Nursing pressure real-time monitoring system and method based on multi-modal data

Abstract

The invention discloses a nursing pressure real-time monitoring system and method based on multi-modal data. The system comprises a data acquisition module, an intelligent analysis module, an analysis module and a real-time feedback module. The data acquisition module synchronously acquires physiological data, behavior activity data and environment situation data of nursing personnel and forms a nursing signal; the analysis module receives the nursing signal from the data acquisition module, receives the dynamic situation information from the intelligent analysis module, and generates a pressure state signal containing pressure event evaluation; the real-time feedback module receives the pressure state signal containing the pressure event evaluation from the analysis module, and generates and outputs a real-time feedback signal based on the pressure state signal. According to the nursing pressure real-time monitoring system and method based on the multi-modal data, the problem that traditional wearable equipment only collects physiological behavior data and cannot be associated with specific nursing situations such as drug administration operation and patient first aid can be solved.

Description

Technical Field

[0001] The present invention relates to the field of medical and health information technology, and particularly to a real-time monitoring system and method for nursing stress based on multimodal data. Background Art

[0002] Nursing work is the high-pressure core of the medical system. Statistics from the World Health Organization show that the burnout rate of nursing staff is over 35%. Existing stress monitoring technologies have significant limitations: the mainstream solutions rely on wearable devices to collect physiological indicators such as heart rate variability, but the frequent movement in nursing work leads to severe interference from motion artifacts. For example, the accelerated heart rate during patient handling is misjudged as psychological stress. Behavior monitoring technologies analyze gait through accelerometers but cannot distinguish normal fast walking from rapid movement caused by anxiety. Environmental sensors record noise levels but are not associated with nursing tasks - the impact of the same decibel value on stress is vastly different in routine inspections and emergency situations. A more prominent contradiction is data siloing: physiological monitoring devices, hospital information systems, and environmental sensing networks operate independently, causing stress assessment to be divorced from the actual work scenario. For example, when the existing system detects a peak in skin conductance, it cannot automatically trace whether the nurse was handling chemotherapy drugs or facing family member queries at that moment.

[0003] Current academic research has attempted multimodal fusion, but there are still two major bottlenecks: one is that simple weighted average is used to fuse electrocardiogram, gait, and acoustic features, and no dynamic mapping rules between nursing operation stages and stress representations are established, resulting in subtle hand tremor signals in the night shift medication preparation scenario being ignored by general algorithms; the other is relying on a centralized cloud computing architecture, with a feedback delay of over 3 minutes from data collection, missing the golden intervention window for stress events. Existing patents have proposed stress assessment in combination with electronic medical records, but only statically associate patient disease types with historical stress data and do not capture the context switching of nurses when moving between wards in real time. Market products such as the "Nursing Angel Bracelet" only achieve vibration reminders for excessive heart rate and do not distinguish the sources of physiological load and psychological stress. Therefore, there is an urgent need to develop a real-time monitoring system that is deeply embedded in the nursing workflow, realizes dynamic association of context - physiology - behavior, accurately identifies the causes of stress events, and implements graded interventions. Summary of the Invention

[0004] In view of the above-mentioned disadvantages of the prior art, the purpose of the present invention is to provide a real-time monitoring system and method for nursing pressure based on multi-modal data, which is used to solve the problem that traditional wearable devices only collect physiological behavior data and cannot be associated with specific nursing situations such as drug administration operations and patient first aid. The present invention synchronously collects physiological data, behavior data, and environmental data in nursing operations through multi-source sensors; the intelligent analysis module analyzes the hospital information system and dynamically generates three-dimensional situation information of task type, regional risk, and patient criticality index; the analysis module establishes a correlation rule between the nursing stage and pressure representation: monitors the correlation between hand tremors and heart rate variability during the drug preparation stage, and analyzes the matching degree between gait and environmental channels during the patient transfer stage; when the situation is dynamically associated with abnormal physiological behavior, a graded pressure event marker is triggered, and precise intervention is implemented through a tactile feedback device and a manager terminal.

[0005] The present invention provides a real-time monitoring system for nursing pressure based on multi-modal data, including: A data acquisition module that synchronously acquires physiological data, behavioral activity data, and environmental situation data of nursing staff and forms a nursing signal; An intelligent analysis module that receives the nursing signal from the data acquisition module and extracts dynamic situation information related to the current working state of the nursing staff from it; An analysis module that receives the nursing signal from the data acquisition module and also receives the dynamic situation information from the intelligent analysis module. Through a situation awareness and pressure event association model based on the nursing workflow, it dynamically associates and fuses physiological data, behavioral activity data, and dynamic situation information to generate a pressure state signal including a pressure event assessment. The pressure event association model establishes a spatio-temporal association map centered on the nursing workflow stage, focuses on monitoring the correlation between hand tremor behavior data and heart rate variability physiological data during the drug administration operation period, and strengthens the analysis of the matching degree between gait rapidity behavior data and the width of the environmental channel during the patient transfer stage; A real-time feedback module that receives the pressure state signal including the pressure event assessment from the analysis module and generates and outputs a real-time feedback signal based on the pressure state signal.

