Intelligent nursing body position adjusting method and system based on respiratory function protection

Abstract

The invention discloses an intelligent nursing body position adjusting method and system based on respiratory function protection, and relates to the technical field of artificial intelligence and medical nursing. The method comprises the steps that body pressure distribution signals and respiratory thoracic impedance signals of a patient are collected in real time, a physiological data sequence is generated through Kalman filtering and wavelet fusion noise reduction, the physiological data sequence is input into a pre-trained deep learning network, and fusion feature vectors and body position performance indexes are output. And if the body position performance index is not lower than a preset threshold value, inputting the fusion feature vector into a pre-trained classification model, and outputting a classification label of the breathing state. A preset parameter optimization mapping table is inquired according to the classification labels, candidate parameter combinations sorted according to priorities are obtained, the parameter group with the highest priority is selected, a multi-motor cooperation instruction is calculated through a feedback control algorithm, and dynamic regulation and control of the bed surface are achieved. The invention aims to realize respiratory function protection and individualized nursing by sensing the respiratory state in real time, intelligently deciding and adjusting parameters, cooperatively executing control and adopting a safety protection mechanism.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of artificial intelligence and medical care technology, and in particular to an intelligent nursing body position adjustment method and system based on respiratory function protection. BACKGROUND

[0002] Intelligent nursing beds are an important research direction in the field of clinical nursing and health monitoring. Currently, existing systems mostly adopt fixed-time turning, fixed-angle adjustment or rely on manual operation, and some support voice, gesture and other interactions, but still cannot be applied to patients with speech or movement disorders, and cannot respond to physiological state changes. In addition, existing systems have many limitations in function: first, they usually only use pressure sensors to prevent pressure ulcers and cannot assess the impact of body position on respiratory function, lacking adjustment mechanisms targeting respiratory improvement; second, adjustment strategies are based on fixed rules and cannot identify respiratory state deterioration or generate optimized parameters based on individual data; third, the execution level lacks multi-motor collaborative control and real-time safety monitoring, making it difficult to achieve precise and smooth body position adjustment, and when local pressure is too high or respiratory abnormalities occur, the system cannot interrupt and return to a safe body position in time, posing a clinical risk. This passive management approach that ignores respiratory function and individual differences not only limits the quality of care, but also increases the risk of respiratory system complications. SUMMARY

[0003] To address the problems of existing systems, such as the disconnection between physiological monitoring and body position control, the lack of intelligence and predictability in decision-making mechanisms, and insufficient control precision and safety, the present application proposes an intelligent nursing body position adjustment method and system based on respiratory function protection, aiming to protect respiratory function and provide individualized care by real-time respiratory state sensing, intelligent adjustment parameter decision-making, collaborative control execution, and safety protection mechanisms.

[0004] The present application achieves the above-mentioned objectives through the following technical solutions:

[0005] The intelligent nursing body position adjustment method based on respiratory function protection comprises:

[0006] Real-time acquisition of patient body pressure distribution signals and respiratory thoracic impedance signals;

[0007] Fusion and denoising processing of the acquired body pressure distribution signals and respiratory thoracic impedance signals using Kalman filtering and wavelet denoising algorithms to generate physiological data sequences, including pressure distribution time series data and respiratory waveform time series data;

[0008] Inputting the physiological data sequences into a pre-trained deep learning network to output fusion feature vectors and body position performance indicators;

[0009] If the body position performance index exceeds or is equal to the preset threshold value, the fusion feature vector is input into a pre-trained classification model, and a classification label of the respiratory state is output, including normal, mild limitation, moderate limitation and severe limitation;

[0010] According to the classification label, a preset parameter optimization mapping table is queried to obtain a plurality of candidate parameter combinations of bed surface adjustment, and an adjustment priority sequence is generated;

[0011] The highest priority parameter group is selected from the adjustment priority sequence, and a feedback control algorithm is used to calculate multi-motor cooperative control instructions to dynamically control the cooperative operation of the multi-motor.

[0012] As a preferred scheme of the present application, the data processing process of the pre-trained deep learning network comprises:

[0013] Based on the pressure distribution time series data, the pressure center offset trajectory and the regional pressure imbalance degree feature are extracted through a spatial attention convolutional network;

[0014] Based on the respiratory waveform time series data, the respiratory harmonic component energy ratio and the instantaneous frequency fluctuation feature are extracted through a time series convolutional network;

[0015] The pressure center offset trajectory and the regional pressure imbalance degree feature, the respiratory harmonic component energy ratio and the instantaneous frequency fluctuation feature are dimensionally aligned and fused through a feature fusion layer to generate a fusion feature vector;

[0016] The fusion feature vector is processed through a fully connected regression network to generate a body position performance index.

[0017] As a preferred scheme of the present application, the construction steps of the preset parameter optimization mapping table comprise:

[0018] According to historical clinical data, statistical analysis of bed surface adjustment parameter intervals corresponding to different respiratory state levels is performed to generate an initial parameter interval set; the bed surface adjustment parameters include a backboard target angle, a leg plate target angle, a bed body overall inclination angle and an adjustment speed constraint;

[0019] According to the initial parameter interval set, a multi-objective optimization algorithm is used for Pareto frontier search to generate an optimized parameter group set composed of a plurality of Pareto optimal parameter combinations; the Pareto optimality is to achieve an optimal balance between respiratory improvement benefit and execution cost; the execution cost is positively correlated with the sum of absolute values of changes of each target angle parameter, and is positively correlated with the inverse of the adjustment speed constraint;

[0020] A comprehensive priority score is calculated for each parameter group in the optimized parameter group set, and the calculation formula is:

[0021] S i =α×Q iβ×P i

[0022] In the formula, S i is the comprehensive priority score of the i th parameter set; Q i is the respiratory improvement benefit score of the i th parameter set; P i is the execution cost score of the i th parameter set; α is the weight factor of the respiratory improvement benefit; β is the weight factor of the execution cost; and α+β=1;

[0023] The set of optimized parameter sets with comprehensive priority scores is verified and labeled in combination with clinical expert experience to generate a preset parameter optimization mapping table containing comprehensive priority scores, which can be used for real-time query.

