Automatic anesthesia administration control method
By analyzing historical anesthesia records, classifying sensitive and stable physiological data, and constructing predictive models, the error problem caused by the reliance on experience in traditional anesthesia drug administration control was solved, enabling personalized control of anesthetic concentration and improving safety and accuracy.
Patent Information
- Application Number
- CN202511675263.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-15
- Publication Date
- 2026-02-10
AI Technical Summary
Traditional methods of controlling anesthetic drug administration rely on the experience of anesthesiologists, which are susceptible to fatigue and distraction, resulting in a high risk of dose fluctuations and making it difficult to achieve precise control.
By analyzing historical anesthesia records, physiological data are categorized into sensitive and stable tolerance types. An anesthesia demand prediction model is constructed, and the amplitude and temporal characteristics of physiological data fluctuations are utilized to reduce errors and achieve personalized anesthesia concentration control.
It reduces errors in anesthetic drug administration, improves the accuracy and safety of anesthetic concentration control, and reduces the risk of deep or superficial anesthesia.
Smart Images
Figure CN121506368A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of anesthesia control technology, and in particular to an automated anesthesia drug delivery control method. Background Technology
[0002] Anesthesia is used to ensure that patients achieve an ideal state of unconsciousness and pain relief, and to ensure a smooth postoperative recovery.
[0003] Traditional techniques rely heavily on subjectivity in anesthetic drug administration control, depending on the anesthesiologist's experience and judgment (such as manually adjusting the concentration / injection rate of the vaporizer). This makes them susceptible to factors like fatigue and distraction, leading to high risks of dosage fluctuations. For example, deep anesthesia may cause hypotension, while shallow anesthesia can result in intraoperative awareness events. This also easily leads to errors in anesthetic drug concentration, resulting in inappropriate anesthetic drug administration control. Summary of the Invention
[0004] To overcome the shortcomings of the prior art, this application provides an automated anesthesia drug delivery control method.
[0005] This application provides an automated anesthesia drug delivery control method, the method comprising:
[0006] Step S1: Based on the time-varying amplitude trend characteristics of the physiological data to be tested detected in real time in historical anesthesia records, classify and determine that the ratio of the physiological data items belonging to sensitive physiological data and stable tolerance physiological data is greater than or equal to one-half. If the ratio is greater than or equal to one-half, then implement the amplitude segmentation scheme. Based on the difference between each physiological data item in the physiological data to be tested and the variable reference benchmark, segment the time sequence to be tested to obtain the variable time sequence. Extract the overlapping time sequence comparison results and the difference time sequence comparison results from the overlapping time sequence comparison results of each physiological data item in the variable time sequence to be tested, and statistically analyze the correlation between each physiological item in the physiological data to be tested to obtain physiological item correlation feature one and physiological item correlation feature two.
[0007] Step S2: If the physiological item association feature one and the physiological item association feature two are the same, then extract the variable physiological data belonging to the equidistant time series from the physiological data to be tested, obtain the first physiological data sampling feature, and construct the anesthesia demand prediction model one.
[0008] Step S3: If physiological item association feature one and physiological item association feature two are different, and the time sequence of physiological item association feature one and physiological item association feature two are opposite, then the second physiological data sampling feature is calculated. If physiological item association feature one and physiological item association feature two are different, and the time sequence of physiological item association feature one and physiological item association feature two are cross-related, then the third physiological data sampling feature is calculated. Based on the second physiological data sampling feature or the third physiological data sampling feature, anesthesia demand prediction model two is constructed, the current anesthesia record data of the target patient is obtained, and the current anesthesia demand prediction data of the target patient is obtained based on anesthesia demand prediction model one or anesthesia demand prediction model two.
[0009] Preferably, the historical anesthesia record data of the target patient is obtained, and the real-time detected physiological data is extracted from the historical anesthesia record data to obtain the physiological data to be tested. Based on the time-varying amplitude trend characteristics of each physiological data in the physiological data to be tested, the physiological data to be tested is divided into sensitive physiological data and stable tolerance physiological data.
