Auxiliary selection system for nursing schemes for orthopedic patients

Abstract

The invention relates to the technical field of nursing decisions, in particular to an auxiliary selection system for nursing schemes of orthopedic patients, which comprises the following steps: in a rehabilitation training process, acquiring a patient moving track in real time, performing deviation analysis on the patient moving track and an ideal path, and establishing a deviation risk fragment set by means of a space coordinate sequence and a Hurst index calculation mode; recognition and positioning of the deviation trend of rehabilitation actions are enhanced, and the precision of walking control stability evaluation is enhanced. The heart rate signal is mapped to the multi-dimensional density space, and the local abnormal interval in the heart rate change is accurately calibrated through the local abnormal factor operation mode under the sliding window, so that the tiny physiological change can be identified. According to the method, the heart rate abnormal interval is subjected to duration time classification processing, a heart rate fluctuation typing system is constructed, transient physiological fluctuation and a continuous stress state can be effectively distinguished, and the recognition degree of pain inducements is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of nursing decision-making, and particularly to a system for assisting in the selection of nursing plans for orthopedic patients. Background Art

[0002] A system for assisting in the selection of nursing plans for orthopedic patients refers to a system technical solution that provides standardized nursing path recommendations and support for combinations of nursing measures based on information such as the disease characteristics, treatment methods, rehabilitation cycles, and individual health conditions of orthopedic patients. Specifically, based on the patient's diagnosis and treatment records, assessment scale data, and vital sign parameters, by setting a list of standard nursing actions and precautions corresponding to the nursing path, it assists nursing staff in making plan selections.

[0003] The existing technology relies on standardized plan matching in the recommendation of nursing paths for orthopedic patients, mainly setting static paths around diagnosis and treatment records and assessment scales, resulting in insufficient ability to identify subtle changes during the recovery process. The identification of mobile path control deviations relies on subjective observation or single-point monitoring methods, making it difficult to depict the trend of continuous behavior changes, and thus accurately identifying potential deviation risk periods. The assessment of heart rate and blood pressure stays at the basic parameter levels such as mean and extreme values, ignoring their dynamic characteristics in the local time domain, resulting in a lag in the identification of risk states such as pain, stress, or hypoperfusion. During the night nursing process, there is a lack of continuous assessment means for venous return, often triggering passive responses only after obvious swelling or symptoms of blood circulation disorders appear in the patient, missing the opportunity for early intervention. In addition, nursing path suggestions often adopt a unified process setting and fail to make precise adjustments according to the patient's physiological fluctuation state and real-time performance. For example, when the heart rate fluctuates abnormally but there are no obvious abnormal behavior manifestations, it is often ignored, delaying pain intervention; when the blood pressure continuously drops at night but there is no intervention in the nursing record, it is not recognized by the system either, increasing nursing blind spots and risk exposure points, and affecting the safety, comfort, and rehabilitation quality of the patient during the recovery process. Summary of the Invention

[0004] The purpose of the present invention is to solve the deficiencies existing in the prior art and propose a system for assisting in the selection of nursing plans for orthopedic patients.

[0005] To achieve the above purpose, the present invention adopts the following technical solutions: The system for assisting in the selection of nursing plans for orthopedic patients includes: A gait deviation monitoring module collects the movement path of the patient during the daytime rehabilitation training, compares it with the ideal movement path, extracts the deviation distance between the paths according to the comparison result, forms a deviation time series, and screens the risk time periods by calculating the Hurst index of the amplitude change of the deviation distance in the deviation time series to form a set of deviation risk segments; The heart rate abnormality recognition module collects the heart rate of the patient during the daytime rehabilitation training process and maps it to a multi-dimensional density space, and marks the continuous local abnormal time points of the heart rate in the multi-dimensional density space as the heart rate local abnormal interval; The heart rate fluctuation classification module divides the interval types according to the duration of the heart rate local abnormal interval in each time period, and establishes a heart rate fluctuation classification set; The blood pressure trend analysis module collects the continuous systolic blood pressure pulse waveform signals of the patient at night after surgery, analyzes the rhythm changes of the systolic blood pressure pulse within a specified period, and generates a blood pressure trend analysis result; The body position intervention discrimination module, according to the blood pressure trend analysis result, links the operation data in the same time period in the nursing record, marks the target cycle segment without nursing intervention, and establishes a record of unresponsive venous return abnormality; The rehabilitation guidance module recommends corresponding nursing plans to the patient according to the offset risk segment set, the heart rate fluctuation classification set, and the record of unresponsive venous return abnormality, and obtains a multi-dimensional nursing path recommendation result.

[0006] As a further solution of the present invention, the obtaining steps of the offset risk segment set are specifically as follows: Collect the movement path of the patient during the daytime rehabilitation training through the wearable device on the patient, convert the longitude and latitude coordinate points recorded per second in the movement path into an equally spaced continuous space coordinate sequence, preset the ideal movement path of the rehabilitation training plan according to the current postoperative recovery period of the patient, and calculate and record the deviation distance between the coordinate points of the continuous space coordinate sequence at the same time and the corresponding position coordinate points of the ideal movement path; Statistically form an offset time series of the deviation distances at each time, and adopt the Hurst exponent algorithm. By analyzing the power-law relationship between the change amplitude of the deviation distances at different times in the offset time series and the length of the entire offset time series, calculate the Hurst exponent of the change amplitude of the deviation distances in the offset time series; Compare the Hurst exponent with a preset reference exponent threshold to evaluate the stability of the direction control during the patient's walking, and mark the corresponding risk time periods in the offset time series that are lower than the reference exponent threshold to form an offset risk segment set.

[0007] As a further solution of the present invention, the obtaining steps of the heart rate local abnormal interval are specifically as follows: Collect the heart rate of the patient per minute during the daytime rehabilitation training through the wearable device on the patient, divide it according to a fixed sliding window, and extract the continuous heart rate within each sliding window to form a heart rate information set; Select each time point in the collected heart rate information in sequence, as well as the heart rate information in the adjacent areas before and after the target time point. By mapping all the heart rate information to a multi-dimensional density space and using the local outlier factor algorithm, calculate the local density of each time point in the corresponding adjacent area within the multi-dimensional density space; Compare the local density with a preset density threshold, screen out the local abnormal time points with local density lower than the density threshold, and mark the continuous local abnormal time points as the heart rate local abnormal intervals.

[0008] As a further solution of the present invention, the steps for obtaining the heart rate fluctuation classification set are specifically as follows: Statistically analyze the duration of each heart rate local abnormal interval in each time period, compare the statistically analyzed durations with a set time threshold respectively, mark the heart rate local abnormal intervals with durations higher than the time threshold as continuous fluctuation sections, and mark the heart rate local abnormal intervals with durations lower than the time threshold as instantaneous fluctuation sections; Classify and mark according to the section types corresponding to the continuous fluctuation sections and the instantaneous fluctuation sections, and establish a heart rate fluctuation classification set.

[0009] As a further solution of the present invention, the steps for obtaining the blood pressure trend analysis result are specifically as follows: Collect the continuous systolic blood pressure pulse waveform signals of the patient at night after the operation through a blood pressure monitoring device, divide them into multiple equal-length time signal windows, and for each segment of the systolic blood pressure pulse waveform in the time signal window, construct a continuous signal curve with the corresponding time as the horizontal axis and the systolic blood pressure pulse as the vertical axis; Perform Hilbert transform on the continuous signal curve to analyze the instantaneous frequency and phase trajectory within each period corresponding to each systolic blood pressure pulse, analyze the rhythm change within each period corresponding to the systolic blood pressure pulse according to the instantaneous frequency and phase trajectory, identify and mark the period segments with a rhythm decline and no rebound trend, and generate the blood pressure trend analysis result.

