An automatic needle insertion method combining a blood flow optimization model with a neural network

Abstract

The application discloses an automatic needle insertion method combining a blood flow optimization model with a neural network. The method first calculates the blood vessel position and blood flow velocity distribution of a needle insertion area according to blood flow information measured by automatic ultrasonic scanning by means of the Doppler ultrasonic principle, then establishes a blood flow relative velocity optimization model and a neural network with a selection function, and decides the optimal needle insertion point and needle insertion angle on the basis of physical measurement results, and finally simulates the needle insertion path by means of a neural network with a learning function according to the needle insertion method of a professional. The method can be used for needle insertion of subcutaneous blood vessels of limbs and other body parts, and overcomes the difficulty of manual needle insertion when the blood vessel is not visible to the naked eye, thereby providing technical assistance for self-rescue and first aid in an environment lacking professional personnel.

Description

Technical Field

[0001] This invention belongs to the fields of artificial intelligence and medicine, specifically relating to an automatic needle insertion method that combines a blood flow optimization model with a neural network. Background Technology

[0002] Vascular insertion is ubiquitous in modern medicine, from simple blood draws during routine checkups to life-saving transfusions and intravenous infusions. However, accurately locating veins or arteries and determining their course is often a challenge for healthcare professionals. For instance, some people have small blood vessels, are tightly covered by subcutaneous fat and muscle layers, or have dense body hair, resulting in very low visibility. Animals, on the other hand, are often completely covered by fur, making it difficult to visually identify their veins and arteries. Delays in resuscitation due to the inability to accurately locate a suitable vein for insertion are common. Poor insertion techniques can lead to repeated unnecessary needle pricks for patients or animals, causing pain, bleeding, and even varying degrees of infection, potentially resulting in medical accidents and serious conflicts between patients' families and doctors.

[0003] Ultrasound can be used to detect superficial and deep tissues invisible to the human eye, and is particularly suitable for accurately detecting and locating blood vessels in humans and animals. In nature, vampire bats use their innate ultrasonic abilities to accurately determine the distribution of blood vessels in animals and livestock that are invisible to the human eye, and to collect blood. In artificially developed medical devices, color Doppler ultrasound imaging can provide two-dimensional or three-dimensional image information such as the direction and velocity of blood flow within a given cross-section.

[0004] However, there is currently no automated needle insertion technique based on Doppler ultrasound principles and blood flow optimization models that does not require imaging. Summary of the Invention

[0005] To address the shortcomings and deficiencies of existing technologies, this invention proposes an automated needle insertion method combining a blood flow optimization model and a neural network. This invention is specifically implemented through the following technical solution: An automated needle insertion method combining a blood flow optimization model and a neural network includes the following steps: Step 1: Acquisition of Ultrasonic Blood Flow Information – Using the Doppler ultrasound principle, the spatial angular relationship between the blood flow feedback signal measured by automatic ultrasound scanning and the location of blood vessels and the distribution of blood flow velocity in the injection area is calculated. Step 2: Establish a blood flow relative velocity optimization model, the instantaneous minimization model expression of which is as follows: (1a) (1b) In equations (1a)-(1b), and It is a constant. Let p be the directional derivative or difference vector of the relative velocity of blood flow along the flow direction. For discriminant functions; A set of optimal solutions for the insertion pin position and angle were obtained through physical calculations and a blood flow relative velocity optimization model; Step 3: The preferred scheme is combined with a neural network with selection function to determine the optimal pin insertion point and pin angle, thus obtaining the optimal pin insertion scheme; Step 4: Based on the optimal needle insertion scheme, and combined with a neural network with learning capabilities, learn and simulate the needle insertion techniques of medical staff to obtain the needle insertion path.

[0006] Furthermore, step one includes an automated ultrasound scan with multiple frequencies and adjustable angles, and calculations of vascular location and blood flow velocity based on Doppler ultrasound. Normally, ultrasound examinations in medical practice are performed manually. However, to obtain the variable values ​​required for the blood flow relative velocity optimization model and achieve automated needle insertion, an ultrasound scan of the area to be inserted is performed on the body, with automatically adjustable emission angles and frequencies, followed by automatic calculations using the Doppler principle. Ultrasonic scanning can be performed using either a motorized or a stationary method. Motorized ultrasound scanning requires the ultrasound probe to be able to rotate automatically at a variable angle within a local area and perform a motorized translational scan along a movable support. Stationary ultrasound scanning uses an electronically controlled phased array mode. Based on the acquired ultrasound information, irrelevant information is filtered out. Using the Doppler ultrasound principle, essential information, including the local spatial distribution of major subcutaneous blood vessels at the needle insertion site and the direction of blood flow, is demodulated and calculated. The method for measuring and annotating the spatial angle of blood flow using the Doppler ultrasound principle is as follows: Assume the erythrocyte swarm moves at a velocity v along the X-axis (the direction of blood flow), and W is the reference axis. This reference axis can be a vertical axis in the direction of gravity. The Z-axis, X-axis, and W-axis are in the same plane, where the Z-axis is perpendicular to the X-axis, and the angle between the Z-axis and W is θ. The ultrasonic probe emits at a frequency of Ultrasound waves travel at the speed of sound in the body medium. The signal propagates to the target and then returns as a scattered signal to the receiving probe at a frequency of [frequency missing]. Let the incident plane of the ultrasound be the plane containing the incident sound beam axis and the Z-axis. , The axis perpendicular to the Z-axis is Y; the angle between the ultrasound incident plane and the blood flow direction is... That is, the angle between the X-axis and the Y-axis is The angle between the incident sound beam axis and the horizontal plane is set as That is, the angle between the incident sound beam axis and the Y-axis is Let the scattering plane containing the scattering acoustic axis of the red blood cells (groups) along the receiver direction and the Z-axis be denoted as . , The axis perpendicular to the Z-axis is , The angle between the axis and the direction of blood flow is And the scattering acoustic axis is The included angle of the axis is The angle between the incident acoustic beam axis and the W-axis is denoted as . The angle between the scattering acoustic axis and the W-axis is ,in and The value can be directly measured.