[0010] The ratio of the number of physiological items belonging to the sensitive physiological data to the number of physiological items belonging to the stable tolerance physiological data is used to obtain the measured proportion value.
[0011] If the measured proportion is greater than or equal to one-half, the measured physiological data is segmented by variation amplitude characteristics, and the variation amplitude segmentation scheme is output.
[0012] Preferably, according to the amplitude segmentation scheme, the variation reference benchmark of physiological data is statistically derived from historical anesthesia record data, and each physiological data in the physiological data to be tested is compared with the variation reference benchmark to obtain the variation amplitude of the physiological data to be tested.
[0013] Based on the different amplitude segments of the physiological changes to be measured, the time series of the physiological data to be measured is segmented accordingly to obtain the time series of the changes to be measured.
[0014] The physiological data in the time series to be tested are compared and overlapped to obtain the comparison results.
[0015] Preferably, overlapping time series are extracted from the comparison results to obtain overlapping time series comparison results, and non-overlapping time series are extracted from the comparison results to obtain different time series comparison results.
[0016] Based on the overlapping time sequence comparison results and the measured physiological variation amplitude, the correlation between each physiological item in the measured physiological data is statistically analyzed to obtain physiological item correlation feature one.
[0017] Based on the time-series comparison results of the differences and the amplitude of the physiological changes to be measured, the correlation between the physiological items in the physiological data to be measured is statistically analyzed to obtain the second physiological item correlation feature.
[0018] Preferably, if the physiological item association feature one and the physiological item association feature two are the same, then the variable physiological data belonging to the equidistant time series are extracted from the physiological data to be tested to obtain the first physiological data sampling feature.
[0019] Obtain the historical concentration values of anesthesia in vivo and the historical supplementary anesthesia concentration of the target patient corresponding to the equidistant time sequence of the first physiological data sampling features, and obtain the historical operation time of the target patient;
[0020] Assess the stability weight of the first physiological data sampling feature, and construct an anesthesia demand prediction model one based on the stability weight of the first physiological data, the historical concentration of anesthesia in vivo, the historical supplementary anesthesia concentration, and the historical operation duration.
[0021] Preferably, if physiological item association feature one and physiological item association feature two are different, and the time series to which physiological item association feature one and physiological item association feature two belong are opposite, then the variable physiological data belonging to the time series of the test data to which the overlapping time series comparison results belong and the test data to which the differential time series comparison results belong are extracted respectively, the second physiological data sampling feature is statistically obtained, and the second physiological data stability weight to which the second physiological data sampling feature belongs is evaluated.
[0022] Preferably, if physiological item association feature one and physiological item association feature two are different, and the time sequence to which physiological item association feature one and physiological item association feature two belong is cross-related, then the time sequence nodes to which physiological item association feature one and physiological item association feature two are different are marked to obtain time sequence marked nodes.
[0023] Based on the time-series marker nodes, the variable physiological data belonging to the time-series marker nodes are extracted from the physiological data to be tested, and the third physiological data sampling features are obtained. The third physiological data stability weight to which the third physiological data sampling features belong is evaluated.
[0024] Based on the first anesthesia demand prediction model and the second or third physiological data stability weights, anesthesia demand prediction model two is constructed.
[0025] The current anesthesia record data of the target patient is obtained. The current anesthesia record data includes current physiological data, the concentration value of anesthesia in the body, and the current operation time. The current anesthesia record data is input into anesthesia demand prediction model one or anesthesia demand prediction model two for testing, and the current anesthesia demand prediction data is output.
