Anesthesia simulation analysis method based on data driving and ultrasonic image

Abstract

The invention provides an anesthesia simulation analysis method based on data driving and ultrasonic images, which makes full use of the advantage of mass data of assisting anesthesia practice in an ultrasonic guidance mode, and is used for an anesthesia prediction model training process based on a particle swarm optimization algorithm through data extraction processing. The trained prediction model can simulate and output the optimal anesthesia effect conforming to the premise of safety and normal metabolism for individuals with different characteristics and medication and retardation areas, so that powerful support can be provided for anesthesia-related theoretical research and practical operation training.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of ultrasound image assisted anesthesia simulation and analysis, and particularly relates to an anesthesia simulation analysis method based on data driving and ultrasound images. BACKGROUND

[0002] Anesthesia ultrasound images have a wide range of applications in the medical field, especially in ultrasound guided regional anesthesia (UGRA), which can assist doctors in precise puncture and local anesthetic injection. In anesthesia operations, ultrasound imaging technology is widely used in scenarios such as tracheal intubation, nerve block, and blood vessel positioning due to its non-invasive and real-time nature. However, due to individual differences in organ and tissue volume, depth of location in the body, and shifts caused by breathing, body posture, and other factors during actual drug administration, the current ultrasound guided method has limited effectiveness, and the specific operation and final results still rely too much on the experience of medical personnel, so there is an urgent need in the field to establish complete statistical data and reliable guidance theory. SUMMARY

[0003] Therefore, in view of the technical problems existing in the field, the present application provides an anesthesia simulation analysis method based on data driving and ultrasound images, which specifically includes the following steps:

[0004] Step one, collect historical data of regional anesthesia process using ultrasound image guided method; the historical data at least includes: patient age, gender, weight, basic value and intraoperative value including blood pressure, heart rate and respiration index, intraoperative blood pressure, intraoperative heart rate, intraoperative respiration index, basic value and intraoperative index of related organs and / or tissues, specific organ and / or tissue type of block region and its ultrasound image coordinates and brightness data, BIS index (bispectral index), frequency and amplitude parameters in EEG (electroencephalogram), drug type, drug dose, and anesthesia time data;

[0005] Step two, perform standardization processing on part of the data collected in step one, including: dividing the specific range of age, weight, basic value of each physiological index and intraoperative change rate, related organ or tissue index change, drug dose and anesthesia time into several segments, and assigning corresponding data labels to each segment; dividing different needle regions according to the ultrasound image coordinates of specific organs and / or tissues, and assigning corresponding region labels; dividing the specific range of brightness of specific organs and / or tissues ultrasound image into several brightness segments, and assigning corresponding brightness labels; at the same time, combining anesthesia time, dividing different brightness change rate segments and assigning corresponding brightness change rate segment labels;

[0006] Step three, use the historical data processed in step two to establish a training sample set;

[0007] Step four, establish an anesthesia effect prediction model based on a particle swarm optimization algorithm, taking patient age, sex, weight, basic values of various physiological indicators, specific organ and / or tissue type, drug type, dose, and needle insertion area as input, and the brightness of the ultrasound image within a predetermined range around the needle insertion area, BIS index, and EEG frequency and amplitude, anesthesia time as output, and comprehensively consider the safety of intraoperative drug use and drug metabolism to establish the corresponding objective function; use the training sample set to train the anesthesia effect prediction model until it converges stably or reaches the maximum training times;

[0008] Step five, use the trained anesthesia effect prediction model to simulate and analyze the anesthesia operation, which outputs the drug effect corresponding to different anesthesia times according to the specific organ and / or tissue to be blocked in the region, the drug to be used and the dose, and the individual characteristics including age, gender, and weight. Among them, the brightness of the ultrasound image within a predetermined range of the needle insertion area is used to represent the degree of local diffusion of the drug, the BIS index is used to represent the depth of anesthesia, and the EEG frequency and amplitude are used to reflect the specific anesthesia effect of different types of drugs.

