Intelligent nursing decision support system for patients undergoing cardiac intervention

Abstract

The present application relates to big data mining technical field, specifically for the intelligent nursing decision support system for patients of cardiology department intervention operation, the system includes: data analysis module, feature calculation module, stress discrimination module, recovery analysis module, decision optimization module.In the present application, through the synchronous analysis of the intervention equipment running state and the multi-source physiological signal, the time boundary identification of the intervention stage is realized, and the stable physiological response feature sequence is constructed in the continuous stage, so that the change trend of electrocardiogram and blood pressure can form a comparable overall relationship, thereby accurately reflecting the myocardial load state and stage difference, and the recovery behavior after the load is ended is associated, so that the nursing instruction can be adjusted in sequence and interval according to the continuous change trend, effectively reducing the instruction conflict between stages, and improving the stable performance of nursing decision in time sequence consistency and individual adaptability.

Description

Technical Field

[0001] This invention relates to the field of big data mining technology, and in particular to an intelligent nursing decision support system for patients undergoing interventional cardiology surgery. Background Technology

[0002] Big data mining technology refers to the collection, processing, and analysis of large amounts of structured and unstructured data to discover potential patterns, trends, and directions, thereby providing data-driven decision support. Core aspects of this field include data acquisition, data cleaning, data analysis, data mining algorithms such as classification, clustering, regression analysis, and association rule mining, as well as the application and optimization of machine learning models. These technologies enable the extraction of valuable information from massive amounts of data, providing precise decision-making support for various industries, and are widely used in finance, healthcare, manufacturing, and many other fields.

[0003] Traditional intelligent nursing decision support systems for patients undergoing interventional cardiology procedures utilize intelligent information systems that combine patient physiological data, surgical records, and historical medical information to assist nursing staff in making decisions during interventional cardiology surgeries. These systems monitor patient vital signs, key indicators during surgery, and postoperative recovery in real time, leveraging data mining and analysis techniques to support clinical nursing decisions. They typically employ rule engines, artificial intelligence models, or reasoning methods based on historical data. The design of these systems usually relies on integrating multiple data sources, combining the professional experience of medical staff with algorithmic models, to provide patients with personalized nursing plans, thereby improving the quality of care and surgical success rates.

[0004] Traditional intelligent nursing decision support systems for patients undergoing interventional cardiology procedures often operate based on static rules or historical experience. They fail to adequately depict the dynamic relationship between the rhythm of instrument operation and the patient's immediate physiological changes during the intervention. Vital signs data are often separated from the surgical stage, making it difficult to achieve precise time alignment and stage division. This leads to nursing judgments relying more on single-point indicators or human experience, making it difficult to identify differences in myocardial load at different interventional stages in a timely manner. This results in a risk of delays or conflicts in the triggering sequence and duration of nursing instructions, thereby affecting the pertinence and continuity of nursing responses and increasing the uncertainty of clinical decision-making. Summary of the Invention

[0005] To address the technical problems existing in the prior art, embodiments of the present invention provide an intelligent nursing decision support system for patients undergoing interventional cardiology procedures. The technical solution is as follows:

[0006] On the one hand, it provides an intelligent nursing decision support system for patients undergoing interventional cardiology procedures. This system includes:

[0007] The data analysis module collects and analyzes the operating status of interventional devices, extracts the state switching time points corresponding to catheter advancement, device dwell and dilation states and divides them into interventional stages, collects electrocardiogram, non-invasive blood pressure and heart rate signals and aligns them in time, generates physiological monitoring sequences and transmits them to the feature calculation module;

[0008] The feature calculation module performs sliding window statistics on the electrocardiogram, non-invasive blood pressure and heart rate signals of the multi-intervention stage based on the physiological monitoring sequence, calculates the systolic blood pressure, RR interval and pulse pressure waveform rise time and removes abnormal windows, generates a physiological response feature set and transmits it to the stress discrimination module.

[0009] The stress discrimination module, based on the physiological response feature set, performs trend alignment on the changes in systolic blood pressure, RR interval and pulse pressure waveform rise time and calculates the trend consistency index to determine the myocardial stress state level, generates myocardial discrimination results and transmits them to the recovery analysis module.

[0010] The recovery analysis module determines the end point of the multi-intervention stage based on the myocardial discrimination results, extracts the heart rate and blood pressure recovery time and the electrocardiogram stability interval, and performs correlation with the corresponding stage myocardial discrimination results, calculates the recovery change trend of continuous stages, and generates a nursing decision instruction set.

[0011] The decision optimization module, based on the nursing decision instruction set and combined with the myocardial discrimination results and the recovery trend of continuous stages, performs rule constraint judgment and logical adjustment calculation on the instruction triggering order, instruction duration interval and adjacent stage instruction conflict relationship in the nursing decision instruction set, and generates a nursing decision optimization instruction set.

[0012] As a further embodiment of the present invention, the physiological monitoring sequence includes a catheter advancement phase, a device dwell phase, and a dilation phase; the physiological response feature set includes systolic blood pressure, RR interval, pulse pressure waveform rise time, and abnormal window markers; the myocardial discrimination result includes myocardial load status and load end time; the nursing decision instruction set includes load end time point, recovery time characteristics, and recovery change trend; and the nursing decision optimization instruction set includes instruction triggering sequence, instruction duration interval, and instruction conflict resolution result.

[0013] As a further aspect of the present invention, the data analysis module includes:

[0014] The status acquisition submodule collects and analyzes the operating status of the interventional device, obtains the catheter advancement status, instrument dwell status and dilation operation status and sorts them by timestamp, detects the status change amplitude between adjacent sampling points, marks the time point at the abrupt position in the continuous sampling points, and generates a set of device status change time points.

[0015] The phase division submodule, based on the set of device state change time points, calls the time corresponding to the catheter advancement state, instrument dwell state and dilation state, performs time difference calculation on the interval between adjacent time points, and divides the continuous time period into intervals in combination with the intervention phase switching benchmark value to generate an intervention phase time interval sequence.

[0016] The physiological alignment submodule acquires electrocardiogram signals, non-invasive blood pressure signals, and heart rate signals. Based on the interventional stage time interval sequence, it calls the multi-signal sampling time axis, performs resampling and time offset correction on multiple sampling frequency signals, and splices the stage identifier with the corresponding physiological signal to generate a physiological monitoring sequence.

[0017] As a further aspect of the present invention, the intervention phase switching benchmark value is obtained by statistically analyzing the obtained sliding update benchmark value, summarizing the duration of adjacent phases, and calculating the sum of the average duration and a preset three times the standard deviation as the benchmark value.

[0018] As a further aspect of the present invention, the feature calculation module includes:

[0019] The signal window molecular module, based on the physiological monitoring sequence, continuously extracts fixed-length data segments from the electrocardiogram, non-invasive blood pressure, and heart rate signals at multiple intervention stages according to the time axis, and binds the window number and stage identifier according to the sampling time sequence to generate a staged sliding window sequence.

[0020] The feature calculation submodule calls the staged sliding window sequence to detect the R wave time point of the electrocardiogram window and calculate the interval between adjacent R waves. It extracts the systolic blood pressure value for the non-invasive blood pressure window and calculates the time difference from the start to the peak of the pulse pressure waveform, generating a set of physiological feature vectors for the window.

[0021] The anomaly removal submodule, based on the window physiological feature vector set, performs statistical distribution summarization on systolic blood pressure, RR interval and pulse rise time, calculates the corresponding mean and standard deviation and constructs the feature value benchmark interval, removes window feature vectors that exceed the interval, and generates a physiological response feature set.

[0022] As a further aspect of the present invention, the stress discrimination module includes:

[0023] The trend alignment submodule, based on the physiological response feature set, extracts sampling points from the systolic blood pressure time series, RR interval time series and pulse pressure waveform rise time series in the order of intervention operation stages, performs translation correction on the start timestamps of multiple sequences and performs stage index matching to generate multi-indicator aligned trend sequences.