[0006] In an embodiment of the present invention, the data acquisition module includes a wearable sensing unit and an environmental perception unit integrated in the nursing staff's work card; the wearable sensing unit continuously acquires electrocardiogram signals and skin electrical response signals through flexible bioelectrodes to form physiological data, and captures limb movement trajectories and posture changes through an inertial measurement unit to form behavioral activity data; the environmental perception unit acquires the sound pressure characteristics of the working area through a distributed noise sensor to form environmental situation data, and receives the nursing task list and patient critical value alarm record pushed by the hospital information system in real time through a near-field communication interface; the wearable sensing unit and the environmental perception unit fuse the three types of data into a nursing signal through a timestamp synchronization mechanism.

[0007] In an embodiment of the present invention, the situation awareness operations performed by the intelligent analysis module include: parsing the data stream of the hospital information system from the nursing signal, and identifying task types such as medicine preparation, critical care, and family member communication currently performed by the nursing staff; combining the positioning beacon coordinates with the ward electronic fence information to determine the stress classification areas such as the surgical preparation area, isolation ward, or nurse station; obtaining the critical illness index and special nursing requirement marks of the currently responsible case in real time by associating with the patient's electronic medical record; finally integrating the task type, the area risk level associated with the nursing signal, and the case critical illness index into a dynamic situation information vector.

[0008] In an embodiment of the present invention, the stress event association model in the analysis module operates according to the following process: establishing a spatio-temporal association map centered on the nursing workflow stage, focusing on monitoring the correlation between hand tremor behavior data and heart rate variability physiological data during the medication administration operation period, and strengthening the analysis of the matching degree between gait rapidity behavior data and the width of the environmental passage during the patient transfer stage; when a sudden increase in skin conductance and abnormal respiratory rate occur in the context of critical patient care, triggering a first-level stress event mark; when continuously high heart rate and irregular activity paths are detected in the context of continuously processing call bells across wards, triggering a second-level stress event mark; finally generating a stress status signal carrying a stress level label and event cause classification.

[0009] In an embodiment of the present invention, the dynamic association and fusion process introduces a situation weighting mechanism: automatically increasing the weight coefficient of the insufficient environmental lighting factor during the night shift workflow stage, and increasing the correction parameter of the restraint feeling of protective equipment on behavior data during the infectious disease isolation area workflow stage; when the system identifies that the nursing staff is continuously processing high-risk medicine preparation tasks, starting a high-sensitivity analysis mode for the subtle hand tremor behavior data collected during the same period; when the situation of communicating with the patient's family members remains extremely high, activating the voice emotion feature extraction algorithm to enhance the data analysis dimension of the environmental situation.

[0010] In an embodiment of the present invention, the real-time feedback module includes multiple output channels: when the stress status signal contains a first-level stress event mark, driving the tactile belt worn by the nursing staff to generate a low-frequency vibration prompt, and projecting a deep breathing guidance animation onto their smart glasses; when the signal contains a second-level stress event mark, playing a predetermined soothing audio through the nurse station broadcast system, and at the same time sending an encrypted warning text to the nursing team leader's terminal; all feedback signals are attached with stress event cause codes, among which events related to medicine verification errors trigger yellow visual warnings, and events related to sudden changes in the patient's condition trigger red visual warnings.

[0011] In one embodiment of the present invention, the analysis module has a built-in personalized calibration engine: baseline physiological parameters are automatically collected during the low-stress period of morning shift handover for nursing staff to establish individualized resting heart rate and gait smoothness baselines; when abnormal ECG signals appear in the midday workflow, the baseline data of the person is called to perform drift compensation calculations; by continuously learning the stress response patterns in the shift cycle, the assessment threshold coefficients of the night shift and the day shift are dynamically adjusted; and a personalized stress sensitivity factor report is generated every week for optimizing the associated model parameters.

[0012] In one embodiment of the present invention, the stress event assessment includes a cause tracing function: when the system detects that an abnormal ECG signal and an abnormal positioning data occur simultaneously, the workflow context data of the previous ten minutes is automatically traced back. If it is found that the person handles three ward emergency calls in succession and passes through a noisy corridor environment, a stress event report with dual attributions of environmental interference and task overload is generated; when the behavioral data shows repeated returns to the pharmacy, the failed drug collection record of the drug management system is associated to form a process-blocking stress event analysis chain.

