Blood vessel model collision force detection method for interventional operation training

Abstract

The invention discloses a blood vessel model collision force detection method for interventional operation training, which comprises the following steps: arranging a blood vessel model in a closed shell, and connecting the blood vessel model with an external flow pump to promote fluid to circularly flow in the blood vessel model; the closed shell outside the blood vessel model is filled with medium liquid, and the blood vessel model is completely immersed in the liquid medium; a plurality of pressure sensors connected with an external data processing unit are orderly distributed on the inner wall surface of the closed shell; during training, the guide wire extends into the blood vessel model for interventional operation, when the guide wire collides with the wall surface of the blood vessel model, the wall surface of the blood vessel model triggers the medium liquid to trigger the medium liquid in the closed shell to generate pressure field change, and the corresponding pressure sensor monitors pressure field data and feeds back the pressure field data to the data processing unit; the data processing unit obtains the pressure value and calculates and outputs the corresponding collision force, objective, quantitative and comprehensive data feedback of the collision force in interventional operation training is achieved, and the operation training effect is greatly improved and guaranteed.

Description

Technical Field

[0001] This invention relates to the field of interventional surgery training devices, and in particular to a method for detecting the collision force of a vascular model used in interventional surgery training. Background Technology

[0002] Interventional surgery is a minimally invasive diagnostic and treatment technique in which doctors use catheters, guidewires, and other instruments to operate inside blood vessels. This technique demands a high level of tactile sensitivity from the doctor, who needs to judge the position of the guidewire tip within the blood vessel, especially its contact and collision with the vessel wall. To train qualified interventional surgeons, extensive simulation training using interventional surgery training devices is typically required.

[0003] Most existing interventional surgery training devices include a vascular model for trainees to practice guidewire / catheter manipulation. During training, the impact force between the guidewire tip and the vessel wall is a key factor determining surgical safety; excessive force can lead to vascular damage or perforation. However, trainees primarily rely on their subjective tactile sense to perceive this force, lacking objective, quantifiable data feedback.

[0004] To address this issue, existing technologies have attempted to directly mount pressure sensors on the outer wall of blood vessel models to measure impact forces. However, this approach has significant drawbacks: (1) Difficult to install and cannot cover all areas: The structure of blood vessel models (especially those simulating coronary arteries, cerebral blood vessels, etc.) is complex and varied. It is almost impossible to accurately install sensors in every area where collisions may occur, which easily leads to measurement blind spots.

[0005] (2) Single information dimension: Pressure sensors can usually only measure the pressure perpendicular to the contact surface and cannot obtain complete three-dimensional vector information of the collision force (such as lateral friction force, shear force, etc.), resulting in incomplete feedback information.

[0006] Therefore, there is an urgent need in this field for a device and method that can monitor the impact force in interventional surgery training in a global, comprehensive, and multi-dimensional real-time manner to achieve better surgical teaching and training results. Summary of the Invention

[0007] To address the aforementioned issues, this application provides a structurally sound method for detecting collision force in a vascular model used for interventional surgery training. This method enables objective, quantitative, and comprehensive data feedback on collision force during interventional surgery training, thereby significantly improving and ensuring the effectiveness of surgical training.

[0008] The technical solution adopted in this invention is as follows: A method for detecting the impact force of a vascular model for interventional surgery training includes setting a vascular model inside a closed shell, the vascular model being connected to an external flow pump to promote fluid circulation within the vascular model; filling the closed shell outside the vascular model with a medium liquid, the vascular model being completely immersed in the liquid medium; and multiple pressure sensors connected to an external data processing unit are arranged in an orderly manner on the inner wall of the closed shell. During training, the guidewire is inserted into the vascular model to perform the interventional operation. When the guidewire collides with the wall of the vascular model, the wall of the vascular model will touch the medium liquid, triggering a change in the pressure field of the medium liquid inside the sealed shell. The corresponding pressure sensor monitors the pressure field data and feeds it back to the data processing unit. The data processing unit obtains the pressure value and calculates and outputs the corresponding collision force.

