by wavelet threshold denoising, wherein the soft threshold function is expressed as: ； ； in, is the signal to be processed; is the threshold, which is automatically selected using the VisuShrink criterion; is the noise standard deviation; is the signal length; The time alignment process uses a dynamic time warping algorithm to find the best alignment path by accumulating the minimum distance. The formula is expressed as: ； in, Represents the local distance, which calculates the distance between the current two points. Indicates the selection of the minimum cumulative distance, where indicates the path from above, represents the path coming from the left, Indicates the path coming from the upper left; Based on the preprocessed multimodal physiological signal data, each signal contains a time series feature sequence, and the LSTM output of each input channel is calculated, which is expressed as: ； in, Indicates The output function of the LSTM unit is No. An input feature sequence, is the input weight matrix, is the circulant weight matrix, is the hidden state at the previous moment, is the bias vector, Indicates that the sigmoid function acts on the cell state; Process the original time series data through LSTM and output the hidden state matrix. As the input of the multi-head attention mechanism, for each attention head , the calculation function is expressed as: ； in, For the The calculation function of the attention head, is the query matrix, is the transposed matrix of the Key matrix, is the Value matrix, is the scaling factor; According to the calculation results of LSTM and multi-head attention mechanism, a deep learning model is constructed, which is expressed as: ； in, Represents the feature vector after fusion of multimodal physiological signals, Indicates The output function of the LSTM unit is For the The calculation function of the attention head, For the The dynamic weight coefficient of each feature, is the number of input features, is the number of attention heads, is the energy regulation coefficient, is the total energy of the system, is the spatial feature adjustment coefficient, is the total amount of spatial features; Based on the attention score of the multi-head attention mechanism, the similarity between the query and the key is extended to the physiological signal correlation, the correlation between the physiological signals is established, and the temporal dependency between the signals is analyzed, which is expressed as: ； in, For signal and The correlation coefficient between For the The signal is The value of the moment, For the The signal is The value of the moment, For signal The mean of For signal The mean of is the length of the time series; On the basis of correlation, the time factor is introduced to calculate the coupling strength, which is expressed as: ； in, For signal and The coupling strength between is the absolute value of the correlation coefficient, is the time attenuation coefficient, For signal and The time delay between By calculating the normal fluctuation range of each signal, identifying the typical signal interaction mode, analyzing the speed characteristics of the signal response, evaluating the amplitude of the system fluctuation, and integrating the system model to establish the patient's individualized physiological system coupling model, it is expressed as: ； in, The signal for prediction exist The value of the moment, is the associated signal, is the individualized parameter set, is a non-linear activation function.

2. The intelligent nursing management method for critically ill patients according to claim 1, characterized in that: Based on the analysis results of the deep learning model, the patient's condition risk assessment value is calculated and expressed as: ； ； in, for The total risk value at the moment, For the The weight coefficient of each indicator is Risk contribution for a single indicator, is the overall stability index of the system. is the risk mapping function, For the predicted The signal is The value of the moment, is the individual baseline value, To allow fluctuation range, is the dynamic weight; According to the patient's condition risk assessment value, set a multi-level warning threshold, including the first threshold, the second threshold, and the third threshold, and the first threshold < the second threshold < the third threshold, to warn the patient's single organ function, specifically: If the overall risk value of a single organ is abnormal, and the overall risk value exceeds its normal upper limit but is lower than the first warning threshold, the warning level is set to level one, indicating that the organ function is slightly abnormal and the fluctuation is beyond the normal range. It is necessary to increase the monitoring frequency and prepare targeted intervention plans; If the risk value of the organ exceeds the first warning threshold but is lower than the second warning threshold, the warning level is set to level 2, indicating that the organ function is moderately abnormal and requires timely intervention. The organ-specific intervention plan is initiated and the status of other organs is closely monitored; If the risk value of the organ exceeds the second warning threshold, the warning level is set to level three, indicating that the organ function is seriously abnormal and there is a risk of functional failure, and the emergency plan needs to be activated immediately.

3. An intelligent nursing management system for critically ill patients, based on the intelligent nursing management method for critically ill patients described in any one of claims 1 to 2, characterized in that: include: The acquisition module sets up multiple physiological signal acquisition channels to synchronously acquire the patient's multi-dimensional physiological signals and pre-processes the acquired signals, including signal filtering, denoising and time alignment operations, to form a standardized multi-modal physiological data stream; A building module is used to input the pre-processed multimodal physiological data into a deep learning model based on LSTM and multi-head attention mechanism, combined with a dynamic weight adaptive algorithm to achieve intelligent fusion analysis of different physiological signals and establish a patient-specific physiological system coupling model; The early warning module calculates the patient's disease risk assessment value in real time based on the analysis results of the deep learning model, and outputs a specific warning level according to the set multi-level warning thresholds.

Images