[0026] Compared with the prior art, the present invention has the following characteristics and beneficial effects:
[0027] By analyzing historical anesthesia injection data of the target patient over a historical period, and based on the trend characteristics of various physiological data, it is possible to determine which physiological items are sensitive and which are stably tolerated. Since sensitive physiological items are more significantly affected by even slight changes in anesthetic concentration, they are more likely to pose a safety risk to the target patient. By statistically analyzing the ratio of sensitive to stably tolerated physiological items, it is possible to determine whether the target patient's individual characteristics lean towards a higher or lower proportion of sensitive physiological items, and to formulate a differentiated analysis plan accordingly. If the majority of physiological items are sensitive, then the data will be segmented based on the amplitude of physiological data changes as the primary reference, and further validated by time-series characteristics. To reduce errors, the time series is segmented in reverse by dividing the amplitude of the variation, resulting in the time series to be measured. By comparing the segmented time series to which each physiological data belongs in the time series to be measured, the correlation and change patterns between the physiological items in the physiological data to be measured within the time series variation segments of the overlapping and differential time series comparison results are known. This further determines the correlation influence of changes in anesthetic concentration. For cases where the statistically identified physiological item correlation feature one and physiological item correlation feature two are the same or different, feature data prediction models are constructed for both cases. This enhances the selectivity of data analysis methods and reduces the probability or degree of misjudgment caused by a single processing scheme performing a unified analysis of diverse situations. Attached Figure Description
[0028] Figure 1 This is a flowchart illustrating the steps of an automated anesthesia drug delivery control method, which is the main feature of this embodiment. Detailed Implementation
[0029] The present invention will be further described in detail below with reference to the following embodiments.
[0030] Reference Figure 1 An automated anesthesia drug delivery control method, the method comprising the following steps:
[0031] Step S1: Based on the time-varying amplitude trend characteristics of the physiological data to be tested detected in real time from historical anesthesia records, classify and determine that the ratio of physiological data items belonging to sensitive physiological data and stable tolerance physiological data is greater than or equal to one-half. If the ratio is greater than or equal to one-half, then implement the amplitude segmentation scheme. Based on the difference between each physiological data item in the physiological data to be tested and the variable reference benchmark, segment the time sequence to which the physiological data to be tested belongs to obtain the variable time sequence to be tested. Extract the overlapping time sequence comparison results and the difference time sequence comparison results from the overlapping time sequence comparison results of each physiological data item in the variable time sequence to be tested, and statistically analyze the correlation between each physiological item in the physiological data to be tested to obtain physiological item correlation feature one and physiological item correlation feature two.
[0032] Step S2: If physiological item association feature one and physiological item association feature two are the same, then extract the variable physiological data belonging to the equidistant time series from the physiological data to be tested, obtain the first physiological data sampling feature, and construct the anesthesia demand prediction model one.
[0033] Step S3: If physiological item association feature one and physiological item association feature two are different, and the time sequence of physiological item association feature one and physiological item association feature two are opposite, then the second physiological data sampling feature is calculated. If physiological item association feature one and physiological item association feature two are different, and the time sequence of physiological item association feature one and physiological item association feature two are cross-related, then the third physiological data sampling feature is calculated. Based on the second physiological data sampling feature or the third physiological data sampling feature, anesthesia demand prediction model two is constructed, the current anesthesia record data of the target patient is obtained, and the current anesthesia demand prediction data of the target patient is obtained based on anesthesia demand prediction model one or anesthesia demand prediction model two.
[0034] Specifically, by analyzing the historical anesthesia injection data of the target patient over a historical period, the initial assessment is made based on the changing trends of various physiological data to determine which physiological parameters are sensitive and which are stably tolerated. Because sensitive physiological parameters are more significantly affected by even slight changes in anesthetic concentration, they are more likely to pose a safety risk to the target patient. By statistically analyzing the ratio of sensitive to stably tolerated physiological parameters, it is possible to determine whether the target patient's individual characteristics lean towards a higher or lower proportion of sensitive physiological parameters, and to develop a differentiated analysis plan accordingly. If the majority of physiological parameters are sensitive, then the data will be segmented based on the amplitude of changes in physiological data as the primary reference, and further analysis will be conducted using time-series characteristics as auxiliary verification. By combining evidence to reduce errors, the time series is segmented in reverse according to the segmentation of the variation amplitude to obtain the time series to be measured. By comparing the segmented time series to which each physiological data belongs in the time series to be measured, the correlation and change patterns between the physiological items in the physiological data to be measured within the time series variation segments of the overlapping time series comparison results and the differential time series comparison results are known. The correlation influence of the change of anesthetic concentration is further judged. For the same or different cases of the statistical physiological item correlation feature one and physiological item correlation feature two, feature data prediction models are constructed for the two cases, which enhances the selectivity of data analysis methods and reduces the probability or degree of misjudgment caused by a single processing scheme to uniformly analyze diverse cases.