[0009] As a further improvement to the above technical solution: The blood vessel model is made of an elastic material. When the wall of the blood vessel model is subjected to force, it will undergo elastic deformation and trigger the medium liquid. The material of the blood vessel model includes, but is not limited to, elastic silicone, hydrogel, and polyurethane. The medium liquid is made of an incompressible material and is filled in a closed shell without air bubbles. The material of the medium liquid includes, but is not limited to, silicone oil, glycerin, and water. The closed shell is made of high-transparency acrylic material.

[0010] The enclosed housing is a sealable cuboid structure equipped with a top cover. Pressure sensors are evenly distributed on each inner wall surface of the enclosed housing. The medium liquid fills the inner wall space of the enclosed housing, and the pressure sensors on each inner wall surface simultaneously monitor the pressure changes of the medium liquid inside the enclosed housing in real time.

[0011] The sealed shell has an inlet and an outlet on its side that are connected to the end of the vascular model. The vascular model is circulated with an external flow pump through the inlet and outlet to simulate the flow of blood in the blood vessels. During training, a guidewire is inserted into the vascular model through the inlet and / or outlet to train for interventional surgery.

[0012] When the guidewire is in normal operation and the intervention is not colliding with the wall of the vascular model, the fluid in the vascular model simulates the flow of blood. The vascular model acts on the medium liquid to form a basic pressure field, which is continuously monitored by various pressure sensors to obtain the dynamic fluctuation baseline of the basic pressure field. When a collision occurs, the pressure field of the medium liquid changes, and the monitoring value of the corresponding pressure sensor will change instantaneously. The data processing unit uses the difference of the instantaneous change as the pressure value.

[0013] The data processing unit includes a database that maps pressure values ​​to impact forces, and the impact force is calculated from the pressure values ​​based on the database.

[0014] The database is configured with n pressure sensors. During training, the pressure values ​​monitored by each pressure sensor constitute n. column vectors , ; in, This represents the pressure value monitored and obtained by the nth pressure sensor. The collision force is defined as a three-dimensional force vector. , constitute 3 Column vectors ; in, , , These represent the three components of the force in a three-dimensional Cartesian coordinate system; Database consists of column vectors and column vectors Constructing a mapping function , .

[0015] The database is generated as follows: In the simulation model, multiple collision sections are set at intervals along the central axis of the blood vessel model. Multiple collision points are set at intervals along the circumference on each collision section. Forces are applied at each collision point along the normal and tangential directions, with the forces ranging from 0.1N to 2.0N in steps of 0.1N to form discrete forces. The pressure values ​​corresponding to each collision point on each collision section under each discrete force are obtained, thereby obtaining the force-pressure value relationship and generating a database corresponding to the mapping between collision force and pressure value.

[0016] Before the database is generated, the simulation model is physically verified and calibrated. The physical verification and calibration method is as follows: a known force is applied to a preset key position of the real prototype, and the pressure sensor feeds back the measured pressure value; a simulation model is constructed, and the pressure simulation value is obtained by applying force under the same conditions; the pressure simulation value is compared with the pressure measurement value, and the verification iteration is repeated by adjusting the parameters in the simulation model until the difference between the pressure simulation value and the pressure measurement value is within the threshold.

[0017] The data processing unit includes a collision force calculation model, which is a neural network model trained based on a database. Given n pressure sensors, construct an input layer with n neurons to receive pressure values ​​monitored by the n pressure sensors. Set 1-3 hidden layers, each with 16-128 neurons; use ReLU as the activation function; set an output layer with 3 neurons to output collision force. The model consists of three components; the dataset from the database is input into the neural network model, and the internal weights and biases are adjusted through the backpropagation algorithm and gradient descent optimizer for model training; after training, the neural network model is deployed to the data processing unit, and the predicted collision force is obtained by inputting the real-time collected pressure data.

