Intelligent nursing intervention system and method for nausea and vomiting after pulmonary operation

Abstract

The invention relates to the technical field of intelligent nursing, in particular to an intelligent nursing intervention system and method for nausea and vomiting after a pulmonary operation. The analysis of nausea and vomiting performance is established on the basis of continuous and quantifiable indexes, the sensitivity of judging a risk trend is improved, a risk change path is identified through a long-short-term memory neural network, an influence rule of intervention intensity on a nausea and vomiting evolution trend is accurately described, stage prediction of a nursing intervention effect is realized, and the risk prediction accuracy is improved. Structural optimization is performed on an intervention time sequence and nursing task allocation and execution rhythm through a convolutional neural network, an intervention rhythm path is recombined in combination with resource adaptation information, the coverage density of intervention measures for risk periods is enhanced, integrated response of nursing resources in a high-sensitivity interval is promoted, and synchronous optimization of task allocation and time sequence matching is realized.

Description

Technical Field

[0001] The present invention relates to the field of intelligent nursing technology, and in particular to an intelligent nursing intervention system and method for nausea and vomiting after lung surgery. Background Art

[0002] The field of intelligent nursing technology aims to improve the intelligence, precision and efficiency of nursing work. Through physiological monitoring, data analysis, intelligent decision-making and automated intervention, it optimizes the nursing process, improves the quality of patient recovery, reduces the workload of nursing staff, and reduces medical risks.

[0003] The purpose of the intelligent nursing intervention system for nausea and vomiting after lung surgery is to reduce the incidence and severity of nausea and vomiting in patients after lung surgery, improve the accuracy and efficiency of postoperative care, and promote faster recovery of patients. By real-time monitoring of physiological parameters, intelligent analysis of disease trends, and automatic delivery of nursing intervention plans, the system can promptly identify the risk of nausea and vomiting, optimize nursing decisions, reduce the workload of nursing staff, improve patient comfort, reduce the risk of complications, and improve the quality of postoperative care and patient experience.

[0004] Existing technologies lack a forward warning mechanism and are unable to achieve a structural match between physiological parameters and symptom manifestations, which can easily lead to delays in risk identification and misjudgment. As a result, some high-risk patients only begin intervention treatment after nausea or vomiting occurs. The formulation of intervention plans lacks serialized parameter support, and it is difficult to adjust the execution frequency and time period according to the dynamic changes in the postoperative status, resulting in a rigid operation path and insufficient execution adaptability. The arrangement of nursing resources is not coupled with the patient's postoperative status, and there are problems such as task duplication, resource mismatch or response time gaps. In addition, a dynamic closed-loop process with symptom risk fluctuations as the core has not been established, which affects the improvement of nursing accuracy and the guarantee of intervention effectiveness. Summary of the Invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose an intelligent nursing intervention system and method for nausea and vomiting after lung surgery.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] Obtaining the postoperative nursing intervention adjustment parameter set through the patient data processing module;

[0008] Obtain the postoperative nausea and vomiting nursing intervention prediction model through the postoperative risk assessment module;

[0009] Obtain individualized nursing intervention implementation plans through the individualized nursing optimization module;

[0010] Obtain nursing abnormality monitoring results through the nursing abnormality monitoring module;

[0011] The dynamic regulation path set of nursing intervention is obtained through the dynamic adjustment module of nursing plan.

[0012] An intelligent nursing intervention system for nausea and vomiting after lung surgery includes:

[0013] Patient data processing module: Based on the key postoperative medical parameters retained by the hospital, calculate the blood oxygen level, heart rate, respiratory rate and pain score, combine the postoperative change rate of the patient's medical history, analyze the trend of postoperative nausea and vomiting, and calculate the nursing intervention adjustment coefficient based on the implementation of nursing interventions to obtain the postoperative nursing intervention adjustment parameter set;

[0014] Postoperative risk assessment module: Based on the postoperative nursing intervention adjustment parameter set, the impact of nursing intervention on the risk change of postoperative nausea and vomiting is calculated. In combination with the occurrence trend of postoperative nausea and vomiting and the nursing intervention response rate, the intervention adaptability of the nursing method is determined. A neural network model is generated using a long-short-term memory neural network. A prediction mapping of nursing adjustment to postoperative nausea and vomiting is established. In combination with the nursing measure adjustment space, the adaptive nursing intervention intensity is calculated to obtain a postoperative nausea and vomiting nursing intervention prediction model.

[0015] Individualized nursing optimization module: Based on the postoperative nausea and vomiting nursing intervention prediction model, the module calculates the impact of the nursing intervention sequence on postoperative nausea and vomiting. Combined with the nursing intervention adjustment coefficient, the module optimizes the nursing intervention timing, frequency, and implementation intensity. The module also calculates the adaptability of nursing measures. Combined with postoperative pain and analgesic use, the module optimizes the nursing intervention plan and obtains an individualized nursing intervention implementation plan.

[0016] Nursing abnormality monitoring module: Based on the individualized nursing intervention execution plan, monitor nursing execution data, calculate nursing measure deviation values, screen abnormal nursing execution status, calculate the degree of abnormal impact based on the occurrence trend of postoperative nausea and vomiting, determine whether nursing abnormalities affect the stability of postoperative nausea and vomiting control, calculate nursing abnormality risk levels based on the nursing intervention adjustment coefficient, and obtain nursing abnormality monitoring results;

[0017] Nursing plan dynamic adjustment module: Based on the abnormal nursing monitoring results, a convolutional neural network is used to adjust the nursing staff tasks, nursing processes and intervention timing, calculate the adaptability of nursing resources, and combine the postoperative nausea and vomiting nursing intervention prediction model to evaluate the adaptability of the adjusted nursing plan. Based on the adjustment space of nursing measures, the nursing intervention plan is optimized to ensure the accuracy of nursing execution and obtain a set of dynamic control paths for nursing intervention.

[0018] As a further solution of the present invention, the patient data processing module includes:

[0019] Postoperative parameter extraction submodule: Based on the key postoperative medical parameters retained by the hospital, it performs unified timestamp alignment and outlier screening for postoperative monitoring data, groups multi-channel data and calculates continuous fluctuation intervals, performs segmented averaging and extreme value identification for blood oxygen, heart rate, respiratory rate, and pain scores, and obtains a basic monitoring data set.

[0020] Medical history feature analysis submodule: Based on the basic monitoring data set, it extracts and cleans the patient's medical history text information fields, performs unified coding processing on abnormal records and maps them to structured data, cross-compares monitoring data with medical history features, extracts physiological fluctuation parameters within the postoperative time window and performs trend segmentation to obtain pathological evolution characteristic factors;

[0021] Physiological trend identification submodule: Based on the pathological evolution characteristic factors, it performs time-series mapping of nursing intervention time points and monitoring data nodes, compares the segments before and after the intervention and marks the inflection points of physiological indicators, calculates the corresponding change rate of each intervention item and performs difference analysis, classifies trend segments according to the abnormal accumulation threshold of indicators, generates a nursing intervention response difference table, and obtains the postoperative nursing intervention adjustment parameter set;

[0022] The key postoperative medical parameters retained by the hospital include postoperative physiological indicators, ventilator parameters, anesthesia methods, types and dosages of medications, and postoperative complications retained by the hospital.

[0023] As a further embodiment of the present invention, the postoperative risk assessment module includes:

[0024] Risk impact calculation submodule: Based on the postoperative nursing intervention adjustment parameter set, the time window between the intervention item parameters and the nausea and vomiting marker events is aligned, the interval change frequency of each parameter dimension is extracted and the intervention response span is calculated, and after extracting the key influencing variables, multivariate weighting is performed to obtain the nursing intervention impact coefficient group;

[0025] Nursing Adaptation Judgment Submodule: Based on the nursing intervention influence coefficient group, a long-short-term memory neural network is used to reconstruct the time series of nausea and vomiting marking intervals, extract intervention execution items with high response rates and classify them into groups, construct an adaptation matrix for each execution group under different patient conditions, and screen high-frequency combinations to obtain nursing adaptability structural characteristics;

[0026] Intervention intensity optimization submodule: Based on the nursing adaptability structural characteristics, the execution order of all intervention combination plans is rearranged and a transformation sequence list is established. The state response trajectory between intervention intensity and nausea and vomiting events is extracted, the intensity fluctuation amplitude and interval of each trajectory segment are extracted and sorted, and the high-intensity response segment identification is aggregated to obtain a postoperative nausea and vomiting nursing intervention prediction model.

