Anesthesia depth monitoring method based on ultrasonic image recognition

Abstract

The invention relates to the technical field of clinical anesthesia, and discloses an anesthesia depth monitoring method based on ultrasonic image recognition, which comprises the following steps of: acquiring a pre-anesthesia ultrasonic image for image processing, obtaining a target nerve anesthesia area contour point and an occlusion tissue area contour point by using a contour tracking algorithm, and respectively calculating the areas of the two; initial conditions and boundary conditions are defined, a Fick's second law is utilized to establish a mathematical model of anesthesia diffusion, the position and the area of a pre-anesthesia area are simulated, a corresponding prediction image is generated and processed, a binary prediction image is compared with a binary pre-anesthesia ultrasonic image, and a pre-anesthesia result is obtained. Calculating to obtain the overlapping area of the target nerve anesthesia area and the pre-anesthesia area, obtaining the ratio of the overlapping area to the area of the target nerve anesthesia area, and judging whether the area of the pre-anesthesia area contains the area of the target nerve anesthesia area or not, thereby determining the anesthesia depth and realizing anesthesia depth monitoring.

Description

Technical Field

[0001] The invention relates to the technical field of clinical anesthesia, and in particular to an anesthesia depth monitoring method based on ultrasonic image recognition. Background Art

[0002] Ultrasound image anesthesia, also known as ultrasound visualization anesthesia technology, is an anesthesia method that closely combines ultrasound imaging technology with anesthetic drug infusion technology. During the anesthesia process, the doctor uses the reflection and propagation characteristics of high-frequency ultrasound to observe the patient's internal anatomical structure, such as blood vessels, nerves, and muscles, in real time, making the injection of anesthetic drugs more accurate and ensuring that the drugs act directly on the target area.

[0003] In Chinese invention patent publication number CN116531089A, a method for processing block anesthesia ultrasound-guided data based on image enhancement is proposed. The method uses an image enhancement algorithm to preprocess the collected ultrasound image to make the image clearer and more recognizable. The preprocessed ultrasound image is then feature extracted and segmented through computer vision technology to identify and locate the relevant anatomical structures required for block anesthesia. The image registration algorithm is then used to achieve accurate calculation of guided positioning.

[0004] The above invention is combined with the existing image registration algorithm. Although it can effectively guide the puncture needle for safe operation, since each patient is configured with a different concentration of anesthetic agents, and at the same time, during the injection process, the position of the tissue blocked by the patient is different, the diffusion direction of the anesthetic agent after injection is unstable, and the anesthetic agent does not fully anesthetize the nerve segment that needs anesthesia, thereby affecting the anesthesia effect. Therefore, an anesthesia depth monitoring method based on ultrasonic image recognition is proposed to solve the above problems. Summary of the invention

[0005] 1. Technical issues to be resolved In view of the shortcomings of the prior art, the present invention provides an anesthesia depth monitoring method based on ultrasonic image recognition, which has the advantages of simulating the anesthesia diffusion position, etc., and solves the problem of poor effect of the existing ultrasonic anesthesia depth monitoring.

[0006] (II) Technical solution To achieve the above object, the present invention provides the following technical solution: a method for monitoring the depth of anesthesia based on ultrasonic image recognition, characterized in that it comprises the following steps: S1. Determine the target area and collect the ultrasound image of the area before anesthesia I original data, and apply image processing techniques to enhance image clarity; S2, based on the image processing data in S1, using the contour tracking algorithm to extract the contour of the target nerve anesthesia area and the contour of the occluded tissue area, respectively, to obtain the contour point g of the target nerve anesthesia area and the contour point g' of the occluded tissue area, and using the contour point g of the target nerve anesthesia area to calculate the area A1 of the target nerve anesthesia area; S3, define initial conditions based on the image processing data and physical properties of anesthetic drugs in S1, define boundary conditions based on the image processing data, physical properties of anesthetic drugs, physiological structure of occluded tissues and contour points g' of occluded tissue regions in S1, establish a mathematical model of anesthetic diffusion based on Fick's second law, substitute the initial conditions and boundary conditions into the mathematical model, simulate the position and area of ​​the pre-anesthesia region, and generate the corresponding predicted image I pre .