[0010] As a further solution of the present invention, the steps for obtaining the record of unresponsive venous return abnormality are specifically as follows: According to the blood pressure trend analysis result, infer whether there is slowdown of lower limb venous return in the patient during the target period segment. Among them, the period segment with a rhythm decline and no rebound trend corresponds to a stable slowdown process with continuously weakened hemodynamics and not interrupted by body position changes, indicating a decrease in venous return efficiency under a long-term static state, and obtain the lower limb venous return analysis result; Based on the lower limb venous return analysis result, identify whether there is a nursing intervention such as raising the lower limb and body position guidance by linking the operation data in the same time period in the nursing record. If there is no nursing intervention, mark the target period segment again to establish a record of unresponsive venous return abnormality.

[0011] As a further solution of the present invention, the steps of obtaining the multi-dimensional maintenance path recommendation result are specifically as follows: According to the set of deviation risk fragments, recommendations are made on the setting of daytime movement paths and guidance methods, including leaning against walls, setting boundary aids, and arranging accompanying personnel to guide movement at a specified distance; According to the heart rate fluctuation classification set, the instantaneous fluctuation segment is regarded as a non-sustained reaction, and no nursing plan adjustment is made. The continuous fluctuation segment is regarded as a painful state, and it can be recommended to increase the cold compress intervention reminder and nursing rounds. Based on the record of abnormal venous return without response, it is recommended to increase the time point setting for raising the lower limbs or guiding turning over within the calibrated night target cycle segment to obtain multi-maintenance care path recommendation results, which are used to provide decision support for auxiliary selection of nursing plans for orthopedic patients.

[0012] Compared with the prior art, the advantages and positive effects of the present invention are: In the present invention, during the rehabilitation training process, the patient's movement trajectory is collected in real time and the deviation analysis is performed with the ideal path. The deviation risk segment set is established with the help of the spatial coordinate sequence and the Hurst index calculation method, so as to strengthen the identification and positioning of the deviation trend of the rehabilitation action and enhance the accuracy of the walking control stability assessment. The heart rate signal is mapped to the multidimensional density space, and the local abnormality interval in the heart rate change is accurately calibrated by the local abnormality factor calculation method under the sliding window, which is conducive to identifying small physiological changes. The duration of the abnormal heart rate interval is classified and the heart rate fluctuation classification system is constructed, which can effectively distinguish between transient physiological fluctuations and continuous stress states, and improve the recognition of pain inducements. In nighttime blood pressure monitoring, based on the continuous curve construction of the systolic pulse waveform, the instantaneous frequency and phase trajectory are obtained by Hilbert transform, so as to realize the dynamic trend capture of the blood pressure rhythm and the venous return hysteresis process. Combined with the nursing record period data, the rhythm decay cycle that is not interrupted by the operation is determined to realize the automatic identification and marking of the missing links of nursing intervention. By integrating the risk shift segment, heart rate fluctuation type and venous return response status, the nursing recommendations are matched according to the recovery stage, such as guidance method, cold compress intervention frequency and body position adjustment settings, to enhance the personalized adaptability and decision rationality of the path recommendation. In summary, through multi-signal linkage analysis and dynamic recognition mechanism, the response speed and execution accuracy of rehabilitation program selection can be effectively improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 It is a system flow chart of the present invention. DETAILED DESCRIPTION

[0014] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0015] Please refer to Figure 1 , the nursing plan assisted selection system for orthopedic patients includes: The gait deviation monitoring module collects the movement path of the patient during the daytime rehabilitation training, compares it with the ideal movement path, extracts the deviation distance between the paths according to the comparison result, forms a deviation time series, and screens the risk time period by calculating the Hurst exponent of the amplitude change of the deviation distance in the deviation time series to form a deviation risk segment set; The heart rate abnormality recognition module collects the heart rate of the patient during the daytime rehabilitation training and maps it to a multi-dimensional density space, and marks the continuous local abnormal time points of the heart rate in the multi-dimensional density space as the heart rate local abnormal interval; The heart rate fluctuation classification module divides the interval types according to the duration of the heart rate local abnormal interval in each time period to establish a heart rate fluctuation classification set; The blood pressure trend analysis module collects the continuous systolic blood pressure pulse waveform signals of the patient at night after surgery, analyzes the rhythm change of the systolic blood pressure pulse within a specified period, and generates a blood pressure trend analysis result; The body position intervention discrimination module, according to the blood pressure trend analysis result, links the operation data in the same time period in the nursing record, marks the target period segment without nursing intervention, and establishes a record of unresponsive venous return abnormality; The rehabilitation guidance module recommends the corresponding nursing plan to the patient according to the deviation risk segment set, the heart rate fluctuation classification set and the record of unresponsive venous return abnormality to obtain a multi-dimensional nursing path recommendation result.

[0016] The specific steps for obtaining the deviation risk segment set are as follows: Collect the movement path of the patient during the daytime rehabilitation training through the wearable device on the patient, convert the longitude and latitude coordinate points recorded per second in the movement path into an equally spaced continuous space coordinate sequence, preset the ideal movement path of the rehabilitation training plan according to the current postoperative recovery period of the patient, and calculate and record the deviation distance between the coordinate points of the continuous space coordinate sequence at the same time relative to the corresponding position coordinate points of the ideal movement path; Collect the movement path of the patient during the daytime rehabilitation training through the wearable device on the patient. The device includes an intelligent wearable terminal with high-precision GNSS positioning ability, which can sample and record the longitude and latitude coordinate points per second during the patient's movement in real time at a frequency of 1Hz. The collected coordinate data is expressed in the WGS-84 coordinate system, and the entire process is uniformly processed using ArcGIS Pro software, including batch converting the original longitude and latitude coordinates into XY plane coordinate points in the projected coordinate system, with the conversion unit being meters to ensure the unity of the spatial scale for distance calculation; subsequently, combined with the patient's postoperative recovery period, rehabilitation experts set rehabilitation training plans for different stages. Among them, the ideal movement path can be determined and formed through statistical analysis of the trajectory data of patients with the same previous surgical type and similar physical signs during this recovery period. Through the standard path layer imported in ArcGIS Pro, the continuous coordinate point sequence of the ideal movement path is marked and output according to each second time point; after exporting the actual trajectory and the ideal path trajectory with the two time points aligned, calculate the deviation distance between the coordinate points of the continuous spatial coordinate sequence at the same time relative to the corresponding position coordinate points of the ideal movement path. The formula is as follows:

[0017] Wherein, represents the deviation distance between the movement path of the patient at the th second in the continuous spatial coordinate sequence and the ideal movement path, with the unit being meters, represents the coordinate of the patient's movement path on the X-axis at the th second in the continuous spatial coordinate sequence, with the unit being meters, sourced from the extraction of projected coordinates in ArcGIS Pro, represents the coordinate of the patient's movement path on the Y-axis at the th second in the continuous spatial coordinate sequence, with the unit being meters, obtained in the same way; represents the coordinate of the patient's ideal movement path on the X-axis at the th second, with the unit being meters, extracted after converting the training trajectory defined by rehabilitation, represents the coordinate of the patient's ideal movement path on the Y-axis at the th second, with the unit being meters, is the time point index, incrementing starting from 1, representing the time series points corresponding to the sampling frequency per second, represents the movement path of the patient in the continuous spatial coordinate sequence, represents the ideal movement path.