[0013] In one embodiment of the present invention, dynamic situational information includes team collaboration indicators: obtaining the number of other caregivers within a five-meter range through Bluetooth beacon ranging, and calculating the team response delay time in combination with voice activity detection; automatically increasing the situational stress weight coefficient when the team personnel density in the critically ill patient transfer situation is lower than a threshold and the call response times out; appending a team collaboration missing mark to the stress state signal and suggesting a request for reinforcement.

[0014] The present invention also provides a real-time monitoring method for nursing pressure based on multimodal data, comprising: S1: Synchronously collect the physiological data, behavioral activity data and environmental situation data of the nursing staff and form nursing signals; S2: receiving nursing signals from the data acquisition module, and extracting dynamic situational information related to the current working status of the nursing staff; S3: Receives nursing signals from the data acquisition module and dynamic situational information from the intelligent analysis module, dynamically associates and fuses physiological data, behavioral activity data and dynamic situational information through a situational awareness and stress event association model based on nursing workflow, and generates a stress state signal containing stress event evaluation; S4: receiving a pressure state signal including a pressure event assessment from the analysis module, and generating and outputting a real-time feedback signal based on the pressure state signal.

[0015] The real-time monitoring system and method for nursing pressure based on multi-modal data provided by the present invention synchronously collect physiological data, behavioral data, and environmental data during nursing operations through multi-source sensors; the intelligent analysis module analyzes the hospital information system and dynamically generates three-dimensional context information of task type, regional risk, and patient critical index; the analysis module establishes a correlation rule between the nursing stage and pressure representation: monitors the correlation between hand tremors and heart rate variability during the drug preparation stage, and analyzes the matching degree between gait and environmental channels during the patient transfer stage; when the context is dynamically associated with abnormal physiological behavior, a graded pressure event marker is triggered, and precise intervention is implemented through a tactile feedback device and a manager terminal. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0017] Figure 1 It is a system architecture diagram of the real-time monitoring system for nursing pressure based on multi-modal data; Figure 2 It is a schematic diagram showing the working process of the real-time monitoring system for nursing pressure based on multi-modal data; Figure 3 It is a method flow chart of the real-time monitoring method for nursing pressure based on multi-modal data. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0018] The following specific examples illustrate the embodiments of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments. Various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that, without conflict, the following embodiments and the features in the embodiments can be combined with each other.

[0019] It should be noted that the diagrams provided in the following embodiments only illustrate the basic concept of the present invention in a schematic manner. Therefore, only the components related to the present invention are shown in the diagrams, rather than being drawn according to the number, shape, and size of the components in actual implementation. The type, quantity, and ratio of each component in actual implementation can be arbitrarily changed, and the component layout type may also be more complex.

[0020] In the following description, numerous specific details are set forth to provide a more thorough explanation of embodiments of the present invention. However, it will be apparent to those skilled in the art that embodiments of the present invention may be practiced without these specific details. In other instances, well-known structures and devices are shown in block diagram form rather than in detail to avoid obscuring embodiments of the present invention.

[0021] Please refer to Figures 1-3 , which shows the real-time monitoring system and method for nursing pressure based on multimodal data of the present invention. The real-time monitoring system for nursing pressure based on multimodal data of the present invention includes a data acquisition module, an intelligent analysis module, an analysis module, and a real-time feedback module. The data acquisition module synchronously acquires the physiological data, behavioral activity data, and environmental context data of the nursing staff and forms a nursing signal; the data acquisition module synchronously acquires the physiological data, behavioral activity data, and environmental context data of the nursing staff and forms a nursing signal; the analysis module receives the nursing signal from the data acquisition module and receives the dynamic context information from the intelligent analysis module, and through a context awareness and stress event association model based on the nursing workflow, dynamically associates and fuses the physiological data, behavioral activity data, and dynamic context information to generate a stress state signal including a stress event assessment; the stress event association model establishes a spatio-temporal association map centered on the nursing workflow stage, focuses on monitoring the correlation between the hand tremor behavior data and the heart rate variability physiological data during the drug administration operation period, and strengthens the analysis of the matching degree between the gait rapidity behavior data and the environmental passage width during the patient transfer stage; the real-time feedback module receives the stress state signal including the stress event assessment from the analysis module and generates and outputs a real-time feedback signal based on the stress state signal.