[0035] The specific step S1 includes the following sub-steps:
[0036] The historical anesthesia records of the target patient are obtained, and real-time physiological data are extracted from the historical anesthesia records to obtain the physiological data to be tested. Based on the time-varying amplitude trend characteristics of each physiological data in the physiological data to be tested, the physiological data to be tested is divided into sensitive physiological data and stable tolerance physiological data.
[0037] The ratio of the number of physiological items belonging to the sensitive physiological data category to the number of physiological items belonging to the stable and tolerable physiological data category is used to obtain the measured proportion value.
[0038] If the measured proportion is greater than or equal to one-half, the measured physiological data is segmented by variation amplitude characteristics, and the variation amplitude segmentation scheme is output.
[0039] According to the amplitude segmentation scheme, the reference benchmark for the variation of physiological data is statistically derived from historical anesthesia records. The physiological data to be measured are compared with the reference benchmark to obtain the amplitude of the variation of the physiological data to be measured.
[0040] Based on the different amplitude segments of the physiological changes to be measured, the time series of the physiological data to be measured is segmented accordingly to obtain the time series of the changes to be measured.
[0041] The physiological data of each variable time series to be tested are compared and overlapped to obtain the comparison results.
[0042] The overlapping time series are extracted from the alignment results to obtain the overlapping time series alignment results, and the non-overlapping time series are extracted from the alignment results to obtain the difference time series alignment results.
[0043] Based on the results of the overlap time sequence comparison and the amplitude of the physiological changes to be measured, the correlation between the physiological items in the physiological data to be measured is statistically analyzed to obtain the first physiological item correlation feature.
[0044] Based on the time-series comparison results of the differences and the amplitude of the physiological changes to be measured, the correlation between the physiological items in the physiological data to be measured is statistically analyzed to obtain the second correlation feature of physiological items.
[0045] Specifically, the physiological data to be measured (including heart rate, blood pressure, blood oxygen saturation, and temperature, obtained through multi-parameter monitoring), sensitive physiological data, and stable-tolerance physiological data (time-varying amplitude trend characteristics refer to whether the fluctuation amplitude of various physiological data at different time points tends to a stable state or a non-stable state; if it tends to a stable state, the corresponding physiological data fluctuation is judged as stable-tolerance physiological data; if it tends to a non-stable state, the corresponding physiological data fluctuation is judged as sensitive physiological data), and amplitude segmentation scheme (if the number of physiological items belonging to sensitive physiological data is greater than or equal to the number of physiological items belonging to stable-tolerance physiological data, it indicates that the target patient's individual characteristics are more sensitive to anesthetic drugs). Therefore, the subsequent analysis and processing methods for controlling anesthetic drug concentrations need to be further refined to reduce errors. Using an amplitude-segmentation scheme: Medical guidelines (such as the ASA standards for difficult airway management) rely on absolute numerical boundaries as the basis for decision-making, which naturally aligns with amplitude logic. Individual patient differences lead to significant variations in baseline heart rate, blood pressure, etc. (e.g., higher resting blood pressure in patients with a history of hypertension). If only time slices are used, chronic elevations may be misjudged as acute reactions, while the amplitude method more accurately reflects the true pathophysiological changes. A reference benchmark for variation (e.g., statistically analyzing the trend fluctuations of various physiological data over different historical periods (within the normal effective period of anesthesia, under no abnormal conditions), calculating the median of the highest and lowest amplitudes, and then combining the median values obtained from different historical periods) is used. The data is analyzed by averaging the position values (i.e., obtaining the reference baseline for variation), the amplitude