Anesthesia depth monitoring method based on multivariate data

Abstract

The invention discloses an anesthesia depth monitoring method based on multivariate data, relates to the technical field of anesthesia depth monitoring, and solves the technical problems of interval distortion and inaccurate sudden change trend judgment and traceability caused by single hemodynamic index anomaly detection. Multi-dimensional information such as patient age, basic medical history and preoperative baseline is synchronously extracted, an index judgment interval is generated in combination with an improved algorithm, the monitoring accuracy is improved, temporary interference data such as pain and stress are removed in haemodynamic index processing, and then individual deviation degree detection and distribution rationality detection are carried out, so that the accuracy of monitoring is improved. The defect of only group adjacent difference detection is overcome, the index accuracy is further improved, sudden change authenticity verification is newly added, invalid alarms caused by equipment errors are effectively filtered, meanwhile, an emergency priority tracing mechanism is established, and emergency delay caused by no priority of tracing in the prior art is avoided.

Description

Technical Field

[0001] This invention relates to the field of anesthesia depth monitoring technology, specifically to a method for anesthesia depth monitoring based on multivariate data. Background Technology

[0002] Anesthesia depth monitoring is a core component of clinical anesthesia, and its accuracy is directly related to patient safety during surgery and postoperative recovery. Currently, the commonly used methods for anesthesia depth monitoring in clinical practice mainly revolve around two types of indicators: one is hemodynamic indicators, which indirectly reflect the depth of anesthesia by monitoring the patient's circulatory status; the other is electroencephalogram (EEG) related indicators, which directly assess the depth of anesthesia through central nervous system activity.

[0003] According to patent application CN202410934996.7, an artificial intelligence-based method and system for anesthesia depth classification is disclosed. First, anesthesia dosage is predicted in a planned manner, and then anesthesia depth is classified using the values ​​given in the personalized anesthesia plan and physiological signal data. A long short-term memory artificial neural network incorporating attention mechanisms is used for planned anesthesia dosage prediction, improving the reliability and accuracy of the prediction. Finally, an encoding / decoding network combining physiological signal component segmentation enhancement and improved channel attention is used for quantitative anesthesia depth prediction, improving data usability and providing a technical reference for the overall automated process of anesthesia depth classification.

[0004] However, in existing technologies, the application of hemodynamic parameters mostly relies on general ranges for the population, without fully considering individual differences such as patient age, underlying medical history, and long-term medication history. Secondly, existing technologies only eliminate outliers by comparing adjacent differences in population data, without verifying individual deviations and distribution rationality. This results in excessively broad or narrow parameter ranges, failing to reflect the fluctuation range of parameters under actual anesthesia. At the same time, existing technologies lack a mechanism for verifying the authenticity of sudden changes, easily misjudging equipment errors or random fluctuations as real sudden changes. Furthermore, the lack of priority in tracing the causes of sudden changes may delay life-threatening emergency treatment or blindly adjust anesthetic drugs, worsening the condition. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a method for monitoring anesthesia depth based on multivariate data, which solves the problems of interval distortion caused by single detection of abnormal hemodynamic indicators and inaccurate judgment and tracing of sudden changes.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for monitoring anesthesia depth based on multivariate data, the method specifically comprising the following steps: Medical data of patients is obtained from medical databases, and data classification information is obtained by grouping them by age based on clustering algorithms; Blood pressure and heart rate data of each group of patients were extracted, and an initial interval was generated based on the maximum and minimum values. After removing outliers, the interval was corrected to obtain the index judgment interval. Based on the judgment interval of the aforementioned indicators, the deviation rate, slope of change, and fluctuation range of the indicators at each time point are calculated, and the stable trend, shallow trend, or deepening trend are obtained by comprehensively analyzing the three factors. The trend was verified by combining the patient's bispectral index of EEG, and warning information containing level 1, level 2 and level 3 warning signals was generated.