[0007] Ultrasonic receiving frequency With incident frequency The relational expression is: (2) In equation (2), Blood flow velocity and ultrasonic velocity in tissue media of human or animal tissues The ratio, where t represents the emission time of the ultrasonic wave; The angle between the incident sound beam and the blood flow direction relative to the incident plane. Let be the angle between the scattered sound beam and the blood flow direction relative to the scattering plane; each spatial angle satisfies the following relationship: (3a) (3b) When the relative velocity of blood flow Furthermore, when the ultrasound transmitter and receiver are located within the same probe, i.e., in the same spatial position, the Doppler frequency shift, relative blood flow velocity, and directional angle have the following relationship: (4) In the formula, For Doppler frequency shift; remember This is the transmission center frequency of the ultrasound. Given the average angular frequency, the average velocity ratio is... Relationship: (5) Based on the effective ultrasound feedback bands acquired and filtered after the scanning angle change, a set of Doppler frequency shifts is obtained by directional demodulation, and the vessel position and blood flow velocity variables are calculated accordingly. That is, the relative flow velocity and relative angle parameters are solved by using the frequency shift change equation set, the corresponding time delay change and equation (3). This allows us to obtain the distribution and orientation of local blood vessels in the area requiring needle insertion relative to the reference coordinates. When the ultrasound transmitter and receiver are located within the same probe, i.e., in the same spatial position, the expression for the frequency shift change equations is as follows: (6) or (7) In the formula, Indicates the change in frequency shift. It is a positive integer; When the parameters are kept in the frequency shift equation system At the same time, set , Then we have: (8a) When the parameters are kept in the frequency shift equation system At the same time, Become , , If is a positive integer, then: (8b) In particular, when , At that time, (8a)-(8b) have a simplified linear computational form. (9) In equations (8a)-(9), The sign ensures that the computational amount on the right side is equal to that on the left side. or Keep positive and negative consistent; and Under precise and controllable conditions, the relative angle parameters are solved, and the local relative velocity is obtained from the frequency shift equation; then, the angle between the Z-axis and the reference axis W is calculated and recorded using equation (3). .

[0008] In addition, while maintaining parameters At the same time, it can also be obtained by calculating the time delay caused by the change in sound path. Value: (10) In equation (10), The distance between the probe and the target detection point is denoted by , and t represents the time of ultrasonic wave emission. express Become and The moment the signal is received while remaining constant.

[0009] Furthermore, in step two, a set of candidate optimal needle coordinates and needle angles are obtained using a blood flow relative velocity optimization model with one or more local optima. The specific construction of the blood flow relative velocity optimization model is as follows: For a cylindrical blood vessel model with radius r, the blood flow rate is... And in blood vessels with constant mass and no bifurcation According to the principle of fluid continuity, ,in Mean blood flow velocity, This represents the change in pressure per unit length along the direction of blood flow. Therefore: (11) in, This represents the change in average relative velocity of blood flow per unit length; this means that blood flow is highly dependent on the radius of the blood vessel, and changes in the radius of the blood vessel depend on changes in the relative velocity of blood flow. Therefore, instantaneous minimization models of relative blood flow velocity are established (1a)-(1b); in the instantaneous minimization model, with position p as the center and distance as... Within the sampling range, the region where the blood flow velocity does not jump or change significantly in the direction of blood flow is the preferred region. In this case, let the discriminant function... ;otherwise This corresponds to situations where blood vessels are bifurcated, blocked, or partially ruptured. Within the limits of computational efficiency and computational cost, a time-varying variable is incorporated into the instantaneous minimization model of relative blood flow velocity, requiring the blood flow velocity optimization model to be stable over time; firstly, for Noise reduction preprocessing is performed, including For m cardiac cycles A set of time columns above, , For positive integers; now consider the time-dependent blood flow velocity optimization model: (12a) in (12b) The blood flow velocity optimization model must satisfy the following constraints: (13a) in, (13b) and (13c) These are estimation formulas for the pulsatility index and resistance index, respectively, which are clinically used to characterize vascular obstruction. For the first Average flow rate within one cardiac cycle; In the blood flow relative velocity model, each norm is relative to the time series. or Norm, or ,in The selection is based on the desired effect, ensuring computational efficiency while reducing the impact of noise on the calculated values; the median of (13a) is used. , , and The settings allow for the preliminary identification and exclusion of blocked or diseased blood vessels, thereby reducing the computational load of the blood flow relative velocity optimization model.