[0027] As a further embodiment of the present invention, the long short-term memory neural network is according to the formula:

[0028]

[0029] Where: R is the patient's nausea and vomiting response rate, α i is the impact coefficient of the intervention execution item, x i is the characteristic of the intervention execution item, f(x i ,w i ,t i ,d i ) is the response function of the long short-term memory network to the intervention characteristics and parameters, n is the number of intervention execution items, w i is the weight coefficient of the intervention execution item, t i is the individual patient's physiological state parameter, d i Provide patient treatment history data.

[0030] As a further embodiment of the present invention, the individualized care optimization module includes:

[0031] Intervention sequence evaluation submodule: Based on the postoperative nausea and vomiting nursing intervention prediction model, all combinations of nursing action sequences are generated and sorted by time, nausea and vomiting reaction nodes within the combination are extracted and corresponding distribution maps are constructed, response durations of key action nodes are recorded and concurrent positions are marked, high-frequency response segments are extracted, and intervention conflict positions between sequences are screened to obtain nursing sequence impact characteristics;

[0032] Nursing rhythm adjustment submodule: Based on the influencing characteristics of the nursing sequence, it performs numerical mapping between the nursing intervention adjustment coefficient and the rhythm parameter list, extracts the frequency fluctuation value within the adjustment interval and reorganizes the distribution block, clusters the intervention items in the frequency jump segment and analyzes the rhythm continuity, extracts the intervention intensity sequence in the frequency compression interval and sorts it by rhythm offset to obtain the intervention rhythm adaptation structure;

[0033] Program structure matching submodule: Based on the intervention rhythm adaptation structure, multivariate alignment of postoperative pain records and analgesic usage data is performed, the intersection blocks of high-value points of intervention intensity and rising drug concentration segments are extracted, the time overlap rate and the overlap position of drug effects in the nursing structure are analyzed, the intervention path is selected according to priority and the complete structure group is output to obtain an individualized nursing intervention execution plan.

[0034] As a further solution of the present invention, the nursing abnormality monitoring module includes:

[0035] Execution data comparison submodule: Based on the individualized nursing intervention execution plan, the nursing execution record time nodes are synchronously matched with the execution intensity data, the actual start and end time of the intervention task are extracted and the duration difference is calculated, the execution delay and advance nodes are identified and the variation forms are marked, the execution distribution matrix between the plan and the actual is constructed, and the nursing execution deviation parameters are obtained;

[0036] Abnormal deviation identification submodule: Based on the nursing execution deviation parameters, it performs connectivity division and intensity layer extraction of the deviation distribution segment, extracts the functional labels of the intervention items to which the deviation nodes belong and records the continuous mismatch intervals, screens the span and fluctuation trend of the abnormal accumulation segment, constructs a combined identification of the deviation form and nursing events, and obtains the characteristics of the abnormal nursing status;

[0037] Intervention stability judgment submodule: Based on the abnormal nursing state characteristics, combined with the occurrence trend of postoperative nausea and vomiting and the nursing intervention adjustment coefficient, state hierarchical mapping is performed, the control parameter blocks corresponding to the abnormal labels are extracted and the impact distribution analysis is performed, the stability decline event segments are constructed and classified into the abnormal intensity matrix, and the impact score table of each nursing task is generated to obtain the nursing abnormality monitoring results.

[0038] As a further solution of the present invention, the nursing plan dynamic adjustment module includes:

[0039] Task structure reorganization submodule: Based on the nursing abnormality monitoring results, a convolutional neural network is used to align the abnormal event marking time points with the nursing task execution logs, extract the task start and end times and the executor identification and establish an event mapping matrix, perform time overlap detection between tasks and screen for conflicts in nursing role responsibilities, reconstruct the task sequence and update the task connection node list to obtain the intervention task adjustment structure;

[0040] Resource load assessment submodule: Based on the intervention task adjustment structure, the task time period is split and the execution frequency is calculated. The nursing position scheduling information and the list of schedulable durations are extracted and matched with the task demand sequence. The task-intensive section load diagram is constructed and the resource utilization peak position is identified. The high-intensity intervention blocks are marked in combination with the postoperative nausea and vomiting nursing intervention prediction model to obtain the resource adaptation parameter model.

[0041] Plan path optimization submodule: Based on the resource adaptation parameter model, the intervention component set in the nursing measure adjustment space is extracted and the nodes are rearranged, the task sorting list is matched with the resource interval diagram and a time conflict evaluation table is established, the low-conflict permutations and combinations in the intervention path are screened and the node numbers are marked, the sequential adaptation task execution roadmap is output, and the nursing intervention dynamic control path set is obtained.

[0042] As a further solution of the present invention, the convolutional neural network is based on the formula:

[0043]

[0044] Where: E t β is the alignment between the task execution time point and the abnormal event marking time point, i The weight coefficient for each task, y i is the task feature, g(y i ,θ i ,δ i ,s i ) is the response function of the convolutional neural network to the task characteristics and other parameters, n is the number of tasks, θ i is the convolution kernel parameter of the convolution layer, δ i is the role identifier of the task executor, s i is the strength of the relationship between tasks and abnormal events.

[0045] An intelligent nursing intervention method for nausea and vomiting after lung surgery is provided. The intelligent nursing intervention method for nausea and vomiting after lung surgery is implemented based on the intelligent nursing intervention system for nausea and vomiting after lung surgery, and comprises the following steps:

[0046] Step 1: Based on the numerical changes in postoperative blood oxygen level, heart rate, respiratory rate, and pain score at six time points (0h, 6h, 12h, 24h, 36h, and 48h) after surgery, perform a difference calculation on each parameter at adjacent time points. The difference is then combined with the anesthesia method, surgery duration, analgesia pump use record, and first water drinking time at the corresponding time points. The risk scores for postoperative nausea and vomiting are then divided into three levels. The manifestations of nausea, vomiting, and loss of appetite within 48h of each level are time-indexed to generate a set of postoperative nursing intervention adjustment parameters.

[0047] Step 2: Based on the postoperative nausea and vomiting risk fluctuation record group, four parameters, namely, vomiting score, number of analgesic adjustments, ginger powder usage, and ginger slice application period, recorded every 4 hours within 48 hours after surgery, are extracted in chronological order to form a time series matrix. The matrix is ​​input into the long-short-term memory neural network structure for iterative calculation, and the node state transition structure in each sequence is extracted. A sequence difference graph is constructed according to the amplitude of state change within the time period. The jump position in the difference graph is then paired and annotated with the time point when the postoperative nausea and vomiting symptoms worsen. The time series corresponding trajectory of the intervention factors and the risk trend is established to obtain a time series graph mapping the nursing intervention and the postoperative nausea and vomiting risk.

[0048] Step 3: Based on the time series diagram of the mapping between the nursing intervention and the risk of postoperative nausea and vomiting, four parameters, namely, nursing frequency, ginger powder dosage, ginger slice replacement times, and nursing period coverage, are extracted at the key jump points. A two-dimensional matrix image is constructed, and the image is input into the convolutional neural network structure. The gradient response value change of the convolution kernel between each segment is calculated, and the parameter segments corresponding to the peak value of the convolution kernel response are identified. The segments are stratified and marked according to the frequency of occurrence of the segments. Intervention items with unstable gradient values ​​are eliminated, and intervention sequences with high frequency stability are sorted to obtain a combination set of nursing intervention segments with high adaptability.