[0007] S4, based on the predicted image I pre The contour tracking algorithm is used to extract the contour of the pre-anesthesia area from the data to obtain the contour point k of the pre-anesthesia area; S5, predicting image I pre Ultrasound image before anesthesia original Compare and calculate the overlapping area A using the contour point k of the pre-anesthesia area and the contour point g of the target nerve anesthesia area overlap , calculate its ratio to the target nerve anesthesia area A1, determine whether the pre-anesthesia area includes the target nerve anesthesia area, if it is included, continue the anesthesia injection, and monitor the anesthesia effect and patient reaction at the same time, if it is not included, adjust the anesthetic injection position, and enter step S3 to adjust the initial conditions.

[0008] Preferably, the image processing technology includes an adaptive histogram equalization algorithm, a bilateral filtering algorithm and image binarization, wherein the adaptive histogram equalization algorithm is used to enhance image contrast, the bilateral filtering algorithm is used to remove image noise and retain edge details, and the image binarization converts the image into a binary image for segmentation and feature extraction.

[0009] Preferably, the method for calculating the target nerve anesthesia region area and the blocked tissue region area in S2 is: The convolutional neural network was used to segment the pre-anesthesia ultrasound image after image processing to obtain the binary image of the target nerve anesthesia area and the occluded tissue area. Each pixel was assumed to represent a unit area. The contour tracking method was used to extract the contour of the target nerve anesthesia area and the occluded tissue area respectively, so as to obtain the contour point g (x g ,y g ) and the occluded tissue area contour point g'(x g’ ,y g’ ), use Green's formula to calculate the target nerve anesthesia area: , Among them, A1 is the area of ​​the target nerve anesthesia area, (x p+1 ,y p+1 ) = (x1, y1), p is the number of points on the contour of the target nerve anesthesia area.

[0010] , Among them, A2 is the area of ​​the blocked tissue region, (x g’ ,y g’ ) is the g'th point on the contour of the occluded tissue area, (x q+1 ,y q+1 ) = (x1, y1), q is the number of points on the contour of the occluded tissue area.

[0011] Preferably, in said S3, a mathematical model of anesthesia diffusion is established to generate a prediction image I pre Way: 1) Define initial conditions Binarized ultrasound image before anesthesia I original , assuming that the anesthetic is injected into a specific position (x0, y0) at time t=0, and the initial concentration is C0, then the initial condition is expressed as: ; 2) Obtain the diffusion time marching equation According to Fick's second law, the diffusion equation of anesthetic drugs is: , Where C(x, y, t) is the concentration at position (x, y) and time t, and D is the diffusion coefficient; Considering the influence of the occluded tissue area, the diffusion coefficient D is corrected to the position-dependent diffusion coefficient D0. At the same time, the space is discretized into grid points, and the time is discretized into the time step ∆t. The grid spacing, that is, the spatial step, is set to ∆x and ∆y. The position of each grid point is represented as (i, j), so the position (x, y) is the same as the grid point position (i, j). Then, the position-dependent diffusion coefficient D0 is substituted into the anesthetic drug diffusion equation, and this equation is discretized to obtain the displayed finite difference formula: , in, is the concentration at position (i∆x, j∆y) at time t; 3) Define boundary conditions Free boundary condition: After the anesthetic is injected, the drug concentration gradient at the boundary is zero, that is, the free boundary condition is ; Boundary condition of blocked tissue: During the diffusion process of anesthetic injection, the boundary of the blocked tissue area cannot be penetrated at all, that is, C(x, y, t) = 0. It is hindered to a certain extent at the boundary of the occluded tissue area, which can be expressed as: , Among them, R represents the penetration resistance, which is a constant, C O is the drug concentration inside the obscured tissue region; 4) Mathematical model simulation process Substitute the initial conditions and boundary conditions into the explicit finite difference formula to calculate the concentration at the position (i∆x, j∆y) at each time t, and set the concentration threshold of each grid point to C th The sum of all grid points exceeding this threshold is regarded as the effective pre-anesthesia area A di , count all concentrations greater than C th The number of grid points N di , thus the area of ​​the effective pre-anesthesia area is: A di =N di ×A p , Among them, A p It is expressed as the area of ​​each grid point; Check the concentration at each grid point Is it greater than C? th ,if , then record the grid point coordinates (i, j), and then create a blank image with the same size as the pre-anesthesia ultrasound image to effectively pre-anesthetize the area A di The grid points of the area are marked as white, and the grid points of other areas are marked as black, thereby generating the predicted image I pred .