[0009] Further, the objective function in step four is specifically established by weighted fusion of the change rate of the physiological indicators relative to the basic value, the change rate of the ultrasound image brightness within a predetermined range of the needle insertion area, and the basic value and intraoperative indicators of the related organs and / or tissues. Among them, the change rate of the physiological indicators relative to the basic value is used to ensure the safety of anesthesia, and the change rate of the ultrasound image brightness within a predetermined range of the needle insertion area and the basic value and intraoperative indicators of the related organs and / or tissues are used to balance the onset and metabolism of anesthetic drugs.

[0010] Further, the basic value and intraoperative indicators of the related organs and / or tissues specifically include related indicator parameters of the liver, kidneys, peripheral blood leukocytes, etc.

[0011] Further, the respiratory indicators used by the method specifically include respiratory cycle, respiratory depth, real-time respiratory displacement and phase data, which are used to realize respiratory offset compensation of the anesthesia effect prediction model output.

[0012] Further, for anesthesia simulation analysis of specific soft tissues such as fascia, three-dimensional reconstructed ultrasound image coordinate data and highlighted image brightness data are specifically used, and intraoperative human body posture data including tissue stretching degree, related joint angle, and supine posture are extracted, which are used to construct a training sample set and as input data during model training.

[0013] Further, the historical data of the regional anesthesia process using the ultrasound image guided method is periodically re-collected, and the anesthesia effect prediction model is re-trained and updated.

[0014] The anesthesia simulation analysis method based on data driving and ultrasonic images provided by the present application fully utilizes the advantages of massive data of ultrasonic guidance mode assisted anesthesia practice, and is subjected to data extraction processing and used for anesthesia prediction model training process based on particle swarm optimization algorithm. The trained prediction model can simulate output the best anesthesia effect meeting the safety and normal metabolism premise for different characteristic individuals, drug use and block regions, thereby being capable of providing strong support for anesthesia related theoretical research and practical operation training. BRIEF DESCRIPTION OF DRAWINGS

[0015] Figure 1 The flowchart of the method provided by the present application. DETAILED DESCRIPTION

[0016] The technical solutions of the present application will be described clearly and completely in combination with the drawings. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the protection scope of the present application.

[0017] The anesthesia simulation analysis method based on data driving and ultrasonic images provided by the present application fully utilizes the advantages of massive data of ultrasonic guidance mode assisted anesthesia practice, and is subjected to data extraction processing and used for anesthesia prediction model training process based on particle swarm optimization algorithm. The trained prediction model can simulate output the best anesthesia effect meeting the safety and normal metabolism premise for different characteristic individuals, drug use and block regions, thereby being capable of providing strong support for anesthesia related theoretical research and practical operation training. Figure 1 As shown in the figure, the method specifically includes the following steps:

[0018] Step one, collecting the historical data of regional anesthesia process using ultrasonic image guided mode; the historical data at least includes: patient age, gender, weight, basic value and intraoperative value of blood pressure, heart rate and respiration index, intraoperative blood pressure, intraoperative heart rate, intraoperative respiration index, basic value and intraoperative index of related organs and / or tissues, these human individual characteristics are important factors for determining drug type, drug dosage difference and anesthesia time in practical operation; since the diffusion of anesthetic in the local needle area will cause obvious brightness distribution change of ultrasonic image, therefore, the specific organ and / or tissue type of block region and its ultrasonic image are selected, the coordinate and brightness data are extracted after performing grid processing on the related image area, and the frequency and amplitude parameters in BIS index (bispectral index) and EEG (electroencephalogram) are selected to represent anesthesia effect;

[0019] Step two, standardize the part of data collected in step one, including: dividing the specific range of age, weight, each physiological index basic value and intraoperative change rate, related organ or tissue index change, drug dosage and anesthesia time into several sections, and assigning corresponding data tags to each section; according to the coordinates of the specific organ and / or tissue ultrasound image, different needle insertion areas are divided, and corresponding area tags are assigned; the specific range of brightness of the specific organ and / or tissue ultrasound image is divided into several brightness sections, and corresponding brightness tags are assigned; at the same time, combined with anesthesia time, different brightness change rate sections are divided and corresponding brightness change rate section tags are assigned; through this data standardization processing method, the calculation overhead during subsequent model training can be effectively reduced, and by adjusting the fineness of each section and data tag, the expected accuracy of the prediction model can be adjusted;

[0020] Step three, using the historical data processed in step two to establish a training sample set;