[0024] The consistency calculation submodule calls the multi-indicator alignment trend sequence to calculate the trend direction sign of the changes in systolic blood pressure, RR interval, and pulse pressure rise time within the same period and constructs a direction vector. It then performs a consistency comparison based on the angle relationship between the direction vectors to generate a trend consistency index.

[0025] The stress determination submodule performs interval judgment on the corresponding index values ​​of multiple interventional operation stages and the preset stress determination benchmark value based on the trend consistency index, completes the stage label mapping and integrates the stage sequence information, and generates myocardial discrimination results.

[0026] As a further aspect of the present invention, the stress determination benchmark value is obtained by summarizing the collected changes in systolic blood pressure, RR interval, and pulse pressure waveform rise time in stages within the complete intervention process, obtaining the numerical distribution range of multiple indicators in multiple stages, forming an indicator set based on the consistency calculation results within the distribution range, performing interval sorting on the indicator set, and selecting the median value for determination.

[0027] As a further aspect of the present invention, the recovery analysis module includes:

[0028] The load termination submodule determines the load termination time point of the multi-intervention stage based on the myocardial discrimination result, detects the switching position from the loaded state to the non-load state for the continuous myocardial discrimination marker sequence, calculates the stage boundary timestamp based on the continuity of the myocardial state before and after the switching, and generates a load termination time point sequence.

[0029] The data recovery submodule obtains the heart rate sampling sequence, blood pressure sampling sequence, and electrocardiogram rhythm marker sequence after the end of the multi-stage process based on the load end time point sequence, calculates the duration for the heart rate and blood pressure to fall back to the stable segment, identifies the stable segment of the electrocardiogram rhythm marker, and obtains the recovery time feature set.

[0030] The instruction generation submodule aligns the execution time of the recovery time feature set with the corresponding stage myocardial load discrimination result, calculates the heart rate recovery time difference, blood pressure recovery time difference, and ECG stable interval length change rate between adjacent stages, analyzes the recovery change trend vector, performs rule mapping, and generates a nursing decision instruction set.

[0031] As a further aspect of the present invention, the decision optimization module includes:

[0032] The instruction constraint submodule, based on the nursing decision instruction set and combined with the myocardial discrimination result, obtains the stage identifier and time sequence mark corresponding to multiple instructions, performs a sequence consistency judgment on the triggering order relationship of adjacent instructions, and performs sequence rearrangement on inconsistent positions to generate an instruction triggering sequence mapping table.

[0033] The interval correction submodule triggers the sequence mapping table according to the instruction, calls the recovery change trend of the continuous stage, obtains the start and end time markers corresponding to multiple instructions, performs interval overlap detection on adjacent stage time segments, performs interval pruning and splicing operations on the overlapping part, and generates a set of instruction continuous intervals.

[0034] The conflict resolution submodule, based on the instruction duration interval set and the myocardial discrimination result, detects multiple nursing instructions existing in parallel within the same time interval, performs logical mutual exclusion judgment and stage priority sorting on the parallel instructions, and generates a nursing decision optimization instruction set.

[0035] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following:

[0036] By synchronously analyzing the operating status of interventional devices and multi-source physiological signals, the time boundary of the interventional stage can be identified, and a stable physiological response feature sequence can be constructed within the continuous stage. This allows the trends of ECG and blood pressure changes to form a comparable overall relationship, thereby accurately reflecting the myocardial load status and stage differences. Combined with the recovery behavior after the load ends, correlation analysis is performed, enabling nursing instructions to be adjusted in sequence and interval according to the continuous change trend. This effectively reduces inter-stage instruction conflicts and improves the stability of nursing decisions in terms of temporal consistency and individual adaptability. Attached Figure Description

[0037] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0038] Figure 1 This is a schematic diagram of the system of the present invention;

[0039] Figure 2 This is a schematic diagram of the system framework of the present invention;

[0040] Figure 3 This is a flowchart of the data analysis module in this invention;

[0041] Figure 4 This is a flowchart of the feature calculation module in this invention;

[0042] Figure 5 This is a flowchart of the stress discrimination module in this invention;

[0043] Figure 6 This is a flowchart of the recovery analysis module in this invention;

[0044] Figure 7This is a flowchart of the decision optimization module in this invention. Detailed Implementation

[0045] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0046] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0047] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0048] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0049] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0050] This invention provides an intelligent nursing decision support system for patients undergoing interventional cardiology procedures, such as... Figure 1-2 The diagram shown illustrates an intelligent nursing decision support system for patients undergoing interventional cardiology procedures. The system includes:

[0051] The data analysis module collects and analyzes the operating status of interventional devices, extracts the state switching time points corresponding to catheter advancement, device dwell and dilation states and divides them into interventional stages, collects electrocardiogram, non-invasive blood pressure and heart rate signals and aligns them in time, generates physiological monitoring sequences and transmits them to the feature calculation module;

[0052] The feature calculation module performs sliding window statistics on electrocardiogram, non-invasive blood pressure and heart rate signals at multiple intervention stages based on physiological monitoring sequences, calculates systolic blood pressure, RR interval and pulse pressure waveform rise time and removes abnormal windows, generates physiological response feature set and transmits it to the stress discrimination module.

[0053] The stress discrimination module, based on the physiological response feature set, performs trend alignment on the changes in systolic blood pressure, RR interval and pulse pressure waveform rise time and calculates the trend consistency index to determine the myocardial stress state level, generates myocardial discrimination results and transmits them to the recovery analysis module.

[0054] The recovery analysis module determines the end point of the multi-intervention stage based on the myocardial discrimination results, extracts the recovery time of heart rate and blood pressure and the electrocardiogram stability interval, and performs correlation with the corresponding stage myocardial discrimination results, calculates the recovery change trend of continuous stages, and generates a nursing decision instruction set.

[0055] The decision optimization module, based on the nursing decision instruction set and combined with the myocardial discrimination results and the recovery trend of continuous stages, performs rule constraint judgment and logical adjustment calculation on the instruction triggering order, instruction duration interval and adjacent stage instruction conflict relationship in the nursing decision instruction set, and generates a nursing decision optimization instruction set.

[0056] The physiological monitoring sequence includes the catheter advancement phase, the device dwell phase, and the dilation phase. The physiological response feature set includes systolic blood pressure, RR interval, pulse pressure waveform rise time, and abnormal window markers. The myocardial discrimination results include myocardial load status and load end time. The nursing decision instruction set includes the load end time point, recovery time characteristics, and recovery trend. The nursing decision optimization instruction set includes instruction triggering sequence, instruction duration interval, and instruction conflict resolution results.

[0057] Specifically, such as Figure 2 , 3 As shown, the data analysis module includes:

[0058] The status acquisition submodule collects and analyzes the operating status of the interventional device, obtains the catheter advancement status, instrument dwell status and dilation operation status and sorts them by timestamp, detects the status change amplitude between adjacent sampling points, marks the time point at the abrupt position in the continuous sampling points, and generates a set of device status change time points.