[0024] As a preferred scheme of the present application, the generating step of the adjustment priority sequence comprises:

[0025] According to the classification label, the preset parameter optimization mapping table is queried to obtain a plurality of groups of candidate parameter combinations for bed surface adjustment, which have been provided with comprehensive priority scores;

[0026] The plurality of groups of candidate parameter combinations for bed surface adjustment obtained are sorted from high to low according to the comprehensive priority scores to generate an adjustment priority sequence.

[0027] As a preferred scheme of the present application, the calculation of the multi-motor collaborative control instruction by using the feedback control algorithm comprises:

[0028] The backboard target angle, the leg plate target angle and the overall bed body inclination angle in the highest priority parameter set are taken as a target angle vector;

[0029] Actual backboard angle, actual leg plate angle and actual bed body inclination angle are obtained according to real-time measurement data of the bed body posture sensor to generate an actual angle vector;

[0030] An angle deviation vector between the target angle vector and the actual angle vector is calculated;

[0031] According to the angle deviation vector, a multi-input multi-output adaptive fuzzy PID controller is used to generate a group of motor control signals for eliminating the angle deviation;

[0032] The group of motor control signals is decomposed and converted into control instructions of each motor by using a kinematics inverse solution algorithm;

[0033] The control instructions of each motor are sent to the corresponding motor driver, and real-time position feedback from the motor encoder and current feedback from the driver are received until the absolute values of each component of the angle deviation vector are less than a preset angle error threshold.

[0034] As a preferred scheme of the present application, the method further comprises a safety protection mechanism:

[0035] The average pressure of any local area in the body pressure distribution signal is calculated in real time, and if the average pressure of any area continuously exceeds the preset pressure threshold, it is determined that the local pressure is out of limit;

[0036] The respiratory thoracic impedance signal is analyzed in real time, the instantaneous respiratory frequency is calculated, and if the instantaneous respiratory frequency continuously falls below the first preset frequency threshold or continuously rises above the second preset frequency threshold, it is determined that the respiratory frequency is abnormal;

[0037] When any of the local pressure out of limit or the respiratory frequency abnormal occurs:

[0038] If in the process of body position adjustment, the current adjustment operation is immediately suspended, the body position is adjusted to the preset reference safe body position, the sound and light alarm is simultaneously sent, and the parameter combination triggering the local pressure out of limit or the respiratory frequency abnormal is recorded;

[0039] If within the preset time period after the body position adjustment is completed, the body position is adjusted to the preset reference safe body position, the sound and light alarm is simultaneously sent, and the parameter combination triggering the local pressure out of limit or the respiratory frequency abnormal is recorded.

[0040] As a preferred scheme of the present application, the method further comprises an optimization mechanism:

[0041] The recorded parameter combination triggering the local pressure out of limit or the respiratory frequency abnormal is added to the disabled parameter group set of the preset parameter optimization mapping table, and the parameter group is excluded in subsequent queries.

[0042] The adjustment system of the intelligent nursing body position adjustment method based on respiratory function protection, the system comprises:

[0043] A data acquisition module for acquiring patient body pressure distribution signals and respiratory thoracic impedance signals in real time;

[0044] A data processing module for fusing and denoising the collected body pressure distribution signals and respiratory thoracic impedance signals using Kalman filtering and wavelet denoising algorithm to generate physiological data sequences;

[0045] An intelligent analysis module for inputting the physiological data sequences into a pre-trained deep learning network to output fusion feature vectors and body position performance indicators;

[0046] A respiratory state classification module for inputting the fusion feature vectors into a pre-trained classification model when the body position performance indicators exceed or equal to the preset threshold to output classification labels of the respiratory state, including normal, mild limitation, moderate limitation and severe limitation;

[0047] The adjustment parameter decision module is configured to query a preset parameter optimization mapping table according to the classification label, obtain a plurality of candidate parameter combinations of bed surface adjustment, and generate an adjustment priority sequence.

[0048] The control execution module is configured to select a parameter group with the highest priority from the adjustment priority sequence, and calculate a multi-motor cooperative control instruction by using a feedback control algorithm, so as to dynamically control the cooperative operation of the multi-motor.

[0049] The safety monitoring module is configured to monitor the body pressure distribution signal and the respiratory thoracic impedance signal in real time, and when it is judged that there is local pressure overrun or abnormal respiratory frequency, trigger the operations of stopping adjustment, restoring to a preset reference safe body position and alarming, and record the parameter combination triggering the local pressure overrun or abnormal respiratory frequency.

[0050] The optimization module is configured to add the recorded parameter combination triggering the local pressure overrun or abnormal respiratory frequency to a disabled parameter group set of the preset parameter optimization mapping table.

[0051] A computer device comprises a memory and a processor, the memory stores a computer program, and the processor implements the steps of the intelligent nursing body position adjustment method based on respiratory function protection when executing the computer program.

[0052] A computer readable storage medium stores a computer program, and the computer program implements the steps of the intelligent nursing body position adjustment method based on respiratory function protection when executed by a processor.