[0018] Suppose a test and training are carried out in the postoperative rehabilitation department of a hospital. The movement data is recorded using a patient-wearing device and matched with the ideal path set by a doctor in ArcGIS Pro. Suppose in the early postoperative rehabilitation stage, the walking speed of the patient is controlled between 0.3 - 0.7 meters per second, and the path fluctuation is small. If the actual coordinates exported from ArcGIS Pro at the 10th second are meters, meters, and the coordinates corresponding to the ideal path are meters, meters, then the calculation is as follows:

[0019] This result indicates that at the 10th second, the actual walking position of the patient is offset by approximately 1.945 meters relative to the preset ideal path.

[0020] The formula obtains the offset at each time point by calculating the distance between the coordinate position of each time point in the actual movement trajectory of the patient and the coordinate position of the ideal movement path at the same time point, reflecting the degree of deviation of the patient's actual position from the ideal path at this moment. This operation process is repeatedly executed at all time points to form a numerical sequence describing the continuous offset status of the patient during training.

[0021] The deviation distances at each time are statistically counted to form an offset time series. Using the Hurst exponent algorithm, by analyzing the power-law relationship between the change amplitude of the deviation distances at different times in the offset time series and the length of the entire offset time series, the Hurst exponent of the change amplitude of the deviation distances in the offset time series is calculated; According to the acquisition condition of a sampling frequency of 1 Hz, all the deviation distances are sorted into a continuous time series in chronological order to form an offset time series ; this series is used to reflect the spatial offset change of the patient during training.

[0022] Suppose the current offset time series is (5 offset distance data): , calculate the average offset distance within the offset time series: meters.

[0023] In the calculation of the Hurst index, the degree of discreteness of each point in the offset time series only reflects static volatility. In order to reveal the fluctuation trend of the patient's gait offset in the time dimension, the offset distance series needs to be centered and the difference accumulated over time, so it is necessary to construct a centralized cumulative deviation series. The construction of the centralized cumulative deviation series is to reveal whether the offset time series has persistent or trend fluctuations based on the overall offset mean. First, the average offset distance of the entire offset time series is counted, and the offset value of each time point is subtracted from the average value to obtain the centralized offset value. Subsequently, the centralized offset values ​​of each time point are accumulated in turn to form a cumulative time series. The construction logic of this sequence is based on: if the offset trend is persistent, the cumulative value will continue to rise or fall, otherwise it will fluctuate around zero.

[0024] Construct a centralized cumulative deviation series based on the average offset distance in the offset time series : ; ; ; ; , the centralized cumulative deviation sequence is .

[0025] Calculate the range of the centered cumulative deviation series : Meters; The range refers to the difference between the maximum and minimum values ​​in the centered cumulative deviation sequence, which is usually used to measure the maximum fluctuation range of the sequence during the observation time. In the Hurst index calculation, the range represents the maximum amplitude of the cumulative change in the deviation trend of the patient during the entire training period. This value can be used to determine whether the deviation accumulates over time and shows significant unidirectional fluctuations, thereby reflecting whether the patient's gait has the ability to control trend deviations. Used in combination with the standard deviation, the range as a numerator can express the degree of trend diffusion and is an important measurement indicator for long-term dependent features.

[0026] Calculate the standard deviation of a time series : Meters. The standard deviation is a statistical indicator used to measure the degree of dispersion of a set of data relative to its mean. In the offset time series, the standard deviation reflects the fluctuation range of the offset distance per second relative to the average offset distance in the entire training. The larger the value, the wider the range of the offset distance and the more drastic the fluctuation; the smaller the value, the more stable the offset distance and the smaller the fluctuation. In the calculation of the Hurst index, the standard deviation is used as the denominator to standardize the range of the centralized cumulative deviation, so that the result is not affected by the absolute size of the offset value, so that it can focus on the trend fluctuation characteristics itself.

[0027] Recovery period weight factor is a dimensionless parameter used to adjust the sensitivity of the Hurst exponent to offset fluctuations, and its setting needs to be based on the current postoperative recovery stage of the patient. The setting process is based on the training data of patients with the same surgical procedure collected in the hospital rehabilitation database. By statistically analyzing the standard deviation range of the offset time series and the range of the cumulative deviation range within different postoperative days, and combining the objective score of the rehabilitation expert on gait control ability (MRC muscle strength rating), the patient's rehabilitation stage can be divided into the following three categories: Early postoperative stage (from the 1st to the 7th day after surgery): Assume that the standard deviation of the offset time series of patients in this stage is concentrated between 0.18 m and 0.35 m, and the centered cumulative deviation range is concentrated between 0.38 m and 0.70 m. Therefore, the weight factor is set to enhance the fluctuation response ability and make the Hurst exponent have a higher offset recognition sensitivity. Middle postoperative stage (from the 8th to the 14th day after surgery): Assume that the offset standard deviation of patients in this stage is generally between 0.10 m and 0.18 m, and the range is between 0.25 m and 0.40 m, which is in the median 50% interval of the full data statistics. The fluctuation amplitude is between the initial stage and the stable stage. Therefore, is set to moderately adjust the fluctuation characteristics. Stable postoperative stage (from the 15th day after surgery and later): Assume that the offset standard deviation drops to the interval of 0.05 m to 0.10 m, and the extreme value is concentrated between 0.12 m and 0.25 m, which is the lowest 25% fluctuation interval in the database. To avoid the Hurst exponent from overresponding to small perturbations, is set to reduce the sensitivity to micro-amplitude fluctuations and ensure that the exponent reflects the actual offset trend rather than data noise. Assume that a certain patient is on the 5th day after surgery, with a muscle strength evaluation grade of M2 and limited mobility. The doctor on-site assesses that there is a visible continuous offset trend in the patient's gait offset. The measured offset standard deviation is 0.24 m, and the cumulative deviation range is 0.52 m. This standard deviation is within the range of [0.18, 0.35] m, and the range is also within the range of [0.38, 0.70] m, both belonging to the preset early postoperative fluctuation interval. After comprehensive judgment, it is in the early postoperative stage and meets the setting conditions. According to the predefined standard, the recovery option weight factor is reasonably set to to enhance the Hurst exponent's ability to identify the offset trend in this stage.

[0028] The Hurst exponent for calculating the change in the deviation distance amplitude in the offset time series using the Hurst exponent algorithm is as follows:

[0029] where is dimensionless and is used to measure the fluctuation persistence or recurrence characteristics of the offset time series, is the logarithmic function with base 10 and is used to express the power-law relationship, is the range of the centralized cumulative deviation sequence, with the unit of meter (m). is the standard deviation of the offset time series, with the unit of meter (m). is the recovery option weight factor (dimensionless) for adjusting the Hurst exponent setting. is the length of the entire offset time series, with the unit of second (s), representing the number of data points in the series.

[0030] Substitute the above data into the Hurst exponent formula for calculation:

[0031] The results show that the Hurst exponent of the deviation distance amplitude change in the offset time series is .

[0032] Compare the Hurst exponent with the preset benchmark index threshold to evaluate the stability of the patient's direction control during walking. Mark the corresponding risk time periods in the offset time series that are lower than the benchmark index threshold to form an offset risk segment set; The benchmark index threshold of the Hurst exponent is used as a reference standard for judging whether the gait offset stability meets the standard. Its setting is based on the historical Hurst exponent data distribution in the training records of orthopedic postoperative patients in the clinical rehabilitation database, and is double-calibrated in combination with the grading evaluation results of rehabilitation experts on the gait control state.