[0059] First, the PCI-1710 data acquisition card is initialized, and the sampling frequency is set to 50 Hz, meaning data is acquired every 20 milliseconds. The displacement signal from the photoelectric encoder and the fluid pressure signal from the pressure sensor in the interventional device are read in real time via the PCI bus. The acquisition card converts the received 0-5 volt analog voltage signal into a 12-bit binary digital quantity, ranging from 0 to 4095, and maps this digital quantity to a physical quantity using a preset linear calibration coefficient. For catheter advancement, the displacement value at the current sampling moment is read and compared with the displacement value at the previous sampling moment. The difference between the two is calculated. If the difference is greater than 0.5 mm, the current moment is determined to be in the catheter advancement state. For instrument dwell, a sliding window of length 5 is established. If the absolute value of the displacement difference between 5 consecutive sampling points within the window is less than 0.1 mm, and the pressure value at the corresponding moment is less than 0.2 atmospheres, the instrument dwell state is determined. For dilation operation, the pressure value is monitored. When the detected pressure value is greater than 2.0 atmospheres and the current pressure value is higher than the previous pressure value, the dilation operation state is determined. The parsed state identifiers are bound to a unified timestamp at the time of data acquisition and sorted in ascending order based on the timestamp values. When detecting the state change amplitude between adjacent sampling points, the absolute value of the difference between the displacement values ​​at the current time and the previous time is calculated as the displacement change amplitude, and the absolute value of the difference between the pressure values ​​is calculated as the pressure change amplitude. Abrupt change detection threshold is set, with the displacement abrupt change threshold set to 2.0 mm and the pressure abrupt change threshold set to 2.0 atmospheres. This threshold is set based on a margin of 0.2 times the maximum peak noise level of the sensor, i.e., 0.2 * 10 = 2.0. When the calculated state change amplitude is greater than the corresponding abrupt change detection threshold, and the current determined state is inconsistent with the previous state, that time is immediately marked as the abrupt change location. For example, at time 10:05:01:200, the duct displacement difference abruptly changes from 0 mm to 5 mm. Since 5 mm is greater than the set threshold of 2.0 mm, and the state changes from stationary to advancing, this time is marked. Table 1 below shows some of the acquired data and analysis results. Finally, all the marked mutation time points are summarized to generate a set of device state change time points, which accurately records the precise moment of each operation switch during the interventional procedure.

[0060] Table 1. Data Acquisition and Analysis of Interventional Device Status

[0061] 1000 150.0 0.0 Stay 0.0 no 1020 150.0 0.0 Stay 0.0 no 1040 155.0 0.0 Advance 5.0 yes 1060 160.0 0.0 Advance 5.0 no 1080 160.0 4.0 expansion 4.0 yes

[0062] As shown in Table 1, at timestamp 1040 ms, a displacement amplitude of 5.0 mm exceeding the threshold of 2.0 mm and a state change were detected, marking the start of the propulsion state; at timestamp 1060 ms, although the amplitude exceeded the threshold, the state remained unchanged, so no marking was made; at timestamp 1080 ms, a pressure amplitude of 4.0 atmospheres exceeding the threshold of 2.0 atmospheres and a state change were detected, marking the start of the expansion state. This result demonstrates the ability to accurately capture the instantaneous switching of equipment operating states, eliminate redundant signals in continuous actions, and provide high-precision discrete-time anchor points for subsequent stage division.

[0063] The phase division submodule, based on the set of time points of equipment status changes, calls the time corresponding to the catheter advancement state, instrument dwell state and dilation state, performs time difference calculation on the interval between adjacent time points, and divides the continuous time period into intervals in combination with the interventional phase switching benchmark value to generate the interventional phase time interval sequence.

[0064] The system receives a set of device state change time points. First, it iterates through this set, retrieving the corresponding original parsed state for each time point. Then, it calculates the time difference between adjacent time points, subtracting the timestamp of the previous time point from the timestamp of the later time point. During this process, a baseline value for intervention phase switching is introduced, set to 3.0 seconds. This baseline value is based on a statistical analysis of 1000 historical surgical operation logs. The statistical results show that the average interval between actions within a single continuous operation step is 1.2 seconds, with a standard deviation of 0.5 seconds. Following the rule of three standard deviations, the average of 1.2 seconds is added to three times the standard deviation of 0.5 seconds, resulting in 2.7 seconds. To ensure coverage, this result is rounded up to 3.0 seconds. Compare the calculated time difference with the baseline value of 3.0 seconds: if the time difference is less than or equal to 3.0 seconds, it is determined that the two adjacent time points belong to the same continuous operation phase, and a merging operation is performed to include the latter time point into the current time interval; if the time difference is greater than 3.0 seconds, it is determined that the operation phase has undergone a substantial switch, and a truncation operation is performed at the previous time point to end the current interval, and a new time interval is started with the latter time point as the starting point. Example description: The extracted mutation time point set fragment contains four key points: 10:00:05 (end of advancement), 10:00:07 (start of expansion), 10:00:20 (end of expansion), and 10:00:30 (start of withdrawal). (1) Calculate the interval between the second time point and the first time point: subtract 10:00:05 from 10:00:07 to get 2 seconds. Since 2 seconds is less than 3.0 seconds, it is determined to be a continuous operation, and the advancement and expansion are merged into the same phase. (2) Calculate the interval between the fourth and third time points: Subtract 10:00:20 from 10:00:30 to get 10 seconds. Since 10 seconds is greater than 3.0 seconds, it is determined to be a stage break. The previous stage ends at 10:00:20, and the next stage begins at 10:00:30. The above logic is executed sequentially for all time points, and finally, an interventional stage time interval sequence containing the start and end times is generated. This sequence clarifies the time boundaries of each independent operation stage during the operation. For example, the interval from 10:00:00 to 10:00:20 is identified as the delivery and expansion stage, and the interval from 10:00:30 to 10:00:45 is identified as the instrument withdrawal stage. This result shows that by dividing the intervals based on statistical benchmark values, discrete operation points can be automatically aggregated into continuous surgical stages with clinical semantics.

[0065] The physiological alignment submodule acquires electrocardiogram signals, non-invasive blood pressure signals, and heart rate signals. Based on the interventional stage time interval sequence, it calls the multi-signal sampling time axis, performs resampling and time offset correction on multiple sampling frequency signals, and splices the stage identifier with the corresponding physiological signal to generate a physiological monitoring sequence.

[0066] The patient's electrocardiogram (ECG), non-invasive blood pressure, and heart rate signals were read via the hospital's HL7 local area network interface. The original sampling frequency for the ECG signal was 500 Hz, the non-invasive blood pressure signal was sampled intermittently and updated every 60 seconds, and the original sampling frequency for the heart rate signal was 1 Hz. Based on the generated time interval sequence of the interventional phase, for example, all physiological signal data within the time period from 10:00:00 to 10:00:20 were selected. Due to clock discrepancies between the signal sources, time offset correction was first performed: the clock difference between the physiological monitor and the interventional device control host was calculated using the Network Time Protocol (NTP). Actual measurements showed that the physiological monitor clock was 200 milliseconds ahead of the interventional device host. To unify the benchmark, 200 milliseconds were subtracted from the timestamps of all physiological signals to obtain the corrected timestamps. Subsequently, a multi-signal resampling operation was performed, with the sampling frequency of the interventional device status at 50 Hz, i.e., a time interval of 20 milliseconds, as the target frequency: (1) ECG signal resampling (downsampling): The 500 Hz signal was processed using a multiphase anti-aliasing filter, and the decimation factor was set to 10, i.e., 500 divided by 50. One point was extracted every 10 original data points, thereby reducing the frequency to 50 Hz. (2) Heart rate signal resampling (linear interpolation): For the 1 Hz low-frequency signal, the data was filled using linear interpolation. The initial heart rate was set to 60, and the next second's heart rate was set to 61. For the target point between these two times, 0.02 seconds away from the initial time, the time ratio was first calculated, and 0.02 seconds was divided by 1 second to get 0.02. Then, the heart rate difference 1 was multiplied by 0.02 to get 0.02. Finally, this increment was added to the initial heart rate of 60, and the interpolated heart rate of the point was calculated to be 60.02. A continuous heart rate sequence of 50 Hz was generated in this way. (3) Non-invasive blood pressure signal resampling (zero-order hold): Since blood pressure is a non-continuous measurement, the zero-order hold method is adopted. Before the next measurement update, the values ​​of all resampling points are forcibly set to the latest blood pressure measurement value, keeping the values ​​unchanged. After completing resampling and time correction, the intervention stage identifier is used as a label column and concatenated with the aligned ECG value, corrected blood pressure value, and heart rate value row by row. Finally, a standardized physiological monitoring sequence matrix containing timestamp, stage identifier, ECG value, blood pressure value, and heart rate value is generated. For example, at the corrected time point 10:00:05:020 milliseconds, the matrix row data are the timestamp, expansion stage identifier, 1.2 mV ECG value, 120 mmHg blood pressure value, and 75.02 beats per minute heart rate value, respectively. This result shows that after alignment processing, high-frequency and low-frequency physiological signals are uniformly mapped to the same time axis of device operation, and the clock error between devices is eliminated, so that the impact of device operation on the patient's physiological function in every millisecond can be accurately quantified and traced.