[0053] The beneficial effects of the present application are: by real-time acquisition of body pressure distribution and respiratory thoracic impedance signals, and real-time identification of respiratory state by combining a deep learning network, the body position can be dynamically adjusted according to the actual respiratory condition of the patient, the respiratory function can be effectively improved, and the risk of complications can be reduced. On this basis, relying on historical clinical data and combining a multi-objective optimization algorithm, a parameter optimization mapping table containing a Pareto optimal parameter combination is constructed, and then verified by clinical expert experience, precise parameter recommendation for different people is realized, and the precision and adaptability of nursing are improved. At the control level, a multi-input multi-output adaptive fuzzy PID controller is used, combined with a multi-motor cooperative control architecture, to realize cooperative decoupling and precise adjustment of multiple variables such as the inclination angle of the back plate, the leg plate and the bed body, which not only improves the tracking accuracy and response speed of the target body position of the patient, but also ensures the stability and comfort of the body position adjustment process. At the same time, a real-time safety monitoring mechanism is introduced to dynamically judge local pressure overrun and abnormal respiratory frequency, and once a risk is identified, the adjustment process is immediately interrupted and automatically restored to a safe body position, while recording and disabling the related parameter combination, thereby enhancing the clinical safety and reliability. In addition, closed-loop control is realized based on physiological signals, without the need for active interaction of the patient, further expanding the application range of intelligent nursing beds in critical, coma and disabled populations. BRIEF DESCRIPTION OF DRAWINGS

[0054] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed to be used in the embodiment description will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without any creative effort on the basis of these drawings.

[0055] Wherein:

[0056] Figure 1 The flow chart of the intelligent nursing body position adjustment method based on respiratory function protection in the embodiment of the present application;

[0057] Figure 2 The modular structure schematic diagram of the adjustment system of the intelligent nursing body position adjustment method based on respiratory function protection in the embodiment of the present application. DETAILED DESCRIPTION

[0058] In order to make the objects, technical solutions and advantages of the embodiments of the present application more clear, the technical solutions of the embodiments of the present application will be described clearly and completely in the following with reference to the drawings of the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, not all the embodiments. Based on the described embodiments of the present application, all other embodiments obtained by those skilled in the art belong to the scope of protection of the present application.

[0059] The existing intelligent nursing system mostly adopts timing, fixed angle adjustment or relies on manual and simple interaction mode, which is difficult to apply to patients with speech or movement disorders, and cannot respond to physiological state changes. The existing technology only relies on pressure sensing to prevent pressure ulcers, cannot evaluate the influence of body position on respiratory function, and lacks adjustment mechanism targeting respiratory improvement. Meanwhile, the adjustment strategy is mostly based on fixed rules, and cannot identify respiratory state deterioration or generate individualized parameters. In addition, there is a lack of multi-motor cooperation and real-time safety monitoring, the adjustment is not accurate, and when the pressure is too high or the respiratory function is abnormal, the system cannot interrupt and restore the safe body position in time, which brings clinical risks. In view of the above problems, the present application proposes an intelligent nursing body position adjustment method and system based on respiratory function protection. The method collects body pressure and respiratory thoracic impedance signals in real time, identifies the respiratory state by using a deep learning network, dynamically adjusts the body position, and effectively improves the respiratory function. Based on historical clinical data and multi-objective optimization algorithm, a parameter optimization mapping table is constructed, and the individualized parameter recommendation is realized through clinical expert verification. Meanwhile, a multi-input multi-output adaptive fuzzy PID controller is adopted, combined with a multi-motor cooperative control architecture, to realize smooth and stable adjustment of the body position. In addition, a real-time safety monitoring mechanism is introduced to dynamically judge the pressure overrun and respiratory abnormalities. Once the risk is identified, the adjustment is interrupted and restored to the safe body position immediately, and the related parameters are recorded and disabled, thereby improving the safety of body position adjustment.

[0060] As shown in Figure 1 An embodiment of the present application provides an intelligent nursing body position adjustment method based on respiratory function protection, which comprises:

[0061] S1, collecting body pressure distribution signals and respiratory thoracic impedance signals of a patient in real time;

[0062] The two types of physiological signals are collected in real time by a sensor system deployed on an intelligent nursing bed.

[0063] In some embodiments, the body pressure distribution signals of the patient are collected in real time by adopting a distributed flexible pressure sensor array, such as the FlexiForce series sensor. The array is embedded in the surface layer of the mattress in the form of an MxN matrix, fully covering the effective load-bearing area of the bed surface. In addition, a bioelectrical impedance measurement module based on the four-electrode method is used to accurately collect the respiratory thoracic impedance signals. In this process, two pairs of electrodes are placed on the mattress at the predetermined positions corresponding to the left and right sides of the patient's chest, respectively, wherein the two pairs of electrodes include excitation electrodes and detection electrodes.

[0064] S2, adopting Kalman filtering and wavelet denoising algorithm to fuse and denoise the collected body pressure distribution signals and respiratory thoracic impedance signals, to generate physiological data sequences, including pressure distribution time series data and respiratory waveform time series data;

[0065] Kalman filter is mainly used to process respiratory thoracic impedance signals, effectively suppress power frequency interference and random noise, and obtain smooth respiratory waveform time series data. Wavelet denoising is mainly used to process body pressure distribution signals, and is used to eliminate high-frequency noise caused by patient micro-movement or sensor itself.

[0066] The denoised body pressure distribution signal and the respiratory thoracic impedance signal are time-aligned. On this basis, the readings of all pressure sensors at the same time are combined into an MxN pressure distribution matrix, and are arranged in time sequence to construct a three-dimensional data structure. At the same time, the output respiratory waveform time series data after Kalman filtering are arranged in time sequence to form a one-dimensional time series. Finally, a synchronized and denoised multi-modal physiological data sequence is generated.