[0049] Step 4: Based on the highly adaptable nursing intervention fragment combination set, extract the intervention structures that meet the following conditions: ginger powder is used twice a day, the total duration of ginger slice application is not less than 6 hours, and the first delay time of ginger application is no more than 2 hours. Perform difference matching between each group of records and the actual ginger powder dosage, each interval time, and the number of ginger slice changes in the patient's actual nursing log. Calculate the offset value vector of each group of parameters, multiply the offset value with the vomiting score of the corresponding time period, and obtain an intervention offset intensity map, thereby obtaining a nursing intervention offset intensity grade structure map.

[0050] Step 5: Based on the nursing intervention offset intensity grade structure diagram, the intervention record numbers with offset intensity greater than the upper quartile of the average offset are eliminated, and the corresponding intervention frequency, ginger powder dosage, and ginger slice application replacement frequency under the remaining numbers are extracted. The sequence path segments under the same number in the nursing intervention and postoperative nausea and vomiting risk mapping time series diagram are combined to perform time axis alignment, and a new sequence head is constructed according to the time interval in the aligned path segment. The position with the appropriate fitting degree in the sequence is used as the entry point to insert the remaining intervention items, reconstruct the optimized intervention arrangement sequence, and obtain the nursing intervention dynamic control path set.

[0051] Compared with the prior art, the advantages and positive effects of the present invention are:

[0052] 1. In this invention, by sequentially calculating key postoperative physiological parameters and combining them with medical history variables to construct a rate of change curve, the analysis of nausea and vomiting symptoms is based on continuous and quantifiable indicators, thereby improving the sensitivity of judging risk trends.

[0053] 2. In this invention, the risk change path is identified through a long-short-term memory neural network, and the influence of intervention intensity on the evolution trend of nausea and vomiting is accurately portrayed, achieving a phased prediction of the effect of nursing intervention. The sequence structure, execution frequency and intervention intensity of the intervention measures are extracted. Through the combined optimization mechanism between the adjustment coefficients, the nursing implementation plan is made more suitable for the individual's postoperative status and medication use, thereby improving the responsiveness and adaptability of the intervention implementation.

[0054] 3. In the present invention, the intervention timing, nursing task allocation and execution rhythm are structurally optimized through convolutional neural networks, and the intervention rhythm path is reorganized in combination with resource adaptation information to enhance the coverage density of intervention measures for risk periods, promote the integrated response of nursing resources in high-sensitivity intervals, and achieve synchronous optimization of task allocation and timing matching. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 is a system flow chart of the present invention;

[0056] Figure 2 Schematic diagram of the system framework of the present invention;

[0057] Figure 3 Schematic diagram of the method steps of the present invention. DETAILED DESCRIPTION

[0058] In order to make the purpose, 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 intended to limit the present invention.

[0059] See also Figure 1 The present invention provides a technical solution: an intelligent nursing intervention system for nausea and vomiting after lung surgery, comprising:

[0060] Obtaining the postoperative nursing intervention adjustment parameter set through the patient data processing module;

[0061] Obtain the postoperative nausea and vomiting nursing intervention prediction model through the postoperative risk assessment module;

[0062] Obtain individualized nursing intervention implementation plans through the individualized nursing optimization module;

[0063] Obtain nursing abnormality monitoring results through the nursing abnormality monitoring module;

[0064] Obtain the dynamic regulation pathway set of nursing intervention through the dynamic adjustment module of nursing plan;

[0065] Patient data processing module: Based on the key postoperative medical parameters retained by the hospital, calculate the blood oxygen level, heart rate, respiratory rate and pain score, combine the postoperative change rate of the patient's medical history, analyze the trend of postoperative nausea and vomiting, and calculate the nursing intervention adjustment coefficient based on the implementation of nursing interventions to obtain the postoperative nursing intervention adjustment parameter set;

[0066] Postoperative risk assessment module: Based on the postoperative nursing intervention adjustment parameter set, the impact of nursing intervention on the risk change of postoperative nausea and vomiting is calculated. The intervention adaptability of the nursing method is determined by combining the occurrence trend of postoperative nausea and vomiting and the nursing intervention response rate. A neural network model is generated using a long-short-term memory neural network. A prediction mapping of nursing adjustment to postoperative nausea and vomiting is established. Combined with the adjustment space of nursing measures, the adaptive nursing intervention intensity is calculated to obtain a postoperative nausea and vomiting nursing intervention prediction model.

[0067] Individualized nursing optimization module: Based on the postoperative nausea and vomiting nursing intervention prediction model, the module calculates the impact of the nursing intervention sequence on postoperative nausea and vomiting. Combined with the nursing intervention adjustment coefficient, the module optimizes the nursing intervention timing, frequency, and implementation intensity. The module also calculates the adaptability of nursing measures and optimizes the nursing intervention plan based on postoperative pain and analgesic use to obtain an individualized nursing intervention implementation plan.

[0068] Nursing abnormality monitoring module: Based on the individualized nursing intervention execution plan, it monitors nursing execution data, calculates nursing measure deviation values, screens abnormal nursing execution status, calculates the degree of abnormal impact based on the occurrence trend of postoperative nausea and vomiting, determines whether nursing abnormalities affect the stability of postoperative nausea and vomiting control, calculates the nursing abnormality risk level based on the nursing intervention adjustment coefficient, and obtains nursing abnormality monitoring results;

[0069] Nursing plan dynamic adjustment module: Based on the results of nursing abnormality monitoring, a convolutional neural network is used to adjust the nursing staff tasks, nursing processes and intervention timing, calculate the adaptability of nursing resources, and combine the postoperative nausea and vomiting nursing intervention prediction model to evaluate the adaptability of the adjusted nursing plan. Based on the adjustment space of nursing measures, the nursing intervention plan is optimized to ensure the accuracy of nursing execution and obtain a set of dynamic control paths for nursing intervention.

[0070] See also Figure 2 , the patient data processing module includes:

[0071] Postoperative parameter extraction submodule: Based on the key postoperative medical parameters retained by the hospital, it performs unified timestamp alignment and outlier screening for postoperative monitoring data, groups multi-channel data and calculates continuous fluctuation intervals, performs segmented averaging and extreme value identification for blood oxygen, heart rate, respiratory rate, and pain scores, and obtains a basic monitoring data set.

[0072] Medical history feature analysis submodule: Based on the basic monitoring data set, it extracts and cleans the patient's medical history text information fields, performs unified coding processing on abnormal records and maps them to structured data, cross-compares monitoring data with medical history features, extracts physiological fluctuation parameters within the postoperative time window and performs trend segmentation to obtain characteristic factors of pathological evolution;

[0073] Physiological trend identification submodule: Based on the characteristic factors of pathological evolution, it performs time-series mapping of nursing intervention time points and monitoring data nodes, compares the segments before and after intervention and marks the inflection points of physiological indicators, calculates the corresponding change rate of each intervention item and performs difference analysis, classifies trend segments according to the abnormal accumulation threshold of indicators, generates a nursing intervention response difference table, and obtains the postoperative nursing intervention adjustment parameter set;

[0074] The key postoperative medical parameters retained by the hospital include postoperative physiological indicators, ventilator parameters, anesthesia methods, medication types and dosages, and postoperative complications retained by the hospital;

[0075] Postoperative parameter extraction submodule: Based on the key postoperative medical parameters retained by the hospital, a sliding window anomaly detection method is used to align the monitoring data with unified timestamps and filter out outliers. The sliding window is set to 60 seconds and the window step is set to 5 seconds. The mean and standard deviation of blood oxygen, heart rate, respiratory rate, and pain score in each window are calculated respectively. The rule for identifying outliers is to eliminate samples whose difference between the current value and the window mean exceeds 3 times the standard deviation. Subsequently, a data channel grouping strategy is adopted for the filtered data. The monitoring channels are divided into two groups according to the record type: physiological parameter category and score category. The extreme difference value retrieval operation is performed in independent channels respectively. The fluctuation range is constructed for the maximum and minimum values ​​of 5 consecutive points in each channel. Then, the segmented mean calculation is performed on each channel, and the time period is divided into every 10 minutes. The segmented mean processing is performed on the blood oxygen, heart rate, respiratory rate, and pain score in each period, and the extreme value identification operation is performed to generate the basic monitoring data set.