[0012] Preferably, the diffusion coefficient D is corrected to a position-dependent diffusion coefficient D0 in the following manner: Since the outline of the occluded tissue area is composed of a series of points g'(x g’ ,y g’ ), set f(x, y) to indicate whether a point is located in the occluded tissue area, D eff is the diffusion coefficient of the occluded tissue region; When the occluded tissue area hinders the diffusion of the anesthetic agent, the diffusion coefficient D eff =0, so according to the binary mask, we get f(x, y)=0 in the non-occluded area and f(x, y)=1 in the occluded area, so we get: ; When the occluded tissue area slows down the diffusion of the anesthetic agent, it is necessary to make the position-dependent diffusion coefficient D0 change according to the position so that D eff , and D is applied in the non-occluded area, which can be expressed by the formula: .

[0013] Preferably, the contour points of the pre-anesthesia area of ​​the predicted image are obtained in the following manner: Use convolutional neural network to generate prediction image I pred Segmentation is performed to obtain the predicted image I pred The contour tracking method is used to extract the contour of the pre-anesthesia area in the predicted image, thereby obtaining the contour point k (x k ,y k ).

[0014] Preferably, the method for determining whether the area of ​​the pre-anesthesia region in S5 includes the area of ​​the target nerve anesthesia region is as follows: Set the ultrasound image before anesthesia to I original Then, the binary predicted image I pred Superimposed on the binary pre-anesthesia ultrasound image I original superior; Generate a binary mask of the pre-anesthesia area, M pre (x k ,y k ) = 1 means it belongs to the point (x k ,y k ) belongs to the pre-anesthesia area, otherwise M pre (x k ,y k ) = 0, Generate a binary mask of the target neural anesthesia area, M targt (x g ,y g ) = 1 means it belongs to the point (x g ,y g ) belongs to the target nerve anesthesia area, otherwise M targt (x g ,y g ) = 0, The contour point k of the pre-anesthesia area and the contour point g of the target nerve anesthesia area are brought into the binary mask to obtain the overlapping area: , where k∩g represents the intersection of the pre-anesthetized region and the target neuroanesthetized region, W and H are the width and height of the image, respectively, and ⋀ is the logical AND operator. For each pixel position (x, y), if M pre (x k,y k ) = 1 and M targt (x g ,y g ) = 1, then M pre (x k ,y k )⋀M targt (x g ,y g ) = 1, otherwise, M pre (x k ,y k )⋀M targt (x g ,y g ) = 0; Therefore, the ratio of the overlapping area to the target nerve anesthesia area A1 is: , When C ratio =1, indicating that the pre-anesthesia area and the target nerve anesthesia area completely overlap. ratio <1, it means that the pre-anesthetized area fails to cover the target nerve anesthesia area.