[0022] As Figure 1As shown, the real-time monitoring system for nursing stress based on multi-modal data includes a cooperative operation mechanism of a data acquisition module, an intelligent analysis module, an analysis module, and a real-time feedback module. The data acquisition module realizes the synchronous capture of three types of data through embedded hardware. The flexible fabric electrode embedded in the nursing staff ID card collects single-lead electrocardiogram signals at a sampling rate of 256 Hz. The electrode uses a silver chloride coating and an elastic fiber substrate to ensure that the skin contact impedance is stable below 10 kΩ. The electrocardiogram signals are processed by an instrumentation amplifier to obtain waveform data with a bandwidth of 0.5 to 100 Hz. The nine-axis inertial measurement unit integrated on the side wall of the ID card collects raw data of three-axis acceleration, three-axis angular velocity, and three-axis magnetic force at a frequency of 100 Hz. The torso pitch angle and roll angle are calculated through a quaternion fusion algorithm, and the operation stability index is calculated by combining the hand movement covariance matrix. The environmental data acquisition is realized through a distributed microelectromechanical system microphone array, whose frequency response range covers 100 Hz to 10 kHz, and the sound pressure level measurement accuracy reaches ±2 dB. At the same time, the hospital information system interface parses the patient admission update and doctor's order execution message flow in real time through the Health Level 7 (HL7) protocol, and extracts key fields including the nursing task type code, the patient early warning score value, and the ward electronic fence coordinates. The above physiological data, behavioral activity data, and environmental context data are wirelessly transmitted to the edge network node through a low-power wide area network, and are synchronously aligned at the millisecond level by the hardware timestamp engine based on the Precision Time Protocol (PTP). Finally, they are encapsulated into a structured nursing signal data packet, which contains a unified time scale reference, 256 electrocardiogram waveform segments, a 15-dimensional behavioral feature vector, and a context tuple triple data entity.

[0023] Further, the hardware architecture of the data acquisition module is physically implemented based on a wearable sensing unit and an environmental perception unit. The wearable sensing unit adopts a stacked design of three-layer printed circuit boards with a size of forty millimeters by sixty millimeters. The physiological sensing layer integrates an electrocardiogram signal conditioning front-end chip, providing input impedance greater than one hundred gigaohms and a common-mode rejection ratio of one hundred and ten decibels. It cooperates with a constant-voltage excitation circuit of 0.5 volts to achieve accurate measurement of skin conductance in the range of 1 to 100 microsiemens. The behavior sensing layer is equipped with a nine-axis motion processing chip to output raw sensing data through an integrated circuit bus. It uses the gradient descent method to fuse accelerometer and gyroscope data to generate attitude quaternions, and calculates the hand operation tremor index within a 30-second window in real time, which is defined as the arithmetic mean of the standard deviations of the three-axis angular velocities. At the same time, the mobile path efficiency parameter is derived as the ratio of the actual displacement distance to the path integral length. The environmental perception unit consists of four microphone nodes deployed on the ward ceiling, which separates human voices and environmental noise within a radius of three meters through a beamforming algorithm. The acoustic feature extraction includes the root mean square value of the sound pressure level and the zero-crossing rate statistic within a 500-millisecond time window. The hospital information system interface module parses the patient management event fields in the medical information exchange standard message, extracts the ward classification code from the patient admission message, obtains the early warning score value from the patient status update message, and parses the current nursing task status code from the doctor's order execution message. Data synchronization is achieved using the Precision Time Protocol to control the master-slave clock deviation within one microsecond. The finally generated nursing signal data packet is encapsulated in the Protocol Buffer format, including a 64-bit timestamp field, an array of repeated floating-point electrocardiogram waveforms, a nested behavior data structure, and a context tuple container. The behavior data structure stores the three-dimensional acceleration peak value, the pitch angle change range, and the hand stability index. The context tuple container contains an enumeration value of the task type, a regional risk level flag associated with the nursing signal, and a floating-point patient critical score.

[0024] Specifically, the intelligent analysis module executes the dynamic context information generation process at the edge computing node. First, it performs semantic parsing on the medical text stream input by the hospital information system, using a biomedical pre-trained language model to perform word segmentation and entity recognition operations on nursing records. For example, it extracts entity elements such as the action subject insulin, the administration route subcutaneous injection, the target of action bed three, and the dose parameter ten units from the sentence "Insulin subcutaneous injection of ten units was performed on bed three at 15:02". At the same time, the ultra-wideband positioning system provides the three-dimensional spatial coordinates of the nursing staff at a refresh rate of 30 Hz, and determines the current area attribute by matching the spatial topological relationship with the ward electronic fence database. When the coordinates fall into the polygon geographical fence of the chemotherapy drug preparation room, the high-risk operation flag is activated. The patient's critical state assessment is calculated through a dynamic weight formula, which integrates 30% of the early warning score weight, 50% of the ventilator dependence flag weight, and 20% of the pain score weight to output the critical index. The context information generation process introduces a spatio-temporal alignment engine to perform associative mapping of the parsed nursing task entities and the positioning trajectory along the time axis. For example, if it is determined that the nursing staff is located at the coordinates of bed three at a certain moment and there is a record of an insulin injection task in the same time period, it is recognized as a medication administration operation context. The finally output dynamic context information is a structured vector updated every 200 milliseconds, and this vector contains six-dimensional data entities: the task type field stores enumeration codes such as medication administration operations or patient transfer, the area risk level field associated with the nursing signal records classification indexes such as general ward or isolation area, the patient critical index field saves floating-point values in the range of zero to ten, the task duration field counts the number of seconds of the current continuous operation, the operation constraint field marks the equipment status such as sterile gloves or protective masks, and the team collaboration field stores the count of the number of other nursing staff within five meters. This vector is directly transmitted to the analysis module through the zero-copy shared memory mechanism for real-time processing.