[0018] Compared with the prior art, the present invention has the following beneficial effects: This invention immerses a vascular model in a medium liquid, using the medium liquid as a force transmission medium. The impact force on the vascular model is transmitted through the medium liquid to a pressure sensor on the inner wall of the sealed shell. This ensures that the pressure sensor can sensitively detect any impact that occurs anywhere inside the vascular model. It is easy to operate, has keen and accurate perception, and completely solves the problems of difficult installation and measurement blind spots in traditional solutions. It realizes objective, quantitative, and comprehensive data feedback of impact force in interventional surgery training, greatly improving and ensuring the effectiveness of surgical training. The present invention also includes the following advantages: The pressure field of the entire medium liquid can be analyzed through the pressure sensor array formed on each wall surface, and the three-dimensional vector of the collision force can be reconstructed by reverse calculation, which provides more comprehensive and accurate force feedback information than traditional one-dimensional pressure sensors. The sealed shell and the vascular model can be separated from each other. When changing vascular models with different structures and different lesions, there is no need to change the layout of the pressure sensors on the inner wall of the sealed shell. The device has strong applicability and helps to reduce the cost of consumables. It provides standardized, quantifiable, and objective data for physician training, helping them quickly establish a precise correspondence between tactile sensation and actual force, effectively improving training quality and efficiency, and is of great significance for ensuring the safety of future clinical surgeries. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the external appearance of the interventional surgery training device of the present invention.

[0020] Figure 2 This is a schematic diagram of the internal structure of the interventional surgical device of the present invention.

[0021] Figure 3 This is a schematic flowchart of the collision force detection method of the present invention.

[0022] The components include: 1. Enclosed shell; 2. Inlet; 3. Outlet; 4. Top cover; 5. Pressure sensor; 6. Blood vessel model; 7. Fixing structure. Detailed Implementation

[0023] The specific embodiments of the present invention will now be described with reference to the accompanying drawings.

[0024] This embodiment proposes a method for detecting the collision force of a vascular model used in interventional surgery training. The interventional surgery training device used is as follows: Figure 1 and Figure 2 As shown, the system includes a blood vessel model 6 set inside a closed housing 1. The blood vessel model 6 is a training model for simulating human blood vessels. The blood vessel model 6 is connected to an external flow pump to cause fluid to circulate within the blood vessel model 6. The closed housing 1 outside the blood vessel model 6 is filled with a medium liquid, and the blood vessel model 6 is completely immersed in the liquid medium. Multiple pressure sensors 5 connected to an external data processing unit are arranged in an orderly manner on the inner wall of the closed housing 1.

[0025] During training, such as Figure 3 As shown in the usage stage, the guidewire is inserted into the blood vessel model 6 to perform the interventional operation. When the guidewire collides with the wall of the blood vessel model 6, the wall of the blood vessel model 6 will touch the medium liquid, triggering a change in the pressure field of the medium liquid inside the sealed shell 1. The pressure field data is monitored by the corresponding pressure sensor 5, forming a real-time dynamic response of the pressure field, and fed back to the data processing unit. The data processing unit obtains the pressure value and calculates and outputs the corresponding collision force.

[0026] In this embodiment, by immersing the blood vessel model 6 in a medium liquid, the medium liquid is used as a force transmission medium to transmit the impact force on the blood vessel model 6 to the pressure sensor 5 on the inner wall of the closed shell 1. This allows the pressure sensor 5 to sensitively detect any impact that occurs inside the blood vessel model 6, making it easy to operate and providing accurate and sensitive perception. It completely solves the problems of difficult installation and blind spots in traditional solutions, and realizes feedback detection of impact forces at specific locations, in specific directions and of specific magnitudes within the blood vessel model 6.

[0027] In this embodiment, the pressure sensor 5 can be a highly sensitive miniature pressure sensor, depending on actual needs.

[0028] In this embodiment, the pressure sensors 5 are arrayed on the inner wall of the enclosed housing 1, rather than on the blood vessel model 6, which simplifies the overall device structure and layout and improves the flexibility of use.

[0029] In this embodiment, the vascular model 6 can be a standardized general model or a personalized model 3D printed based on the patient's CT / MRI data; it can be immersed in the medium liquid contained in the closed shell 1.

[0030] In this embodiment, the blood vessel model 6 is made of an elastic material with a certain degree of elasticity to simulate the mechanical properties of real blood vessels. When the wall of the blood vessel model 6 is subjected to force, it will undergo elastic deformation and touch the medium liquid. The material of the blood vessel model 6 includes, but is not limited to, elastic silicone, hydrogel, and polyurethane.

[0031] In this embodiment, the material selection for the blood vessel model 6 can be set according to actual operational needs. Based on training experience, its Shore A hardness needs to be controlled within the range of 20HS-50HS.