[0015] (III) Beneficial effects Compared with the prior art, the present invention provides a method for monitoring the depth of anesthesia based on ultrasonic image recognition, which has the following beneficial effects: The anesthesia depth monitoring method based on ultrasonic image recognition collects pre-anesthesia ultrasonic images for image processing, uses a contour tracking algorithm to obtain the position and area data of the target nerve anesthesia area and the occluded tissue area in the pre-anesthesia ultrasonic image, uses Fick's second law to establish a mathematical model of anesthesia diffusion, generates a corresponding predicted image and performs image processing, compares the binarized predicted image with the binarized pre-anesthesia ultrasonic image, determines the overlap between the pre-anesthesia area and the target nerve anesthesia area, and realizes anesthesia depth monitoring, wherein the contour tracking algorithm, Fick's second law and the anesthesia diffusion equation are used to quickly process a large amount of image data and generate prediction results, which can accurately control the diffusion process of anesthetic drugs, monitor the target nerve anesthesia state in real time, make anesthesia more accurate, and at the same time reduce the risk of complications such as anesthetic nerve damage to ensure patient safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a schematic diagram of the process of the present invention; Figure 2 This is a schematic diagram of the anesthetic agent diffusion simulation of the present invention. DETAILED DESCRIPTION

[0017] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0018] A method for monitoring anesthesia depth based on ultrasonic image recognition comprises the following steps: S1. Determine the target area and collect the ultrasound image of the area before anesthesia I original data, and apply image processing techniques to enhance image clarity; S2, based on the image processing data in S1, using the contour tracking algorithm to extract the contour of the target nerve anesthesia area and the contour of the occluded tissue area, respectively, to obtain the contour point g of the target nerve anesthesia area and the contour point g' of the occluded tissue area, and using the contour point g of the target nerve anesthesia area to calculate the area A1 of the target nerve anesthesia area; S3, define initial conditions based on the image processing data and physical properties of anesthetic drugs in S1, define boundary conditions based on the image processing data, physical properties of anesthetic drugs, physiological structure of occluded tissues and contour points g' of occluded tissue regions in S1, establish a mathematical model of anesthetic diffusion based on Fick's second law, substitute the initial conditions and boundary conditions into the mathematical model, simulate the position and area of ​​the pre-anesthesia region, and generate the corresponding predicted image I pre .

[0019] S4, based on the predicted image I pre The contour tracking algorithm is used to extract the contour of the pre-anesthesia area from the data to obtain the contour point k of the pre-anesthesia area; S5, predicting image I pre Ultrasound image before anesthesia original Compare and calculate the overlapping area A using the contour point k of the pre-anesthesia area and the contour point g of the target nerve anesthesia area overlap , calculate its ratio to the target nerve anesthesia area A1, determine whether the pre-anesthesia area includes the target nerve anesthesia area, if it is included, continue the anesthesia injection, and monitor the anesthesia effect and patient reaction at the same time, if it is not included, adjust the anesthetic injection position, and enter step S3 to adjust the initial conditions.

[0020] In this embodiment, the image processing technology includes an adaptive histogram equalization algorithm, a bilateral filtering algorithm and image binarization. The adaptive histogram equalization algorithm is used to enhance image contrast, the bilateral filtering algorithm is used to remove image noise and retain edge details, and the image binarization converts the image into a binary image for segmentation and feature extraction.

[0021] In this embodiment, the calculation method of the target nerve anesthesia area and the blocked tissue area in S2 is: The convolutional neural network was used to segment the pre-anesthesia ultrasound image after image processing to obtain the binary image of the target nerve anesthesia area and the occluded tissue area. Each pixel was assumed to represent a unit area. The contour tracking method was used to extract the contour of the target nerve anesthesia area and the occluded tissue area respectively, so as to obtain the contour point g (x g ,y g ) and the occluded tissue area contour point g'(x g’ ,y g’ ), use Green's formula to calculate the target nerve anesthesia area: , Among them, A1 is the area of ​​the target nerve anesthesia area, (x p+1 ,y p+1 ) = (x1, y1), p is the number of points on the contour of the target nerve anesthesia area.

[0022] , Among them, A2 is the area of ​​the blocked tissue region, (x g’ ,y g’ ) is the g'th point on the contour of the occluded tissue area, (x q+1 ,y q+1 ) = (x1, y1), q is the number of points on the contour of the occluded tissue area.