[0010] Furthermore, in situations where rapid measurement is possible In the case of the discriminant function The expression is: (14a) (14b) In equations (14a)-(14b), and They are respectively in the direction of blood flow and in the opposite direction of blood flow. Distance sampling points, For the threshold, For positive integers, The value is approximately equal to and not less than the needle insertion depth; in the optimization scheme of blood flow relative velocity, that is, in equations (1a)-(1b) and (12a)-(12b), the range of blood flow relative velocity at the ideal needle insertion position is further limited based on clinical experience, that is, set .

[0011] Furthermore, in the time-dependent blood flow velocity optimization model, equation (12a) can be replaced by: (15a) And satisfy the constraints. (15b) This further simplifies the calculation.

[0012] Furthermore, in step three, a selection-enabled neural network is trained using the optimal scheme obtained from physical calculations and the blood flow relative velocity optimization model, along with actual clinical performance scores. This network is integrated into a learning chip. The selection-enabled neural network contains multiple intermediate layers, using parameters corresponding to the blood flow relative velocity and coordinates of candidate injection points. ,…, The sequence numbers (1,2,3,4,5) serve as the input data for the input layer, and experienced healthcare professionals provide evaluation feedback based on the physically calculated coordinates and angles. Blood flow velocity and ultrasonic velocity in tissue media of human or animal tissues The ratio, ( The angle between the incident sound beam and the blood flow direction is determined; the evaluation feedback of the training group includes the preferred injection sites pre-marked on the experimental samples by medical staff using other medical devices and the actual vascular direction during needle insertion, as well as the sequence number ranking given according to priority; the preferred injection sites and vascular directions marked in the experiment are scanned and calculated by the system and fed back to the neural network with selection function in the form of data; after multiple use evaluations, the trained neural network with selection function can, according to the input data, feed back the data in the form of parameters. Output the optimal pin coordinates and pin angle.

[0013] Furthermore, the pin path planning in step four is as follows: If the calculated local blood flow direction at the optimal insertion point is... Corresponding pin angle Therefore, the initial pusher angle should be selected as follows: Then gradually adjust the needle to align it with the local blood flow direction. It can be set directly based on experience.

[0014] The process of gradually adjusting the needle body is guided by the neural network with learning capabilities, and is reflected in the insertion angle during the process of the needle contacting the skin and being inserted. Insertion speed Subtle changes over time; these changes, reflecting professional pin insertion techniques, are input and used to train a learning neural network; the training of the learning neural network is completed independently of the overall method steps by professionals.

[0015] The beneficial effects of this invention are as follows: (1) This invention combines the blood flow optimization model and the learning and discrimination functions of neural networks to automatically select the optimal needle insertion position on the blood vessel and simulate the needle insertion technique of professionals to design the needle insertion path, so as to overcome the difficulty of finding blood vessels with the naked eye in special circumstances and improve the survival probability of self-rescue and first aid in environments lacking professionals.

[0016] (2) The present invention is beneficial to avoid or reduce the pain caused to humans and animals by repeated failed needle pricks.

[0017] (3) This invention is based on the principle of ultrasound Doppler, but only requires the calculation and optimization of spatial coordinates in a local area using relevant data of blood flow velocity. It does not require three-dimensional ultrasound imaging, thus eliminating the huge amount of calculation and imaging equipment required for three-dimensional ultrasound imaging. This provides a new technical basis for realizing an automatic needle insertion device that is small in size, light in weight and easy to carry. Attached Figure Description

[0018] Figure 1 A schematic diagram of the overall process of an automatic needle insertion method that combines a blood flow optimization model with a neural network.

[0019] Figure 2 This is a schematic diagram showing the relationship between various angular parameters in the Doppler frequency shift calculation formula. Detailed Implementation

[0020] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. The specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0021] like Figure 1 As shown, the automatic needle insertion method combining blood flow optimization model and neural network includes three main parts in its overall implementation process: ultrasound blood flow information acquisition, establishment of blood flow relative velocity optimization model and neural network decision-making, and automatic needle insertion. The specific steps are as follows: Step 1: Acquisition of Ultrasonic Blood Flow Information – Using the Doppler ultrasound principle, the location of blood vessels and the distribution of blood flow velocity in the injection area are calculated from the blood flow feedback signal obtained by automatic ultrasound scanning.