of the physiological variation to be measured (by comparing each physiological data point in the data to be measured with the reference baseline and calculating the difference), the time series of the variation to be measured (for example, different amplitude segments in the amplitude of the physiological variation to be measured are w1, w2, w3, w4, w5, where the maintenance time series of w1 is t1, the maintenance time series of w2 is t2, and so on, then t1, t2, t3, t4, t5 are the time series of the variation to be measured, and the same process is applied to the remaining three physiological data points), and the comparison results (e.g., the time series of temperature variation is t1, t2, t3, t4, t5, and the time series of heart rate variation is i1, i2, i3, i4, i5, i6). The time sequence of changes in blood oxygen saturation is: r1, r2, r3, r4; the time sequence of changes in blood pressure is: l1, l2, l3, l4, l5. The time sequences of these four physiological data are compared to determine if they overlap, in order to facilitate subsequent analysis of the relationships between the four physiological data. The results of the overlapping time sequence comparisons (e.g., the first three time segments overlap: (t1, t2, t3), (i1, i2, i3), (r1, r2, r3), (l1, l2, l3)) can be unified as t1, t2, t3, which is the overlapping time sequence comparison result). The results of the difference time sequence comparisons (e.g., the remaining time segments do not overlap: (t4, t5), (i4, i5, i6), (r4), (l4, l5)).This refers to the difference in time series comparison results), physiological item association feature one (e.g., in the overlapping time series comparison results: the association relationship between the four physiological items in the first three time series t1, t2, t3 (e.g., based on the variation amplitude of the four physiological items in the same time series t1, t2, t3, real-time curve statistics are performed; these four physiological items are interrelated, and the order of association between them will change at different times; based on the sensitive item data with the largest variation amplitude at the same time as the initial associated physiological item, the association order is arranged and statistically analyzed, such as the association relationship feature of heart rate value-blood pressure value-blood oxygen saturation-temperature value, if it is T1)), physiological item association feature two (e.g., based on t4, t5, the variation amplitude of the physiological data in the corresponding t4, t5 time series of the other three physiological items is statistically analyzed; the four physiological items in the time series t4, t5, real-time curve statistics are performed; the four physiological items in the time series t1, t2, t3 ... The variation amplitude within the same time period t4 and t5 is statistically analyzed in real time. This process is repeated for subsequent data points. This yields the correlation between the four physiological data points within t4 and t5. If the variation is u1, similarly, based on i4, i5, and i6, the variation amplitude of the physiological data within the corresponding time periods i4, i5, and i6 is statistically analyzed for the other three physiological data points. The variation amplitude of the four physiological data points within the same time period i4, i5, and i6 is statistically analyzed in real time. This process is repeated for subsequent data points. This yields the correlation between the four physiological data points within i4, i5, and i6. If the variation is u2, similarly, based on r4, l4, and l5, the correlation between the four physiological data points within r4 and l4 and l5 is obtained. If the variation is u3 and u4 respectively, then u1, u2, u3, and u4 constitute the second physiological data point correlation feature.
[0046] The specific step S2 includes the following sub-steps:
[0047] If physiological item association feature one and physiological item association feature two are the same, then the variable physiological data belonging to the equidistant time series are extracted from the physiological data to be tested to obtain the first physiological data sampling feature.
[0048] Obtain the historical concentration values of anesthesia and the historical supplementary anesthesia concentration of the target patient corresponding to the equidistant time sequence of the first physiological data sampling feature, and obtain the historical operation duration of the target patient.
[0049] Assess the stability weight of the first physiological data sampling characteristics, and construct an anesthesia demand prediction model one based on the stability weight of the first physiological data, the historical concentration of anesthesia in vivo, the historical supplementary anesthesia concentration, and the historical operation duration.