[0025] In an embodiment of the present invention, the pressure event correlation model in the analysis module achieves accurate discrimination by constructing a spatio-temporal correlation map. The map establishes a multi-dimensional analysis space with the nursing workflow stage as the spatio-temporal coordinate axis. A dynamic coupling model of hand movement trajectory and physiological indicators is established during the medication administration operation stage. The original data of the three-axis angular velocity of the wrist is captured by a nine-axis inertial measurement unit, and the tremor power spectral density eigenvalue in the 3-8 Hz frequency band is extracted through four-layer wavelet decomposition. The standard deviation of the RR interval of the electrocardiogram signal is synchronously analyzed as a characterization of heart rate variability. The dynamic correlation degree between tremor energy and the low-frequency band energy of heart rate variability is calculated using the sliding window Pearson correlation coefficient. When the correlation coefficient exceeds 0.7 within five consecutive sampling windows (window width: 8 seconds) and the standard deviation of heart rate variability is lower than 30% of the baseline value, an operation anxiety event marker is triggered; during the patient transfer stage, environmental constraints and behavior pattern analysis are integrated. The direction angle of the displacement vector is calculated based on the position update data of the ultra-wideband positioning system at 30 times per second, and the current corridor width parameter is obtained in combination with the building information model database. The gait fitness function is defined as the non-linear mapping relationship between the coefficient of variation and the corridor width. When the function output value exceeds the dynamic threshold and lasts for more than 10 seconds, a path compression warning signal is generated; in the intensive care situation, a multi-modal anomaly detection protocol is initiated. The skin conductance signal is subjected to first-order differential processing to capture the mutation point. When the differential value exceeds three times the standard deviation and the respiratory rate is simultaneously detected to be greater than 25 times per minute, combined with the condition that the patient's early warning score is greater than 7 points, a first-level pressure event marker is activated; during the cross-ward task processing stage, a behavior pattern portrait is constructed. The ratio of the turning-back distance to the straight-line distance within five minutes is calculated through the path integral algorithm. When the ratio is greater than 2.5 and the synchronous electrocardiogram signal shows a continuous heart rate exceeding 110 beats per minute, a second-level pressure event marker is triggered. The finally generated pressure state signal includes a four-level pressure level code, six types of event cause classification labels, and a confidence score ranging from 0.0 to 1.0.

[0026] As Figure 2As shown, the dynamic association and fusion process uses a context-adaptive weighting mechanism to achieve precise analysis, automatically adjusting the weight coefficients of multi-modal data for different workflow stages. During the night shift, the weight of the environmental light sensor data is increased to 1.4 times the base value, and a compensation coefficient of 0.3 is added to the correction parameter of the behavior data due to the restraint feeling of protective equipment during operations in the infectious disease isolation area. When the system identifies a chemotherapy drug preparation task (task code CHEMO), it activates a high-sensitivity analysis mode to expand the hand micro-tremor monitoring frequency band to 2 - 12 Hz and shorten the sampling time window to 0.5 seconds to capture transient abnormalities. When the communication context with the patient's family times out, a voice emotion enhancement protocol is triggered. The dominant voice stream within a 3-meter radius is extracted through the beamforming technology of the distributed microphone array, and the joint analysis technology of Mel Frequency Cepstral Coefficients and fundamental pitch period is used to identify features such as increased speech rate and elevated pitch, generating an emotional tension index from 0 to 10 and integrating it as a new dimension into the environmental context data. The cross-module data verification mechanism ensures the reliability of the analysis. When the physiological data shows a sudden increase in skin conductance but no abnormal movement is detected in the behavior data, the environmental sensor is automatically called to detect whether the environmental temperature exceeds 28 degrees Celsius or the sudden noise is greater than 70 decibels. If the association model outputs a warning of the risk of medication error, the hand operation trajectory data for the previous two minutes is forcibly traced back, and the inverse kinematic solution is used to verify whether the grasping angle of the medicine bottle deviates from the standard value by more than 15 degrees. The finally formed dynamic weight matrix is updated every 200 milliseconds, containing a four-tuple parameter of physiological data weight, behavior data weight, environmental data weight, and the new dimension.