[0022] To obtain the variable values ​​required for the blood flow relative velocity optimization model and achieve automated needle insertion, an ultrasound scan of the insertion site is performed using an instrument capable of automatically adjusting the emission angle and frequency. The Doppler principle is then used for automatic calculation. Ultrasound scanning can be either mobile or stationary. Taking arm vein needle insertion as an example, a mobile ultrasound probe can automatically translate along the arm axis within a certain distance around the injection site, parallel to the skin plane, and can also rotate axially within a certain angle range. This requires a support of appropriate length to allow for limited probe movement. A stationary probe can use a phased array, with electronically controlled phase difference for focusing, making it more suitable for body parts with limited space. Furthermore, the ultrasound transmitter must be able to automatically adjust and emit multiple sets of ultrasound at different frequencies within a certain range (e.g., above 7.5MHz). Since the insertion site is generally located superficially under the skin, a relatively high ultrasound frequency band can be used to improve sampling accuracy. Ultrasound scanning should be performed when the heart rate is stable to ensure a relatively stable cardiac cycle and to avoid or reduce blood flow disturbances.

[0023] To protect the privacy of the recipient, sampling must be completed within a limited space. The entire insertion device can be fixed to a localized area of ​​the body using securing devices, maintaining relative stillness with respect to the area requiring insertion to ensure the required accuracy for ultrasound detection and insertion. Selecting a suitable sampling volume is crucial and challenging for obtaining accurate blood flow information. For example, increasing the number of phased arrays can improve accuracy by increasing the density of the sampling points; however, precisely selecting the sampling volume is not within the scope of the method described in this invention—the following calculations assume that the sampling volume at each location can be automatically determined quickly and accurately to obtain sufficiently precise measurement results.

[0024] Based on the installation location of the insertion device, a reference coordinate is selected, and variables such as blood vessel position and blood flow velocity are automatically calculated using the Doppler ultrasound principle. The following describes a method for calculating and marking the spatial angle of blood flow direction using the Doppler ultrasound principle: like Figure 2 As shown, assume the erythrocyte swarm moves at a velocity v along the X-axis (blood flow direction), with W as the reference axis, which can be a vertical axis in the direction of gravity. The Z-axis, X-axis, and W-axis are in the same plane, where the Z-axis is perpendicular to the X-axis, and the angle between the Z-axis and W is θ. The ultrasonic probe emits at a frequency of Ultrasound waves travel at the speed of sound in the body medium. The signal propagates to the target and then returns as a scattered signal to the receiving probe at a frequency of [frequency missing]. Let the incident plane of the ultrasound be the plane containing the incident sound beam axis and the Z-axis. , The axis perpendicular to the Z-axis is Y; the angle between the ultrasound incident plane and the blood flow direction is... That is, the angle between the X-axis and the Y-axis is The angle between the incident sound beam axis and the horizontal plane is set as That is, the angle between the incident sound beam axis and the Y-axis is Let the scattering plane containing the scattering acoustic axis of the red blood cells (groups) along the receiver direction and the Z-axis be denoted as . , The axis perpendicular to the Z-axis is , The angle between the axis and the direction of blood flow is And the scattering acoustic axis is The included angle of the axis is The angle between the incident acoustic beam axis and the W-axis is denoted as . The angle between the scattering acoustic axis and the W-axis is ,in and The value can be directly measured. An excessively large angle between the incident sound beam axis and the blood flow direction should be avoided. , , and Although the initial value is a quantity to be determined, it should be ensured that it does not exceed sixty degrees to avoid inaccurate measurement results.

[0025] Ultrasonic receiving frequency With incident frequency The relational expression is: (1) In equation (1), Blood flow velocity and ultrasonic velocity in human tissue (or animal tissue) media The ratio, where t represents the emission time of the ultrasonic wave. The angle between the incident sound beam and the blood flow direction relative to the incident plane. Let be the angle between the scattered sound beam and the blood flow direction relative to the scattering plane. The spatial angles satisfy the following relationship: (2a) (2b) When the relative velocity of blood flow The value, i.e., the blood flow velocity and the ultrasonic velocity in the medium of human tissue (or animal tissue). The ratio is much smaller than At that time, the Doppler frequency shift obtained by ultrasound scanning The relative velocity of blood flow and its direction / angle have the following relationship: (3) The average speed ratio The relationship is as follows: (4) In equation (4), The frequency of ultrasound transmission is the center frequency, while the average angular frequency is... It can be obtained using the autocorrelation algorithm of Doppler echo signals.

[0026] Based on the measured Doppler frequency shift values, the precise blood flow velocity distribution and direction at different locations can be obtained through directional demodulation—that is, by using the frequency shift change equations, the corresponding time delay changes, and equation (2), the relative flow velocity and relative angle parameters can be solved. This allows us to obtain the distribution and orientation of local blood vessels in the area requiring needle insertion relative to the reference coordinates. When the ultrasound transmitter and receiver are located within the same probe (i.e., in the same spatial position), the frequency shift equations are expressed as follows: (5) or (6) In the formula, Indicates the change in frequency shift. It is a positive integer.