[0050] Specifically, the sampling characteristics of the first physiological data include: (if physiological correlation feature one and physiological correlation feature two are the same, it indicates that the sensitive and stable tolerance physiological data of the target patient are regularly affected by changes in anesthetic drug concentration. Then, the equidistant time intervals of the variable physiological data are extracted from the physiological data to be tested. Here, the equidistant time intervals need to be set according to the predetermined anesthesia duration. If the predetermined anesthesia duration is 1 hour, then the equidistant time interval is 10 minutes. The physiological data at each time point of the equidistant time interval are statistically analyzed, which constitutes the sampling characteristics of the first physiological data.) Historical concentrations of anesthetics in vivo (e.g., through gas chromatography-mass spectrometry, which can accurately determine residual metabolites of ancient anesthetics in biological samples (blood / tissue). Headspace gas chromatography is also used to analyze volatiles above liquids / solids.) Gas components, which can directly detect the concentration of anesthetic in exhaled air), first physiological data stability weight (currently, in clinical practice, the stability weight of various physiological indicators is mainly dynamically evaluated by integrating multifunctional monitoring equipment and algorithm models), anesthesia demand prediction model one (such as comparing the first physiological data stability weight with the historical concentration of anesthesia in the body (for example, comparing the stability weight of the four physiological data with the historical concentration of anesthesia in the body respectively, and averaging the obtained ratios), as the value on the y-axis, comparing the historical supplementary anesthesia concentration with the historical operation time, and the obtained ratio as the value on the x-axis, all of which are obtained from multiple equidistant time series time points to construct the correlation curve characteristics of xy. If it is Q1, it is the anesthesia demand prediction model one).
[0051] The specific step S3 includes the following sub-steps:
[0052] If physiological item association feature one and physiological item association feature two are different, and the time series to which physiological item association feature one and physiological item association feature two belong are opposite, then the variable physiological data belonging to the time series of the test data to which the overlapping time series comparison results belong and the test data to which the differential time series comparison results belong are extracted respectively, the second physiological data sampling feature is statistically obtained, and the second physiological data stability weight to which the second physiological data sampling feature belongs is evaluated.
[0053] If physiological item association feature one and physiological item association feature two are different, and the time sequence to which physiological item association feature one and physiological item association feature two belong is cross-related, then the time sequence nodes to which physiological item association feature one and physiological item association feature two are different are marked to obtain time sequence marked nodes.
[0054] Based on the time-series marker nodes, the variable physiological data belonging to the time-series marker nodes are extracted from the physiological data to be tested, and the sampling features of the third physiological data are obtained. The stable weight of the third physiological data to which the sampling features of the third physiological data belong is evaluated.
[0055] Based on the first anesthesia demand prediction model and the stable weights of the second or third physiological data, anesthesia demand prediction model two is constructed.
[0056] Obtain the current anesthesia record data of the target patient. The current anesthesia record data includes the current physiological data, the concentration value of anesthesia in the body, and the current operation time. Input the current anesthesia record data into either anesthesia demand prediction model one or anesthesia demand prediction model two for testing, and output the current anesthesia demand prediction data.