[0033] Suppose that from multiple orthopedic postoperative patients collected from the hospital rehabilitation department, extract the training trajectory data of each patient on the 15th, 30th, and 45th days after surgery, and calculate the Hurst exponent of the corresponding offset time series, forming more than 1000 groups of valid sample data. On the day of collection, each group of training data was independently judged by two experienced rehabilitation doctors through video evaluation and functional scoring tools (such as the Tinetti score form) whether the gait was stable. Stable was defined as the patient being able to independently control the gait direction, with no more than 2 obvious direction drifts and an average trajectory error of less than 2 meters. Extract the Hurst exponent from the training records rated as stable in the above samples, and its statistical distribution range is mainly concentrated between [0.60, 0.78], with a mean of 0.68, a median of 0.66, and a standard deviation of 0.045. Its lower quartile (25th percentile) is 0.60. To ensure the recognition sensitivity and avoid excessive misjudgment, the lower quartile 0.60 of the Hurst exponent distribution in the stable state is used as the lowest stability limit and set as the benchmark index threshold. The Hurst exponent below this value represents that the gait offset trend fluctuation exceeds the normal range, and there may be a risk of unstable direction control or getting worse over time. Therefore, the current benchmark index threshold is reasonably set as .

[0034] The core of the comparison process is to compare the Hurst exponent calculated from the current patient's offset time series Compare segment by segment with a preset reference index threshold to identify whether there are periods with insufficient directional control stability. The whole process is divided into the following steps: Assume that the entire offset time series is divided into consecutive non-overlapping analysis windows of a fixed time length (for example, each segment is 60 seconds long), ensuring that each segment of data has a sufficient number of samples (such as n = 60) for calculating the Hurst exponent to ensure statistical stability. The setting of this window length is based on previous empirical analysis, and it has been verified that the trajectory changes within 60 seconds can better reflect the characteristics of small-scale trend fluctuations. Independently calculate the Hurst exponent for the offset distance series within each window, and record the corresponding time period number and its start and end timestamps. For each segment of the Hurst exponent Compare with the reference index threshold for numerical comparison: If : It is judged that the gait direction control has trend stability during this time period and is not marked as a risk; if : It is judged that there is a risk of unstable gait direction control during this time period and needs to be marked. Integrate all time period numbers and timestamps to form an offset risk segment set. This segment set consists of the start and end times of the time period, the Hurst value, the recovery period label, etc., and can be used for retrospective analysis and subsequent training plan adjustment. For example, the current training process is 300 seconds, divided into 5 windows, each 60 seconds: Assume that the calculated , , , , ; then the Hurst exponents in windows 3 and 4 are lower than the reference value of 0.60, identified as risk periods, and the time range is 120 - 180 seconds and 180 - 240 seconds; finally, the offset risk segment set is recorded as: ; This comparison process not only ensures continuous analysis on the time scale but also retains the quantitative judgment ability of gait offset trends, making the identification of risk time periods more targeted and accurate, providing a key reference basis for the rehabilitation training monitoring system.

[0035] By continuously statistically analyzing the spatial deviation between the patient's moving coordinates per second and the ideal path and using the trend of the deviation magnitude for exponential analysis, it is possible to dynamically identify the time periods with unstable direction control during walking, discover potential gait abnormalities or fall risks in advance, and improve the perception accuracy of behavioral deviations and the timeliness of nursing interventions during rehabilitation training.

[0036] The specific steps for obtaining the local abnormal interval of heart rate are as follows: Collect the patient's heart rate per minute during the day's rehabilitation training through the wearable device on the patient, and divide it according to a fixed sliding window to extract the continuous heart rate within each sliding window to form a heart rate information set; The heart rate data per minute during the daytime rehabilitation training is collected in real time through an intelligent wearable terminal on the patient. The photoplethysmogram sensor built into the terminal can continuously obtain heart rate change information. All the collected data is synchronously transmitted to the edge processing unit. The Matlab software is used to process the entire heart rate sequence. The original heart rate sequence is segmented according to the set sliding window parameters, and the sliding window parameters include the window length and the sliding step size. The program automatically traverses the original data and constructs multiple continuous subsequences to form a sliding window sequence set. Subsequently, the Pandas library in Python is used to classify, number, and parallelly arrange and integrate the heart rate data within each window, and finally output it as a heart rate information set for subsequent density analysis and anomaly detection operations.

[0037] In the heart rate information set, each time point is sequentially selected, as well as the heart rate information in the adjacent regions before and after the target time point. By mapping all the heart rate information to a multi-dimensional density space and using the local outlier factor algorithm, the local density of each time point in the corresponding adjacent region within the multi-dimensional density space is calculated. First, the heart rate data is reconstructed by sliding window in chronological order. Each time point corresponds to a heart rate vector with a fixed dimension, representing the continuous heart rate change characteristics within a certain number of minutes before and after that time point. Subsequently, with each time point as the center, its heart rate vector and the heart rate vectors of surrounding time points are combined to form a data set. The spatial embedding process of the multi-dimensional vector is completed through MATLAB, that is, all heart rate vectors are represented as data points located in a unified multi-dimensional Euclidean space to achieve the spatial expression of heart rate information.

[0038] Calculate each time point in the multi-dimensional density space The formula for the local density of the corresponding adjacent region is as follows:

[0039] Among them, represents the local density of the target time point in the corresponding adjacent region within the multi-dimensional density space, ' is the target time point currently being analyzed in the multi-dimensional density space, and its heart rate vector comes from the sliding window in the heart rate information set, is the target time point in the multi-dimensional density space The corresponding adjacent region set, indicating the set composed of the nearest time points in its front and back time sequences, and the neighborhood size is set through experiments (for example, when selecting it means taking the nearest 3 time points), represents the neighborhood point compared with the target time point in the adjacent region set, , Indicates the summation over all neighborhood points in the set of adjacent regions by traversing, where the traversal variable is , and the object of summation is each , is the target time point to the neighborhood point The reachable distance is defined as: , represents the target time point and the neighborhood point The Euclidean distance between them is the distance between two heart rate sliding window vectors in a multi-dimensional space. represents the Euclidean distance of the th nearest neighbor of the neighborhood point, which means that after calculating the distances between and other points in the dataset and sorting them, the distance ranked is selected.

[0040] Assume that the original 5-dimensional sliding window vectors of each heart rate time point have been standardized by Z-score normalization, which is calculated based on the mean and standard deviation of the daily heart rate full sample data. After normalization, all data are dimensionless values, centered at 0, and the standard deviation is 1.

[0041] Define the heart rate vectors of the analysis target time point and neighborhood points: Assume that the standardized heart rate vector at the target time point is , assume that the standardized heart rate vector at the neighborhood point is , assume that the standardized heart rate vector at the neighborhood point is , assume that the standardized heart rate vector at the neighborhood point is , where the neighborhood size , that is .

[0042] Calculate the Euclidean distance between the target time point in the multi-dimensional density space and each neighborhood point :

[0043] where is the th standardized heart rate vector at the target time point is the th standardized heart rate vector at the neighborhood point is the heart rate vector index after standardization, and its value range is , representing the th heart rate value, is the dimension of the heart rate vector after standardization, equal to the width of the sliding window. The acquisition method is to set the sliding window time range (such as 5 minutes), is the difference between the th heart rate vectors after standardization of the target time point and the neighborhood points.