[0027] For example, if the parameters are kept constant in the frequency shift equation system At the same time, Become , , If the integer is positive, then the parameter can be obtained by measuring the minute change in the Doppler frequency shift. Value: (7a) or (7b) Similarly, if the parameters are kept constant At the same time, Become , Then there is (8) In particular, when , At that time, there is a simplified linear calculation form. (9) In equations (7b)-(10), The sign ensures that the computational amount on the right side is equal to that on the left side. or The positive and negative signs remain consistent. Then, calculate and record the angle between the Z-axis and the reference axis W using equation (2). .

[0028] In addition, while maintaining parameters At the same time, it can also be obtained by calculating the time delay caused by the change in sound path. Value: (10) In equation (10), The distance between the probe and the target detection point is denoted by , and t represents the time of ultrasonic wave emission. express Become and The moment the signal is received while remaining constant.

[0029] exist and Under precise and controllable conditions, the relative angle parameters can be solved, and the local relative velocity can be obtained from the frequency shift equation. To achieve the above calculations... and For precise control, a feasible approach is to emit a set of ultrasound pulses with fixed relative angles at each location where blood flow velocity needs to be detected (precise control using a phased array is relatively convenient). In setting the frequency shift change equations, the selected frequency shift intensity should be higher than a certain threshold, meaning that a large amount of low-intensity frequency shift information corresponds to non-blood flow related information and can be excluded in advance.

[0030] Step 2: Establish a blood flow relative velocity optimization model. Based on physical calculations and the blood flow relative velocity optimization model, obtain a set of optimal solutions for the insertion position and angle.

[0031] It can be considered as being close to the superficial layer of the skin Ideal candidate insertion points are those where the blood flow value changes slowly in the direction of blood flow and where the blood flow is relatively large (corresponding to thicker veins without bifurcations). (Assuming there are five such points, let's denote them as follows:) .in The optimal point obtained from physical calculations. The next best point is the second best point, and so on.

[0032] For the selection of candidate injection sites, a universally applicable optimization model can be considered, meaning one that is as applicable as possible to different parts of the human body and different animals. According to fluid dynamics principles, if a blood vessel is simplified to a cylinder of radius r, then the maximum blood flow velocity... and average flow velocity They are respectively and Blood flow ,in This represents the change in pressure per unit length along the direction of blood flow. This refers to the blood viscosity coefficient; on the other hand, according to the principle of fluid continuity, in a blood vessel with constant mass and no bifurcation... ,because Therefore, And thus (11) in, This represents the change in average relative velocity of blood flow per unit length.

[0033] This means that blood flow is highly dependent on the vessel radius, and changes in the vessel radius depend on changes in the relative velocity of blood flow. Therefore, the following instantaneous minimization model of the relative velocity of blood flow can be established: (12a) (12b) in , and An appropriate constant set based on actual measurement results; This refers to the directional derivative or difference vector of the relative velocity of blood flow at point p along the blood flow direction, depending on actual needs and desired effects. It can be set to the maximum relative velocity or the average relative velocity. This is a discriminant function used to determine the distance before and after a point p along the blood flow direction. Within the region, does the blood flow velocity exhibit any abrupt changes? Regions where there are no abrupt or significant changes in blood flow velocity are considered preferred regions; in this case, let... ;otherwise This corresponds to conditions such as blood vessel bifurcation, blockage, or partial rupture. For example, in situations where rapid measurement is possible... In the case of discriminant function It can be set as: (13a) (13b) in and They are respectively in the direction of blood flow and in the opposite direction of blood flow. Distance sampling points, For the threshold, For positive integers, The value is approximately equal to and not less than the ideal pin depth.

[0034] Without incurring significant additional computational costs (or where computational efficiency and workload permit), the discriminant function can be constructed by incorporating other key ultrasound information that effectively reflects vascular and blood flow characteristics, besides the relative blood flow velocity variable. In the relative blood flow velocity optimization model, an additional constraint is placed on the ideal range of relative blood flow velocity at the needle insertion site based on clinical experience; that is, a constraint is set... Locations where the relative blood flow velocity is too fast or too slow are considered undesirable. This applies to injections or blood draws from the arm or lower limb. The value can be set between 1 and 2 cm, and should not be less than the depth required for the insertion pin.