[0067] Through Kalman filtering and wavelet denoising algorithm, the collected signals are fused and denoised, and high-quality physiological data sequence is generated for subsequent analysis.

[0068] S3, inputting the physiological data sequence into the pre-trained deep learning network to output a fusion feature vector and a body position performance index;

[0069] The data processing process of the pre-trained deep learning network includes:

[0070] Based on the pressure distribution time series data, the pressure center offset trajectory and the regional pressure imbalance degree feature are extracted through the spatial attention convolutional network.

[0071] By introducing a spatial attention mechanism in the convolutional network, the network can focus more on the areas with concentrated or rapidly changing pressure. The pressure center offset trajectory represents the time series feature vector of the movement law of the pressure center point, and the regional pressure imbalance degree feature quantifies the scalar feature of the pressure difference in each region.

[0072] Processing the time series data of the pressure distribution through the spatial attention convolutional network can effectively extract the spatial features related to the pressure ulcer risk and body position stability.

[0073] Based on the respiratory waveform time series data, the respiratory harmonic component energy ratio and the instantaneous frequency fluctuation feature are extracted through the time series convolutional network.

[0074] The time series convolutional network contains 4 layers, each layer is provided with a causal dilated convolution, and the dilated coefficient is multiplied by 2 layer by layer. This structure can effectively capture the long-term dependence of the respiratory waveform. The energy ratio of the respiratory harmonic component is obtained by calculating the energy ratio of a specific frequency band through fast Fourier transform, which is used to represent the regularity of respiration. The instantaneous frequency fluctuation feature is obtained by calculating the standard deviation of the instantaneous frequency, which is used to represent the stability of respiration.

[0075] The extracted features are dimensionally aligned and fused through a feature fusion layer to generate a fused feature vector;

[0076] The specific steps of generating the fused feature vector are as follows:

[0077] The extracted features are mapped to a unified feature dimension through a fully connected layer;

[0078] The mapped feature vectors are spliced in the feature dimension to form a fused feature vector.

[0079] The fused feature vector is processed through a fully connected regression network to generate a body position performance indicator.

[0080] The body position performance indicator is a scalar value between 0 and 1. The closer the value is to 0, the more ideal the current body position is; otherwise, the closer the value is to 1, the more the body position needs to be adjusted.

[0081] The historical clinical data containing rich annotation information are used to train the deep learning network. Each training sample includes a segment of synchronous pressure and respiration data, and a "body position performance" label evaluated by a clinical expert according to the real-time state of the patient. Specifically, 0 represents an ideal state, and 1 represents an undesirable state. During training, the Adam optimizer is used, and the mean square error is selected as the loss function. The training is iterated continuously until the performance of the model on the validation set converges.

[0082] The fused feature vector is extracted through the deep learning network, and the body position performance indicator is generated to quantify the suitability of the current body position for the patient.

[0083] S4, if the body position performance indicator exceeds or is equal to a preset threshold, the fused feature vector is input into a pre-trained classification model to output a classification label of the respiratory state, including normal, mild restriction, moderate restriction, and severe restriction;

[0084] The preset threshold is an empirical value determined through clinical verification, and the value range is 0 to 1. In some embodiments, the typical recommended value of the preset threshold is 0.6. If the body position performance indicator does not exceed 0.6, it is determined that the current body position is generally good and does not need to be adjusted; otherwise, it is determined that the current body position is not good and needs to be determined whether the reason is respiratory function restriction.

[0085] In some embodiments, the pre-trained classification model uses a relatively lightweight fully connected neural network classifier, including an input layer, a hidden layer, and an output layer. The number of neurons in the input layer matches the dimension of the fused feature vector; the hidden layer can be set to one or two layers, with the number of neurons decreasing step by step, using a ReLU activation function, and introducing a dropout layer with a dropout rate of 0.4 to prevent overfitting; the output layer outputs four neurons, corresponding to four respiratory state classification labels.

[0086] The training is based on a historical data set, and the label of each data is determined by a clinical expert according to the respiratory waveform, blood oxygen saturation and other parameters synchronously acquired, and the respiratory state grade, including normal, mild, moderate and severe. In the training process, the Adam optimizer is selected, and the cross-entropy loss function is used for model training to ensure that the probability distribution output by the model is as close as possible to the real expert annotation distribution.

[0087] For the four respiratory state classification labels, specifically, the normal state refers to the regular respiratory waveform, and the frequency and amplitude are within the normal physiological range, for example, 10-25 times / minute; the mild restriction refers to slight abnormality of respiratory frequency or short-term irregularity of waveform, but the blood oxygen saturation does not decrease significantly; the moderate restriction refers to obvious rapid or slow breathing or amplitude reduction, which may be accompanied by mild blood oxygen saturation reduction; the severe restriction refers to severe difficulty in breathing, such as respiratory arrest, extremely shallow and slow breathing or obvious three concave signs, etc., which requires immediate intervention.

[0088] The result of comparing the body position performance index with the preset threshold value is used as a trigger condition, which avoids the tedious classification calculation for each frame of data, thereby improving the processing efficiency. Only when an abnormal condition is indeed detected, a more detailed respiratory state classification diagnosis process is started to accurately determine the severity and specific type of the problem. In addition, the output classification label is directly input to the subsequent step, and the corresponding adjustment strategy is selected according to different labels to ensure the effectiveness of the countermeasures.