[0044] Calculate : the Euclidean distance between the target time point and the neighborhood points ; ; Calculate : the Euclidean distance between the target time point and the neighborhood points ; ; Calculate : the Euclidean distance between the target time point and the neighborhood points ; .

[0045] The results show that , , .

[0046] Determine the th nearest neighbor distance of each neighborhood point : The th nearest neighbor distance formula: ; Calculate the Euclidean distance between the neighborhood point and each time point except the neighborhood point : :

[0047] where is the th standardized heart rate vector in each time point except the neighborhood point .

[0048] Suppose in the neighborhood size , take the second smallest Euclidean distance between each point and other points as its th nearest neighbor distance.

[0049] Calculate : the Euclidean distance between the current neighborhood point and the target time point ; ; Calculate : The current neighborhood point and the neighborhood point The Euclidean distance between ; Calculate : The neighborhood point and the neighborhood point The Euclidean distance between ; Sort and take the second smallest value: .

[0050] Calculate : The neighborhood point and the target time point The Euclidean distance between ; Calculate : The neighborhood point and the neighborhood point The Euclidean distance between ; Calculate : The neighborhood point and the neighborhood point The Euclidean distance between ; Sort and take the second smallest value: .

[0051] Calculate : The neighborhood point and the target time point The Euclidean distance between ; Calculate : The neighborhood point and the neighborhood point The Euclidean distance between ; Calculate : The neighborhood point and the neighborhood point The Euclidean distance between ; Sort and take the second smallest value: .

[0052] The results show that , , .

[0053] It is known that , , ; , , .

[0054] According to the formula:

[0055] Calculate : the target time point and the reachable distance to the neighborhood point ; ; Calculate : the target time point and the reachable distance to the neighborhood point ; ; Calculate : the target time point and the reachable distance to the neighborhood point ; .

[0056] Calculate the local density of the target time point in the corresponding adjacent region ; The result shows that in the standardized heart rate data, the local density of the neighborhood where the target time point is located is .

[0057] Calculate the local density of the th neighborhood point in the multi-dimensional density space to the corresponding adjacent region except the target neighborhood point : :

[0058] Among them, is the neighborhood size, is the set of corresponding adjacent regions of the th neighborhood point in the multi-dimensional density space, represents the other time points compared with the th neighborhood point in the set of adjacent regions, is the reachable distance from the th neighborhood point to the corresponding time point , defined as: is the th neighborhood point and the Euclidean distance between other time points , represents the Euclidean distance of the th nearest neighbor of other time points ', which means that after calculating the distances between and other points in the dataset and sorting them, the The distance, is the neighborhood point index.

[0059] According to the formula For : where z are respectively 、 、 ; Obtain : The Euclidean distance between the neighborhood point and the target time point is ; The set of Euclidean distances between the target point and other points is: ; After sorting, take the second smallest value: ; According to the formula: .

[0060] Obtain : The Euclidean distance between the neighborhood point and the neighborhood point ; : The distance between the point and other points is: ; After sorting, take the second smallest value: ; According to the formula: .

[0061] Obtain : The Euclidean distance between the neighborhood point and the neighborhood point ; : The distance between the point and other points is: ; After sorting, take the second smallest value: ; According to the formula: .

[0062] Calculate : The local density of the target neighborhood point is: .

[0063] For : where z are respectively 、 、 ; Obtain : The Euclidean distance between the neighborhood point and the target point is: ; Obtain : The Euclidean distance between the target point and other points is: ; After sorting, take the second smallest value: ; Substitute into the formula: .

[0064] Obtain : neighborhood points and the neighborhood points ' Euclidean distance: ; Obtain : the point ' distance from other points is: ; After sorting, take the second smallest value: ; Substitute into the formula: .

[0065] Obtain : neighborhood points and the neighborhood points ' Euclidean distance: ; Obtain : the point ' distance from other points is: ; After sorting, take the second smallest value: ; Substitute into the formula: .

[0066] Calculate : the local density of the neighborhood points : .

[0067] For : where z are respectively , , ; Obtain : the Euclidean distance between the neighborhood points and the target point : ; Obtain : ; Substitute into the formula: .

[0068] Obtain : the Euclidean distance between the neighborhood points and the neighborhood points : ; Obtain : ; Substitute into the formula: .

[0069] Obtain : the Euclidean distance between the neighborhood points and the neighborhood points : ; Obtain : ; Substitute into the formula: .

[0070] Calculate : Neighborhood points Local density of: .

[0071] The local outlier factor is used to measure the target point The relative ratio of the density of its neighborhood points , and the definition formula is as follows:

[0072] Where: , , , , is the neighborhood size, is the neighborhood point index.

[0073] Calculate each ratio item by item: , , .

[0074] Result: ; The result shows that the target time point The local density in the corresponding adjacent area is .

[0075] LOF (Local Outlier Factor) is an anomaly detection method based on density ratio. Its core logic does not directly rely on a certain absolute threshold or distance size, but judges whether it is abnormal by comparing the local density difference between the target point and its neighborhood. First, the reachable distance between the target point and the neighborhood points is obtained through the Euclidean distance, and then the dense degree of each point in its local space (i.e., the local reachable density LRD) is reflected. Then, LOF takes the ratio of the density of the target point to the average density of the neighborhood points: if the density of the target point is significantly lower than the neighborhood average density (i.e., the LOF value is greater than 1), it means that the point is relatively isolated and is identified as a local outlier; on the contrary, if LOF is close to or less than 1, it means that the point is in a density-consistent or crowded area and belongs to a normal point. Its effect is to be able to identify those time points that are inconsistent with the surrounding environment density, so as to accurately discover the local abnormal segments in the heart rate sliding sequence, rather than relying only on global statistical features or single-point fluctuations.

[0076] Compare the local density with the preset density threshold, screen out the local abnormal time points with local density lower than the density threshold, and mark the continuous local abnormal time points as the heart rate local abnormal interval; Specifically, the LOF value is equal to the ratio of the average local density of neighboring points to the local density of the current point itself. If the density of a point is much lower than that of its surrounding points, then this ratio will be greater than 1, and vice versa, it will be less than 1. Therefore: LOF = 1 indicates that the density of this point is comparable to that of its surrounding neighbors, belonging to ordinary points in the density structure; LOF > 1 indicates that the density of this point is less than the average density of the neighborhood, and it may fall in a relatively sparse or isolated position, with a risk of local anomalies; LOF < 1 indicates that this point is in a dense area, and its density is higher than that of the neighborhood, and it is usually not regarded as an anomaly. Based on this characteristic, when the LOF value at a certain time point is greater than 1, it indicates that its relative density is low, and it can be regarded as a potential local anomaly time point. By scanning the entire time series, all time points that satisfy LOF > 1 are selected and initially used as local anomaly candidate points. These points may correspond to mutations, disorders, or other atypical fluctuations in the heart rate signal. Further, in order to enhance the stability of detection, these discrete anomaly points can be analyzed by connecting them in the time dimension: if multiple anomaly points are adjacent in time or have a very short interval, they are merged into a continuous heart rate local anomaly interval and uniformly marked as the heart rate local anomaly interval.

[0077] By performing a sliding window process on the continuous heart rate data during rehabilitation training and combining multi-dimensional density analysis with the local outlier factor algorithm, it is possible to accurately identify the continuous abnormal fluctuations of the heart rate in a local time period, effectively distinguish short-term anomalies from physiological fluctuations in the stable state, improve the ability to identify the patient's stress response or potential abnormal physiological state, and enhance the sensitivity and discrimination accuracy of heart rate monitoring.