[0035] During needle insertion, blood flow velocity depends not only on the spatial distribution of blood vessels but also varies to some extent over time. This means that, considering the time factor comprehensively, i.e., assuming... As a time-dependent variable, the blood flow relative velocity optimization model needs to have a certain degree of stability over time. Therefore, where computational efficiency and cost allow, a time-dependent variable can be incorporated into the above blood flow relative velocity optimization model, i.e., a time-dependent assumption can be made. A set of time columns ( (where N is a large positive integer), corresponding to a certain ultrasonic emission frequency, where It contains m cardiac cycles. First, regarding... After performing noise reduction preprocessing, consider the following time-dependent blood flow relative velocity optimization model: (14a) in (14b) This reflects the stability of blood flow across multiple cardiac cycles; the blood flow velocity optimization model must satisfy the following constraints: (15a) in, (15b) and (15c) These are the estimation formulas for the pulsatility index and resistance index, respectively, which characterize vascular obstruction. For the first Average flow rate over one cardiac cycle. (Based on the median of (15a)) , , and The settings can initially identify and rule out blocked blood vessels, thereby reducing the computational load of the optimization model.

[0036] In the blood flow relative velocity optimization model, the norms can be selected according to the desired effect, such as those related to time series. or Norms, or more generally This approach aims to reduce the impact of noise on calculated values ​​while maintaining computational efficiency. To ensure high accuracy in the blood flow velocity model, consistent selection of sampling volume is crucial. Under stable heart rate conditions, the cardiac cycle... It can be determined by the time interval between the occurrence of adjacent peaks.

[0037] The time-dependent blood flow velocity optimization model can be used in Item changed to constraint condition To further simplify the calculation, equation (14a) can be replaced with (16a) And satisfy the constraints. (16b) An optimized model with multiple local optima or near-optimal solutions is considered ideal, corresponding to multiple candidate sites in clinical practice that can be used for needle insertion.

[0038] Step 3: Combine the preferred solution obtained in Step 2 with a neural network with selection function to determine the optimal pin insertion point and pin angle, thus obtaining the optimal pin insertion solution; To improve clinical outcomes, a selection-enabled neural network (Network 1) is trained using optimal values ​​obtained from physical measurements and mathematical optimization models, along with actual clinical performance scores, and integrated into a learning chip. A simple example is generating a convolutional neural network with multiple intermediate layers, using parameters corresponding to candidate injection points. ,…, The sequence numbers (1, 2, 3, 4, 5) serve as input data for the input layer, and experienced professionals provide evaluation feedback based on physically calculated coordinates and angles. The evaluation feedback from the training group may include preferred needle insertion locations pre-marked on the experimental samples by professionals or using other equipment, the actual blood vessel direction during needle insertion, and a ranking based on priority, such as (2, 1, 3, 5, 4). The preferred needle insertion locations and blood vessel directions marked in the experiment are scanned and calculated by the system and fed back to the neural network as data. After multiple evaluations, the trained neural network with selection capabilities can provide a better output scheme based on the input data, determining the best blood vessel for needle insertion, the needle insertion location p, and the precise direction of blood flow at that point. ).

[0039] The physical calculations, blood flow optimization model calculations, and neural network training described above must be integrated into a small chip or small computing device (rather than on an external standalone computer or large computing device), thus allowing the entire pin device to be easily portable. Therefore, the selection of the optimization model and the number of neural network nodes and layers should meet the actual computing capabilities.

[0040] It should also be noted that although the equipment required for this invention utilizes the Doppler ultrasound principle, it differs fundamentally from a color Doppler blood flow imaging system in the following ways: This invention utilizes a blood flow optimization model and Doppler principle to calculate blood flow velocity distribution and measure the precise position and angle of the sampling volume relative to the reference coordinates, preparing for needle insertion. It eliminates the need for imaging, resulting in a small size, light weight, and easy portability. In contrast, color Doppler blood flow imaging is a direct qualitative method. Currently, color Doppler blood flow imaging systems require a monitor and computer to work together, and the acquired ultrasound signals must be processed by digital ultrasound (DSC) to generate color images, involving a significant amount of time spent processing the large amount of data required to generate three-dimensional images.

[0041] Step 4: Based on the optimal pin insertion scheme, control the movement of the pin tip to accurately contact the optimal pin insertion point and begin pin insertion; during the pin insertion process, combine with a neural network with learning function to learn and simulate the pin insertion technique of professionals, automatically adjust the pin insertion angle and speed, and obtain the pin insertion path.

[0042] The adjustment device used in the pin insertion process has translation and rotation functions, and the optimal pin insertion scheme obtained in step three is transmitted via parameter commands. The output is sent to the adjustment device. The adjustment device then moves the needle body horizontally to approximately 2 cm above the corresponding insertion point, and rotates and adjusts the needle body at a specified angle. The needle body is then lowered parallel to the target point so that the needle tip precisely contacts the optimal insertion point, initiating the insertion process. The insertion path is planned as follows: [The calculation of the local blood flow direction at the optimal insertion point is missing from the original text.] Corresponding pin angle Therefore, the initial pin angle should be selected as follows: Then gradually adjust the needle to align it with the local blood flow direction.

[0043] Professional pin insertion techniques are reflected in the angle during the pin insertion process. and insertion speed Subtle changes occur over time. It is necessary to introduce another neural network model with learning capabilities (Network 2), such as a convolutional neural network, to learn these changes; the insertion process is guided by this learning-capable neural network. The training of Network 2 can be pre-arranged by professionals.