[0032] In this embodiment, the medium liquid is filled into the closed shell 1 in an incompressible material without air bubbles. The medium liquid material includes, but is not limited to, silicone oil, glycerin, and water, thereby effectively ensuring the reliability of the medium liquid force transmission.

[0033] The enclosed shell 1 is made of high-transparency acrylic material, which helps to ensure visual observation and improve visibility during training.

[0034] The enclosed housing 1 is a sealable cuboid structure equipped with a top cover 4. Pressure sensors 5 are evenly distributed on each inner wall surface of the enclosed housing 1. The medium liquid fills the inner wall space of the enclosed housing 1, and the pressure changes of the medium liquid inside the enclosed housing 1 are monitored in real time by the pressure sensors 5 on each inner wall surface.

[0035] In this embodiment, the pressure sensors 5 can be arranged in an array on the inner wall of each side of the cuboid structure. The density, number, and actual arrangement of the pressure sensors 5 can be set and adjusted according to the accuracy required for actual detection.

[0036] In this embodiment, various mounting slots for accommodating the pressure sensor 5 can be opened on the inner wall of the cuboid structure to facilitate the installation of the pressure sensor 5. Other fixing methods such as adhesive can also be used.

[0037] In this embodiment, the pressure sensor 5 can be a non-contact sensor such as a piezoresistive type, pressure point type, capacitive type, or fiber optic grating type.

[0038] The sealed shell 1 has an inlet 2 and an outlet 3 on its side that are connected to the end of the vascular model 6. The vascular model 6 is circulated with an external flow pump through the inlet 2 and outlet 3 to simulate the flow of blood in the blood vessels. During training, the guide wire is inserted into the vascular model 6 through the inlet 2 and / or outlet 3 to train interventional surgery.

[0039] In this embodiment, in order to ensure the airtightness of the sealed shell 1 and the training effect, the inlet 2 and outlet 3 for the blood vessel model 6 and the outlet for the connection line of the pressure sensor 5 are all filled and sealed with waterproof material.

[0040] In this embodiment, the blood vessel model 6 can be set with several fulcrums, and fixed to the inner cavity of the closed shell 1 by means of the fixing structure 7 at the fulcrums. The fixing structure 7 can be a conventional clamp, fastener, etc., which can realize the quick installation and support of the blood vessel model 6 in the closed shell 1 and the quick removal and switching.

[0041] When the guidewire, during normal operation and intervention, does not collide with the wall of the vascular model 6, the fluid in the vascular model 6 simulates blood flow. The vascular model 6 acts on the medium liquid to form a basic pressure field, which is continuously monitored by each pressure sensor 5 to obtain the dynamic fluctuation baseline of the basic pressure field. When a collision occurs, the pressure field of the medium liquid changes, and the monitoring value of the corresponding pressure sensor 5 will change instantaneously. The data processing unit uses the difference of instantaneous changes as the pressure value. This effectively eliminates the interference of slowly changing factors such as simulated blood flow and accurately extracts the pressure signal caused by the collision.

[0042] In practice, instantaneous changes can be set based on experience and accuracy requirements. For example, the monitoring value that is instantaneously greater than the preset value can be set as the detection target, and the difference in instantaneous changes can be used as the pressure value input of the data processing unit.

[0043] The data processing unit in this embodiment includes a database that maps pressure values ​​to collision forces. The collision force is calculated based on the pressure values ​​mapped from the database. The detection feedback from each pressure sensor 5 can be used as the database input, and the corresponding collision force can be calculated and output.

[0044] The database specifies that the number of pressure sensors 5 is n. During the training process, the pressure values ​​monitored by each pressure sensor 5 constitute n. column vectors , which are input variables; ; in, This represents the pressure value monitored and obtained by the nth pressure sensor, 5. The collision force is defined as a three-dimensional force vector. , constitute 3 Column vectors, used as output variables; ; in, , , These represent the three components of the force in a three-dimensional Cartesian coordinate system; Database consists of column vectors and column vectors Constructing a mapping function , This enables the mapping relationship between pressure values ​​and collision forces in the database.

[0045] In this embodiment, the database can be generated as follows: In the simulation model, multiple collision sections are set at intervals along the central axis of the blood vessel model 6, and multiple collision points are set at intervals along the circumference on each collision section. Forces are applied at each collision point along the normal and tangential directions, with the forces ranging from 0.1N to 2.0N in steps of 0.1N to form discrete forces. The pressure values ​​corresponding to each collision point on each collision section under each discrete force are obtained, thereby obtaining the relationship between force and pressure values ​​and generating a database corresponding to the mapping between collision force and pressure value.