[0027] Furthermore, the real-time feedback module is configured with multiple levels of output channels to achieve precise intervention. When the pressure status signal contains a first-level event marker and the confidence level exceeds 0.8, it drives the tactile belt worn by the caregiver to generate an intermittent vibration mode of 5 Hz (vibrating for 3 seconds and pausing for 2 seconds), and simultaneously projects a deep breathing guidance animation within a 30-degree field of view of the augmented reality glasses. This animation includes a 3D model of lung expansion and an 8-second breathing rhythm light band; when the signal contains a second-level event marker and the confidence level is greater than 0.6, it switches the color temperature of the ambient light in the nurse station to 4700 Kelvin warm white through a Digital Addressable Lighting Interface, and the gradual adjustment of illuminance takes no more than 3 seconds. At the same time, it sends an event summary message encrypted by the Advanced Encryption Standard algorithm to the handheld terminal of the nursing team leader. The message includes the pressure event type code, the grid coordinates of the occurrence location, and the duration; all feedback signals are attached with visual encoding of the event cause. For events of drug check errors, a yellow pulse border warning is triggered in the augmented reality glasses, lasting for 6 seconds and accompanied by the flashing of a medicine bottle icon; for events of sudden changes in the patient's condition, a red full-screen fade-out warning is triggered, and an electrocardiogram symbol fluctuation animation is superimposed synchronously; for the event of continuous standing operation exceeding the limit (more than 45 minutes), the system automatically retrieves the position of the nearest buffer seat and generates a 3D navigation path, and indicates the turning direction through the intensity difference of the vibration motors on the left and right sides of the tactile belt; the multi-channel feedback protocol sets a priority arbitration mechanism. When a first-level and a second-level event occur simultaneously, tactile feedback and breathing guidance are preferentially executed, and the ambient light adjustment is delayed by 5 seconds to start. The manager alert message is sent immediately under the condition of available bandwidth.

[0028] As Figure 3 shown, it is a real-time monitoring method for nursing pressure based on multi-modal data of the present invention. S1: Synchronously collect the physiological data, behavioral activity data, and environmental context data of the caregiver and form a nursing signal; S2: Receive the nursing signal from the data collection module and extract the dynamic context information related to the current working state of the caregiver therefrom; S3: Receive the nursing signal from the data collection module and receive the dynamic context information from the intelligent analysis module. Through a context-aware and pressure event association model based on the nursing workflow, dynamically associate and fuse the physiological data, behavioral activity data, and the dynamic context information to generate a pressure status signal including a pressure event assessment; S4: Receive the pressure status signal including a pressure event assessment from the analysis module and generate and output a real-time feedback signal based on the pressure status signal.

[0029] Furthermore, the analysis module is built with a personalized calibration engine to achieve dynamic optimization. During the low-pressure period of the morning shift handover for nursing staff (ten minutes after the system automatically recognizes the sign-in), the baseline parameter collection protocol is initiated, continuously monitoring the electrocardiogram signal in the resting state for three minutes, calculating the average heart rate as the individualized baseline value and storing the standard deviation range. Meanwhile, the trunk angular velocity data in the standing posture is collected through the inertial measurement unit to establish a gait stability baseline matrix. When abnormal fluctuations in heart rate occur during the daytime workflow (exceeding 20% of the baseline value for 30 consecutive seconds), the morning baseline data of this person is called to calculate the drift compensation amount. The compensation formula is: corrected heart rate = real-time heart rate × (morning average heart rate / group average heart rate). The shift mode adaptation mechanism dynamically adjusts the evaluation threshold. The heart rate variability threshold during the night shift (from 10 pm to 6 am) is increased by 0.2 times, and the gait disorder determination standard is relaxed by 15%. The continuous learning module updates the model parameters through the analysis of the stress response mode. When the system detects that a certain nursing staff continuously shows hand tremors but no abnormal physiological indicators during the handling of chemotherapy drugs, the person is automatically marked as behaviorally sensitive and the splitting gain of the tremor feature in the decision tree is reduced. A personalized stress-sensitive factor report is generated weekly, including three-dimensional data on physiological index offset statistics, abnormal behavior feature frequencies, and environmental factor response intensities. This report drives the associated model for incremental parameter optimization, using the stochastic gradient descent method to update the classifier weight matrix. The optimization process is completed at the edge computing node and takes no more than 15 minutes, ensuring that the system adapts to the stress representation modes of different nursing staff.