[0007] As a further aspect of the present invention, the method for obtaining data classification information is as follows: Retrieve medical data for all patients within a time range T from a medical database, wherein the medical data is independently identified for each individual patient. Based on clustering algorithms, patients are classified according to age as the core criterion. They are grouped according to preset age segmentation rules, and the corresponding medical data of each group are associated to generate data classification information.

[0008] As a further aspect of the present invention, the medical data includes medical records and examination results, with the core extracted information being the patient's age; the data classification information includes the number of patients in each age group, the medical dataset, and key features, with the key features including the distribution of disease types and the average value of examination indicators.

[0009] As a further aspect of the present invention, the initial interval is generated as follows: Process patients in age groups sequentially and obtain their blood pressure and heart rate for all patients in the current group. Extract all blood pressure measurements, and filter the minimum and maximum values ​​to form an initial blood pressure range; Extract all heart rate measurements, and filter the minimum and maximum values ​​to form the initial heart rate interval.

[0010] As a further aspect of the present invention, the method for generating the indicator judgment interval includes: Sort all measured values ​​of the current indicator, take the first value after sorting as the initial minimum value, and calculate the difference between it and its adjacent values; If the difference is greater than a preset threshold, the first and second values ​​are determined to be outliers and removed, and the adjacent values ​​are used as the new minimum values. The maximum value is verified and corrected in the same way to obtain the indicator judgment range.

[0011] As a further aspect of the present invention, the deviation rate, the slope of change, and the fluctuation range are calculated as follows: Deviation rate = (actual value - interval midpoint) / interval midpoint × 100%, where the interval midpoint is the average of the upper and lower limits of the interval; Slope of change = (Indicator value at current time point - Indicator value at previous time point) / Time interval; 15-minute fluctuation range = maximum value of indicator within 15 minutes - minimum value of indicator within 15 minutes.

[0012] As a further aspect of the present invention, the time interval is dynamically set according to the anesthesia stage: The 0-30 minute induction period is set to 1 minute; The maintenance period from 30 minutes to 30 minutes before the end of the procedure is set at 5 minutes; The recovery period from 30 minutes before the end of the procedure to full consciousness is set at 3 minutes.

[0013] As a further aspect of the present invention, the method for determining the trend includes: If the following conditions are met simultaneously: deviation rate < ±5%, absolute slope value < 3, and fluctuation range < 50% of the interval width, it is judged as a stable trend. If the deviation rate is greater than +10% for three consecutive time points and the slope is positive, it is judged as a shallowing trend. If the deviation rate is less than -10% for three consecutive time points and the slope is negative, it is judged as a deepening trend; If the absolute value of the slope is greater than 5, it is determined to be a sudden change trend and a sudden change analysis signal is generated.

[0014] As a further aspect of the present invention, the analysis method for the sudden change analysis signal is as follows: If the sudden change occurs only once and there are no other abnormal indicators, it is judged as a false positive sudden change, and only intensified monitoring is required. If two consecutive time points show sudden changes, or multiple indicators show synchronous abnormalities, it is determined to be a real sudden change and an abnormal warning message is generated.

[0015] As a further aspect of the present invention, after determining the actual sudden change, a 2-minute stratified source tracing is also included: Prioritize screening for non-anesthesia-related acute risk factors, including circulatory emergencies, respiratory emergencies, drug allergies or toxic reactions, and surgical procedure irritation; After excluding acute risk factors, anesthesia-related precipitating factors were analyzed, including abnormal dosage of anesthetic drugs, drastic fluctuations in the depth of anesthesia, and the effects of adjuvant medications.

[0016] As a further aspect of the present invention, the method for verifying the trend includes: If the hemodynamics are determined to be stable and the BIS value is within the range of 40-60, the depth of anesthesia is confirmed to be normal. If hemodynamics indicate a tendency toward superficial anesthesia and BIS > 60, then the anesthesia is confirmed to be too superficial. If hemodynamics indicate a deepening trend and BIS < 40, the anesthesia is confirmed to be too deep. If the hemodynamic trend conflicts with the BIS value, the BIS value is used to generate anesthesia depth information.