[0046] In this embodiment, the settings of the collision cross section, the collision point, and the force step size in the simulation model are all closely related to the accuracy and precision of the database; these settings can be made according to actual detection requirements.

[0047] In this embodiment, the simulation model needs to be physically verified and calibrated before the database is generated to ensure the accuracy of the data relationships in the database.

[0048] The physical verification and calibration method is as follows: a known force is applied to a preset key position of the real prototype, and the pressure sensor 5 feeds back the measured pressure value; a simulation model is constructed, and the pressure simulation value is obtained by applying force under the same conditions; the pressure simulation value is compared with the pressure measurement value, and the verification iteration is repeated by adjusting the parameters in the simulation model until the difference between the pressure simulation value and the pressure measurement value is within the threshold.

[0049] In practice, relying solely on data relationships in a database is relatively limited, especially considering the complexity and nonlinearity of real-world systems. Therefore, neural network models can be used to obtain more accurate and comprehensive data relationships.

[0050] The data processing unit includes a collision force calculation model, which is a neural network model trained on a database.

[0051] The number of pressure sensors 5 is set to n, and an input layer with n neurons is constructed to receive the pressure values ​​monitored by the n pressure sensors 5. The system has n components; it has 1-3 hidden layers, each with 16-128 neurons; it uses ReLU (Rectified Linear Unit) as the activation function; and it has an output layer with 3 neurons that outputs the collision force. The three components constitute a three-dimensional force vector.

[0052] The dataset from the database is input into the neural network model, and the model is trained and iterated by continuously adjusting the internal weights and biases through the backpropagation algorithm and gradient descent optimizer (such as the Adam optimizer).

[0053] The training objective (loss function L) can be set to minimize the mean squared error (MSE) between the predictive power and the true power: in The mapping function representing the neural network, For the first in the dataset One collision force, The total number of data points in the dataset. for The values ​​in the table. Training continues until the loss function L converges to an acceptable low level.

[0054] After training, the neural network model is deployed to the data processing unit, and the predicted collision force is obtained by inputting the real-time collected pressure data.

[0055] Of course, in practice, other forms of machine learning models can also be used for data training and application, which can construct the mapping relationship between pressure value and collision force, and map the corresponding collision force from the input of pressure value.

[0056] like Figure 3 As shown, the vascular model collision force detection method for interventional surgery training in this embodiment includes a calibration training stage and a usage stage. The calibration training stage is mainly used for the establishment of the database and collision force calculation model in the data processing unit; the usage stage is the training and use combined with the training device.

[0057] 1. Calibration and training phase: A complete device model was established using a virtual simulation platform to simulate the insertion of guidewires or catheters into blood vessels. Finite element analysis was used to synchronously acquire data from the corresponding sensor arrays. By adjusting the material parameters within the model, consistency between the virtual and real data was ensured. Based on this, numerous experiments were conducted on the virtual platform to build a large database. This database records the pressure readings generated by the sensor arrays on the inner wall of the vessel when three-dimensional forces of different magnitudes and directions are applied at different locations within the blood vessel model. Based on this database, the final force reconstruction algorithm model was trained and generated.

[0058] a. Physical verification and calibration: To ensure consistency between simulation and reality, physical verification is required.

[0059] In practice, a robotic arm equipped with a high-precision six-dimensional force sensor can be used to apply a known force (such as a pushing force of 1N) to several key locations on a real prototype, and record the readings of the real sensor (an array of internal wall pressure sensors). This reading is then compared with the predicted value of the simulation model under the same conditions. If the error exceeds the range, the material parameters in the simulation model (such as liquid viscosity and model elastic modulus) are fine-tuned and re-verified, iteratively optimizing until the model is accurate.

[0060] b. Database Generation: A digital twin model completely identical to the physical device is established in a virtual simulation platform. A collision section is set every 5 mm along the central axis of the blood vessel model. On each section, a collision point is simulated every 45 degrees (i.e., 8 directions) along the circumference. At each collision point, discrete forces ranging from 0.1 N to 2.0 N (in 0.1 N increments) are applied in the normal and two tangential directions. Finite element analysis is used to record the sensor array pressure readings corresponding to each force application, thereby generating a high-density "force-pressure response" database covering the entire workspace.