[0076] Medical history feature parsing submodule: Based on the basic monitoring data set, a regular expression field extraction method is used for the patient's medical history text information. The text scanning operation is performed according to the keyword list, and the extraction results are cleaned. The punctuation marks are replaced and meaningless words are deleted through Python functions to construct structured text records. A unified coding mapping strategy is used to map disease nouns to code values. The mapping rule uses the disease name as the key and the standard ICD code as the value to construct a dict mapping table. The mapping table is then called to convert the text field into structured data. The structured data is then joined with the basic monitoring data set. The key-value pair binding method is used to connect the medical history item code with each physiological data record according to the patient number. The trend segmentation judgment function is used on the connected data to extract data within 24 hours after surgery. The data segments with physiological fluctuation range values ​​greater than 15% per hour are labeled, and the start and end positions of the segments with continuous values ​​greater than the threshold are extracted to generate pathological evolution characteristic factors.

[0077] Physiological trend identification submodule: Based on the pathological evolution characteristic factors, a time series mapping operation is performed on the nursing intervention time points and the monitoring data nodes. The key frame correspondence algorithm is used to perform intersection matching on the intervention record time list and the monitoring data time axis. The matching tolerance value is set to ±30 seconds. After mapping, the physiological indicator sequences are extracted from the 30 minutes before and 30 minutes after the intervention respectively. The inflection point identification function is performed. The local extreme value judgment condition is used, the detection interval is set to 5 sampling points, the first-order derivative difference is calculated, and the points with alternating positive and negative changes are determined as inflection points. They are marked in the original sequence as an inflection point set. The corresponding inflection point identification is performed once for each intervention number. Then, for each intervention segment, the difference between its before and after means is calculated and divided by the duration to obtain the change rate value. Pairwise difference calculation is performed on the change rate of all intervention items. The difference matrix is ​​constructed, and the above segments are classified using the segment clustering method. The change rate threshold interval is set to ±10%. The segments are labeled according to their affiliation. Finally, a nursing intervention response difference table is generated, and a postoperative nursing intervention adjustment parameter set is obtained.

[0078] See also Figure 2 , the postoperative risk assessment module includes:

[0079] Risk impact calculation submodule: Based on the postoperative nursing intervention adjustment parameter set, the time window between the intervention item parameters and the nausea and vomiting marker events is aligned, the interval change frequency of each parameter dimension is extracted and the intervention response span is calculated. After extracting the key influencing variables, multivariate weighting is performed to obtain the nursing intervention impact coefficient group;

[0080] Nursing Adaptation Judgment Submodule: Based on the nursing intervention impact coefficient group, a long-short-term memory neural network is used to reconstruct the time series of nausea and vomiting marker intervals, extract intervention execution items with high response rates and classify them into groups, construct an adaptation matrix for each execution group under different patient conditions, and screen high-frequency combinations to obtain the structural characteristics of nursing adaptability;

[0081] Intervention intensity optimization submodule: Based on the structural characteristics of nursing adaptability, the execution order of all intervention combination plans is rearranged and a transformation sequence list is established. The state response trajectory between intervention intensity and nausea and vomiting events is extracted, the intensity fluctuation amplitude and interval of each trajectory segment are extracted and sorted, and high-intensity response segment identification and aggregation are performed to obtain a nursing intervention prediction model for postoperative nausea and vomiting.

[0082] Risk impact calculation submodule: Based on the postoperative nursing intervention adjustment parameter set, the parameters corresponding to each intervention number are aligned in time windows. A sliding time synchronization matching algorithm is used, and the reference time axis is set as the nausea and vomiting mark event record table. The time window length is defined as 120 minutes, and the step length is 15 minutes. The intervention time point execution interval within the window is matched with the symptom mark time point according to the timestamp to generate the intervention response segment within the window. The change frequency of the intervention parameters within the segment is calculated within the window. The frequency is defined as the number of times the intervention intensity is greater than 1.2 times the mean of the previous period divided by the total number of interventions within the window. Subsequently, the maximum sustained response duration within the window length is calculated for all intervention numbers and used as the intervention response span. The intervention numbers with an average response span value greater than 90 minutes are selected as key variables. The multivariate weight scattering method is used to construct the feature weight vector of the selected variables in matrix form. The weight vector of each column variable is normalized, and the orthogonal projection value operation is performed on each row vector. The first 70% principal components are retained to form a combined intervention impact coefficient to generate a nursing intervention impact coefficient group.

[0083] Nursing adaptation judgment submodule: Based on the nursing intervention influence coefficient group, a long short-term memory neural network is used to reconstruct the time series of nausea and vomiting mark intervals. The network structure is set to two-layer stacked units, each layer contains 128 hidden nodes, and the input feature dimensions are intervention number, intervention intensity, and time point index. The learning rate is set to 0.001, the number of training rounds is 100 rounds, and the batch size is 32. After training, the response rate array output by the network is extracted with each intervention number as the sequence label. After sorting by response value, the intervention number with a response rate greater than 0.7 is selected as the high response group. The execution group set is constructed by classification according to the response number. The response difference of each intervention number in the group under the state variables of different patients is compared. A multidimensional adaptation matrix generation strategy is adopted to construct a two-dimensional matrix with intervention number as row and patient status label as column. The value is the average response rate. The combination with an average response rate greater than 0.65 is screened to generate the nursing adaptability structural characteristics.

[0084] Intervention intensity optimization submodule: Based on the structural characteristics of nursing adaptability, all intervention combination plans are reordered, and the full permutation sequence generation method is used. Taking the number of intervention numbers in the combination as the benchmark, an ordered index number generation is performed for each permutation combination to construct a transformation sequence list. The execution order of each intervention number in each sequence on the time axis is mapped, and the state response trajectory between the intervention intensity value on the corresponding time axis and the nausea and vomiting mark event time point is extracted to complete the time series smoothing processing. The intensity fluctuation amplitude is calculated for each continuous interval in the trajectory by subtracting the sliding mean of the past four points from the current value, with a sliding window of 4 and a step size of 1. At the same time, the trajectory interval time is extracted, the time index is subtracted, and a fluctuation and interval data list is generated. The segments with amplitude values ​​greater than the top 75 percentile values ​​are marked as high-response segments, and the marked segments are numbered and aggregated to generate a postoperative nausea and vomiting nursing intervention prediction model.

[0085] Long short-term memory neural network, according to the formula:

[0086]

[0087] Where: R is the patient's nausea and vomiting response rate, α i is the impact coefficient of the intervention execution item, x i is the characteristic of the intervention execution item, f(x i ,w i ,t i ,d i ) is the response function of the long short-term memory network to the intervention characteristics and parameters, n is the number of intervention execution items, w i is the weight coefficient of the intervention execution item, t i is the individual patient's physiological state parameter, d i Provide patient treatment history data;

[0088] Implementation process: First, the patient's nausea and vomiting response rate R needs to be calculated by comprehensively evaluating the effects of different nursing intervention items. For each intervention item x i , determine the influence coefficient α according to the characteristics of the intervention item i , reflecting the specific impact of the intervention item on the patient's nausea and vomiting symptoms. Then the long short-term memory network combines the characteristics of each intervention item x through its unique temporal processing ability. i and the patient's individual physiological state parameter t i , treatment history data d i , and the weight coefficient w of the intervention term i, the intervention items are responded to through the deep learning model, the effectiveness of each intervention item is calculated, and finally the overall response rate R of the patient is obtained by weighted summation of the response values ​​of all intervention items, which reflects the adaptive effect of the current nursing strategy for the nausea and vomiting symptoms of patients after lung surgery, and helps to further optimize the nursing intervention strategy.

[0089] See also Figure 2 , the personalized care optimization module includes:

[0090] Intervention sequence evaluation submodule: Based on the postoperative nausea and vomiting nursing intervention prediction model, all combinations of nursing action sequences are generated and sorted by time. Nausea and vomiting reaction nodes within the combination are extracted and corresponding distribution maps are constructed. The response duration of key action nodes and concurrent position marking are recorded. High-frequency response segments are extracted and conflicting intervention positions between sequences are screened to obtain the impact characteristics of the nursing sequence.