[0044] This invention utilizes ultrasound to accurately measure local blood flow information, combines a blood flow velocity optimization model and a neural network to automatically adjust the needle position and angle, and simulates the needle insertion technique of a professional to complete the automatic needle insertion. The automatic needle insertion technology of this invention provides a technical foundation for the manufacture of portable automatic needle insertion devices, facilitating injection and blood collection in environments lacking medical personnel.

[0045] It will be understood by those skilled in the art that the above descriptions are merely preferred examples of the invention and are not intended to limit the invention. Although the invention has been described in detail with reference to the foregoing examples, those skilled in the art can still modify the technical solutions described in the foregoing examples or make equivalent substitutions for some of the technical features. All modifications and equivalent substitutions made within the spirit and principles of the invention should be included within the scope of protection of the invention.

Claims

1. An automatic needle insertion method combining a blood flow optimization model and a neural network, characterized in that, Includes the following steps: Step 1: Acquisition of Ultrasonic Blood Flow Information – Using the Doppler ultrasound principle, the spatial angular relationship between the location of the blood vessels and the distribution of blood flow velocity in the area to be needled is calculated from the blood flow feedback signal measured by automatic ultrasound scanning. Step 2: Establish a blood flow relative velocity optimization model. The instantaneous minimization expression of the blood flow relative velocity optimization model is as follows: (1a) (1b) In equations (1a)-(1b), c, and It is a constant. Let p be the directional derivative or difference vector of the relative velocity of blood flow along the flow direction. The discriminant function is used to determine the distance before and after a point p along the blood flow direction. Does the blood flow velocity inside the body exhibit abrupt changes? In the blood flow relative velocity optimization model, each norm is relative to the time series. or Norm, or , where q≥1; A set of optimal solutions for the insertion pin position and angle were obtained through physical calculations and a blood flow relative velocity optimization model; Step 3: The preferred scheme is combined with a neural network with selection function to determine the optimal pin insertion point and pin angle, thus obtaining the optimal pin insertion scheme; Step 4: Based on the optimal pin insertion scheme, and combined with a neural network with learning capabilities, learn and simulate the pin insertion techniques of professionals to obtain the pin insertion path; In step one, in order to obtain the values ​​of each variable required for the blood flow relative velocity optimization model and achieve the purpose of automatic needle insertion, an ultrasound scan of the body at the site to be needled is performed, which can automatically adjust the emission angle and frequency, and then the Doppler principle is used to automatically calculate. Ultrasonic scanning can be performed using either a motorized or a stationary method. Motorized ultrasound scanning requires the ultrasound probe to be able to rotate automatically at a variable angle within a local area and perform a motorized translational scan along a movable support. Stationary ultrasound scanning uses an electronically controlled phased array mode. The method for measuring and marking the spatial angle of blood flow using the Doppler ultrasound principle is as follows: The erythrocyte population moves at a velocity v along the X-axis, which is the direction of blood flow. W is the reference axis. The Z-axis, X-axis, and W-axis are all in the same plane, with the Z-axis perpendicular to the X-axis and the angle between the Z-axis and W being θ. The ultrasonic probe emits at a frequency of Ultrasound waves travel at the speed of sound in the body medium. The signal propagates to the target and then returns as a scattered signal to the receiving probe at a frequency of [frequency missing]. Let the ultrasound incident plane be the plane P1 containing the incident sound beam axis and the Z-axis, and the axis perpendicular to the Z-axis in P1 be Y; the angle between the ultrasound incident plane and the blood flow direction is α, i.e., the angle between the X-axis and the Y-axis is α, and the angle between the incident sound beam axis and the horizontal plane is β, i.e., the angle between the incident sound beam axis and the Y-axis is β; let the scattering plane containing the scattered sound axis of the red blood cell population along the receiver direction and the Z-axis be P2, and the axis perpendicular to the Z-axis in P2 be Y', the angle between the Y' axis and the blood flow direction is α', and the angle between the scattered sound axis and the Y' axis is β'; let the angle between the incident sound beam axis and the W-axis be γ, and the angle between the scattered sound axis and the W-axis be γ', where the values ​​of γ and γ' can be directly measured; Ultrasonic receiving frequency With incident frequency The relational expression is: (7) In equation (7), μ is the relative velocity of blood flow, that is, the blood flow velocity and the ultrasonic velocity in the tissue medium of human or animal tissue. The ratio, where t represents the emission time of the ultrasonic wave; the spatial angles satisfy the following relationship: (8a) (8b) Doppler shift The relative velocity μ of blood flow has the following relationship with the direction angle: (9) remember This is the transmission center frequency of the ultrasound. If the average angular frequency is given, then the average velocity ratio is... Relationship: (10) Based on the effective ultrasound feedback bands acquired and filtered after the scanning angle change, a set of Doppler frequency shifts is obtained by directional demodulation, and the vessel position and blood flow velocity variables are calculated accordingly. That is, the relative flow velocity and relative angle parameters are solved by using the frequency shift change equation set, the corresponding time delay change and equation (8). This allows us to obtain the distribution and orientation of local blood vessels in the area requiring needle insertion relative to the reference coordinates. When the ultrasound transmitter and receiver are located within the same probe, i.e., in the same spatial position, the expression for the frequency shift change equations is as follows: (11) or (12) In the formula, Indicates the change in frequency shift. It is a positive integer; While keeping the parameter α in the frequency shift equation system, set , Then there is (13a) While keeping parameter β in the frequency shift equations, α becomes , , If is a positive integer, then: (13b) In particular, when , At that time, (13a)-(13b) has a simplified linear computational form. (14) In equations (13a)-(14), the ± sign ensures that the calculated value on the right side is consistent with the sign of cosβ or cosα on the left side; then, the angle between the Z-axis and the reference axis W is calculated and recorded by equation (8). .