[0057] Specifically, if physiological correlation feature one and physiological correlation feature two are different, and the time series of physiological correlation feature one and physiological correlation feature two are in opposition, it indicates that the distribution characteristics of the sensitive physiological data and stable tolerance physiological data of the target patient affected by changes in anesthetic drug concentration are divided into the first half and the second half. Here, the opposition means that the time series of physiological correlation feature one belongs to the first part of the historical operation time, and the time series of physiological correlation feature two belongs to the second part of the historical operation time. The sum of the two is the historical operation time, and there is no overlap. In order to reduce the error judgment, equidistant time series are extracted from the physiological data to be tested according to the overlapping time series comparison results (based on the overlapping time series comparison results and the difference time series comparison). The results are divided into equal-interval time-series sampling points according to the ratio. For example, if the ratio of overlapping time-series comparison results to differential time-series comparison results is 2:1, then the overlapping time-series comparison results contain 4 equal-interval time-series sampling points, and the differential time-series comparison results contain 2 equal-interval time-series sampling points. The resulting changes in physiological data are then used to obtain the second physiological data sampling feature. The time-series marker node (if physiological item association feature one and physiological item association feature two are different, and the time series of physiological item association feature one and physiological item association feature two are intersecting, the intersecting relationship means that the time series of physiological item association feature one and the time series of physiological item association feature two are adjacent in time, then the time series of physiological item association feature one and the time series of physiological item association feature two are intersecting in time.) The sequence segment is marked with time nodes for adjacent time points, thus obtaining time-series marked nodes. Anesthesia demand prediction model two (correspondingly obtaining the historical in vivo anesthesia concentration, historical supplementary anesthesia concentration, and historical surgical duration corresponding to the second or third physiological data stable weight at each equidistant time point, based on anesthesia demand prediction model one, and so on, if it is Q2), current anesthesia demand prediction data (e.g., substituting current anesthesia record data into Q1 or Q2 for feature data matching comparison, performing the same conditional statistical analysis on the current physiological data as the physiological item association feature one and physiological item association feature two of the above-mentioned anesthesia demand prediction model one; if it matches, then perform feature data matching comparison through Q1, based on the time-series data). The regular development and dynamic changes of the anesthesia demand are used to predict the current anesthesia concentration requirement. Similarly, the current physiological data is subjected to a statistical analysis based on the following conditions: First, the physiological correlation features of the second physiological item are different, and their time sequences are opposite; or, second, they are different, and their time sequences are overlapping. If either condition is met, feature data matching is performed using Q2. This involves evaluating the stability weight of the current physiological data, and then matching the obtained stability weight evaluation results, the concentration of anesthesia in vivo, and the current surgical duration with Q2.Based on the regular dynamic changes in time-series data, the current anesthetic demand concentration can be predicted. To improve the accuracy of the prediction, a "three-compartment model" can be used to simulate drug distribution: the central compartment (blood), the rapidly equilibrating peripheral compartment (vascularized tissues), and the slowly equilibrating peripheral compartment (fat / muscle). This model describes the drug distribution process in the body, quantifies the differences in uptake / release rates in different tissues through a system of differential equations, and predicts the trend of total systemic drug levels over time, thereby validating the rationality of the current anesthetic demand prediction data.
[0058] The above are all preferred embodiments of this application, and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.
Claims
1. An automated anesthesia drug delivery control method, characterized in that, Includes the following steps: Step S1: Based on the time-varying amplitude trend characteristics of the physiological data to be tested detected in real time in historical anesthesia records, classify and determine that the ratio of the physiological data items belonging to sensitive physiological data and stable tolerance physiological data is greater than or equal to one-half. If the ratio is greater than or equal to one-half, then implement the amplitude segmentation scheme. Based on the difference between each physiological data item in the physiological data to be tested and the variable reference benchmark, segment the time sequence to be tested to obtain the variable time sequence. Extract the overlapping time sequence comparison results and the difference time sequence comparison results from the overlapping time sequence comparison results of each physiological data item in the variable time sequence to be tested, and statistically analyze the correlation between each physiological item in the physiological data to be tested to obtain physiological item correlation feature one and physiological item correlation feature two. Step S2: If the physiological item association feature one and the physiological item association feature two are the same, then extract the variable physiological data belonging to the equidistant time series from the physiological data to be tested, obtain the first physiological data sampling feature, and construct the anesthesia demand prediction model one. Step S3: If physiological item association feature one and physiological item association feature two are different, and the time sequence of physiological item association feature one and physiological item association feature two are opposite, then the second physiological data sampling feature is calculated. If physiological item association feature one and physiological item association feature two are different, and the time sequence of physiological item association feature one and physiological item association feature two are cross-related, then the third physiological data sampling feature is calculated. Based on the second physiological data sampling feature or the third physiological data sampling feature, anesthesia demand prediction model two is constructed, the current anesthesia record data of the target patient is obtained, and the current anesthesia demand prediction data of the target patient is obtained based on anesthesia demand prediction model one or anesthesia demand prediction model two.