[0091] Nursing rhythm adjustment submodule: Based on the impact characteristics of the nursing sequence, it performs numerical mapping between the nursing intervention adjustment coefficient and the rhythm parameter list, extracts the frequency fluctuation values ​​within the adjustment interval and reorganizes the distribution blocks, clusters the intervention items in the frequency jump segment and analyzes the rhythm continuity, extracts the intervention intensity sequence in the frequency compression interval and sorts it by rhythm offset to obtain the intervention rhythm adaptation structure;

[0092] The program structure matching submodule: Based on the intervention rhythm adaptation structure, multivariate alignment of postoperative pain records and analgesic use data is performed, and the intersection blocks of high-value intervention intensity points and rising drug concentration segments are extracted. The time overlap rate and drug action overlap positions in the nursing structure are analyzed, and the intervention path is selected according to the priority and the complete structure group is output to obtain the individualized nursing intervention implementation plan;

[0093] Intervention sequence evaluation submodule: Based on the postoperative nausea and vomiting nursing intervention prediction model, a full permutation and combination generation algorithm is used for the nursing measure number list. The parameter r is the sequence length, ranging from 3 to 6, to construct all possible combination schemes. Then, the execution time field in each item in the generated combination is sorted in ascending time order to generate a time sequence list. All time nodes that intersect with the vomiting mark event timestamp in each combination are extracted, and a response node index set is constructed. The frequency of occurrence of each response node is statistically analyzed, and a histogram is constructed as a distribution representation of the intervention event. Then, the response duration is recorded for the intervention items within 10 minutes before the vomiting mark node in each sequence. Concurrent marking is performed on continuous intervention items with a timestamp difference of less than 5 minutes. The continuous response segment with the highest frequency in the sequence is extracted. The set intersection function is called for screening the positions with overlapping time but different intervention numbers in multiple sequences to generate the nursing sequence impact characteristics.

[0094] Nursing rhythm adjustment submodule: Based on the impact characteristics of the nursing sequence, a parameter mapping matrix generation method is adopted, and the mapping benchmark is set as the intervention adjustment coefficient interval [0.5, 2.0] and the rhythm parameter interval [10 minutes, 60 minutes]. The mapping values ​​are bound to the intervention numbers respectively, and interpolation operations are performed to generate a complete rhythm parameter list. The positions in the list where the frequency changes by more than 30% appear three times in a row are extracted as frequency fluctuation segments. The segments are grouped using the interval reorganization function, and the frequency jump values ​​within the group are calculated. At the same time, the K-means clustering algorithm is used to cluster the intervention numbers corresponding to the jump segments. The number of clusters is set to 3, and the initial center is the first three values ​​of the jump amplitude. After clustering, the intervention items are grouped according to the label, the rhythm standard deviation is calculated, the continuous stable segment is judged, and the intervention intensity values ​​corresponding to the continuous segments with a frequency standard deviation less than 0.1 are extracted to generate the intervention rhythm adaptation structure;

[0095] Plan structure matching submodule: Based on the intervention rhythm adaptation structure, a multivariate sliding alignment algorithm was used to construct a sliding window for the analgesic concentration curve and postoperative pain score data. The window length was set to 30 minutes and the step size was set to 10 minutes. The drug concentration change value and the mean pain score within each window were jointly recorded to establish a sliding statistical array. Then, the nodes in the intervention intensity list in the intervention rhythm with a value greater than 1.5 times the mean were defined as high-intensity points. The time index corresponding to the high-intensity point was extracted and the above index was time-aligned with the drug concentration rising interval. The parameters were set to backward matching mode and the tolerance value was set to ±5 minutes to obtain cross-blocks. Subsequently, the overlap between the intervention execution time and the high-value drug concentration segment in each nursing structure was analyzed. The overlap rate was calculated by the union length, and the blocks with an overlap rate greater than 0.6 were marked as priority path candidate segments. The blocks were sorted from high to low according to the overlap rate, and the top three path segments were extracted to form an output sequence to generate an individualized nursing intervention implementation plan.

[0096] See also Figure 2 , the nursing abnormality monitoring module includes:

[0097] Execution data comparison submodule: Based on the individualized nursing intervention execution plan, it synchronizes the nursing execution record time nodes with the execution intensity data, extracts the actual start and end time of the intervention task and calculates the duration difference, identifies the execution delay and advance nodes and marks the variation form, constructs the execution distribution matrix between the plan and the actual, and obtains the nursing execution deviation parameters;

[0098] Abnormal deviation identification submodule: Based on the nursing execution deviation parameters, it performs connectivity division and intensity layer extraction of the deviation distribution segment, extracts the functional labels of the intervention items to which the deviation nodes belong and records the continuous mismatch intervals, screens the span and fluctuation trend of the abnormal accumulation segment, constructs a combined identification of the deviation form and nursing events, and obtains the characteristics of the abnormal nursing status;

[0099] Intervention stability determination submodule: Based on the abnormal nursing status characteristics, combined with the occurrence trend of postoperative nausea and vomiting and the nursing intervention adjustment coefficient, the state hierarchical mapping is performed. The control parameter blocks corresponding to the abnormal labels are extracted and the impact distribution analysis is performed. The stability decline event segments are constructed and classified into the abnormal intensity matrix. The impact score table of each nursing task is generated to obtain the nursing abnormality monitoring results.

[0100] Execution data comparison submodule: Based on the individualized nursing intervention execution plan, a bidirectional time series alignment algorithm is used to synchronously match the actual execution time records in the nursing log with the planned execution time series. The matching tolerance is set to plus or minus 2 minutes, and the matching direction is the two-way nearest time. The actual start time, planned start time, actual end time, and planned end time of each intervention task are extracted from the matched data records. The starting deviation is calculated by subtracting the planned start time from the actual start time, and the ending deviation is calculated by subtracting the planned end time from the actual end time. The duration deviation is equal to the starting deviation plus the ending deviation and divided by 2 as the task duration deviation. Task records with a starting deviation greater than 5 minutes are marked as delayed, and records with a starting deviation less than minus 5 minutes are marked as early. A two-dimensional matrix is ​​constructed for all tasks according to the intervention number, with rows representing the intervention number and columns representing the task time and deviation value. The value of each cell in the matrix is ​​the offset in seconds of the current intervention task, and the nursing execution deviation parameter is generated;

[0101] Abnormal deviation identification submodule: Based on the nursing execution deviation parameters, a sliding window connectivity segmentation method is used. After arranging the task records in chronological order, the window length is set to 3 records and the step length is 1 record. The difference between consecutive deviation values ​​in the window is calculated. If the absolute value of the difference is greater than 10 minutes, the current window is marked as a breakpoint window. The data segments between the breakpoints are defined as continuous offset segments. The 25%, 50%, and 75% percentile values ​​of the deviation values ​​in each segment are extracted. The three intervals are divided into low-intensity segment, medium-intensity segment, and high-intensity segment. The corresponding intervention task number is extracted for each abnormal segment, and the functional label corresponding to the number is obtained from the nursing plan. Abnormal time periods with the same task number that appear consecutively are grouped together. The start and end times of the group are recorded and labeled. Then, the linear trend analysis method is used to divide the trend of each abnormal segment, extracting the trend change direction and fluctuation degree. If the fluctuation amplitude is greater than the preset threshold, it is classified as a fluctuating abnormal segment. Combination identification items are constructed by combining the deviation level and the functional label. Each item is composed of the deviation level and the functional label. Frequency statistics of all combination identification items are performed to generate nursing abnormal status features.

[0102] Intervention stability judgment submodule: Based on the characteristics of abnormal nursing status, combined with the trend sequence of postoperative nausea and vomiting and the nursing intervention adjustment coefficient table, the state mapping analysis method is used to construct a three-dimensional structure. The dimensions are abnormal label, abnormal start and end time, and intervention adjustment coefficient interval. The nausea and vomiting score sequence of each label in the specified time period and specific adjustment coefficient interval is analyzed by mean. The time points with score changes greater than 2 points in the abnormal segment are extracted as potential unstable nodes. Nodes with continuous change trends are classified as a type of stability decline fragments. Each type of fragment is arranged in chronological order and grouped together. For each group, the associated intervention task number and functional classification are extracted to construct a two-dimensional matrix. The rows represent the intervention task number and the columns represent the stability decline event number. Each cell records the score change value of the task number in the event segment multiplied by the adjustment coefficient of the segment. The values ​​of each row are summarized to construct the impact score table of each nursing task to obtain the nursing abnormality monitoring results.