[0021] Step four, establish an anesthesia effect prediction model based on particle swarm optimization algorithm, taking patient age, sex, weight, each physiological index basic value, specific organ and / or tissue type, drug type, dosage and needle insertion area as model input, and taking the brightness of ultrasound image within the predetermined range around the needle insertion area, BIS index, EEG frequency and amplitude and anesthesia time as model output, and considering the safety of intraoperative drug use and drug metabolism to establish the corresponding objective function; using the training sample set to train the anesthesia effect prediction model until it converges stably or reaches the maximum training times;

[0022] Step five, using the trained anesthesia effect prediction model to simulate and analyze the anesthesia operation, so as to predict the drug effect corresponding to different anesthesia time according to the specific organ and / or tissue to be blocked in the region, the drug to be used and the dosage, and the individual characteristics including age, gender and weight, wherein the brightness of ultrasound image within the predetermined range of needle insertion area is used to represent the degree of local diffusion of the drug, and the BIS index is used to represent the anesthesia depth level; Since the EEG spectrum will show obvious frequency band difference and amplitude change for different drugs, for example, the drug effect of dexmedetomidine is highly related to the change of slow wave / δ wave in EEG spectrum, while the drug effect of ketamine is highly related to β / γ wave, based on this consideration, the present application uses EEG frequency and amplitude data and BIS index to more accurately reflect the drug effect of specific drugs.

[0023] In a preferred embodiment of the present application, the objective function in step four is specifically established by weighting the change rate of the physiological index relative to the baseline value, the brightness change rate of the ultrasound image in the predetermined range of the needle insertion area, and the baseline value and intraoperative index of the related organs and / or tissues; wherein the change rate of the physiological index relative to the baseline value is used to ensure safety in anesthesia; since drug effect and metabolic failure are also factors that need to be considered in practice, improper use of drugs can cause many risks and health hazards, therefore in this embodiment, the brightness change rate of the ultrasound image in the predetermined range of the needle insertion area and the baseline value and intraoperative index of the related organs and / or tissues are used to balance the diffusion onset speed and metabolism of anesthetic drugs, so that the model prediction result is more meaningful.

[0024] In a preferred embodiment of the present application, the baseline value and intraoperative index of the related organs and / or tissues specifically include related indicator parameters of the liver, kidneys, peripheral blood leukocytes, etc. to assist in representing the onset and metabolism of drugs, which can further ensure the prediction accuracy of the model in the face of different individual differences.

[0025] In specific practice, the respiratory cycle, respiratory depth, real-time respiratory phase in the respiratory index are highly related to the respiratory offset during ultrasound guidance. Therefore, in a preferred embodiment of the present application, these indicators are selected to achieve better model prediction of respiratory offset compensation. For some anesthesia operations that are less affected by respiration, relatively rough respiratory indicator parameters can be selected to reduce computational overhead.

[0026] In anesthesia for minimally invasive surgery such as arthroscopy, laparoscopy, etc., the drug block area is often in soft tissues such as peripheral nerve fascia, which is prone to large tissue deformation and traction with body posture changes. At this time, if two-dimensional ultrasound images are used, it will seriously affect the accuracy of the needle insertion position and depth. Therefore, in a preferred embodiment of the present application, for anesthesia simulation analysis of these special soft tissues, appropriate coordinate system transformation and three-dimensional reconstruction of ultrasound image coordinate data are used, and the image brightness data is appropriately highlighted for easy extraction of intraoperative human body posture data including tissue stretching degree, joint bending angle, pitch supine posture, etc. to construct a training sample set and use as input data during model training.

[0027] In a preferred embodiment of the present application, to ensure that the prediction accuracy of the model is not weakened in the process of long-term changes in the population, the historical data of regional anesthesia guided by ultrasound images can be periodically re-collected, and the anesthesia effect prediction model can be re-trained to realize the continuous updating of the prediction model.

[0028] By using the method provided by the application, related theory research and statistical analysis can be supported, the defects that the existing anesthesia operation excessively depends on personnel experience are overcome, and the intelligentization and automation level in teaching and practical training is improved.

[0029] It should be understood that the size of the serial number of each step in the embodiments of the application does not mean the order of execution, and the execution order of each process should be determined according to its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the application.