[0067] Specifically, such as Figure 2 ,4 As shown, the feature calculation module includes:

[0068] The signal window molecular module, based on physiological monitoring sequences, continuously extracts fixed-length data segments from ECG, non-invasive blood pressure, and heart rate signals at multiple interventional stages along the time axis, and binds window numbers and stage identifiers according to the sampling time sequence to generate a staged sliding window sequence.

[0069] The generated physiological monitoring sequence is invoked. This matrix includes timestamps, stage identifiers (such as "expansion stage"), electrocardiogram values, non-invasive blood pressure values, and heart rate values. A fixed time window length of 5.0 seconds is set, corresponding to 250 data sampling points based on a 50 Hz sampling rate; a sliding step size of 2.5 seconds is set, corresponding to 125 data sampling points, to achieve a 50% data overlap rate and ensure the capture of transient features across windows. The monitoring sequence is traversed and truncated along the time axis. For each truncated candidate window, the "stage identifier" corresponding to all sampling points within the window is checked for consistency. If all data points within a window belong to the same stage (e.g., all are "expansion stage"), the window is retained and assigned a unique window number based on the truncating order; if the data within a window contains stage switching points (i.e., data from both the "expansion stage" and the "stagnation stage"), the window is discarded to ensure feature purity. The retained data segments are strongly bound to the corresponding stage identifiers. Example: The start time of the "expansion stage" in the physiological monitoring sequence is set to 10:00:00 and the end time is set to 10:00:10. (1) The first window intercepts the time period from 10:00:00:00 to 10:00:05:00. After inspection, all samples in this segment are in the "expansion stage", which is deemed valid and retained. (2) The second window slides along the time axis for 2.5 seconds and intercepts the time period from 10:00:02:50 to 10:00:07:50. All sample identifiers are consistent, which is deemed valid and retained. (3) The fifth window intercepts the time period from 10:00:08 to 10:00:13. Since the stage ends at 10:00:10, the window contains two states, which is deemed invalid and removed. Finally, a phased sliding window sequence as shown in Table 2 is generated, providing a standardized data container for subsequent feature extraction.

[0070] Table 2. Staged sliding window sequence fragment table (Table 2)

[0071] W-Exp-01 10:00:00.000 10:00:05.000 Expansion phase 250 efficient W-Exp-02 10:00:02.500 10:00:07.500 Expansion phase 250 efficient W-Exp-03 10:00:05.000 10:00:10.000 Expansion phase 250 efficient W-Mix-04 10:00:07.500 10:00:12.500 mix 250 invalid

[0072] The feature calculation submodule calls the staged sliding window sequence to detect the R wave time point of the electrocardiogram window and calculate the interval between adjacent R waves. It extracts the systolic blood pressure value for the non-invasive blood pressure window and calculates the time difference from the start to the peak of the pulse pressure waveform, generating a set of physiological feature vectors for the window.

[0073] The effective windows in the phased sliding window sequence are called one by one, and feature extraction is performed on the physiological signal data within each window. For electrocardiogram signals, the threshold difference method is used to detect R waves: the first derivative of the signal is calculated, and when the absolute value of the derivative exceeds 0.5 mV / s and the original signal amplitude exceeds 1.0 mV, it is marked as the peak time of the R wave. The time points of all R waves within the window are extracted, the time interval between two adjacent R waves is calculated, and their arithmetic mean is taken as the average RR interval of the window. For non-invasive blood pressure signals, the systolic blood pressure value held within the window is directly read as the systolic blood pressure feature value. For the calculation of the time difference from the start to the peak of the pulse pressure waveform, in the pulse waveform data (from the synchronously acquired high-frequency pulse wave channel), the minimum point (wave foot) time and the maximum point (wave peak) time of the waveform within a single cardiac cycle are searched. The pulse pressure rise time is obtained by subtracting the wave foot time from the peak time. If the window contains multiple cardiac cycles, the rise time of all cycles is calculated and averaged. Finally, the calculated systolic blood pressure, mean RR interval, and pulse pressure rise time are combined. Example: Taking window W-Exp-01 as an example: (1) ECG processing: Three R wave peak time points were detected at 1.0 seconds, 1.8 seconds, and 2.6 seconds. The first interval was calculated as 1.8-1.0=0.8 seconds, and the second interval was calculated as 2.6-1.8=0.8 seconds. The mean RR interval was the sum of the two 0.8 seconds divided by 2, which equals 0.8 seconds (i.e., 800 milliseconds). (2) Blood pressure extraction: The systolic blood pressure value corresponding to this window was read and set to 120 mmHg. (3) Pulse pressure time calculation: In the first cardiac cycle, the pulse wave start point was detected at 1.10 seconds, and the peak point was at 1.22 seconds. The rise time was calculated as 1.22-1.10=0.12 seconds (i.e., 120 milliseconds). (4) Vector generation: The physiological feature vector of the window was generated as [120 mmHg, 800 ms, 120 ms]. This result shows that the complex waveform data was successfully transformed into a low-dimensional numerical feature vector.

[0074] The anomaly removal submodule performs statistical distribution summarization on systolic blood pressure, RR interval and pulse pressure rise time based on the window physiological feature vector set, calculates the corresponding mean and standard deviation and constructs the feature value benchmark interval, removes window feature vectors that exceed the interval, and generates a physiological response feature set;

[0075] Based on the window physiological feature vector set, the vectors are grouped according to the stage identifier (e.g., all vectors of the "expansion stage" are grouped together). For each group of data, statistical distribution analysis is performed on the three dimensions of systolic blood pressure, RR interval and pulse pressure rise time, and the arithmetic mean and standard deviation of each dimension are calculated. The feature value benchmark interval is constructed according to the statistical principle of three times the standard deviation, that is, the lower limit of the interval is the mean minus three times the standard deviation, and the upper limit is the mean plus three times the standard deviation. If any feature vector has a value in any dimension that exceeds the corresponding benchmark interval, the window data is judged to be abnormal (possibly caused by body motion artifacts or poor electrode contact), and the vector is removed from the set. Only vectors in which all dimensions are within the normal range are retained to generate the final physiological response feature set. Example description: 100 feature vectors of the "expansion stage" are collected, and the statistical parameters are calculated as follows: (1) Systolic blood pressure: the mean is 125 mmHg and the standard deviation is 5 mmHg. The lower limit of the baseline interval is 125 minus 3 multiplied by 5 equals 110, and the upper limit is 125+3*5=140, that is, [110, 140] mmHg. (2) RR interval: the mean is 800 ms, and the standard deviation is 20 ms. The lower limit of the baseline interval is 800-3*20=740, and the upper limit is 800+3*20=860, that is, [740, 860] ms. (3) Pulse rise time: the mean is 120 ms, and the standard deviation is 10 ms. The lower limit of the baseline interval is 120 minus 3 multiplied by 10 equals 90, and the upper limit is 120+3*10=150, that is, [90, 150] ms. Screening Execution: Check feature vector A [135, 810, 130]: Systolic blood pressure 135 is within [110, 140], RR interval 810 is within [740, 860], and rise time 130 is within [90, 150]. Decision: Keep. Check feature vector B [145, 800, 120]: Systolic blood pressure 145 exceeds the upper limit 140. Decision: Remove. Check feature vector C [120, 700, 120]: RR interval 700 is below the lower limit 740. Decision: Remove. This step automatically filters out outlier data caused by non-pathological factors (such as interference), ensuring that the subsequently generated physiological response feature set can truly and objectively reflect the actual impact of interventional procedures on human physiological functions.