[0023] In this embodiment, a mathematical model of anesthesia diffusion is established in S3 to generate a prediction image I pre Way: 1) Define initial conditions Binarized ultrasound image before anesthesia I original , assuming that the anesthetic is injected into a specific position (x0, y0) at time t=0, and the initial concentration is C0, then the initial condition is expressed as: ; 2) Obtain the diffusion time marching equation According to Fick's second law, the diffusion equation of anesthetic drugs is: , Where C(x, y, t) is the concentration at position (x, y) and time t, and D is the diffusion coefficient; Taking into account the influence of the occluded tissue area, the diffusion coefficient D is corrected to the position-dependent diffusion coefficient D0. Since the contour of the occluded tissue area is composed of a series of points g'(x g’ ,y g’ ), set f(x, y) to indicate whether a point is located in the occluded tissue area, Deff is the diffusion coefficient of the occluded tissue region; When the occluded tissue area hinders the diffusion of the anesthetic agent, the diffusion coefficient D eff =0, so according to the binary mask, we get f(x, y)=0 in the non-occluded area and f(x, y)=1 in the occluded area, so we get: ; When the occluded tissue area slows down the diffusion of the anesthetic agent, it is necessary to make the position-dependent diffusion coefficient D0 change according to the position so that D eff , and D is applied in the non-occluded area, which can be expressed by the formula: ; At the same time, the space is discretized into grid points, and the time is discretized into the time step ∆t. The grid spacing, that is, the space step, is set to ∆x and ∆y. The position of each grid point is represented as (i, j), so the position (x, y) is the same as the grid point position (i, j). Then, the position-dependent diffusion coefficient D0 is substituted into the anesthetic drug diffusion equation, and this equation is discretized to obtain the explicit finite difference formula: , in, is the concentration at position (i∆x, j∆y) at time t; 3) Define boundary conditions Free boundary condition: After the anesthetic is injected, the drug concentration gradient at the boundary is zero, that is, the free boundary condition is ; Boundary condition of blocked tissue: During the diffusion process of anesthetic injection, the boundary of the blocked tissue area cannot be penetrated at all, that is, C(x, y, t) = 0. It is hindered to a certain extent at the boundary of the occluded tissue area, which can be expressed as: , Among them, R represents the penetration resistance, which is a constant, C O is the drug concentration inside the obscured tissue region; 4) Mathematical model simulation process Substitute the initial conditions and boundary conditions into the explicit finite difference formula to calculate the concentration at the position (i∆x, j∆y) at each time t, and set the concentration threshold of each grid point to C th The sum of all grid points exceeding this threshold is regarded as the effective pre-anesthesia area A di , count all concentrations greater than C th The number of grid points N di , thus the area of ​​the effective pre-anesthesia area is: Adi =N di ×A p , Among them, A p It is expressed as the area of ​​each grid point; Check the concentration at each grid point Is it greater than C? th ,if , then record the grid point coordinates (i, j), and then create a blank image with the same size as the pre-anesthesia ultrasound image to effectively pre-anesthetize the area A di The grid points of the area are marked as white, and the grid points of other areas are marked as black, thereby generating the predicted image I pred .

[0024] It should be noted that the purpose of obtaining the diffusion time marching equation in the above step 2 is to make the anesthetic agent follow a certain period of time and let the diffusion state evolve over time. For image generation, time marching can be achieved through an iterative update mechanism, in which each time step is adjusted according to the state of the previous moment; the definition of boundary conditions in the above step 3 can help constrain the image generation process, making the output more stable and closer to real-world phenomena; the initialization of the above step 1 refers to setting a starting state for the mathematical model, that is, defining the state of the anesthetic agent diffusion at the beginning. For image generation, this usually means providing an initial input image or random noise as a starting point.