[0103] See also Figure 2 , the dynamic adjustment module of nursing plan includes:

[0104] Task structure reorganization submodule: Based on the results of nursing abnormality monitoring, a convolutional neural network is used to align the time points of abnormal event markings with the nursing task execution logs, extract the task start and end times and the executor identification, and establish an event mapping matrix. It also detects time overlaps between tasks and screens conflicts between nursing role responsibilities. It reconstructs the task sequence and updates the task connection node list to obtain the intervention task adjustment structure.

[0105] Resource load assessment submodule: Based on the intervention task adjustment structure, the task time period is split and the execution frequency is calculated. The nursing position scheduling information and the list of schedulable durations are extracted and matched with the task demand sequence. The load diagram of the task-intensive section is constructed and the peak location of resource utilization is identified. The high-intensity intervention blocks are marked in combination with the postoperative nausea and vomiting nursing intervention prediction model to obtain the resource adaptation parameter model.

[0106] Plan path optimization submodule: Based on the resource adaptation parameter model, it extracts the intervention component set in the nursing measure adjustment space and rearranges the nodes, matches the task sorting list with the resource interval diagram and establishes a time conflict evaluation table, screens low-conflict permutations and combinations in the intervention path and marks the node numbers, outputs the sequential adaptation task execution roadmap, and obtains the nursing intervention dynamic control path set;

[0107] Task structure reorganization submodule: Based on the results of nursing abnormality monitoring, a convolutional neural network is used to align the abnormal event marking time points with the nursing task execution logs one by one, and a three-dimensional input matrix is ​​constructed. The input channels are three types of values: abnormal event label, nursing task number, and execution timestamp. The convolution kernel size is set to 1×3, the stride is set to 1, and the number of channels is 32. The input matrix is ​​slid in the time dimension to extract the feature window. The task start time, end time and executor number are arranged in sequence in the window. The maximum and minimum differences are calculated for the time span in each window, and the window output is mapped to the task node sequence to generate a task time block list. All task time blocks are compared pairwise to determine whether there is an intersection between the time intervals. If there is overlap and the corresponding executor numbers are the same, they are marked as responsibility conflict nodes. All marked nodes are numbered and summarized, and a task connection graph is established. All edges in the graph are sorted from small to large by timestamp. After updating the connection order, the intervention task adjustment structure is generated.

[0108] Resource load assessment submodule: Based on the intervention task adjustment structure, the task timeline execution interval is split using segment division and frequency mapping algorithms. Each intervention task time period is divided into two equal segments. The actual execution times in each segment are recorded and counted into the frequency vector. The frequency sequence of each task is constructed. The frequency sequence is matched with the daily scheduling records of the nursing position. The schedulable time periods of each position are extracted and a position time period table is constructed. After matching each task frequency sequence with the position time period table, a position time occupancy list is established. The task density is counted by hourly segment and recorded as the task density distribution value. All distribution values ​​are sorted vertically and the maximum density point is taken to construct the load peak marking area. Compared with the high intervention intensity task numbers marked in the postoperative nausea and vomiting nursing intervention prediction model, the task sequence coinciding with the load peak time period is extracted and a resource occupancy mapping table is established. The task number is associated with its occupied time block and mapped to the nursing position number to generate a resource adaptation parameter model.

[0109] Plan path optimization submodule: Based on the resource adaptation parameter model, the sequence rearrangement and conflict screening algorithm is used to rearrange the node order of the intervention component set in the nursing measure adjustment space, and the intervention measure number and the original task execution time node are combined into a double-index structure table. After renumbering the intervention node set, an optional path set is generated, and a task sorting list is constructed. The task node time period in the sorting list is matched point by point with the position available time period in the resource interval diagram. It is calculated whether there is an intersection mark between each task node and the resource time interval. If the start time or end time is in the unschedulable section of the position, it is marked as a conflict point. A time conflict evaluation table is constructed, and the conflict status of each task node and its corresponding position is listed. Conflict screening operations are performed on all permutations and combinations in the task path, and the path combination with the least number of conflicts is extracted and the new sequence number of each intervention node is marked. A task path structure diagram after sequence adjustment is established, and the diagram is output as a nursing intervention dynamic control path set.

[0110] Convolutional neural network, according to the formula:

[0111]

[0112] Where: E t β is the alignment between the task execution time point and the abnormal event marking time point, i The weight coefficient for each task, y i is the task feature, g(y i ,θ i ,δ i ,s i ) is the response function of the convolutional neural network to the task characteristics and other parameters, n is the number of tasks, θ i is the convolution kernel parameter of the convolution layer, δ i is the role identifier of the task executor, s i is the strength of the relationship between the task and the abnormal event;

[0113] Execution process: First, each nursing task y is trained through a convolutional neural network. i The task features include task type, execution time, and execution conditions, which help to identify the potential correlation between tasks and abnormal events. i Represents the weight coefficient of each task, which is used to adjust the importance of the task to time alignment, ensuring that the key tasks prioritize matching the time points of abnormal events. The convolution kernel parameter θ of the convolution layer i Control the network's response to task features. The network optimizes θ during training. i Improve the accurate recognition of task characteristics and the role identification of task performersδ i This further enhances the judgment of the priority of different nurses in performing tasks, and finally the strength of the relationship between tasks and abnormal events.i Helps determine the matching degree between tasks and abnormal events, s i The larger the value, the stronger the correlation between the task and the abnormal event, and the task should be adjusted first. Through the comprehensive analysis of the task characteristics and these parameters by the convolutional neural network, the alignment degree E of the task execution time point and the abnormal event marking time point is obtained. t , providing an accurate basis for task adjustment and structural reorganization.

[0114] See also Figure 3 An intelligent nursing intervention method for nausea and vomiting after lung surgery is provided. The intelligent nursing intervention method for nausea and vomiting after lung surgery is implemented based on the intelligent nursing intervention system for nausea and vomiting after lung surgery, and includes the following steps:

[0115] Step 1: Based on the numerical changes in postoperative blood oxygen level, heart rate, respiratory rate, and pain score at six time points (0h, 6h, 12h, 24h, 36h, and 48h) after surgery, perform a difference calculation on each parameter at adjacent time points. The difference is then combined with the anesthesia method, surgery duration, analgesia pump use record, and first water drinking time at the corresponding time points. The risk scores for postoperative nausea and vomiting are then divided into three levels. The manifestations of nausea, vomiting, and loss of appetite within 48h of each level are time-indexed to generate a set of postoperative nursing intervention adjustment parameters.

[0116] Step 2: Based on the postoperative nausea and vomiting risk fluctuation record group, four parameters, namely, vomiting score, number of analgesic adjustments, ginger powder usage, and ginger slice application period, recorded every 4 hours within 48 hours after surgery, were extracted in chronological order to form a time series matrix. The matrix was input into the long-short-term memory neural network structure for iterative calculation. The node state transition structure in each sequence was extracted, and a sequence difference graph was constructed according to the amplitude of state change within the time period. The jump position in the difference graph was then paired and annotated with the time point when the postoperative nausea and vomiting symptoms worsened. The time series corresponding trajectory of the intervention factors and the risk trend was established to obtain a time series graph mapping the nursing intervention and the postoperative nausea and vomiting risk.

[0117] Step 3: Based on the time series diagram mapping nursing interventions to the risk of postoperative nausea and vomiting, four parameters were extracted at key transition points: nursing frequency, ginger powder dosage, ginger slice replacement times, and nursing period coverage. A two-dimensional matrix image was constructed and input into the convolutional neural network structure. The gradient response value changes of the convolution kernel between each segment were calculated. The parameter segments corresponding to the peak of the convolution kernel response were identified and labeled according to the frequency of occurrence of the segments. Intervention items with unstable gradient values ​​were eliminated, and intervention sequences with high frequency stability were sorted to obtain a combination set of nursing intervention segments with high adaptability.