[0078] The steps for obtaining the heart rate fluctuation classification set are specifically as follows: Statistically analyze the duration of each heart rate local anomaly interval in each time period, and compare the statistically obtained durations with the set time threshold respectively. The heart rate local anomaly intervals higher than the time threshold are marked as continuous fluctuation sections, and the heart rate local anomaly intervals lower than the time threshold are marked as instantaneous fluctuation sections; According to the obtained local anomaly time points and their specific timestamp positions in the heart rate time series, establish a time interval array , where represents the th sampling time point determined to be locally abnormal. Traverse the interval differences between the time points in in turn. If the interval between adjacent time points is one sampling period , then merge them into the same heart rate local anomaly interval, and record the start timestamp and end timestamp of each heart rate local anomaly interval. For each heart rate local anomaly interval, calculate its duration according to the following formula: ; where Indicates the duration (in seconds) of the th local abnormal heart rate interval, where is a single sampling period (read from device configuration as and are the start and end timestamps (in seconds) of the local abnormal heart rate interval respectively. The durations of all segments form a duration set in sequence. Subsequently, the time threshold set in the system parameters (obtained from clinical experience or statistical analysis of training samples, also in seconds) is extracted, and a comparison operation is performed. When the condition: ; is met, then the th segment is marked as continuously fluctuating, otherwise it is marked as instantaneously fluctuating. The comparison process is completed in the segment index loop. The segment marking results and time index are saved in an array mapping in key-value pair structure for subsequent heart rate fluctuation type classification processing.

[0079] Classify and mark according to the segment types corresponding to continuously fluctuating segments and instantaneously fluctuating segments, and establish a heart rate fluctuation classification set; Extract the time range of each local abnormal segment and the corresponding fluctuation type label, construct a segment classification comparison table, traverse each item in this comparison table, read the start timestamp and end timestamp of the current segment, and according to the fluctuation label type corresponding to the segment, uniformly classify all heart rate data segments within this time interval. If the segment label is continuously fluctuating, add the corresponding data segments to the continuously fluctuating classification set; if the segment label is instantaneously fluctuating, add its data segments to the instantaneously fluctuating classification set. Each time when classifying, the data segments need to be numbered to mark their corresponding segment sequence numbers and sources. After finally completing the traversal of all segments and data classification, assemble the two fluctuation sets into a unified data set structure according to the structure definition method, and name it the heart rate fluctuation classification set.

[0080] By grading the duration of the local abnormal heart rate interval and dividing it into two categories: continuously fluctuating and instantaneously fluctuating, the classification management of different types of heart rate abnormalities is realized, which helps to distinguish between short-term emotional fluctuations and potential pathological states, provides a hierarchical basis for subsequent nursing responses, and improves the accuracy of abnormal heart rate processing and the pertinence of intervention.

[0081] The specific steps for obtaining the blood pressure trend analysis result are as follows: Collect the continuous systolic blood pressure pulse waveform signals of the patient at night after surgery through a blood pressure monitoring device, and divide them into multiple equal-length time signal windows. For each segment of the systolic blood pressure pulse waveform in the time signal window, construct a continuous signal curve with the corresponding time as the horizontal axis and the systolic blood pressure pulse as the vertical axis; Collect the continuous systolic blood pressure pulse waveform signals of the patient at night after surgery through a blood pressure monitoring device, export the signals in a standard format as a continuous data stream with timestamps, call NumPy and Pandas to resample the timestamps at equal intervals to ensure the continuity of the time axis, and then segment the entire signal sequence into units of every 10 seconds according to the set window duration. Use a rolling window method to divide and generate multiple signal segments of equal length, each segment containing a fixed number of sampling points. During the division process, retain the start time index of each segment for subsequent identification processing. After completing the window division, use the Matplotlib and SciPy libraries to plot a continuous curve graph with the time axis of each segment of signal data as the abscissa and the pulse signal amplitude as the ordinate. Before constructing the curve, uniformly perform mean removal and standardization processing on each segment of the signal to improve the smoothness of the drawing and reduce the interference of baseline fluctuations.

[0082] Perform the Hilbert transform on the continuous signal curve to analyze the instantaneous frequency and phase trajectory within each cycle corresponding to the systolic blood pressure pulse. Analyze the rhythm changes within the cycle corresponding to the systolic blood pressure pulse based on the instantaneous frequency and phase trajectory, identify and calibrate the cycle segments with a decreasing rhythm and no rebound trend, and generate the blood pressure trend analysis result; After the continuous signal curve is preprocessed, use the SciPy library in Python to perform the Hilbert transform operation. During the transformation process, the original real-valued systolic blood pressure pulse waveform signal is converted into an analytic signal in complex form, where the real part is the original signal itself and the imaginary part is the Hilbert transform component corresponding to the original signal. The analytic signal generated after the transformation can further extract the instantaneous frequency and phase trajectory. The instantaneous frequency reflects the rapid change of the pulse signal oscillation per unit time, with the unit of Hertz, and the instantaneous phase represents the phase position of the waveform on the time axis, with the unit of radian. SciPy obtains the phase sequence and the corresponding instantaneous frequency sequence by performing angle function and derivative processing on the analytic signal. Statistical analysis of the instantaneous frequency change of each segment of the signal can identify the segments where the frequency decreases over time. After the Hilbert transform is completed, the obtained instantaneous frequency represents: the rapidity of the systolic blood pressure pulse signal oscillation per unit time at each moment, with its numerical unit being Hz, corresponding to the number of pulse waveform cycles contained per second; and the phase trajectory represents: the instantaneous phase corresponding to the signal at each moment, with the unit of radian, which is an indicator of the position of the waveform evolution on the time axis and is commonly used to describe rhythmic progression and waveform alignment. For example, assume that in the analysis of blood pressure waveform data, after applying the Hilbert transform to each signal window segment, extract its corresponding instantaneous frequency and phase sequence. For the analysis of rhythm changes, the system sets the physiologically reasonable interval of the instantaneous frequency to , that is, the normal heart rhythm frequency should be between 0.5 and 2.0 beats per second. Rhythm maintenance segment: If within the time interval , the instantaneous frequency sequence remains within range, and the fluctuation amplitude does not exceed ±0.1 Hz. At the same time, the phase trajectory increases linearly with time, and its first derivative remains in the positive interval, indicating that this section is a rhythm stable section; Rhythm fluctuation section: If within the time period there are multiple rises and falls in the instantaneous frequency, showing interval rebounds, such as: dropping from 1.4 Hz to 1.0 Hz and then rising back to 1.3 Hz. Combining with multiple inflection points in the phase trajectory, it is determined that this section is a rhythm fluctuation section; Rhythm decline section (calibration object): Within the time period , the instantaneous frequency continuously decreases, the initial value is , the final value is , and there is no rise point exceeding during the process (according to the preset rebound judgment threshold), which is regarded as having no rebound trend. And the phase trajectory remains continuously increasing throughout the interval, without phase flipping or interruption, and its derivative is always greater than zero but the absolute value gradually decreases. Then this paragraph is calibrated as a periodic section with rhythm decline and no rebound trend according to the standard.

[0083] By performing time segmentation and signal modeling on the continuous systolic blood pressure pulse waveform signal at night, and introducing instantaneous frequency and phase trajectory analysis, it is possible to identify abnormal cycles with rhythm decline and no recovery trend, effectively capture early signs of weakened hemodynamics in the static state, and enhance the recognition ability and warning value of night blood pressure monitoring for chronic hidden risks.