[0017] As a further aspect of the present invention, the method for generating early warning information is as follows: The anesthesia depth information is compared with the preset safety range and combined with dynamic trend analysis; If the anesthesia depth information is within a safe range but shows a tendency to become shallower, an early warning prompt will be triggered and classified: A Level 1 warning signal is generated if only a single indicator deviates slightly. A level-two warning signal is generated when multiple indicators show slight abnormalities. If the anesthesia depth information exceeds the safe range, a level three warning signal will be generated directly.

[0018] This invention provides a method for monitoring anesthesia depth based on multivariate data. Compared with existing technologies, it has the following advantages: This invention employs a two-tiered grouping system, consisting of a primary age group and a subgroup of patients with underlying medical conditions, to simultaneously extract multi-dimensional information such as patient age, basic medical history, and preoperative baseline. Combined with an improved algorithm, it generates indicator judgment intervals to enhance monitoring accuracy. In hemodynamic indicator processing, temporary interference data such as pain and stress are first removed. Then, individual deviation detection and distribution rationality detection compensate for the shortcomings of simply detecting adjacent differences within a group, further improving indicator accuracy. A new verification of sudden changes is added to effectively filter invalid alarms caused by equipment errors. Simultaneously, an emergency priority tracing mechanism is established to avoid delays in emergency care caused by the lack of priority in existing tracing technologies. Attached Figure Description

[0019] Figure 1 This is a diagram illustrating the overall steps and methods of the present invention; Figure 2 This is a simplified flowchart of the first embodiment; Figure 3 This is a simplified flowchart of the second embodiment; Figure 4 This is a simplified flowchart of the third embodiment; Figure 5 This is a simplified flowchart of the fourth embodiment. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Please see Figure 1 This application provides a method for monitoring the depth of anesthesia based on multivariate data, which specifically includes the following steps: First Embodiment Retrieves medical data of all patients within a time range T from a designated medical database. The specific value of time T is set by the operator. The time range traces backward from the current operation time. The acquired medical data is independently identified by each patient, denoted as i, where i = 1, 2, ..., j, and j represents the total number of patients within that time range. By default, each patient corresponds to a set of independent medical data, which includes medical records, examination results, etc. The personal information of the above patients is obtained simultaneously. The core information extracted is the patient's age. Other personal information, such as gender and basic medical history, can be supplemented according to the analysis needs. The acquired medical data is analyzed based on clustering algorithms. The analysis process uses patient age as the core classification criterion and groups patients according to preset age segmentation rules, such as children, youth, middle-aged, and elderly groups. The segmentation criteria can be flexibly configured, and the medical data corresponding to each group of patients is associated with them. Finally, data classification information is generated, which must include: the number of patients in each age group, the medical data set corresponding to each group, and the key features of the medical data of each group. The key features include disease type distribution, average examination indicators, etc. For example, among the acquired medical data, there is patient A, aged 35; patient B, aged 50; and patient C, aged 20. Then, the system classifies all medical patients according to their age. Assuming that the age range is set as follows: 0-20 years old is the adolescent group, 21-60 years old is the middle-aged and young group, and 61 years old and above is the elderly group, then patient C in the example above would be classified as the adolescent group, patient A as the middle-aged and young group, and patient B as the middle-aged and young group.

[0022] Second Embodiment As a second embodiment of the present invention, it is implemented based on the first embodiment, and the difference from the first embodiment is as follows: Based on the generated age group data classification information, the data is processed sequentially by group, with each group being an independent analysis object. The medical data within the current analysis group are labeled as n (n=1, 2, ..., m), where m is the total number of patients in that group. Each group, by default, contains medical data from m patients. Hemodynamic parameters for all patients in the current group are obtained, including blood pressure and heart rate, and processed independently. Statistical analysis is performed, and all blood pressure measurements are extracted, with the minimum value (BP) selected. min初 and maximum value BP max初 This forms the initial blood pressure range [BP] min初 BP max初 Similarly, extract all heart rate measurements and filter out the minimum HR value. min初 and maximum value HR max初 This forms the initial heart rate zone [HR] min初 HRmax初 ]; For example, the first patient's blood pressure fluctuated between 90-130 mmHg and heart rate between 70-100 beats / minute; the second patient's blood pressure was 85-120 mmHg and heart rate between 65-95 beats / minute, and so on. Hemodynamic parameters were obtained for all patients. Statistical analysis revealed that the minimum blood pressure was 80 mmHg and the maximum was 140 mmHg. Therefore, the corresponding blood pressure index range of 80-140 mmHg was generated, and the heart rate index range was generated similarly.