[0030] Although the embodiments of the application have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and variations can be made to the embodiments without departing from the principles and spirit of the application, and the scope of the application is defined by the appended claims and their equivalents.

Claims

1. A data-driven and ultrasound image-based anesthesia simulation analysis method, characterized by: The specific steps include: Step 1: Collect historical data of regional anesthesia procedures guided by ultrasound images; the historical data include at least: patient age, gender, weight, baseline and intraoperative values ​​of blood pressure, heart rate, and respiratory indicators, intraoperative blood pressure, intraoperative heart rate, intraoperative respiratory indicators, baseline and intraoperative indicators of relevant organs and / or tissues, specific organ and / or tissue types in the blockade area and their ultrasound image coordinates and brightness data, BIS index, bispectral index, EEG frequency and amplitude parameters, drug type, drug dosage, and anesthesia time data; Step 2: Standardize some of the data collected in Step 1, including: dividing the specific ranges of age, weight, baseline values ​​of various physiological indicators and intraoperative change rates, changes in related organ or tissue indicators, medication dosage, and anesthesia time into several segments, and assigning corresponding data labels to each segment; dividing different acupuncture areas according to the coordinates of the ultrasound images of specific organs and / or tissues, and assigning corresponding regional labels; dividing the specific range of the brightness of the ultrasound images of specific organs and / or tissues into several brightness segments, and assigning corresponding brightness labels; and simultaneously, dividing different brightness change rate segments based on anesthesia time and assigning corresponding brightness change rate segment labels; Step 3: Use the historical data processed in step 2 to establish a training sample set; Step 4: Establish an anesthetic effect prediction model based on the particle swarm optimization algorithm. The model inputs are the patient's age, sex, weight, baseline values ​​of various physiological indicators, specific organ and / or tissue type, medication type, dosage, and acupuncture area. The ultrasound image brightness, BIS index, EEG frequency and amplitude, and anesthesia time within a predetermined range around the acupuncture area are used as the model outputs. The corresponding objective function is established by comprehensively considering intraoperative medication safety and drug metabolism. The anesthetic effect prediction model is trained using the training sample set until it converges stably or reaches the maximum number of training times. Step 5. Use the trained anesthetic effect prediction model to perform anesthetic operation simulation analysis, so that it can predict and output the drug effects corresponding to different anesthesia times based on the specific organs and / or tissues to be subjected to regional block anesthesia, the drugs and dosages to be used, and individual characteristics including age, gender, and weight. Among them, the brightness of the output ultrasound image within the predetermined range of the needle insertion area is used to characterize the degree of local drug diffusion, the BIS index is used to characterize the depth of anesthesia, and the EEG frequency and amplitude are used to reflect the specific anesthetic effects of different types of drugs.

2. The method according to claim 1, wherein: The objective function described in step 4 is specifically established by weighted fusion of the rate of change of physiological indicators relative to the baseline value, the rate of change of ultrasound image brightness within the predetermined range of the needle insertion area, and the baseline values ​​and intraoperative indicators of related organs and / or tissues; among them, the rate of change of physiological indicators relative to the baseline value is used to ensure safety during anesthesia, and the rate of change of ultrasound image brightness within the predetermined range of the needle insertion area, and the baseline values ​​and intraoperative indicators of related organs and / or tissues are used to balance the onset and metabolism of anesthetic drugs.

3. The method according to claim 1, wherein: The baseline values ​​and intraoperative indicators of the relevant organs and / or tissues specifically include relevant indicator parameters of the liver, kidneys, and peripheral blood leukocytes.

4. The method according to claim 1, wherein: The respiratory indicators used in the method include respiratory cycle, respiratory depth, real-time respiratory displacement and phase data, which are used to achieve respiratory offset compensation in the output of the anesthesia effect prediction model.

5. The method according to claim 1, wherein: For the anesthesia simulation analysis of specific soft tissues, three-dimensional reconstructed ultrasound image coordinate data and highlighted image brightness data are used, and intraoperative human body posture data including tissue stretching degree, joint bending angle, and supine and supine positions are extracted to construct a training sample set and serve as input data when training the prediction model.

6. The method according to claim 1, wherein: Historical data of regional anesthesia procedures guided by ultrasound images are collected again regularly, and the anesthesia effect prediction model is retrained and updated.

Images