[0076] Specifically, such as Figure 2 , 5 As shown, the stress discrimination module includes:

[0077] The trend alignment submodule, based on the physiological response feature set, extracts sampling points from the systolic blood pressure time series, RR interval time series, and pulse pressure waveform rise time series in the order of intervention operation stages. It performs translation correction on the start timestamps of multiple sequences and performs stage index matching to generate multi-indicator aligned trend sequences.

[0078] The generated physiological response feature set, containing filtered effective window feature vectors, is invoked. First, all feature vectors within each stage are extracted sequentially according to the order of the interventional procedures. For each specific interventional stage, the feature vectors are decomposed, and the systolic blood pressure value, mean RR interval value, and pulse pressure rise time value are extracted. These are then recombined into three independent time series. To address the issue of different absolute values ​​at the start times of different stages, a timestamp shift correction operation is performed. First, the start timestamp of the first effective window of the current stage is locked as the zero-point reference time for that stage. Then, each data point within that stage is traversed, and the difference between its original timestamp and the zero-point reference time is calculated. This difference is used as the new relative time index, with the unit uniformly set to seconds. After time axis normalization, stage index matching is performed to check the data integrity of the three series under the same relative time index. The corrected relative time index is used as the primary key, and the systolic blood pressure value is then used as the primary key. The values ​​of systolic blood pressure, RR interval, and rise time are aligned and merged. Taking the expansion phase as an example, the data of the first three windows of this phase are extracted. First, the timestamp of the first window, 10:00:00, is locked as the baseline zero point. For data point 1 with the original time of 10:00:00, the relative time is calculated to be 0.0 seconds, corresponding to a systolic blood pressure of 120 mmHg, an RR interval of 800 ms, and a rise time of 120 ms. For data point 2 with the original time of 10:00:02:500 ms, the relative time is calculated to be 2.5 seconds, corresponding to a systolic blood pressure of 125 mmHg, an RR interval of 760 ms, and a rise time of 115 ms. For data point 3 with the original time of 10:00:05:000 ms, the relative time is calculated to be 5.0 seconds, corresponding to a systolic blood pressure of 130 mmHg, an RR interval of 730 ms, and a rise time of 112 ms. The above data are integrated to generate the multi-indicator aligned trend sequence shown in Table 3.

[0079] Table 3. Multi-indicator aligned trend sequence segment table (Table 3)

[0080] 0.0 120.0 800.0 120.0 Expansion phase 2.5 125.0 760.0 115.0 Expansion phase 5.0 130.0 730.0 112.0 Expansion phase

[0081] The results indicate that by establishing a relative time axis, the difference in absolute time was successfully eliminated, enabling the physiological changes at different surgical stages to be quantitatively analyzed on a unified time scale.

[0082] The consistency calculation submodule calls the multi-indicator alignment trend sequence, calculates the trend direction sign of the changes in systolic blood pressure, RR interval, and pulse pressure rise time within the same period, and constructs a direction vector. It then performs a consistency comparison based on the angle relationship between the direction vectors to generate a trend consistency index.

[0083] This study uses a multi-indicator aligned trend sequence to quantify whether the changes in three physiological indicators conform to a typical stress response pattern. First, a standardized stress reference vector is defined, consisting of three components representing positive changes in systolic blood pressure, negative changes in the RR interval, and negative changes in rise time. The reference vector is set as a unit-direction vector, with its numerical components set as follows: systolic blood pressure component +1, RR interval component -1, and rise time component -1. The trend sequence is iterated, and the changes between adjacent time points are calculated. The change in systolic blood pressure is obtained by subtracting the previous systolic blood pressure from the current systolic blood pressure. The change in RR interval is obtained by subtracting the previous RR interval from the current RR interval, and the change in rise time is obtained by subtracting the previous rise time from the current rise time. To ensure the balance of vector calculation, the above changes are normalized using the standard deviations of each statistical indicator, i.e., the change is divided by the corresponding standard deviation to obtain dimensionless standardized change values. The observed change vector is constructed from these three standardized change values. Subsequently, spatial geometric calculations are performed to calculate the cosine similarity between the observed change vector and the stress reference vector. The calculation process involves first calculating the dot product of the two vectors, i.e., multiplying the corresponding components. Summing is performed, then the magnitudes of the two vectors are calculated separately. Finally, the dot product is divided by the product of the two magnitudes to obtain the cosine similarity value. Based on the data in Table 3 above, the consistency index at a relative time of 2.5 seconds is calculated. The systolic blood pressure change is 125 minus 120 equals 5, the RR interval change is 760-800=-40, and the rise time change is 115-120=-5. Substituting these values ​​into the preset standard deviations (systolic blood pressure standard deviation 5, RR interval standard deviation 20, and rise time standard deviation 10) for normalization, the standardized value of systolic blood pressure is 1.0, the standardized value of RR interval is -2.0, and the standardized value of rise time is... The normalization value is -0.5. The constructed observation vectors are coordinates 1.0, -2.0, and -0.5, and the reference vectors are coordinates 1.0, -1.0, and -1.0. The dot product is 3.5. The magnitude of the observation vectors is calculated as the square root of the sum of the squares of 1, -2, and -0.5, which is approximately √5.25, equal to 2.291. The magnitude of the reference vectors is calculated as the square root of the sum of the squares of 1, -1, and -1, which is approximately √3, equal to 1.732. The final cosine value is 3.5 / 2.291*1.732=0.882, generating the trend consistency index.

[0084] The stress determination submodule performs interval judgment on the corresponding indicator values ​​of multiple interventional operation stages and the preset stress determination benchmark value based on the trend consistency index, completes the stage label mapping and integrates the stage sequence information to generate myocardial discrimination results.

[0085] The system receives the generated trend consistency index sequence and performs a comprehensive interpretation based on the current interventional procedure stage information. A preset stress threshold of 0.75 is used, set based on retrospective case analysis. The calculated trend consistency index is compared with this threshold. If the consistency index at a given time is greater than 0.75, the patient is considered to be in a state of significant stress. If the index is between 0 and 0.75, it is considered weak stress or no specific pattern. If the index is less than 0, it is considered non-stressful reverse fluctuation. The system further calculates the proportion of time points identified as significantly stressed during the entire expansion phase. If this proportion exceeds 50%, the phase is marked as a positive myocardial load phase. Subsequently, logical inference is performed by integrating the phase sequence information. If the expansion phase is marked as positive and the trend consistency is consistent during the subsequent withdrawal or stay phase... If the indicator rapidly drops to a negative value, a judgment result of reversible myocardial ischemia risk is generated. If the indicator remains at a high positive value during the withdrawal phase, a warning of persistent damage risk is generated. Taking the aforementioned example data, the calculated trend consistency index is 0.882. It is compared with the preset benchmark value of 0.75. Since 0.882 is greater than 0.75, the current moment is determined to be a significant stress state. During the expansion phase lasting 10 seconds, a total of 4 consistency index points were calculated, namely 0.882, 0.910, 0.850, and 0.780. All 4 points are greater than 0.75, and the calculation rate is 100%, which exceeds the 50% judgment threshold. The expansion phase is officially labeled as a positive myocardial load phase, and a myocardial discrimination result containing the content of detecting significant physiological stress consistency induced by the expansion operation is output.