[0023] Anomaly detection is performed on the minimum and maximum values ​​of the initial interval. All measured values ​​of the current indicator are sorted in ascending order to obtain x1≤x2≤…≤x k Let k be the total number of measurements for the indicator. Take the first value after sorting, x1, which is the initial minimum value. Calculate the difference Δx = x2 - x1 between x1 and its adjacent value x2. Compare Δx with a preset threshold, which is set by the operator based on clinical standards or historical data. If Δx ≤ the preset threshold: x1 is determined to be a normal fluctuation value and retained as the effective minimum value; if Δx > the threshold: x1 is determined to be an abnormal value, x1 is removed, and x2 is taken as the new minimum value. Similarly, the maximum value is verified, and the initial interval is corrected to obtain the indicator judgment interval. The indicator judgment interval here includes the blood pressure and heart rate intervals. Third Embodiment As a third embodiment of the present invention, it is implemented based on the second embodiment, and the difference from the second embodiment is as follows: Based on the obtained indicator judgment interval, calculate the deviation rate, change slope and 15-minute fluctuation range of the indicator at each time point. The deviation rate is calculated as (actual value - interval median) / interval median × 100%, and the interval median is calculated as (upper limit of interval + lower limit of interval) / 2 (e.g., for blood pressure intervals of 85-135 mmHg, the median is 110 mmHg). The real-time indicator values ​​must be valid measurements at the same time point. The slope of change = (current time point index value - previous time point index value) / time interval, and the time interval is dynamically set according to the anesthesia stage. The interval is 1 minute during the induction period from 0 to 30 minutes, 5 minutes during the maintenance period from 30 minutes to 30 minutes before the end of the operation, and 3 minutes during the awakening period from 30 minutes before the end of the operation. The 15-minute fluctuation range = the maximum value of the indicator within 15 minutes - the minimum value of the indicator within 15 minutes. The data is traced back 15 minutes from the current time point (if it is less than 15 minutes, the actual monitoring time shall prevail). Simultaneously, the obtained parameters are compared with the indicator judgment interval. If the deviation rate is < ±5%, the absolute value of the slope is < 3 times / min / minute or < 10 mmHg / minute, the fluctuation range is < 50% of the interval width, and the interval width = upper limit of the interval - lower limit of the interval (e.g., blood pressure interval width = 135 - 85 = 50 mmHg), it is judged as a stable trend. If the deviation rate is > +10% for three consecutive time points and the slope is positive, it is judged as a shallowing trend, specifically indicating that the anesthesia may be too shallow. If the deviation rate is < -10% for three consecutive time points and the slope is negative, it is judged as a deepening trend, specifically indicating that the anesthesia may be too deep. If the absolute value of the slope is > 5 times / min / minute or > 15 mmHg / minute, it is judged as a sudden change trend, and a sudden change analysis signal is generated and analyzed. The validity of the data corresponding to the sudden change is analyzed. If the sudden change occurs only once and there are no other abnormal indicators, it is judged as a false positive sudden change. No intervention is required, only intensified monitoring is needed. If the sudden change is shown at two consecutive time points, or multiple indicators are abnormal at the same time, it is judged as a real sudden change and an abnormal warning information is generated. It also includes a 2-minute stratified source tracing: priority is given to investigating non-anesthesia related acute risk factors, including circulatory system emergencies (major bleeding, arrhythmia), respiratory system emergencies (tracheal tube dislodgement, ventilation interruption), drug allergy or toxic reaction and surgical procedure stimulation. After excluding acute risk factors, anesthesia-related precipitating factors were analyzed, including abnormal dosage of anesthetic drugs, drastic fluctuations in the depth of anesthesia, and the effects of adjuvant medications.