[0025] Among them, when the anesthetic enters the obstructing tissue, considering the influence of the obstructing tissue area on the anesthetic drug diffusion equation, the diffusion coefficient D needs to be corrected according to the position to reflect the different degrees of resistance of different tissues to drug diffusion.

[0026] In this embodiment, the method for obtaining the contour points of the pre-anesthesia area of ​​the predicted image is as follows: Use convolutional neural network to generate prediction image I pred Segmentation is performed to obtain the predicted image I pred The contour tracking method is used to extract the contour of the pre-anesthesia area in the predicted image, thereby obtaining the contour point k (x k ,y k ).

[0027] In this embodiment, the method for determining whether the area of ​​the pre-anesthesia region in S5 includes the area of ​​the target nerve anesthesia region is as follows: Set the ultrasound image before anesthesia to I original Then, the binary predicted image I pred Superimposed on the binary pre-anesthesia ultrasound image I original superior; Generate a binary mask of the pre-anesthesia area, M pre (xk ,y k ) = 1 means it belongs to the point (x k ,y k ) belongs to the pre-anesthesia area, otherwise M pre (x k ,y k ) = 0, Generate a binary mask of the target neural anesthesia area, M targt (x g ,y g ) = 1 means it belongs to the point (x g ,y g ) belongs to the target nerve anesthesia area, otherwise M targt (x g ,y g ) = 0, The contour point k of the pre-anesthesia area and the contour point g of the target nerve anesthesia area are brought into the binary mask to obtain the overlapping area: , , where k∩g represents the intersection of the pre-anesthetized region and the target neuroanesthetized region, W and H are the width and height of the image, respectively, and ⋀ is the logical AND operator. For each pixel position (x, y), if M pre (x k ,y k ) = 1 and M targt (x g ,y g ) = 1, then M pre (x k ,y k )⋀M targt (x g ,y g ) = 1, otherwise, M pre (x k ,y k )⋀M targt (x g ,y g ) = 0; Therefore, the ratio of the overlapping area to the target nerve anesthesia area A1 is: , When C ratio =1, indicating that the pre-anesthesia area and the target nerve anesthesia area completely overlap. ratio <1, it means that the pre-anesthetized area fails to cover the target nerve anesthesia area.

[0028] It should be noted that C ratio=1, indicating that the pre-anesthetized area and the target nerve anesthetized area completely overlap, indicating that the diffusion of the anesthetic during the injection process can cover the target nerve anesthetized area. In the actual injection process, the stretching of the nerve under anesthesia can be visually seen from the ultrasound image. When the nerve is completely stretched and there is still some anesthetic left, the injection can be stopped to avoid overinjection. ratio <1, it means that the pre-anesthetized area fails to cover the target nerve anesthesia area, indicating that the diffusion of the anesthetic agent cannot cover the nerve that needs to be anesthetized during the simulated diffusion process. The anesthetic agent is quantitatively adapted, so it is necessary to adjust the initial injection conditions and repeat the simulation to ensure that the anesthetic agent injection diffusion fully covers the target nerve anesthesia and avoid incomplete analgesia or complete lack of anesthesia.

[0029] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for monitoring anesthesia depth based on ultrasonic image recognition, characterized in that: The following steps are involved: S1. Determine the target area and collect the ultrasound image of the area before anesthesia I original data, and apply image processing techniques to enhance image clarity; S2, based on the image processing data in S1, using the contour tracking algorithm to extract the contour of the target nerve anesthesia area and the contour of the occluded tissue area, respectively, to obtain the contour point g of the target nerve anesthesia area and the contour point g' of the occluded tissue area, and using the contour point g of the target nerve anesthesia area to calculate the area A1 of the target nerve anesthesia area; S3, define initial conditions based on the image processing data and physical properties of anesthetic drugs in S1, define boundary conditions based on the image processing data, physical properties of anesthetic drugs, physiological structure of occluded tissues and contour points g' of occluded tissue regions in S1, establish a mathematical model of anesthetic diffusion based on Fick's second law, substitute the initial conditions and boundary conditions into the mathematical model, simulate the position and area of ​​the pre-anesthesia region, and generate the corresponding predicted image I pre . S4, based on the predicted image I pre The contour tracking algorithm is used to extract the contour of the pre-anesthesia area from the data to obtain the contour point k of the pre-anesthesia area; S5, predicting image I pre Ultrasound image before anesthesia original Compare and calculate the overlapping area A using the contour point k of the pre-anesthesia area and the contour point g of the target nerve anesthesia area overlap , calculate its ratio to the target nerve anesthesia area A1, determine whether the pre-anesthesia area includes the target nerve anesthesia area, if it is included, continue the anesthesia injection, and monitor the anesthesia effect and patient reaction at the same time, if it is not included, adjust the anesthetic injection position, and enter step S3 to adjust the initial conditions.