2. The method for automatically administering anesthesia according to claim 1, characterized in that, Step S1 includes: The historical anesthesia record data of the target patient is obtained, and the real-time physiological data is extracted from the historical anesthesia record data to obtain the physiological data to be tested. Based on the time-varying amplitude trend characteristics of each physiological data in the physiological data to be tested, the physiological data to be tested is divided into sensitive physiological data and stable tolerance physiological data. The ratio of the number of physiological items belonging to the sensitive physiological data to the number of physiological items belonging to the stable tolerance physiological data is used to obtain the measured proportion value. If the measured proportion is greater than or equal to one-half, the measured physiological data is segmented by variation amplitude characteristics, and the variation amplitude segmentation scheme is output.
3. The method for automatically administering anesthesia according to claim 2, characterized in that, Step S1 also includes: According to the amplitude segmentation scheme, the reference benchmark for the variation of physiological data is statistically derived from historical anesthesia records. The physiological data to be tested are compared with the reference benchmark to obtain the amplitude of the variation of the physiological data to be tested. Based on the different amplitude segments of the physiological changes to be measured, the time series of the physiological data to be measured is segmented accordingly to obtain the time series of the changes to be measured. The physiological data in the time series to be tested are compared and overlapped to obtain the comparison results.
4. The method for automatically administering anesthesia according to claim 3, characterized in that, Step S1 also includes: The overlapping time series are extracted from the comparison results to obtain the overlapping time series comparison results, and the non-overlapping time series are extracted from the comparison results to obtain the difference time series comparison results. Based on the overlapping time sequence comparison results and the measured physiological variation amplitude, the correlation between each physiological item in the measured physiological data is statistically analyzed to obtain physiological item correlation feature one. Based on the time-series comparison results of the differences and the amplitude of the physiological changes to be measured, the correlation between the physiological items in the physiological data to be measured is statistically analyzed to obtain the second physiological item correlation feature.
5. The method for automatically administering anesthesia according to claim 4, characterized in that, Step S2 includes: If the physiological item association feature one and the physiological item association feature two are the same, then the variable physiological data belonging to the equidistant time series are extracted from the physiological data to be tested to obtain the first physiological data sampling feature. Obtain the historical concentration values of anesthesia in vivo and the historical supplementary anesthesia concentration of the target patient corresponding to the equidistant time sequence of the first physiological data sampling features, and obtain the historical operation time of the target patient; Assess the stability weight of the first physiological data sampling feature, and construct an anesthesia demand prediction model one based on the stability weight of the first physiological data, the historical concentration of anesthesia in vivo, the historical supplementary anesthesia concentration, and the historical operation duration.
6. The method for automatically administering anesthesia according to claim 5, characterized in that, Step S3 includes: If physiological item association feature one and physiological item association feature two are different, and the time series to which physiological item association feature one and physiological item association feature two belong are opposite, then the variable physiological data belonging to the time series of the test data to which the overlapping time series comparison results belong and the test data to which the differential time series comparison results belong are extracted respectively, the second physiological data sampling feature is statistically obtained, and the second physiological data stability weight to which the second physiological data sampling feature belongs is evaluated.
7. The method for automatically administering anesthesia according to claim 6, characterized in that, Step S3 also includes: If physiological item association feature one and physiological item association feature two are different, and the time sequence to which physiological item association feature one and physiological item association feature two belong is cross-related, then the time sequence nodes to which physiological item association feature one and physiological item association feature two are different are marked to obtain time sequence marked nodes. Based on the time-series marker nodes, the variable physiological data belonging to the time-series marker nodes are extracted from the physiological data to be tested, and the third physiological data sampling features are obtained. The third physiological data stability weight to which the third physiological data sampling features belong is evaluated. Based on the first anesthesia demand prediction model and the second or third physiological data stability weights, anesthesia demand prediction model two is constructed. The current anesthesia record data of the target patient is obtained. The current anesthesia record data includes current physiological data, the concentration value of anesthesia in the body, and the current operation time. The current anesthesia record data is input into anesthesia demand prediction model one or anesthesia demand prediction model two for testing, and the current anesthesia demand prediction data is output.