[0030] Specifically, the dynamic context information expands the dimensions of team collaboration metrics. It realizes the perception of collaboration status through a low-power Bluetooth beacon network. A Bluetooth 5.0 chipset is built into the nursing staff's work badges, which broadcast device identifiers containing role codes at one-second intervals. When the received signal strength indication value of other work badges within a five-meter radius is greater than -70 dBm, the number of collaborating personnel is counted; the voice activity detection subsystem is implemented through a distributed microphone array, and a dual-threshold endpoint detection algorithm is used to identify valid voice segments. The defined team response delay time is the time difference between the first frame of voice triggered by the call bell and the first frame of voice answered by the nearest nursing staff; in the context of transporting critically ill patients, a collaboration stress analysis model is activated. When the transport task is activated (context code PT_TRANSFER), the team personnel density parameter (number of nursing staff per unit area) and the response time at key nodes are monitored in real time: If it is detected through Bluetooth ranging that there are less than two collaborating personnel within three meters and the response to the ventilator tube disconnection alarm times out for 20 seconds, the stress weight coefficient of this context is automatically increased by 50%; a team collaboration missing mark (label TEAM_GAP) is attached to the stress status signal, and at the same time, a reinforcement request message is generated and sent to the nursing scheduling system. The message includes the required professional qualifications (such as respiratory therapists), the expected arrival time window (within 60 seconds), and the optimal path recommendation (based on the indoor navigation database); the long-term collaboration effectiveness evaluation module constructs a team stress profile, counting the average number of weekly collaboration missing events for each nursing staff. When the value of a certain person exceeds twice the standard deviation of the team average, a team support sensitivity factor is added to the personalized calibration engine, and the decision tree splitting threshold is lowered by 15%; all collaboration data is encapsulated into a context information sub-structure, including three dynamic parameters: real-time collaborating personnel count, average response delay in seconds, and collaboration requirement satisfaction rate, which are synchronously updated with the main context vector every 200 milliseconds.

[0031] The real-time nursing stress monitoring system and method based on multi-modal data of the present invention synchronously collect physiological data, behavioral data, and environmental data during nursing operations through multi-source sensors; the intelligent analysis module parses the hospital information system and dynamically generates three-dimensional context information of task type, regional risk, and patient criticality index; the analysis module establishes the association rules between the nursing stage and stress characterization: monitors the correlation between hand tremors and heart rate variability during the drug preparation stage, and analyzes the matching degree between gait and environmental channels during the patient transfer stage; when the context is dynamically associated with abnormal physiological behavior, a graded stress event mark is triggered, and precise intervention is implemented through the tactile feedback device and the manager terminal.

[0032] Therefore, through the real-time nursing stress monitoring system and method based on multi-modal data of the present invention, the problem that traditional wearable devices only collect physiological and behavioral data and cannot be associated with specific nursing contexts such as medication administration operations and patient first aid is solved.

[0033] The above embodiments are only illustrative of the principles and effects of the present invention and are not intended to limit the present invention. Any person familiar with this technology can modify or change the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or changes made by those of ordinary skill in the art within the spirit and technical idea disclosed by the present invention should still be covered by the claims of the present invention.

Claims

1. A real-time monitoring system for nursing stress based on multimodal data, characterized in that, Including: A data acquisition module that synchronously acquires the physiological data, behavioral activity data, and environmental context data of nursing staff and forms a nursing signal; An intelligent analysis module that receives the nursing signal from the data acquisition module and extracts dynamic context information related to the current working state of the nursing staff therefrom; An analysis module that receives the nursing signal from the data acquisition module and receives the dynamic context information from the intelligent analysis module. Through a context awareness and stress event correlation model based on the nursing workflow, it dynamically correlates and fuses the physiological data, behavioral activity data with the dynamic context information to generate a stress state signal including a stress event assessment. The stress event correlation model establishes a spatio-temporal correlation map centered on the nursing workflow stage, focuses on monitoring the correlation between hand tremor behavior data and heart rate variability physiological data during the drug administration operation period, and strengthens the analysis of the matching degree between the gait rapidity behavior data and the width of the environmental passage during the patient transfer stage; A real-time feedback module that receives the stress state signal including a stress event assessment from the analysis module and generates and outputs a real-time feedback signal based on the stress state signal.

2. The real-time monitoring system for nursing pressure based on multi-modal data according to claim 1, characterized in that The data acquisition module includes a wearable sensing unit and an environmental perception unit integrated in the nursing staff's work ID card; the wearable sensing unit continuously acquires electrocardiogram signals and skin electroresponse signals through flexible bioelectrodes to form physiological data, and captures limb movement trajectories and posture changes through an inertial measurement unit to form behavioral activity data; the environmental perception unit acquires the sound pressure characteristics of the working area through a distributed noise sensor to form environmental context data, and receives the nursing task list and patient critical value alarm records pushed by the hospital information system through a near-field communication interface in real time; the wearable sensing unit and the environmental perception unit fuse the three types of data into a nursing signal through a timestamp synchronization mechanism.

3. The real-time monitoring system for nursing pressure based on multi-modal data according to claim 1, characterized in that The context awareness operations performed by the intelligent analysis module include: parsing the hospital information system data stream from the nursing signal to identify task types such as drug preparation, critical care, and family member communication currently performed by the nursing staff; combining the positioning beacon coordinates and ward electronic fence information to determine whether it is in a stress classification area such as a surgical preparation area, isolation ward, or nurse station; obtaining the disease severity index and special nursing requirement marks of the currently responsible case in real time by associating with the patient's electronic medical record; finally integrating the task type, the regional risk level associated with the nursing signal, and the case severity index into a dynamic context information vector.