[0118] Step 4: Based on the set of highly adaptable nursing intervention fragment combinations, extract the intervention structures that meet the following conditions: ginger powder is used twice a day, the total duration of ginger slice application is not less than 6 hours, and the first delay time of ginger application is no more than 2 hours. Difference-pair the records of each group with the actual ginger powder dosage, each interval time, and the number of ginger slice changes in the patient's actual nursing log. Calculate the offset value vector of each group of parameters, multiply the offset value with the vomiting score of the corresponding time period, and obtain the intervention offset intensity map, thereby obtaining the nursing intervention offset intensity level structure map.

[0119] Step 5: Based on the nursing intervention offset intensity grade structure diagram, the intervention record numbers with offset intensity greater than the upper quartile of the average offset were eliminated, and the corresponding intervention frequency, ginger powder dosage, and ginger slice patch replacement frequency under the remaining numbers were extracted. The sequence path segments under the same numbers in the time series diagram of the nursing intervention and postoperative nausea and vomiting risk mapping were aligned on the time axis. A new sequence head was constructed based on the time interval in the aligned path segment, and the position with the appropriate fitting degree in the sequence was used as the entry point to insert the remaining intervention items, reconstruct the optimized intervention arrangement sequence, and obtain the dynamic control path set of the nursing intervention.

[0120] The above are merely preferred embodiments of the present invention and do not limit the present invention in any other form. Any technician familiar with the profession may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall still fall within the scope of protection of the technical solution of the present invention.

Claims

1. An intelligent nursing intervention system for nausea and vomiting after lung surgery, characterized in that: include: Obtaining the postoperative nursing intervention adjustment parameter set through the patient data processing module; Obtain the postoperative nausea and vomiting nursing intervention prediction model through the postoperative risk assessment module; Obtain individualized nursing intervention implementation plans through the individualized nursing optimization module; Obtain nursing abnormality monitoring results through the nursing abnormality monitoring module; The dynamic regulation path set of nursing intervention is obtained through the dynamic adjustment module of nursing plan.

2. The intelligent nursing intervention system for nausea and vomiting after lung surgery according to claim 1, characterized in that: The system comprises: Patient data processing module: Based on the key postoperative medical parameters retained by the hospital, calculate the blood oxygen level, heart rate, respiratory rate and pain score, combine the postoperative change rate of the patient's medical history, analyze the trend of postoperative nausea and vomiting, and calculate the nursing intervention adjustment coefficient based on the implementation of nursing interventions to obtain the postoperative nursing intervention adjustment parameter set; Postoperative risk assessment module: Based on the postoperative nursing intervention adjustment parameter set, the impact of nursing intervention on the risk change of postoperative nausea and vomiting is calculated. In combination with the occurrence trend of postoperative nausea and vomiting and the nursing intervention response rate, the intervention adaptability of the nursing method is determined. A neural network model is generated using a long-short-term memory neural network. A prediction mapping of nursing adjustment to postoperative nausea and vomiting is established. In combination with the nursing measure adjustment space, the adaptive nursing intervention intensity is calculated to obtain a postoperative nausea and vomiting nursing intervention prediction model. Individualized nursing optimization module: Based on the postoperative nausea and vomiting nursing intervention prediction model, the module calculates the impact of the nursing intervention sequence on postoperative nausea and vomiting. Combined with the nursing intervention adjustment coefficient, the module optimizes the nursing intervention timing, frequency, and implementation intensity. The module also calculates the adaptability of nursing measures. Combined with postoperative pain and analgesic use, the module optimizes the nursing intervention plan and obtains an individualized nursing intervention implementation plan. Nursing abnormality monitoring module: Based on the individualized nursing intervention execution plan, monitor nursing execution data, calculate nursing measure deviation values, screen abnormal nursing execution status, calculate the degree of abnormal impact based on the occurrence trend of postoperative nausea and vomiting, determine whether nursing abnormalities affect the stability of postoperative nausea and vomiting control, calculate nursing abnormality risk levels based on the nursing intervention adjustment coefficient, and obtain nursing abnormality monitoring results; Nursing plan dynamic adjustment module: Based on the abnormal nursing monitoring results, a convolutional neural network is used to adjust the nursing staff tasks, nursing processes and intervention timing, calculate the adaptability of nursing resources, and combine the postoperative nausea and vomiting nursing intervention prediction model to evaluate the adaptability of the adjusted nursing plan. Based on the adjustment space of nursing measures, the nursing intervention plan is optimized to ensure the accuracy of nursing execution and obtain a set of dynamic control paths for nursing intervention.

3. The intelligent nursing intervention system for nausea and vomiting after lung surgery according to claim 1, characterized in that: The patient data processing module includes: Postoperative parameter extraction submodule: Based on the key postoperative medical parameters retained by the hospital, it performs unified timestamp alignment and outlier screening for postoperative monitoring data, groups multi-channel data and calculates continuous fluctuation intervals, performs segmented averaging and extreme value identification for blood oxygen, heart rate, respiratory rate, and pain scores, and obtains a basic monitoring data set. Medical history feature analysis submodule: Based on the basic monitoring data set, it extracts and cleans the patient's medical history text information fields, performs unified coding processing on abnormal records and maps them to structured data, cross-compares monitoring data with medical history features, extracts physiological fluctuation parameters within the postoperative time window and performs trend segmentation to obtain pathological evolution characteristic factors; Physiological trend identification submodule: Based on the pathological evolution characteristic factors, it performs time-series mapping of nursing intervention time points and monitoring data nodes, compares the segments before and after the intervention and marks the inflection points of physiological indicators, calculates the corresponding change rate of each intervention item and performs difference analysis, classifies trend segments according to the abnormal accumulation threshold of indicators, generates a nursing intervention response difference table, and obtains the postoperative nursing intervention adjustment parameter set; The key postoperative medical parameters retained by the hospital include postoperative physiological indicators, ventilator parameters, anesthesia methods, types and dosages of medications, and postoperative complications retained by the hospital.

4. The intelligent nursing intervention system for nausea and vomiting after lung surgery according to claim 1, characterized in that: The postoperative risk assessment module includes: Risk impact calculation submodule: Based on the postoperative nursing intervention adjustment parameter set, the time window between the intervention item parameters and the nausea and vomiting marker events is aligned, the interval change frequency of each parameter dimension is extracted and the intervention response span is calculated, and after extracting the key influencing variables, multivariate weighting is performed to obtain the nursing intervention impact coefficient group; Nursing Adaptation Judgment Submodule: Based on the nursing intervention influence coefficient group, a long-short-term memory neural network is used to reconstruct the time series of nausea and vomiting marking intervals, extract intervention execution items with high response rates and classify them into groups, construct an adaptation matrix for each execution group under different patient conditions, and screen high-frequency combinations to obtain nursing adaptability structural characteristics; Intervention intensity optimization submodule: Based on the nursing adaptability structural characteristics, the execution order of all intervention combination plans is rearranged and a transformation sequence list is established. The state response trajectory between intervention intensity and nausea and vomiting events is extracted, the intensity fluctuation amplitude and interval of each trajectory segment are extracted and sorted, and the high-intensity response segment identification is aggregated to obtain a postoperative nausea and vomiting nursing intervention prediction model.

5. The intelligent nursing intervention system for nausea and vomiting after lung surgery according to claim 2, characterized in that: The long short-term memory neural network is based on the formula: Where: R is the patient's nausea and vomiting response rate, α i is the impact coefficient of the intervention execution item, x i is the characteristic of the intervention execution item, f(x i ,w i ,t i ,d i ) is the response function of the long short-term memory network to the intervention characteristics and parameters, n is the number of intervention execution items, w i is the weight coefficient of the intervention execution item, t i is the individual patient's physiological state parameter, d i Provide patient treatment history data.