[0024] Fourth embodiment As a fourth embodiment of the present invention, it is implemented based on the third embodiment, and the difference from the third embodiment is as follows: Based on the stable trend, shallowing trend, and deepening trend obtained from the analysis, the patient's BIS value is used for verification analysis. If the hemodynamics is determined to be stable and the BIS is within the range of 40-60, it means that the two are consistent and the anesthesia depth is confirmed to be normal. If the hemodynamics is determined to be shallowing and the BIS is >60, it means that the two are consistent and the anesthesia is too shallow. If the hemodynamics is determined to be deepening and the BIS is <40, it means that the two are consistent and the anesthesia is too deep. If the hemodynamics and BIS conflict, such as hemodynamics being stable but BIS=65, the BIS is taken as the standard to determine the corresponding anesthesia depth and anesthesia depth information is generated. The generated anesthesia depth information is compared with the preset safe range of anesthesia depth. The preset safe range of anesthesia depth is determined based on a large amount of clinical data and expert experience. At the same time, it is combined with dynamic trends for comprehensive analysis. If the anesthesia depth information is within the safe range but shows a trend of shallowing, an early warning is triggered. At the same time, the early warning response is graded and analyzed. If there is a slight deviation of a single indicator, a first-level early warning signal is generated. If multiple indicators are slightly abnormal, a second-level early warning signal is generated. If the anesthesia depth information is not within the safe range, a level three warning signal will be generated directly; After generating a Level 1 warning signal, the monitoring frequency is doubled and the warning time and related data are recorded. After generating a Level 2 warning signal, an audible and visual alarm is issued and a suggested intervention plan is displayed. The patient's recent monitoring data is pushed to the anesthesiologist's workstation. After generating a Level 3 warning signal, a high-decibel audible and visual alarm is issued, the emergency call device outside the operating room is triggered, the infusion of non-essential anesthetic drugs is automatically suspended, and the emergency treatment procedure is displayed.

[0025] Fifth embodiment As a fifth embodiment of the present invention, the focus is on implementing all of the above embodiments in combination.

[0026] Some of the data in the above formulas are numerical calculations with dimensions removed, and the contents not described in detail in this specification are all prior art known to those skilled in the art.

[0027] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A method for monitoring depth of anesthesia based on multivariate data, characterized by, The method specifically comprises the following steps: Obtaining medical data of patients from a medical database, grouping the data according to age based on a clustering algorithm to obtain data classification information; Extracting blood pressure and heart rate of patients in each group, generating an initial interval according to maximum and minimum values, and correcting the initial interval after removing abnormal values to obtain an index judgment interval; Based on the index judgment interval, calculating a deviation rate, a change slope and a fluctuation range of an index at each time point, and comprehensively analyzing the three to obtain a stable trend, a shallow trend or a deep trend; Combining the brain electrical double frequency index of the patient to verify the trend, and generating early warning information containing first, second and third early warning signals.

2. The method for monitoring depth of anesthesia based on multi-element data according to claim 1, characterized in that, The data classification information is obtained in the following manner: Obtaining medical data of all patients within a time range T from a medical database, wherein the medical data is independently identified according to patients; Grouping according to a preset age segmentation rule based on a clustering algorithm taking patient age as the core classification basis, and associating the medical data of each group to generate data classification information.

3. The method for monitoring anesthesia depth based on multivariate data according to claim 2, characterized in that, The medical data includes diagnosis and treatment records and examination results, and the core extraction information is patient age; the data classification information includes the number of patients, medical data sets and key features of each age group, and the key features include disease type distribution and examination index mean value.