[0086] Specifically, such as Figure 2 , 6 As shown, the recovery analysis module includes:

[0087] The load termination submodule determines the load termination time point of multiple intervention stages based on the myocardial discrimination results, detects the switching position from the load state to the non-load state for continuous myocardial discrimination marker sequences, calculates the stage boundary timestamps based on the continuity of myocardial state before and after the switching, and generates a load termination time point sequence.

[0088] The generated myocardial discrimination result sequence is invoked. This sequence consists of binary state markers arranged in chronological order, where a value of 1 represents significant myocardial overload and a value of 0 represents no overload or baseline state. A sliding detection window of length 5 is initialized and moved forward along the time axis of the discrimination result sequence. At each time step, five consecutive state marker values ​​covered within the window are read. The summation operation is performed on the marker values ​​within the window to obtain the window state sum. This sum is compared with a preset switching judgment threshold, which is set to 0, meaning that all markers within the window must be 0 to confirm the entry into the non-overload state. When the state sum of the window at the previous time step is greater than 0 and the state sum of the window at the current time step is equal to 0, the switching event from overload state to non-overload state is identified. The time index corresponding to the starting position of the current window is then locked. To eliminate misjudgments caused by signal jitter, the time index is further retrieved. The myocardial discriminant markers of 10 consecutive sampling points are verified one by one to ensure that all markers remain at 0. If the verification passes, the physical timestamp corresponding to the time index is determined as the candidate end point of the load. If the verification fails, the current candidate point is abandoned and the sliding window continues. After confirming that the candidate point is valid, the timestamp mapping table in the original monitoring data is called to find the precise millisecond-level time value corresponding to the index. For example, if the 500th to 505th positions of the sequence index are all 0 and the preceding position is 1, the timestamp 10:15:00 corresponding to the 500th position is extracted as the end time of the load. The above scanning and verification process is repeated for all intervention stages. The end times of the load corresponding to the expansion stage, delivery stage and withdrawal stage are recorded respectively. These discrete time points are sorted from early to late according to the order of occurrence. Repeated false trigger points with an interval of less than 5 seconds are removed. Finally, a sequence of end time points containing multiple precise moments of the load is constructed.

[0089] The data recovery submodule obtains the heart rate sampling sequence, blood pressure sampling sequence, and electrocardiogram rhythm marker sequence after the end of multiple stages based on the load end time point sequence, calculates the duration for the heart rate and blood pressure to fall back to the stable segment, identifies the stable segment of the electrocardiogram rhythm marker, and obtains the recovery time feature set.

[0090] Based on the time sequence of the end of the interventional procedure, physiological signal data segments were extracted after each interventional stage. First, a baseline stable interval for the physiological parameters was established, constructed based on the average resting state values ​​collected before the intervention. For example, the heart rate stable interval was set to ±5% of the average resting heart rate, and the systolic blood pressure stable interval was set to ±10% of the average resting systolic blood pressure. Starting from the end of the interventional procedure, subsequent heart rate sampling sequences were scanned point by point. The inclusion relationship between each sampling point value and the heart rate stable interval was calculated. The starting time when the heart rate value first entered and remained within the stable interval for 30 consecutive seconds was recorded. The time difference between this starting time and the end of the interventional procedure was calculated, thus determining the duration of the heart rate returning to the stable range. Similarly, the same interval scanning and time difference calculation operations are performed on the blood pressure sampling sequence to obtain the duration of blood pressure falling back to the stable segment. For the ECG rhythm marker, the segments in the ECG rhythm marker sequence where normal sinus rhythm markers appear continuously are checked. The starting point from the end of the load to the first continuous normal rhythm segment with a length of more than 60 seconds is identified, and the time span of this period is calculated as the ECG stabilization recovery time. If the parameters fail to recover to the stable segment within the preset 300-second timeout limit, the recovery time is directly recorded as 300 seconds to prevent numerical overflow. The calculated heart rate recovery time, blood pressure recovery time and ECG stabilization recovery time of each stage are associated and stored, as shown in Table 4, generating a recovery time feature set containing multiple sets of recovery indicators.

[0091] Table 4. Data on the characteristics of phased physiological recovery (Table 4)

[0092] Phase-01 10:15:00.000 45.5 52.0 55.0 Delayed recovery Phase-02 10:25:30.000 30.2 35.5 32.0 Normal recovery Phase-03 10:35:10.000 60.0 75.8 68.5 Significant delay

[0093] The instruction generation submodule aligns the execution time with the recovery time feature set and the corresponding stage myocardial load discrimination result, calculates the heart rate recovery time difference, blood pressure recovery time difference and the rate of change of the length of the electrocardiogram stable interval between adjacent stages, analyzes the recovery change trend vector and performs rule mapping to generate a nursing decision instruction set.

[0094] The system reads the recovery time feature set and the corresponding stage myocardial load discrimination results. First, a time alignment operation is performed, pairing stages belonging to the same intervention type but in different time sequences. For example, the current expansion stage is compared with the previous expansion stage, or the current withdrawal stage is compared with the baseline stage. The heart rate recovery time difference between adjacent stages is calculated, i.e., the heart rate recovery time of the current stage is subtracted from the heart rate recovery time of the previous stage. Similarly, the blood pressure recovery time difference is calculated. For the electrocardiogram (ECG) signal, the rate of change of the ECG stable interval length between adjacent stages is calculated, and this rate of change is mapped to an equivalent recovery time difference through a preset conversion coefficient. Then, the recovery trend vector analysis logic is invoked. Based on the differences in heart rate, blood pressure, and ECG values, a multi-dimensional feature space is constructed. The weighted fusion distance of the differences in each indicator is calculated to quantify the degree of decline in recovery ability. Based on the calculation results, the recovery trend is divided into three levels: "rapid compensation," "delayed recovery," and "risk of decompensation." In the formula... The eigenvalue represents the feature value of the recovery trend vector calculated based on the recovery time difference between adjacent stages. This represents the time difference in heart rate recovery between two adjacent interventional phases (e.g., the current expansion phase and the previous expansion phase) after time alignment. This represents the time difference in blood pressure recovery between corresponding stages. This represents converting the rate of change of the length of the ECG stable interval into an equivalent recovery time difference value over time. This represents the dimensionless coefficient used to adjust the weighting of the ECG recovery time difference. 1. Obtaining: The increment of heart rate recovery time is calculated through subtraction, i.e. Seconds. 2. Obtaining: The increment of blood pressure recovery time is calculated through subtraction, i.e. Seconds. 3. Quantitative acquisition: The length of the ECG stable interval is not a direct measure of time delay; it needs to be converted using a "loss rate-time mapping." First, the shortening rate of the stable interval length is calculated: (That is, it shortens by 25%). A preset "maximum tolerable recovery delay constant" is used. This constant is set at 100 seconds based on medical statistics, meaning that if the stable interval completely disappears (shortens by 100%), it is equivalent to extending the recovery time by 100 seconds. Therefore, Seconds. 4. By selecting data from a historical high-risk case database, the mean time variance from the occurrence of electrocardiographic abnormalities to the diagnosis of ischemia was calculated. (e.g., 30) and the variance from heart rate abnormality to diagnosis. (For example, 20). Settings The rationale for this setting is that minute fluctuations in electrocardiographic rhythm are often more specific than drifts in heart rate values; therefore, assigning it a 1.5x weight can improve the model's sensitivity to occult ischemia. Substituting into the formula... :first step, The second step is to calculate the squared difference between the ECG equivalent time and the heart rate recovery time: The third step is to perform a weighted processing on the second item: Step 4: Calculate the sum of the numerators (weighted fusion distance squared): Step 5: Calculate the denominator (normalization factor): Step 6: Perform division to obtain the mean squared error. Step 7: Perform the square root operation to obtain the final feature values. The final numerical result obtained from the above example The results were compared with the preset "instruction generation baseline interval": the interval [0, 10] was set as the "compensatory adaptation zone", [10, 30] as the "functional decoupling warning zone", and >30 as the "high-risk failure zone". The results showed that although the patient's heart rate, blood pressure, and ECG recovery time were significantly prolonged compared to the previous stage (by 18s, 23s, and 25s, respectively), the calculated trend vector eigenvalue of 6.277 still fell within the low range of [0, 10]. This implies that the patient's current physiological decline is a compensatory response occurring simultaneously, the cardiovascular regulatory coupling mechanism remains intact, and no single, independent deterioration has occurred. A low-risk nursing instruction of "continue observation, postpone increasing the intensity of the procedure" will be generated instead of triggering an emergency shutdown, thus avoiding unnecessary surgical interruption while ensuring safety.