4. The real-time monitoring system for nursing stress based on multimodal data according to claim 1, wherein The stress event association model in the analysis module operates according to the following process: establishing a spatiotemporal association map with the nursing workflow stage as the axis, focusing on monitoring the correlation between hand tremor behavior data and heart rate variability physiological data during the medication operation period, and strengthening the analysis of the matching degree between gait urgency behavior data and environmental channel width during the patient transfer stage; when a sudden increase in skin conductance and abnormal respiratory rate occur in the critically ill patient monitoring situation, a first-level stress event marker is triggered; when a continuous high heart rate and irregular activity path are detected in the cross-ward continuous processing call bell situation, a second-level stress event marker is triggered; and finally a stress state signal carrying a stress level label and event cause classification is generated.

5. The real-time monitoring system for nursing pressure based on multi-modal data according to claim 1, wherein, The dynamic association fusion process introduces a context weighting mechanism: the weight coefficient of the insufficient ambient light factor is automatically increased in the night shift workflow stage, and the correction parameter of the restraint of protective equipment on the behavioral data is increased in the infectious disease isolation area workflow stage; When the system recognizes that nursing staff are continuously handling high-risk drug preparation tasks, it activates a high-sensitivity analysis mode for the subtle hand tremor behavior data collected during the same period; when the communication situation with the patient's family continues to be extremely high, the voice emotion feature extraction algorithm is activated to enhance the environmental context data analysis dimension.

6. The real-time monitoring system for nursing pressure based on multi-modal data according to claim 1, characterized in that The real-time feedback module includes a multi-level output channel: when the pressure state signal includes a first-level pressure event marker, the tactile belt worn by the caregiver is driven to generate a low-frequency vibration prompt, and a deep breathing guidance animation is projected onto the caregiver's smart glasses; When the signal contains a secondary stress event marker, the predetermined soothing audio is played through the nurse station broadcasting system, and an encrypted warning text is sent to the nursing team leader's terminal; all feedback signals are accompanied by stress event cause codes, among which events related to drug verification errors trigger yellow visual warnings, and events related to sudden changes in patients' conditions trigger red visual warnings.

7. The real-time monitoring system for nursing pressure based on multimodal data according to claim 1, characterized in that The analysis module has a built-in personalized calibration engine: it automatically collects baseline physiological parameters during the low-stress period of morning shift handover for nursing staff, and establishes an individualized resting heart rate and gait stability baseline; When an abnormal ECG signal appears in the midday workflow, the baseline data of the person is called to calculate the drift compensation; by continuously learning the stress response mode in the shift cycle, the evaluation threshold coefficients of the night shift and the day shift are dynamically adjusted; A personalized stress sensitivity factor report is generated weekly to optimize the correlation model parameters.

8. The real-time monitoring system for nursing pressure based on multi-modal data according to claim 1, characterized in that The stress event assessment includes a cause tracing function: when the system detects that an abnormal ECG signal and an abnormal positioning data occur simultaneously, the workflow scenario data of the previous ten minutes is automatically traced back. If it is found that the person handles emergency calls from three wards in succession and passes through a noisy corridor environment, a stress event report with dual attributions of environmental interference and task overload is generated; when the behavioral data shows repeated returns to the pharmacy, the failed drug collection record of the drug management system is associated to form a process-blocking stress event analysis chain.

9. The real-time monitoring system for nursing pressure based on multimodal data according to claim 1, characterized in that, The dynamic context information includes team collaboration metrics: obtaining the number of other caregivers within five meters through Bluetooth beacon ranging, and calculating the team response delay time in combination with voice activity detection; when the team personnel density in the critical patient transfer scenario is lower than the threshold and the call response times out, automatically increase the stress weight coefficient of this scenario; attach a team collaboration missing marker to the stress status signal and suggest a reinforcement request.

10. The real-time monitoring method for nursing pressure using the real-time monitoring system for nursing pressure based on multimodal data according to any one of claims 1-9, characterized in that, Including: S1: Synchronously collect the physiological data, behavioral activity data and environmental context data of the caregiver and form a care signal; S2: Receive the care signal from the data collection module and extract the dynamic context information related to the current working state of the caregiver from it; S3: Receive the care signal from the data collection module and receive the dynamic context information from the intelligent analysis module. Through the context awareness and stress event association model based on the care workflow, dynamically associate and fuse the physiological data, behavioral activity data with the dynamic context information to generate a stress status signal including stress event assessment; S4: Receive the stress status signal including stress event assessment from the analysis module, and generate and output a real-time feedback signal based on the stress status signal.

Images