2. The method for monitoring anesthesia depth based on ultrasonic image recognition according to claim 1, characterized in that: The image processing technology includes an adaptive histogram equalization algorithm, a bilateral filtering algorithm and image binarization. The adaptive histogram equalization algorithm is used to enhance image contrast, the bilateral filtering algorithm is used to remove image noise and retain edge details, and the image binarization converts an image into a binary image for segmentation and feature extraction.

3. The method for monitoring anesthesia depth based on ultrasonic image recognition according to claim 1, characterized in that: The calculation method of the target nerve anesthesia area and the blocked tissue area in S2 is: The convolutional neural network was used to segment the pre-anesthesia ultrasound image after image processing to obtain the binary image of the target nerve anesthesia area and the occluded tissue area. Each pixel was assumed to represent a unit area. The contour tracking method was used to extract the contour of the target nerve anesthesia area and the occluded tissue area respectively, so as to obtain the contour point g(x g ,y g ) and the contour point g'(x g’ ,y g’ ), and the target nerve anesthesia area was calculated using Green's formula: Among them, A1 is the area of ​​the target nerve anesthesia area, (x p+1 ,y p+1 )=(x1, y1), where p is the number of points on the contour of the target nerve anesthesia area. Among them, A2 is the area of ​​the blocked tissue region, (x g’ ,y g’ ) is the g'th point on the contour of the occluded tissue area, (x q+1 ,y q+1 )=(x1, y1), q is the number of points on the contour of the occluded tissue area.

4. The method for monitoring anesthesia depth based on ultrasonic image recognition according to claim 3, characterized in that: In S3, a mathematical model of anesthesia diffusion is established to generate a prediction image I pre Way: 1) Define initial conditions Binarized ultrasound image before anesthesia I original , assuming that the anesthetic is injected into a specific position (x0, y0) at time t = 0, and the initial concentration is C0, then the initial condition is expressed as: 2) Obtain the diffusion time marching equation According to Fick's second law, the diffusion equation of anesthetic drugs is: Where C(x, y, t) is the concentration at position (x, y) and time t, and D is the diffusion coefficient; Considering the influence of the occluded tissue area, the diffusion coefficient D is corrected to the position-dependent diffusion coefficient D0. At the same time, the space is discretized into grid points, and the time is discretized into the time step Δt. The grid spacing, that is, the space step, is set to Δx and Δy. The position of each grid point is represented as (i, j), so the position (x, y) is the same as the grid point position (i, j). Then, the position-dependent diffusion coefficient D0 is substituted into the anesthetic drug diffusion equation, and this equation is discretized to obtain the displayed finite difference formula: in, is the concentration at position (iΔx, jΔy) at time t; 3) Define boundary conditions Free boundary condition: After the anesthetic is injected, the drug concentration gradient at the boundary is zero, that is, the free boundary condition is Boundary condition of blocked tissue: During the diffusion process of anesthetic injection, the boundary of the blocked tissue area cannot be penetrated at all, that is, C(x, y, t) = 0. It is hindered to a certain extent at the boundary of the occluded tissue area, which can be expressed as: Among them, R represents the penetration resistance, which is a constant, C O is the drug concentration inside the obscured tissue region; 4) Mathematical model simulation process Substitute the initial conditions and boundary conditions into the explicit finite difference formula, calculate the concentration at the position (iΔx, jΔy) at each time t, and set the concentration threshold of each grid point to C th The sum of all grid points exceeding this threshold is regarded as the effective pre-anesthesia area A di , count all concentrations greater than C th The number of grid points N di , thus the area of ​​the effective pre-anesthesia area is: A di =N di ×A p , Among them, A p It is expressed as the area of ​​each grid point; Check the concentration at each grid point Is it greater than C? th ,if The grid point coordinates (i, j) are recorded, and then a blank image with the same size as the pre-anesthesia ultrasound image is created to effectively pre-anesthetize the area A. di The grid points of the area are marked as white, and the grid points of other areas are marked as black, thereby generating the predicted image I pred .