[0095] Specifically, such as Figure 2 , 7 As shown, the decision optimization module includes:

[0096] The instruction constraint submodule, based on the nursing decision instruction set and combined with the myocardial discrimination result, obtains the stage identifier and time sequence mark corresponding to multiple instructions, performs a sequence consistency judgment on the triggering relationship of adjacent instructions, and performs sequence rearrangement on inconsistent positions to generate an instruction triggering sequence mapping table.

[0097] Based on the generated initial nursing decision instruction set, which contains raw instruction data without time-series verification, the process begins by traversing each record in the instruction set and parsing out two key fields: "Phase Identifier Code" and "Instruction Generation Timestamp." The Phase Identifier Code uses an incremental numeric encoding format; for example, "Phase-001" represents the first intervention phase, and "Phase-002" represents the subsequent operation phase. An index list is created with the Phase Identifier Code as the key value, and all instructions are initially sorted according to the natural number order of the identifier codes to establish a logical sequential execution relationship. Subsequently, a strict time-series consistency scan is performed, comparing the generation timestamps of adjacent instructions pair by pair. A "minimum logical time interval" of 1000 milliseconds is set to accommodate processing delays. During the scan, if a logically subsequent instruction (such as the instruction corresponding to Phase-002) has a physical timestamp earlier than or equal to that of a preceding instruction (such as the instruction corresponding to Phase-001), a "time-series inversion" anomaly is identified. To address this anomaly, a forced rearrangement and timestamp rewriting operation is performed: the timestamp of the preceding instruction remains unchanged, while the timestamp of the subsequent instruction is modified to the timestamp of the preceding instruction plus the minimum logical time interval. For example, if the time of the "Phase-001" instruction is detected as 10:00:00, while the "Phase-002" instruction is marked as 09:59:58 due to network latency, the timestamp of "Phase-002" will be forcibly corrected to 10:00:01. Furthermore, for multiple concurrent instructions that may be generated within the same phase, a microsecond-level time shift is performed based on the ASCII code order of the instruction type codes to ensure that no two instructions have the exact same timestamp index. After all verification and correction are completed, the adjusted data is repackaged to generate an instruction trigger sequence mapping table with strictly monotonically increasing time attributes.

[0098] The interval correction submodule triggers the sequence mapping table according to the instruction, calls the recovery change trend of the continuous stage, obtains the start and end time markers corresponding to multiple instructions, performs interval overlap detection on adjacent stage time segments, performs interval pruning and splicing operations on the overlapping part, and generates a set of instruction continuous intervals.

[0099] The system invokes the time-series-cleaned instruction trigger sequence mapping table and associates it with the previously calculated recovery trend vector feature value (i.e., the value 6.277 obtained in the previous example). This aims to define a reasonable duration window for each instantaneously triggered instruction. First, a baseline duration is set: 300 seconds for "observation" instructions and 600 seconds for "intervention" instructions. Then, the baseline time is dynamically extended using the recovery trend vector feature value. The calculation logic is as follows: divide the feature value 6.277 by a preset adjustment factor of 10 to obtain an extension coefficient of 0.6277, then add 1 to this coefficient and multiply by the baseline time. For example, for observation instructions, the calculated duration is 300 * 1.6277 = 488 seconds. Using the corrected trigger time of the instruction as the starting point, the calculated duration is accumulated to obtain the end time marker, thus constructing the original time interval for the instruction. Based on this, overlap detection of adjacent time segments is performed. The time-sorted list of intervals is traversed, comparing the end time of the current instruction interval with the start time of the next instruction interval. If the calculation finds that the end time of the current instruction is later than the start time of the next instruction, it indicates a time overlap, with the overlap duration being the difference between the two. For the overlapping portion, a "last-order priority" pruning principle is applied, forcibly truncating the end time of the current instruction to the start time of the next instruction to ensure that the new instruction takes effect on time and without interference. Simultaneously, if the gap time between two adjacent intervals is detected to be less than a preset "micro-fragment threshold" (e.g., 10 seconds), a splicing operation is performed, directly extending the end time of the previous interval to the start time of the next interval, filling the monitoring blind spot without instruction coverage. After the above pruning and splicing processes, a series of consecutive or independent instruction interval sets are generated.

[0100] The conflict resolution submodule, based on the instruction duration interval set and combined with the myocardial discrimination result, detects multiple nursing instructions that exist in parallel within the same time segment, performs mutual exclusion judgment and stage priority sorting on the execution logic of parallel instructions, and generates a nursing decision optimization instruction set.

[0101] The system receives the processed instruction interval set and, combined with the current myocardial discrimination result, performs final semantic cleaning to address potential logical conflicts within the same time window. Although interval correction resolves time overlap, some mutually exclusive instruction markers may still remain in certain merged long intervals. First, a "semantic exclusion matrix" is established, defining the coexistence relationships between different instruction types. For example, the "stop operation" instruction and the "continue pressurization" instruction are marked as mutually exclusive with a mutual exclusion coefficient of 1. Each instruction interval is scanned, retrieving all original instruction action codes contained within it. If an instruction pair with a mutual exclusion coefficient of 1 is detected within the same interval, a priority sorting algorithm is immediately initiated. A preset priority weight table is used: emergency instructions have a weight of 10, treatment instructions have a weight of 7, and monitoring instructions have a weight of 3. For example, if both the "continuous monitoring" instruction (weight 3) and the "immediately stop balloon dilation" instruction (weight 10) exist simultaneously within an overlapping interval, the weight values ​​are compared. The higher-weighted "immediate stop" instruction is retained, while the lower-weighted "continuous monitoring" instruction is marked as "suppressed" and removed from the current execution queue. Furthermore, dynamic verification is performed based on the myocardial discrimination results. If the myocardial state corresponding to the current interval is marked as "positive load," the temporary priority of all "oxygen supply improvement" instructions is automatically increased by 5 points, thus giving them a greater advantage in conflict arbitration. After the above semantic exclusion judgment and priority screening, redundant and conflicting items are eliminated, and the final retained instruction sequence and its corresponding optimization time interval are standardized and encoded to generate the nursing decision optimization instruction set as shown in Table 5.

[0102] Table 5 Results of Nursing Decision-Making Instruction Conflict Resolution and Optimization (Table 5)

[0103] INST-082 Continuous monitoring 10:05:00 INST-083 3 Suppressed none INST-083 Stop expanding 10:05:05 INST-082 10 implement [10:05:05,10:12:00] INST-084 Prepare nitroglycerin 10:12:00 none 7 implement [10:12:00,10:20:00]

[0104] These results demonstrate that the logical conflict between INST-082 and INST-083 was successfully identified and resolved, and the high-risk blocking instruction was prioritized and retained based on clinical safety principles, ensuring the accuracy and safety of nursing decisions.