[0084] The steps for obtaining the record of abnormal non - response of venous return are specifically as follows: Based on the blood pressure trend analysis results, it is speculated whether there is slowdown of lower limb venous return in the target cycle section of the patient. Among them, the cycle section with rhythm decline and no rebound trend corresponds to a stable slow - down process in which hemodynamics continuously weakens and is not interrupted by body position changes, indicating a decrease in venous return efficiency during a long - term static state, and the lower limb venous return analysis result is obtained; Select the cycle segment that is calibrated as having a decreasing rhythm and no rebound trend as the target segment. Based on the continuous decrease characteristic of the instantaneous frequency and the linear evolution state of the phase trajectory within this segment, it is speculated that this segment represents a hemodynamic stasis process that is not interrupted by body position changes. In specific analysis, if the instantaneous frequency value within a certain cycle segment gradually decreases from 1.5 Hz to 0.7 Hz, the system detects the derivative sequence of continuous fluctuations. The criterion for judging the downward trend is that the frequency value satisfies the negative derivative in no less than 5 consecutive windows, and the maximum frequency increase amplitude within this segment does not exceed 0.05 Hz. This threshold is determined based on the normal fluctuation standard range of postoperative nocturnal blood pressure rhythm, that is, the normal frequency fluctuation range should not exceed ±0.1 Hz. If the actual fluctuation amplitude is less than the lower limit of this range, it is regarded as having no significant rebound. At the same time, the derivative of the phase trajectory always remains positive throughout the segment and there is no local slope mutation, and its change curvature fluctuation does not exceed 0.05 rad / s², which is used as the criterion for judging the stability of the phase trajectory. This threshold is derived from the standard deviation statistical results of the stable fluctuation segments in the same batch of resting samples. If the target cycle segment simultaneously meets the above two criteria of instantaneous frequency decrease and phase trajectory stability, and the duration of this stasis state is greater than 120 seconds, this duration threshold refers to the minimum duration requirement in clinical standards for the maintenance of non-active body positions that may affect venous return efficiency, then it is inferred that there is a risk state of reduced venous return efficiency in this segment. The system accordingly labels this cycle segment as a suspected state segment with slow venous return.

[0085] Based on the analysis results of lower limb venous return, by linking the operation data in the nursing record for the same time period, identify whether there is any nursing intervention such as raising the lower limbs or body position guidance. If no nursing intervention is performed, mark the target cycle segment again and establish a record of non-response to abnormal venous return. Read the start and end times of the cycle segment determined to have slow venous return, and call the nursing operation record data of the corresponding patient on the same day. Screen out the operation entries whose recording times overlap with the target cycle segment. Compare the recording time field in the nursing record with the target cycle segment one by one to determine whether the recording time is between the start and end times of the cycle segment. Extract the matching recorded content fields and identify whether there is any operation description containing expressions such as "raising the lower limbs", "body position adjustment", "leg elevation pad", "elevating both lower limbs", etc. This identification process is precisely matched through a preset keyword dictionary. If the match is successful, it is regarded that there has been a nursing intervention behavior within this cycle segment and no further processing is required. If no nursing intervention keyword record is identified throughout the cycle segment, this cycle segment is determined to be "non-responsive to abnormal venous return". The system marks this cycle segment as abnormal again, adds it to the non-response record set, and at the same time records the cycle segment number, the corresponding timestamp, and the content of the nursing intervention field that was not matched, for subsequent analysis and responsibility filing.

[0086] By analyzing the cycle of blood pressure rhythm decline without rebound, identify the possible problem of slow lower limb venous return in patients at rest, and combine the nursing records during the same period to determine whether intervention measures such as leg elevation or position guidance have been implemented, so as to mark the risk periods that have not been responded to, improve the recognition accuracy of abnormal nocturnal venous return and the perception ability of nursing deficiencies, and help prevent complications such as thrombosis.

[0087] The specific steps for obtaining the recommended results of the multi-dimensional nursing path are as follows: According to the set of offset risk segments, recommend the daytime movement path settings and guiding methods, including leaning against the wall, setting boundary aids, and arranging escorts to guide the movement at a specified distance; Extract the start and end time information of all time periods marked as having risk offsets, and obtain the monitoring room number and patient number information corresponding to the corresponding segments. Call the corresponding room structure drawings in the on-site environment layout data and load the patient's current bed number and passage route information. Generate the initial movement trajectory line segments according to the actual positioning points of each offset risk segment, and mark the potential offset direction angles. Based on the relationship between the direction angle and the room boundary, judge whether it is close to boundary areas such as doorways, corridors, and corners. If the direction points to an open area, set the auxiliary recommendation in the system as "arrange an escort to follow within one meter of the target position". If the direction is close to the wall area, set the auxiliary method as "move along the wall". If the direction points to the bedside junction position, set it as "place an aid at the boundary". At the same time, record the time label, target point coordinates, recommended guiding method, and auxiliary description of each offset risk path in the system. Finally, output the summary table of the recommended guiding methods for all offset segments as the auxiliary recommendation content for daytime path settings.

[0088] According to the heart rate fluctuation classification set, regard the instantaneous fluctuation section as a non-persistent reaction and do not adjust the nursing plan. Regard the continuous fluctuation section as a state of pain, and it is recommended to increase the cold compress intervention reminder and the nursing patrol frequency; Extract the serial numbers of all time periods marked as instantaneous fluctuations and the corresponding start and end times in the classification set, and directly label the sections of this category as reference areas that do not require intervention, skipping the subsequent nursing configuration processing flow. Subsequently, extract all sections marked as continuous fluctuations in the classification set, count the duration of each section and the patient numbers to which they belong, and call the cold compress intervention recommendation rule to determine whether the set prompt condition is met. This condition is that the duration of continuous fluctuations is greater than or equal to 180 seconds. This threshold is obtained by analyzing the corresponding relationship between the self-reported pain of postoperative clinical patients and continuous heart rate changes. If the condition is met, mark the current time period as the cold compress prompt time period. At the same time, retrieve the record of the patient's previous daily inspection frequency in the nursing log, and compare whether the daily inspection interval exceeds 30 minutes. If the inspection interval is greater than this duration, add a recommendation item to the record form, prompting to add a nursing inspection operation before and after the start time of the current fluctuation section. Write the cold compress prompt and the nursing inspection adjustment recommendation into the nursing plan recommendation list for manual review and confirmation.

[0089] Based on the records of unresponsive venous return abnormalities, it is recommended to increase the time point settings for elevating the lower limbs or guiding turning over within the calibrated night target cycle segments, and obtain the multi-dimensional nursing path recommendation results to provide decision support for the auxiliary selection of nursing plans for orthopedic patients; Read the start and end times of all calibrated night target cycle segments, group them by patient number, arrange the abnormal cycle segments of each patient in chronological order, establish a cycle segment list and filter out duplicate marks, and call the preset nursing intervention time interval standard. It is stipulated that between 23:00 and 05:00 at night, if there is no body position adjustment record for more than 120 consecutive minutes and there are abnormal unresponsive cycle segments, an operation of elevating the lower limbs or guiding turning over needs to be added in this segment. According to this standard, the system calculates the duration of each cycle segment and compares it with the interval between the previous and next cycles. If the interval between the current cycle segment and the previous segment is greater than 60 minutes, insert a record of the recommendation to elevate the lower limbs at the start time of the current cycle segment. If the duration of the current cycle segment itself is greater than 180 seconds, insert a record of the recommendation to guide turning over at the midpoint of the cycle. Package all the generated time points, operation types, and corresponding patient numbers into a recommended action set and write it into the nursing path recommendation file to form the multi-dimensional nursing path recommendation results.