6. The intelligent nursing intervention system for nausea and vomiting after lung surgery according to claim 2, characterized in that: The individualized care optimization module includes: Intervention sequence evaluation submodule: Based on the postoperative nausea and vomiting nursing intervention prediction model, all combinations of nursing action sequences are generated and sorted by time, nausea and vomiting reaction nodes within the combination are extracted and corresponding distribution maps are constructed, response durations of key action nodes are recorded and concurrent positions are marked, high-frequency response segments are extracted, and intervention conflict positions between sequences are screened to obtain nursing sequence impact characteristics; Nursing rhythm adjustment submodule: Based on the influencing characteristics of the nursing sequence, it performs numerical mapping between the nursing intervention adjustment coefficient and the rhythm parameter list, extracts the frequency fluctuation value within the adjustment interval and reorganizes the distribution block, clusters the intervention items in the frequency jump segment and analyzes the rhythm continuity, extracts the intervention intensity sequence in the frequency compression interval and sorts it by rhythm offset to obtain the intervention rhythm adaptation structure; Program structure matching submodule: Based on the intervention rhythm adaptation structure, multivariate alignment of postoperative pain records and analgesic usage data is performed, the intersection blocks of high-value points of intervention intensity and rising drug concentration segments are extracted, the time overlap rate and the overlap position of drug effects in the nursing structure are analyzed, the intervention path is selected according to priority and the complete structure group is output to obtain an individualized nursing intervention execution plan.

7. The intelligent nursing intervention system for nausea and vomiting after lung surgery according to claim 2, characterized in that: The nursing abnormality monitoring module includes: Execution data comparison submodule: Based on the individualized nursing intervention execution plan, the nursing execution record time nodes are synchronously matched with the execution intensity data, the actual start and end time of the intervention task are extracted and the duration difference is calculated, the execution delay and advance nodes are identified and the variation forms are marked, the execution distribution matrix between the plan and the actual is constructed, and the nursing execution deviation parameters are obtained; Abnormal deviation identification submodule: Based on the nursing execution deviation parameters, it performs connectivity division and intensity layer extraction of the deviation distribution segment, extracts the functional labels of the intervention items to which the deviation nodes belong and records the continuous mismatch intervals, screens the span and fluctuation trend of the abnormal accumulation segment, constructs a combined identification of the deviation form and nursing events, and obtains the characteristics of the abnormal nursing status; Intervention stability judgment submodule: Based on the abnormal nursing state characteristics, combined with the occurrence trend of postoperative nausea and vomiting and the nursing intervention adjustment coefficient, state hierarchical mapping is performed, the control parameter blocks corresponding to the abnormal labels are extracted and the impact distribution analysis is performed, the stability decline event segments are constructed and classified into the abnormal intensity matrix, and the impact score table of each nursing task is generated to obtain the nursing abnormality monitoring results.

8. The intelligent nursing intervention system for nausea and vomiting after lung surgery according to claim 2, characterized in that: The nursing plan dynamic adjustment module includes: Task structure reorganization submodule: Based on the nursing abnormality monitoring results, a convolutional neural network is used to align the abnormal event marking time points with the nursing task execution logs, extract the task start and end times and the executor identification and establish an event mapping matrix, perform time overlap detection between tasks and screen for conflicts in nursing role responsibilities, reconstruct the task sequence and update the task connection node list to obtain the intervention task adjustment structure; Resource load assessment submodule: Based on the intervention task adjustment structure, the task time period is split and the execution frequency is calculated. The nursing position scheduling information and the list of schedulable durations are extracted and matched with the task demand sequence. The task-intensive section load diagram is constructed and the resource utilization peak position is identified. The high-intensity intervention blocks are marked in combination with the postoperative nausea and vomiting nursing intervention prediction model to obtain the resource adaptation parameter model. Plan path optimization submodule: Based on the resource adaptation parameter model, the intervention component set in the nursing measure adjustment space is extracted and the nodes are rearranged, the task sorting list is matched with the resource interval diagram and a time conflict evaluation table is established, the low-conflict permutations and combinations in the intervention path are screened and the node numbers are marked, the sequential adaptation task execution roadmap is output, and the nursing intervention dynamic control path set is obtained.

9. The intelligent nursing intervention system for nausea and vomiting after lung surgery according to claim 2, characterized in that: The convolutional neural network is based on the formula: Where: E t β is the alignment between the task execution time point and the abnormal event marking time point, i The weight coefficient for each task, y i is the task feature, g(y i ,θ i ,δ i ,s i ) is the response function of the convolutional neural network to the task characteristics and other parameters, n is the number of tasks, θ i is the convolution kernel parameter of the convolution layer, δ i is the role identifier of the task executor, s i is the strength of the relationship between tasks and abnormal events.

10. An intelligent nursing intervention method for nausea and vomiting after lung surgery, characterized in that: The intelligent nursing intervention system for nausea and vomiting after lung surgery according to any one of claims 1 to 9 comprises the following steps: Step 1: Based on the numerical changes in postoperative blood oxygen level, heart rate, respiratory rate, and pain score at six time points (0h, 6h, 12h, 24h, 36h, and 48h) after surgery, perform a difference calculation on each parameter at adjacent time points. The difference is then combined with the anesthesia method, surgery duration, analgesia pump use record, and first water drinking time at the corresponding time points. The risk scores for postoperative nausea and vomiting are then divided into three levels. The manifestations of nausea, vomiting, and loss of appetite within 48h of each level are time-indexed to generate a set of postoperative nursing intervention adjustment parameters. Step 2: Based on the postoperative nausea and vomiting risk fluctuation record group, four parameters, namely, vomiting score, number of analgesic adjustments, ginger powder usage, and ginger slice application period, recorded every 4 hours within 48 hours after surgery, are extracted in chronological order to form a time series matrix. The matrix is ​​input into the long-short-term memory neural network structure for iterative calculation, and the node state transition structure in each sequence is extracted. A sequence difference graph is constructed according to the amplitude of state change within the time period. The jump position in the difference graph is then paired and annotated with the time point when the postoperative nausea and vomiting symptoms worsen. The time series corresponding trajectory of the intervention factors and the risk trend is established to obtain a time series graph mapping the nursing intervention and the postoperative nausea and vomiting risk. Step 3: Based on the time series diagram of the mapping between the nursing intervention and the risk of postoperative nausea and vomiting, four parameters, namely, nursing frequency, ginger powder dosage, ginger slice replacement times, and nursing period coverage, are extracted at the key jump points. A two-dimensional matrix image is constructed, and the image is input into the convolutional neural network structure. The gradient response value change of the convolution kernel between each segment is calculated, and the parameter segments corresponding to the peak value of the convolution kernel response are identified. The segments are stratified and marked according to the frequency of occurrence of the segments. Intervention items with unstable gradient values ​​are eliminated, and intervention sequences with high frequency stability are sorted to obtain a combination set of nursing intervention segments with high adaptability. Step 4: Based on the highly adaptable nursing intervention fragment combination set, extract the intervention structures that meet the following conditions: ginger powder is used twice a day, the total duration of ginger slice application is not less than 6 hours, and the first delay time of ginger application is no more than 2 hours. Perform difference matching between each group of records and the actual ginger powder dosage, each interval time, and the number of ginger slice changes in the patient's actual nursing log. Calculate the offset value vector of each group of parameters, multiply the offset value with the vomiting score of the corresponding time period, and obtain an intervention offset intensity map, thereby obtaining a nursing intervention offset intensity grade structure map. Step 5: Based on the nursing intervention offset intensity grade structure diagram, the intervention record numbers with offset intensity greater than the upper quartile of the average offset are eliminated, and the corresponding intervention frequency, ginger powder dosage, and ginger slice application replacement frequency under the remaining numbers are extracted. The sequence path segments under the same number in the nursing intervention and postoperative nausea and vomiting risk mapping time series diagram are combined to perform time axis alignment, and a new sequence head is constructed according to the time interval in the aligned path segment. The position with the appropriate fitting degree in the sequence is used as the entry point to insert the remaining intervention items, reconstruct the optimized intervention arrangement sequence, and obtain the nursing intervention dynamic control path set.

Images