4. The method of claim 1, wherein, The initial interval is generated in the following manner: Processing in sequence according to age grouping, obtaining blood pressure and heart rate of all patients in the current grouping; Extracting all blood pressure measurement values, selecting minimum and maximum values to form an initial blood pressure interval; Extracting all heart rate measurement values, selecting minimum and maximum values to form an initial heart rate interval.

5. The method for monitoring depth of anesthesia based on multi-element data according to claim 1, characterized in that, The index judgment interval is generated in the following manner: Sorting all measurement values of the current index, taking the first value after sorting as the initial minimum value, and calculating the difference between the initial minimum value and the adjacent value; If the difference is greater than a preset threshold, the first value is determined to be an abnormal value and is removed, and the adjacent value is taken as a new minimum value; Verifying and correcting the maximum value in the same way to obtain the index judgment interval.

6. The method for monitoring depth of anesthesia based on multi-element data according to claim 1, wherein, The deviation rate, change slope and fluctuation range are calculated in the following manner: Deviation rate=(actual value-interval median value) / interval median value*100%, wherein the interval median value is the average of the upper and lower limits of the interval; Change slope=(current time point index value-previous time point index value) / time interval; 15-minute fluctuation range=15-minute maximum index value-15-minute minimum index value.

7. The method for monitoring anesthesia depth based on multivariate data according to claim 6, characterized in that, The time interval is dynamically set according to the anesthesia stage: The induction period of 0-30 minutes is set to 1 minute; The maintenance period of 30 minutes to 30 minutes before the end of the operation is set to 5 minutes; The recovery period from 30 minutes before the end of the operation to wake up is set to 3 minutes.

8. The method of claim 1, wherein, The trend is determined in the following manner: If the deviation rate is less than ±5%, the absolute value of the slope is less than 3, and the fluctuation range is less than 50% of the interval width, it is determined to be a stable trend; If the deviation rate is greater than +10% and the slope is positive at three consecutive time points, it is determined to be a shallow trend; If the deviation rate is less than -10% and the slope is negative at three consecutive time points, it is determined to be a deep trend; If the absolute value of the slope is greater than 5, it is determined to be a sudden change trend and a sudden change analysis signal is generated.

9. The method of claim 8, wherein, The sudden change analysis signal is analyzed in the following manner: If the sudden change occurs only once and there is no other index abnormality, it is determined to be a false positive sudden change and only encrypted monitoring is performed; If both of the two time points show sudden changes, or multiple indicators show abnormal changes simultaneously, it is determined as a real sudden change and an abnormal warning information is generated.

10. The method of claim 9, wherein, After determining the real sudden change, 2-minute hierarchical traceability inducement is further included: Prioritize the investigation of non-narcotic related acute risk factors, including circulatory system emergencies, respiratory system emergencies, drug allergies or toxicity reactions, and surgical operation stimulation; After excluding acute risk factors, analyze the anesthesia related inducements, including abnormal anesthesia drug dosage, anesthesia depth fluctuation, and auxiliary drug influence.

11. The method of claim 1, wherein, The manner of verifying the trend includes: If the hemodynamics is determined to be stable and the BIS value is in the range of 40-60, it is confirmed that the anesthesia depth is normal; If the hemodynamics is determined to be shallow and BIS>60, it is confirmed that the anesthesia is too shallow; If the hemodynamics is determined to be deep and BIS<40, it is confirmed that the anesthesia is too deep; If the hemodynamics trend conflicts with the BIS value, the anesthesia depth information is generated according to the BIS value.

12. The method of claim 1, wherein, The generation manner of the warning information is: Compare the anesthesia depth information with the preset safety range, and combine with dynamic trend analysis; If the anesthesia depth information is within the safety range but shows a shallow trend, trigger a warning prompt and classify: Only a single indicator slightly deviates to generate a first-level warning signal; Multiple indicators show slight abnormalities to generate a second-level warning signal; If the anesthesia depth information exceeds the safety range, a third-level warning signal is directly generated.

Images