[0090] According to the multi-source data results such as gait deviation, heart rate fluctuation type, and venous return abnormalities, formulate specific nursing recommendations respectively, such as adjusting the walking guidance method, increasing the frequency of cold compress and inspection, arranging night turning over intervention, etc., to achieve personalized, time-segmented, and targeted recommendations for nursing plans, enhance the matching degree between nursing measures and patient status, and improve the effectiveness of nursing intervention and the initiative of risk response.

[0091] The above are only the preferred embodiments of the present invention, and do not limit the present invention in other forms. Any person skilled in the art may use the technical content disclosed above to make changes or modifications into equivalent embodiments with equivalent changes and apply them to other fields. However, as long as it does not depart from the technical solution content of the present invention, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention still fall within the protection scope of the technical solution of the present invention.

Claims

1. A nursing plan auxiliary selection system for orthopedic patients, characterized in that: The system includes: The gait deviation monitoring module collects the movement path of the patient during the daytime rehabilitation training, compares it with the ideal movement path, extracts the deviation distance between the paths according to the comparison result, forms a deviation time series, and screens the risk time period by calculating the Hurst index of the deviation distance amplitude change in the deviation time series to form a deviation risk segment set; The heart rate abnormality recognition module collects the heart rate of the patient during the daytime rehabilitation training and maps it to the multidimensional density space, marking the continuous local abnormal time points of the heart rate in the multidimensional density space as the local abnormal heart rate interval; The heart rate fluctuation classification module divides the interval types according to the duration of the local abnormal heart rate interval in each time period and establishes a heart rate fluctuation classification set; The blood pressure trend analysis module collects the patient's continuous systolic blood pressure pulse waveform signal at night after surgery, analyzes the rhythm changes of the systolic blood pressure pulse within a specified period, and generates blood pressure trend analysis results; The postural intervention discrimination module links the operation data of the same time period in the nursing records based on the blood pressure trend analysis results, marks the target period segments that have not undergone nursing intervention, and establishes a record of abnormal venous return and no response; The rehabilitation guidance module recommends corresponding nursing plans to patients based on the deviation risk segment set, heart rate fluctuation classification set, and venous return abnormality unresponsiveness record, and obtains multi-dimensional nursing path recommendation results.

2. The nursing plan auxiliary selection system for orthopedic patients according to claim 1, characterized in that: The steps for obtaining the offset risk segment set are specifically as follows: The patient's movement path during the daytime rehabilitation training is collected through the wearable device on the patient, and the longitude and latitude coordinate points recorded every second in the movement path are converted into a continuous spatial coordinate sequence with equal intervals. According to the patient's current postoperative recovery period, the ideal movement path of the rehabilitation training plan is preset, and the deviation distance of the coordinate points of the continuous spatial coordinate sequence at the same time relative to the coordinate points of the corresponding position of the ideal movement path is calculated and recorded; The deviation distance at each time is counted to form a migration time series, and the Hurst exponent algorithm is used to calculate the Hurst exponent of the deviation distance amplitude change in the migration time series by analyzing the power law relationship between the variation amplitude of the deviation distance at different times in the migration time series and the length of the entire migration time series; The Hurst index is compared with a preset benchmark index threshold to evaluate the stability of the patient's directional control during walking, and the corresponding risk time periods in the deviation time series that are lower than the benchmark index threshold are marked to form a deviation risk segment set.

3. The nursing plan auxiliary selection system for orthopedic patients according to claim 1, characterized in that: The steps for obtaining the local abnormal heart rate interval are specifically as follows: The patient's heart rate is collected every minute during the daytime rehabilitation training through the wearable device on the patient, and divided into fixed sliding windows, and the continuous heart rate in each sliding window is extracted to form a heart rate information set; Selecting each time point and the heart rate information of the adjacent areas before and after the target time point in the heart rate information set in turn, mapping all the heart rate information to the multidimensional density space, and using the local anomaly factor algorithm to calculate the local density of each time point in the corresponding adjacent area in the multidimensional density space; The local density is compared with a preset density threshold, local abnormal time points where the local density is lower than the density threshold are screened, and continuous local abnormal time points are marked as local abnormal heart rate intervals.

4. The nursing plan auxiliary selection system for orthopedic patients according to claim 1, characterized in that: The steps for obtaining the heart rate fluctuation classification set are specifically as follows: The duration of the local abnormal heart rate interval in each time period is counted, and the counted duration is compared with the set time threshold, and the local abnormal heart rate interval higher than the time threshold is marked as a continuous fluctuation segment, and the local abnormal heart rate interval lower than the time threshold is marked as an instantaneous fluctuation segment; Classification and labeling are performed according to the segment types corresponding to the continuous fluctuation segment and the instantaneous fluctuation segment to establish a heart rate fluctuation classification set.

5. The nursing plan auxiliary selection system for orthopedic patients according to claim 1, characterized in that: The steps for obtaining the blood pressure trend analysis results are specifically as follows: The continuous systolic blood pressure pulse waveform signal of the patient at night after surgery is collected by a blood pressure monitoring device and divided into multiple equal-time signal windows. For each systolic blood pressure pulse waveform in the time signal window, a continuous signal curve is constructed with the corresponding time as the horizontal axis and the systolic blood pressure pulse as the vertical axis; A Hilbert transform is performed on the continuous signal curve to parse the instantaneous frequency and phase trajectory within the corresponding cycle of each systolic blood pressure pulse, and the rhythm changes within the corresponding cycle of the systolic blood pressure pulse are analyzed according to the instantaneous frequency and phase trajectory, and the cycle segments with a rhythm decline and no rebound trend are identified and calibrated to generate a blood pressure trend analysis result.

6. The nursing plan auxiliary selection system for orthopedic patients according to claim 5, characterized in that: The specific steps for obtaining the abnormal venous return unresponsive record are: According to the blood pressure trend analysis results, it is inferred whether the patient has a slowing of lower extremity venous return in the target cycle segment. The cycle segment with a decreasing rhythm and no rebound trend corresponds to a stable hysteresis process in which the hemodynamics is continuously weakened and is not interrupted by changes in body position, indicating that the venous return efficiency is reduced in a long-term static state, and the lower extremity venous return analysis results are obtained; Based on the lower limb venous return analysis results, by linking the operation data of the same time period in the nursing records, it is identified whether there is nursing intervention such as raising the lower limbs and guiding the posture. If no nursing intervention is performed, the target cycle segment is marked again and a record of abnormal venous return and no response is established.

7. The nursing plan auxiliary selection system for orthopedic patients according to claim 1, characterized in that: The steps for obtaining the multi-dimensional management path recommendation result are specifically as follows: According to the set of deviation risk fragments, recommendations are made on the setting of daytime movement paths and guidance methods, including leaning against walls, setting boundary aids, and arranging accompanying personnel to guide movement at a specified distance; According to the heart rate fluctuation classification set, the instantaneous fluctuation segment is regarded as a non-sustained reaction, and no nursing plan adjustment is made. The continuous fluctuation segment is regarded as a painful state, and it can be recommended to increase the cold compress intervention reminder and nursing rounds. Based on the record of abnormal venous return without response, it is recommended to increase the time point setting for raising the lower limbs or guiding turning over within the calibrated night target cycle segment to obtain multi-maintenance care path recommendation results, which are used to provide decision support for auxiliary selection of nursing plans for orthopedic patients.

Images