5. The method for monitoring anesthesia depth based on ultrasonic image recognition according to claim 4, characterized in that: The method of correcting the diffusion coefficient D to the position-dependent diffusion coefficient D0 is: Since the outline of the occluded tissue area is composed of a series of points g'(x g’ ,y g’ ) is given, and f(x, y) is set to indicate whether a point is located in the occluded tissue area. eff is the diffusion coefficient of the occluded tissue region; When the occluded tissue area hinders the diffusion of the anesthetic agent, the diffusion coefficient D eff =0, so according to the binary mask, we get f(x, y)=0 in the non-blocked area and f(x, y)=1 in the blocked area, so we get: When the occluded tissue area slows down the diffusion of the anesthetic agent, it is necessary to make the position-dependent diffusion coefficient D0 change according to the position so that D eff , and D is applied in the non-occluded area, which can be expressed by the formula: D0(x,y)=(1-f(x,y))·D+f(x,y)D eff 。 6. The method for monitoring anesthesia depth based on ultrasonic image recognition according to claim 4, characterized in that: The method for obtaining the contour points of the pre-anesthesia area of ​​the predicted image is as follows: Use convolutional neural network to generate prediction image I pred Segmentation is performed to obtain the predicted image I pred The contour tracking method is used to extract the contour of the pre-anesthesia area in the predicted image, thereby obtaining the contour point k(x k ,y k ).

7. The method for monitoring anesthesia depth based on ultrasonic image recognition according to claim 5, characterized in that: Method for determining whether the pre-anesthesia area in S5 includes the target nerve anesthesia area: Set the ultrasound image before anesthesia to I original Then, the binary predicted image I pred Superimposed on the binary pre-anesthesia ultrasound image I original superior; Generate a binary mask of the pre-anesthesia area, M pre (x k ,y k )=1 means it belongs to point (x k ,y k ) belongs to the pre-anesthesia area, otherwise M pre (x k ,y k )=0, Generate a binary mask of the target neural anesthesia area, M targt (x g ,y g )=1 means it belongs to point (x g ,y g ) belongs to the target nerve anesthesia area, otherwise M targt (x g ,y g )=0, The contour point k of the pre-anesthesia area and the contour point g of the target nerve anesthesia area are brought into the binary mask to obtain the overlapping area: A overlap =∑(k∩g), Where k∩g represents the intersection of the pre-anesthetized area and the target neural anesthetized area, W and H are the width and height of the image, respectively, and ∧ is the logical AND operator. For each pixel position (x, y), if M pre (x k ,y k )=1 and M targt (x g ,y g )=1, then M pre (x k ,y k )∧M targt (x g ,y g )=1, otherwise, M pre (x k ,y k )∧M targt (x g ,y g )=0; Therefore, the ratio of the overlap area to the target nerve anesthesia area is: When C ratio =1, indicating that the pre-anesthesia area and the target nerve anesthesia area completely overlap. ratio <1, it means that the pre-anesthetized area fails to cover the target nerve anesthesia area.

Images