[0105] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. An intelligent nursing decision support system for patients undergoing interventional cardiology surgery, characterized in that: The system includes: The data analysis module collects and analyzes the operating status of interventional devices, extracts the state switching time points corresponding to catheter advancement, device dwell and dilation states and divides them into interventional stages, collects electrocardiogram, non-invasive blood pressure and heart rate signals and aligns them in time, generates physiological monitoring sequences and transmits them to the feature calculation module; The feature calculation module performs sliding window statistics on the electrocardiogram, non-invasive blood pressure and heart rate signals of the multi-intervention stage based on the physiological monitoring sequence, calculates the systolic blood pressure, RR interval and pulse pressure waveform rise time and removes abnormal windows, generates a physiological response feature set and transmits it to the stress discrimination module. The stress discrimination module, based on the physiological response feature set, performs trend alignment on the changes in systolic blood pressure, RR interval and pulse pressure waveform rise time and calculates the trend consistency index to determine the myocardial stress state level, generates myocardial discrimination results and transmits them to the recovery analysis module. The recovery analysis module determines the end point of the multi-intervention stage based on the myocardial discrimination results, extracts the heart rate and blood pressure recovery time and the electrocardiogram stability interval, and performs correlation with the corresponding stage myocardial discrimination results, calculates the recovery change trend of continuous stages, and generates a nursing decision instruction set. The physiological monitoring sequence includes the catheter advancement phase, the device dwell phase, and the dilation phase. The physiological response feature set includes systolic blood pressure, RR interval, pulse pressure waveform rise time, and abnormal window markers. The myocardial discrimination results include myocardial load status and load end time. The nursing decision instruction set includes load end time point, recovery time characteristics, and recovery change trend. The data analysis module includes: The status acquisition submodule collects and analyzes the operating status of the interventional device, obtains the catheter advancement status, instrument dwell status and dilation operation status and sorts them by timestamp, detects the status change amplitude between adjacent sampling points, marks the time point at the abrupt position in the continuous sampling points, and generates a set of device status change time points. The phase division submodule, based on the set of device state change time points, calls the time corresponding to the catheter advancement state, instrument dwell state and dilation state, performs time difference calculation on the interval between adjacent time points, and divides the continuous time period into intervals in combination with the intervention phase switching benchmark value to generate an intervention phase time interval sequence. The physiological alignment submodule acquires electrocardiogram signals, non-invasive blood pressure signals, and heart rate signals. Based on the interventional stage time interval sequence, it calls the multi-signal sampling time axis, performs resampling and time offset correction on multiple sampling frequency signals, and splices the stage identifier with the corresponding physiological signal to generate a physiological monitoring sequence. The stress discrimination module includes: The trend alignment submodule, based on the physiological response feature set, extracts sampling points from the systolic blood pressure time series, RR interval time series and pulse pressure waveform rise time series in the order of intervention operation stages, performs translation correction on the start timestamps of multiple sequences and performs stage index matching to generate multi-indicator aligned trend sequences. The consistency calculation submodule calls the multi-indicator alignment trend sequence to calculate the trend direction sign of the changes in systolic blood pressure, RR interval, and pulse pressure rise time within the same period and constructs a direction vector. It then performs a consistency comparison based on the angle relationship between the direction vectors to generate a trend consistency index. The stress determination submodule performs interval judgment on the corresponding index values ​​of multiple interventional operation stages and the preset stress determination benchmark value based on the trend consistency index, completes the stage label mapping and integrates the stage sequence information, and generates myocardial discrimination results.

2. The intelligent nursing decision support system for patients undergoing interventional cardiology surgery according to claim 1, characterized in that, The intervention phase switching benchmark value is calculated by statistically analyzing the obtained sliding update benchmark values, summarizing the durations of adjacent phases, and using the sum of the average duration and a preset three times the standard deviation as the benchmark value.

3. The intelligent nursing decision support system for patients undergoing interventional cardiology surgery according to claim 1, characterized in that, The feature calculation module includes: The signal window molecular module, based on the physiological monitoring sequence, continuously extracts fixed-length data segments from the electrocardiogram, non-invasive blood pressure, and heart rate signals at multiple intervention stages according to the time axis, and binds the window number and stage identifier according to the sampling time sequence to generate a staged sliding window sequence. The feature calculation submodule calls the staged sliding window sequence to detect the R wave time point of the electrocardiogram window and calculate the interval between adjacent R waves. It extracts the systolic blood pressure value for the non-invasive blood pressure window and calculates the time difference from the start to the peak of the pulse pressure waveform, generating a set of physiological feature vectors for the window. The anomaly removal submodule, based on the window physiological feature vector set, performs statistical distribution summarization on systolic blood pressure, RR interval and pulse rise time, calculates the corresponding mean and standard deviation and constructs the feature value benchmark interval, marks the window feature vectors that exceed the interval as anomalies, and generates a physiological response feature set.

4. The intelligent nursing decision support system for patients undergoing interventional cardiology surgery according to claim 1, characterized in that, The stress determination benchmark is obtained by summarizing the collected changes in systolic blood pressure, RR interval, and pulse pressure waveform rise time in stages within the complete intervention process, obtaining the numerical distribution range of multiple indicators in multiple stages, forming an indicator set based on the consistency calculation results within the distribution range, performing interval sorting on the indicator set, and selecting the median value for determination.

5. The intelligent nursing decision support system for patients undergoing interventional cardiology surgery according to claim 1, characterized in that, The recovery analysis module includes: The load termination submodule determines the load termination time point of the multi-intervention stage based on the myocardial discrimination result, detects the switching position from the loaded state to the non-load state for the continuous myocardial discrimination marker sequence, calculates the stage boundary timestamp based on the continuity of the myocardial state before and after the switching, and generates a load termination time point sequence. The data recovery submodule obtains the heart rate sampling sequence, blood pressure sampling sequence, and electrocardiogram rhythm marker sequence after the end of the multi-stage process based on the load end time point sequence, calculates the duration for the heart rate and blood pressure to fall back to the stable segment, identifies the stable segment of the electrocardiogram rhythm marker, and obtains the recovery time feature set. The instruction generation submodule aligns the execution time of the recovery time feature set with the corresponding stage myocardial load discrimination result, calculates the heart rate recovery time difference, blood pressure recovery time difference, and ECG stable interval length change rate between adjacent stages, analyzes the recovery change trend vector, performs rule mapping, and generates a nursing decision instruction set.

6. The intelligent nursing decision support system for patients undergoing interventional cardiology surgery according to claim 1, characterized in that, The system also includes: The decision optimization module, based on the nursing decision instruction set and combined with the myocardial discrimination results and the recovery change trend of continuous stages, performs rule constraint judgment and logical adjustment calculation on the instruction triggering order, instruction duration interval and adjacent stage instruction conflict relationship in the nursing decision instruction set, and generates a nursing decision optimization instruction set. The nursing decision optimization instruction set includes the instruction triggering order, instruction duration interval, and instruction conflict resolution results.

7. The intelligent nursing decision support system for patients undergoing interventional cardiology surgery according to claim 6, characterized in that, The decision optimization module includes: The instruction constraint submodule, based on the nursing decision instruction set and combined with the myocardial discrimination result, obtains the stage identifier and time sequence mark corresponding to multiple instructions, performs a sequence consistency judgment on the triggering order relationship of adjacent instructions, and performs sequence rearrangement on inconsistent positions to generate an instruction triggering sequence mapping table. The interval correction submodule triggers the sequence mapping table according to the instruction, calls the recovery change trend of the continuous stage, obtains the start and end time markers corresponding to multiple instructions, performs interval overlap detection on adjacent stage time segments, performs interval pruning and splicing operations on the overlapping part, and generates a set of instruction continuous intervals. The conflict resolution submodule, based on the instruction duration interval set and the myocardial discrimination result, detects multiple nursing instructions existing in parallel within the same time interval, performs logical mutual exclusion judgment and stage priority sorting on the parallel instructions, and generates a nursing decision optimization instruction set.

Images