[0089] S5, according to the classification label, querying a preset parameter optimization mapping table to obtain a plurality of candidate parameter combinations of bed surface adjustment, and generating an adjustment priority sequence;

[0090] The construction steps of the preset parameter optimization mapping table include:

[0091] According to the historical clinical data, the statistical analysis of the bed surface adjustment parameter intervals corresponding to different respiratory state grades is performed to generate an initial parameter interval set; the bed surface adjustment parameters include the target angle of the back plate, the target angle of the leg plate, the overall inclination angle of the bed body and the adjustment speed constraint;

[0092] Further, the historical clinical data is derived from long-term accumulated intelligent bed usage data from multiple medical institutions, covering patient information of different age groups, different respiratory disease types and different severity, to ensure that the generated parameter interval set has wide representativeness. Each data sample includes patient respiratory state monitoring records, bed surface adjustment parameter records and corresponding clinical effect evaluation data. These data need to be strictly cleaned and preprocessed to exclude outliers and invalid data, ensuring data quality and consistency. In addition, the initial parameter interval set is not fixed and will be regularly updated and corrected with the accumulation of more clinical data and the enrichment of expert experience, to maintain its accuracy and applicability.

[0093] According to the initial parameter interval set, a multi-objective optimization algorithm is used to perform a Pareto front search to generate an optimized parameter group set composed of multiple Pareto optimal parameter combinations; the Pareto optimality is to achieve the optimal balance between respiratory improvement benefit and execution cost; the execution cost is positively correlated with the sum of the absolute values of the changes in each target angle parameter, and is positively correlated with the inverse of the adjustment speed constraint;

[0094] Further, the multi-objective optimization algorithm commonly uses NSGA-II, MOEA / D, and a method combining Pareto algorithm and Monte Carlo tree search.

[0095] A comprehensive priority score is calculated for each parameter group in the optimized parameter group set, and the calculation formula is:

[0096] S i = α × Q i - β × P i

[0097] In the formula, S i is the comprehensive priority score of the i-th parameter group; Q i is the respiratory improvement benefit score of the i-th parameter group; P i is the execution cost score of the i-th parameter group; α is the weight factor of respiratory improvement benefit; β is the weight factor of execution cost; and α + β = 1;

[0098] Further, the values of the weight factors α and β need to consider the experience of clinical experts and the actual needs of patients. For example, for patients with severe respiratory disorders, a higher α weight factor may be set to highlight the benefit of respiratory improvement; and for patients with relatively stable respiratory status, a lower α weight factor may be set to consider the overall execution cost.

[0099] Based on a specific parameter combination, a prediction model trained based on historical clinical data is used to evaluate the expected improvement degree of the current respiratory status, and output a normalized respiratory improvement benefit score Q i , wherein the respiratory improvement benefit score Qi The higher the score, the more significant the respiratory improvement benefit. In some embodiments, the prediction model employs random forest or gradient boosting tree, etc.

[0100] The execution cost score P i The "cost" required to execute the parameter combination is quantified, including factors such as energy consumption, equipment wear and tear, and adjustment time. Among them, the execution cost score P i The higher the score, the greater the execution cost.

[0101] The set of optimization parameter combinations with comprehensive priority scores is verified and labeled in combination with the clinical expert experience, generating a preset parameter optimization mapping table containing comprehensive priority scores that can be used for real-time query.

[0102] Further, the clinical expert, with his rich experience, carefully reviews the set of optimization parameter combinations with comprehensive priority scores. For obviously unreasonable parameter combinations, they are excluded or their weights are adjusted to ensure the safety and clinical effectiveness of the mapping table. This verification process usually requires multiple iterations until the clinical expert is satisfied with all parameter combinations. Among them, the clinical expert can be a respiratory therapist or a rehabilitation physician.

[0103] The preset parameter optimization mapping table is a structured data table that records the recommended bed surface adjustment parameter combinations, comprehensive priority scores, and expert labeling information under different respiratory status levels. In addition, this mapping table is not static and unchangeable, but has a dynamic updating mechanism.

[0104] The generation steps of the adjustment priority sequence include:

[0105] According to the classification label, the preset parameter optimization mapping table is queried to obtain the corresponding candidate parameter combinations for bed surface adjustment with comprehensive priority scores;

[0106] The obtained multiple candidate parameter combinations are sorted from high to low according to the comprehensive priority scores to generate the adjustment priority sequence.

[0107] In this way, it can ensure that those parameter combinations that obtain the maximum respiratory improvement benefit with the minimum execution cost are preferentially selected. In addition, the body position adjustment method based on the preset parameter optimization mapping table has multiple advantages. First, it quantifies clinical experience through data-driven means, reducing the dependence on subjective experience of medical staff. Second, it effectively balances the improvement benefit and execution cost through multi-objective optimization algorithm, avoiding excessive adjustment that leads to energy consumption and equipment wear and tear. Third, the query mechanism of the mapping table can achieve real-time response, ensuring that the intelligent bed can quickly adapt to changes in the patient's respiratory status, thereby improving nursing efficiency and quality.

[0108] S6, selecting the highest priority parameter group from the adjusted priority sequence, using a feedback control algorithm to calculate the multi-motor collaborative control instruction to dynamically regulate the multi-motor collaborative operation.

[0109] The backboard target angle, the leg plate target angle and the overall bed body inclination angle in the highest priority parameter group are taken as the target angle vector.

[0110] According to the real-time measurement data of the bed posture sensor, the actual backboard angle, the actual leg plate angle and the actual bed body inclination angle are obtained to generate an actual angle vector.

[0111] In some embodiments, the bed posture sensor adopts a high-precision, multi-axis inertial measurement unit, which is respectively installed on the backboard, the leg plate and the bed frame body, and is used to measure the absolute angle in three-dimensional space.