2. The automatic needle insertion method combining a blood flow optimization model and a neural network according to claim 1, characterized in that, In step two, a set of candidate, relatively optimal needle coordinates are obtained using a blood flow relative velocity optimization model with local optima. The specific construction of the blood flow relative velocity optimization model is as follows: For the cylindrical blood vessel model, blood flow... And in blood vessels with constant mass and no bifurcation The average relative velocity of blood flow was obtained. The functional relationship between the vessel diameter and the blood flow direction: (2) Where r is the radius of the blood vessel. This represents the change in pressure per unit length along the direction of blood flow. Let the unit length change be the average relative velocity of blood flow; thus, establish instantaneous minimization models of relative velocity of blood flow (1a)-(1b); in the instantaneous minimization model, with position p as the center and a distance of Within the sampling range, the region where the blood flow velocity does not jump or change significantly in the direction of blood flow is the preferred region. In this case, let the discriminant function... ;otherwise This corresponds to situations where blood vessels bifurcate, become blocked, or partially rupture. When computational efficiency and computational cost allow, a time-varying variable is incorporated into the instantaneous minimization model of relative blood flow velocity, requiring the blood flow relative velocity optimization model to be stable over time; at this point, first... Noise reduction preprocessing is performed, including Given a time series over m cardiac cycles [0, mT], N is a positive integer; then, a time-dependent blood flow relative velocity optimization model is established: (3a) in (3b) The blood flow relative velocity optimization model must satisfy the following constraints: (4a) In equation (4a) (4b) (4c) These are estimation formulas for the pulsatility index and resistance index, respectively, which are clinically used to characterize vascular obstruction. The average flow rate during the (n+1)th cardiac cycle; Through (4a) median I P,min I P,max I R,min and I R,max The settings allow for the preliminary identification and exclusion of blocked or diseased blood vessels, thereby reducing the computational load of the blood flow relative velocity optimization model.

3. The automatic needle insertion method combining the blood flow optimization model and neural network according to claim 2, characterized in that, In order to be able to measure quickly In the case of the discriminant function The expression is: (5a) (5b) In equations (5a)-(5b), p n With p -n Distance from position p in the direction of blood flow and in the opposite direction of blood flow, respectively. sampling points, For the threshold, For positive integers, The value should be close to and not less than the needle insertion depth; in equations (1a)-(1b) and (3a)-(3b), additional limitations are made on the range of relative blood flow velocity at the ideal needle insertion position based on clinical experience, i.e., a setting is made. .

4. The automatic needle insertion method combining the blood flow optimization model and neural network according to claim 2, characterized in that, In the time-dependent blood flow velocity optimization model, equation (3a) can be replaced by: (6a) And satisfy the constraints. (6b) This further simplifies the calculation.

5. The automatic needle insertion method combining the blood flow optimization model and neural network according to claim 1, characterized in that, The neural network with selection function in step three contains multiple intermediate layers, using parameters corresponding to the relative blood flow velocity and coordinates of candidate insertion points. The input data, including the serial number, is used as the input layer, and professional personnel provide evaluation feedback based on the physically calculated coordinates and angles. Blood flow velocity and ultrasonic velocity in tissue media of human or animal tissues The ratio, The angle between the incident sound beam and the blood flow direction is determined; the evaluation feedback from the training group includes the preferred needle insertion positions pre-marked on the experimental samples by professionals using other equipment, the actual blood vessel direction during needle insertion, and the sequence number ranking given according to priority; the trained neural network with selection function uses the input data in parametric form. Output the optimal pin coordinates and pin angle.

6. The automatic needle insertion method combining the blood flow optimization model and neural network according to claim 1, characterized in that, The pin path planning in step four is as follows: If the calculated local blood flow direction at the optimal insertion point is... Corresponding pin angle Therefore, the initial pin angle should be selected as follows: Then, the needle body is gradually adjusted to align with the local blood flow direction; this gradual adjustment process is guided by the neural network with learning capabilities, which is trained using professional needle insertion techniques, specifically learning the needle insertion angle during the insertion process. and insertion speed Changes over time.

Images