[0112] The angle deviation vector between the target angle vector and the actual angle vector is calculated.

[0113] According to the angle deviation vector, a multi-input multi-output adaptive fuzzy PID controller is used to generate a set of motor control signals for eliminating the angle deviation.

[0114] Through the kinematics inverse solution algorithm, the set of motor control signals is decomposed and converted into control instructions for each motor.

[0115] The control instructions of each motor are sent to the corresponding motor driver, and real-time position feedback and current feedback from the motor encoder are received until the absolute value of each component of the angle deviation vector is less than the preset angle error threshold.

[0116] Further, by continuously reading the actual position feedback from the motor encoder and accurately driving the motor through the multi-input multi-output adaptive fuzzy PID controller, the motor can be accurately controlled to adjust the bed surface to the target position. In addition, through the current feedback mechanism, the motor load state is monitored in real time, and once the current of the motor exceeds the preset safety threshold, the driver will immediately trigger an alarm and stop running. In addition, when the absolute value of each component of the angle deviation vector is less than the preset angle error threshold, it is determined that the adjustment process has been completed, and then enters the holding state. In some embodiments, the recommended value of the preset angle error threshold is 0.5°.

[0117] The intelligent nursing body position adjustment method based on respiratory function protection further includes a safety protection mechanism, which is as follows:

[0118] The average pressure of any local area in the body pressure distribution signal is calculated in real time, and if the average pressure of any area continuously exceeds the preset pressure threshold, it is determined that the local pressure is out of limit.

[0119] In some embodiments, the preset pressure threshold is set to 30mmHg, which is a critical threshold to avoid long-term pressure ischemia of tissues, and beyond which there is a risk of pressure ulcer.

[0120] The respiratory thoracic impedance signal is analyzed in real time to calculate the instantaneous respiratory frequency, and if the instantaneous respiratory frequency is continuously below a first preset frequency threshold or continuously above a second preset frequency threshold, it is determined that the respiratory frequency is abnormal.

[0121] Further, the first preset frequency threshold is a preset minimum frequency threshold, and the second preset frequency threshold is a preset maximum frequency threshold.

[0122] In some embodiments, the first preset frequency threshold is set to 10 times / minute for adult patients, and below this value may indicate respiratory depression. The second preset frequency threshold is set to 30 times / minute, and above this value may indicate respiratory distress, anxiety or pain.

[0123] The algorithm implementation of "continuous exceeding" adopts a sliding time window counting method, that is, the controller calculates the average pressure of all local areas every 100 milliseconds. For any area, if more than 80% of the average pressure exceeds 30mmHg within 3 seconds, a local pressure overrun flag is triggered. Similarly, for the instantaneous respiratory frequency, if the instantaneous respiratory frequency calculated within 15 seconds is continuously below 10 times / minute or above 30 times / minute, a respiratory frequency abnormality flag is triggered. This design can effectively avoid false triggering caused by transient interference.

[0124] When either of the local pressure overrun or the respiratory frequency abnormality occurs:

[0125] If it is in the process of body position adjustment, the current adjustment operation is immediately suspended, the body position is adjusted to a preset reference safe body position, a sound and light alarm is simultaneously issued, and the parameter combination triggering the local pressure overrun or the respiratory frequency abnormality is recorded.

[0126] If it is within a preset time period after the body position adjustment is completed, the body position is adjusted to a preset reference safe body position, a sound and light alarm is simultaneously issued, and the parameter combination triggering the local pressure overrun or the respiratory frequency abnormality is recorded.

[0127] In some embodiments, the preset reference safe body position recommended value is a backboard of 20°, a legboard of 0°, and an inclination of 0°, and the preset time period recommended value is 2 minutes. The sound and light alarm mode is that the red LED light at the head of the bed flashes and the medium volume of the buzzer sounds, and at the same time, a message of "safety alarm: pressure overrun, adjustment has been suspended" pops up on the nurse station interface.

[0128] The intelligent nursing body position adjustment method based on respiratory function protection further includes an optimization mechanism, which is as follows:

[0129] The parameter combination triggering local pressure overrun or abnormal respiratory frequency is added to the disabled parameter group set of the preset parameter optimization mapping table, and the parameter group is excluded in subsequent queries.

[0130] As Figure 2 shown, another embodiment of the present application provides an adjustment system for an intelligent nursing body position adjustment method based on respiratory function protection, which comprises a data acquisition module, a data processing module, an intelligent analysis module, a respiratory state classification module, an adjustment parameter decision module, a control execution module, a safety monitoring module, and an optimization module.

[0131] The data acquisition module is configured to acquire patient body pressure distribution signals and respiratory thoracic impedance signals in real time.

[0132] The data processing module is configured to perform fusion and denoising processing on the acquired body pressure distribution signals and respiratory thoracic impedance signals using Kalman filtering and wavelet denoising algorithms to generate physiological data sequences.

[0133] The intelligent analysis module is configured to input the physiological data sequences into a pre-trained deep learning network to output fusion feature vectors and body position performance indicators.

[0134] The respiratory state classification module is configured to input the fusion feature vectors into a pre-trained classification model to output classification labels of respiratory states, including normal, mild restriction, moderate restriction, and severe restriction, when the body position performance indicators exceed or equal to a preset threshold.

[0135] The adjustment parameter decision module is configured to query a preset parameter optimization mapping table according to the classification labels, obtain multiple candidate parameter combinations for bed surface adjustment, and generate an adjustment priority sequence.

[0136] The control execution module is configured to select the highest priority parameter group from the adjustment priority sequence and calculate multi-motor collaborative control instructions using a feedback control algorithm to dynamically control the multi-motor collaborative operation.

[0137] The safety monitoring module is configured to monitor the body pressure distribution signals and respiratory thoracic impedance signals in real time, and when it is determined that there is local pressure overrun or abnormal respiratory frequency, trigger the adjustment to be stopped, the body position to be restored to a preset baseline safe position, and an alarm operation to be performed, and record the parameter combination triggering the local pressure overrun or abnormal respiratory frequency.

[0138] The optimization module is configured to add the recorded parameter combination triggering the local pressure overrun or abnormal respiratory frequency to the disabled parameter group set of the preset parameter optimization mapping table.

[0139] The embodiment also provides a computer device suitable for the intelligent nursing body position adjustment method based on respiratory function protection, comprising a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions to realize all or part of steps of the method provided by the embodiment.

[0140] The embodiment also provides a storage medium on which a computer program is stored, and the computer program is executed by a processor to execute the method in any optional implementation manner of the above-mentioned embodiment. The storage medium can be realized by any type of volatile or non-volatile storage device or a combination thereof, such as a static random access memory (SRAM), an electrically erasable programmable read-only memory (EEPROM), an erasable programmable read-only memory (EPROM), a programmable read-only memory (PROM), a read-only memory (ROM), a magnetic storage, a flash memory, a magnetic disk or an optical disk.

[0141] The storage medium provided by the embodiment belongs to the same inventive concept as the data storage method provided by the above-mentioned embodiment, and the technical details not described in the embodiment can be referred to the above-mentioned embodiment, and the embodiment has the same beneficial effects as the above-mentioned embodiment.

[0142] In summary, the present application collects body pressure distribution signals and respiratory chest impedance signals in real time, dynamically identifies the respiratory state level by combining deep learning network, realizes body position adaptive adjustment based on the actual physiological state of the patient, effectively improves respiratory comfort and ventilation efficiency, and reduces the risk of respiratory system complications. On this basis, combined with historical clinical data and multi-objective optimization algorithm, a Pareto optimal parameter set is generated, and verified by clinical experts to construct a parameter optimization mapping table, realize truly individualized precise parameter recommendation, and improve the accuracy of nursing. At the same time, the adaptive fuzzy PID controller with multiple inputs and multiple outputs is used to analyze the deviation between the target angle vector and the actual angle vector, generate the motor rotation angle and speed command through the kinematics inverse algorithm, and realize the synchronous control of multi-degree-of-freedom position. By real-time monitoring of local pressure and respiratory rate, once the limit or abnormal situation is identified, the adjustment is immediately interrupted and automatically restored to the safe body position, and the parameter combination triggering the abnormality is recorded, enhancing the clinical safety and reliability. By recording abnormal events and parameter use effect, updating the disabled parameter set, realizing long-term adaptation and performance improvement, and further improving the practicality and effectiveness in real scenes. In addition, based on physiological signals, closed-loop control without interaction is realized, which expands the application range in critical, coma and disabled patients. The present application realizes respiratory function protection and individualized nursing.

[0143] The above merely describes a specific embodiment of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think of various changes or replacements within the technical range disclosed by the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for intelligent positioning adjustment of nursing based on respiratory function protection, characterized in that, The method comprises: Real-time acquisition of body pressure distribution signals and respiratory thoracic impedance signals of a patient; Fusion and denoising processing of the acquired body pressure distribution signals and respiratory thoracic impedance signals by using Kalman filtering and wavelet denoising algorithm to generate physiological data sequences, including pressure distribution time series data and respiratory waveform time series data; Inputting the physiological data sequences into a pre-trained deep learning network to output a fusion feature vector and a body position performance index; If the body position performance index exceeds or equals a preset threshold, inputting the fusion feature vector into a pre-trained classification model to output a classification label of the respiratory state, including normal, mild restriction, moderate restriction and severe restriction; According to the classification label, querying a preset parameter optimization mapping table to obtain multiple candidate parameter combinations of bed surface adjustment and generating an adjustment priority sequence; Selecting the highest priority parameter group from the adjustment priority sequence and calculating a multi-motor collaborative control instruction by using a feedback control algorithm to dynamically control the collaborative operation of the multi-motor.

2. The intelligent body position adjustment method based on respiratory function protection of claim 1, wherein, The data processing process of the pre-trained deep learning network comprises: Based on the pressure distribution time series data, extracting pressure center offset trajectory and regional pressure imbalance degree features by using a spatial attention convolutional network; Based on the respiratory waveform time series data, extracting respiratory harmonic component energy ratio and instantaneous frequency fluctuation features by using a time series convolutional network; Aligning and fusing the pressure center offset trajectory and regional pressure imbalance degree features, the respiratory harmonic component energy ratio and the instantaneous frequency fluctuation features by using a feature fusion layer to generate a fusion feature vector; Processing the fusion feature vector by using a fully connected regression network to generate a body position performance index.

3. The intelligent body position adjustment method based on respiratory function protection of claim 1, wherein, The construction steps of the preset parameter optimization mapping table comprise: According to historical clinical data, statistically analyzing bed surface adjustment parameters corresponding to different respiratory state levels to generate an initial parameter interval set; the bed surface adjustment parameters include a backboard target angle, a leg plate target angle, a bed body overall inclination angle and an adjustment speed constraint; according to the initial parameter interval set, performing Pareto front search by using a multi-objective optimization algorithm to generate an optimized parameter group set composed of multiple Pareto optimal parameter combinations; the Pareto optimality is to achieve an optimal balance between respiratory improvement benefit and execution cost; the execution cost is positively correlated with the sum of absolute values of changes of each target angle parameter and is positively correlated with the inverse of the adjustment speed constraint; Calculating a comprehensive priority score for each parameter group in the optimized parameter group set, and the calculation formula is: S i = a x Q i - β x P i wherein S i is the overall priority score for the i-th parameter set; Q i is the respiratory improvement benefit score for the i-th parameter set; P i is the execution cost score for the i-th parameter set; a is a weight factor for the respiratory improvement benefit; b is a weight factor for the execution cost; and a + b = 1. Combining clinical expert experience to verify and label the optimized parameter group set with the comprehensive priority score to generate a preset parameter optimization mapping table containing the comprehensive priority score which can be used for real-time query.

4. The intelligent body position adjustment method based on respiratory function protection of claim 3, wherein, The generation steps of the adjustment priority sequence comprise: According to the classification label, querying the preset parameter optimization mapping table to obtain multiple candidate parameter combinations of bed surface adjustment which have been provided with the comprehensive priority score; According to the comprehensive priority score from high to low, sorting the obtained multiple candidate parameter combinations of bed surface adjustment to generate an adjustment priority sequence.

5. The intelligent body position adjustment method based on respiratory function protection according to claim 4, characterized in that, The calculation of the multi-motor collaborative control instruction by using the feedback control algorithm comprises: The backboard target angle, the leg plate target angle, and the bed body overall inclination angle in the highest priority parameter group are taken as a target angle vector; Actual backboard angle, actual leg plate angle, and actual bed body inclination angle are obtained according to real-time measurement data of the bed body posture sensor, and an actual angle vector is generated; An angle deviation vector between the target angle vector and the actual angle vector is calculated; According to the angle deviation vector, a multi-input multi-output adaptive fuzzy PID controller is used to generate a group of motor control signals for eliminating the angle deviation; The group of motor control signals is decomposed and converted into control instructions for each motor through a kinematics inverse solution algorithm; The control instructions for each motor are sent to the corresponding motor driver, and real-time position feedback from the motor encoder and current feedback from the driver are received until the absolute values of each component of the angle deviation vector are less than a preset angle error threshold.

6. The intelligent body position adjustment method based on respiratory function protection of claim 1, wherein, It also includes a safety protection mechanism: The average pressure of any local area in the body pressure distribution signal is calculated in real time, and if the average pressure of any area continuously exceeds the preset pressure threshold, it is determined that the local pressure is out of limit; The respiratory thoracic impedance signal is analyzed in real time, and the instantaneous respiratory frequency is calculated, and if the instantaneous respiratory frequency continuously falls below the first preset frequency threshold or continuously rises above the second preset frequency threshold, it is determined that the respiratory frequency is abnormal; When any of the local pressure out of limit or the respiratory frequency abnormal occurs: If during the body position adjustment process, the current adjustment operation is immediately suspended, the body position is adjusted to a preset reference safe body position, a sound and light alarm is simultaneously sent out, and the parameter combination triggering the local pressure out of limit or the respiratory frequency abnormal is recorded; If within a preset time period after the body position adjustment is completed, the body position is adjusted to a preset reference safe body position, a sound and light alarm is simultaneously sent out, and the parameter combination triggering the local pressure out of limit or the respiratory frequency abnormal is recorded.

7. The intelligent body position adjustment method based on respiratory function protection of claim 6, wherein, It also includes an optimization mechanism: The recorded parameter combination triggering the local pressure out of limit or the respiratory frequency abnormal is added to the disabled parameter group set of the preset parameter optimization mapping table, and this group of parameters is excluded in subsequent queries.

8. An adjustment system for the intelligent nursing body position adjustment method based on respiratory function protection according to any one of claims 1 to 7, characterized in that, The system includes: A data acquisition module for acquiring patient body pressure distribution signals and respiratory thoracic impedance signals in real time; A data processing module for performing fusion and noise reduction processing on the acquired body pressure distribution signals and respiratory thoracic impedance signals using Kalman filtering and wavelet denoising algorithms to generate physiological data sequences; An intelligent analysis module for inputting the physiological data sequences into a pre-trained deep learning network to output fusion feature vectors and body position performance indicators; A respiratory state classification module for inputting the fusion feature vectors into a pre-trained classification model when the body position performance indicators exceed or equal to a preset threshold to output classification labels of the respiratory state, including normal, mild restriction, moderate restriction, and severe restriction; An adjustment parameter decision module for querying a preset parameter optimization mapping table according to the classification labels to obtain multiple candidate parameter combinations for bed surface adjustment and generating an adjustment priority sequence; A control execution module for selecting the highest priority parameter group from the adjustment priority sequence and calculating multi-motor cooperative control instructions using a feedback control algorithm to dynamically control the multi-motor cooperative operation. A safety monitoring module is configured to monitor the body pressure distribution signal and the respiratory thoracic impedance signal in real time, and trigger suspension of adjustment, return to a preset reference safe body position, and alarm operation when it is determined that there is local pressure overrun or abnormal respiratory frequency, and record the parameter combination triggering the local pressure overrun or abnormal respiratory frequency; An optimization module is configured to add the recorded parameter combination triggering the local pressure overrun or abnormal respiratory frequency to a disabled parameter group set of a preset parameter optimization mapping table. 9.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is characterized in that: The processor executes the computer program to implement the steps of the intelligent nursing body position adjustment method based on respiratory function protection according to any one of claims 1 to 7.

10. A computer readable storage medium having stored thereon a computer program, characterized in that: The computer program is executed by the processor to implement the steps of the intelligent nursing body position adjustment method based on respiratory function protection according to any one of claims 1 to 7.

Images