[0061] c. Model Training: Based on the generated database, a backpropagation neural network (BPNN) was selected as the force reconstruction algorithm model. Pressure readings were used as the input layer of the network, and three-dimensional force vectors were used as the output layer. The model was trained using 80% of the database as the training set and 20% as the test set until the error on the test set converged to below a preset threshold.

[0062] In practice, a model library management system can also be established. That is, for each different vascular model, the aforementioned AC process is repeated to generate its own database and algorithm model, which are then stored in the data processing unit. When using the software, the user simply selects the currently installed vascular model model in the software interface, and the system automatically loads the corresponding algorithm model, thus achieving rapid adaptation to different vascular models.

[0063] 2. Usage phase: a. The trainee physician inserts a guidewire / catheter into a vascular model inside a closed container to perform the procedure.

[0064] b. When the guidewire tip collides with the blood vessel wall, the blood vessel wall undergoes a slight deformation.

[0065] c. This deformation pushes the surrounding incompressible conductive fluid, causing an instantaneous change in the pressure field throughout the sealed cavity.

[0066] d. An array of pressure sensors installed on the inner wall of the container captures the distribution changes of this pressure field in real time and sends the multi-point pressure data to the data processing unit.

[0067] e. The data processing unit calls the preset force reconstruction algorithm model to solve the collected real-time pressure data and instantly calculate the three-dimensional vector (magnitude, direction) and the location of the point of application of the collision force.

[0068] f. The calculated force information is fed back to the trained physicians in real time via a display screen, sound, or force feedback device, providing them with objective guidance.

[0069] In this embodiment, the pressure field of the entire medium liquid can be analyzed by the array of pressure sensors 5 formed on each wall surface of the closed shell 1, and the three-dimensional vector of the collision force can be calculated and reconstructed, which provides more comprehensive and accurate force feedback information than the traditional one-dimensional pressure sensor. In this embodiment, the enclosed shell 1 and the blood vessel model 6 can be separated from each other. When replacing blood vessel models 6 with different structures and different lesions, there is no need to change the layout of pressure sensors 5 on the inner wall of the enclosed shell 1. The device has strong applicability and helps to reduce the cost of consumables. This embodiment provides standardized and quantifiable objective data for doctors' training, helping them to quickly establish a precise correspondence between feel and actual force, effectively improving training quality and efficiency, and is of great significance for ensuring the safety of future clinical surgeries.

[0070] This invention enables objective, quantitative, and comprehensive data feedback on collision forces during interventional surgery training, greatly improving and ensuring the effectiveness of surgical training.

[0071] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0072] The above description is an explanation of the present invention and not a limitation thereof. The scope of the present invention is defined by the claims. Within the scope of protection of the present invention, any form of modification may be made.

Claims

1. A method for detecting the impact force of a vascular model used in interventional surgery training, characterized in that: The system includes a vascular model (6) set inside a closed housing (1), the vascular model (6) being connected to an external flow pump to cause fluid to circulate within the vascular model (6); a medium liquid is filled inside the closed housing (1) outside the vascular model (6), and the vascular model (6) is completely immersed in the liquid medium; multiple pressure sensors (5) connected to an external data processing unit are arranged in an orderly manner on the inner wall of the closed housing (1). During training, the guidewire is inserted into the vascular model (6) to perform interventional operations. When the guidewire collides with the wall of the vascular model (6), the medium liquid will be touched by the wall of the vascular model (6), triggering a change in the pressure field of the medium liquid inside the closed shell (1). The pressure field data is monitored by the corresponding pressure sensor (5) and fed back to the data processing unit. The data processing unit obtains the pressure value and calculates and outputs the corresponding collision force.

2. The method for detecting collision force of a vascular model for interventional surgery training as described in claim 1, characterized in that: The blood vessel model (6) is made of elastic material. When the wall of the blood vessel model (6) is subjected to force, it will undergo elastic deformation and touch the medium liquid. The material of the blood vessel model (6) includes, but is not limited to, elastic silicone, hydrogel, and polyurethane. The medium liquid is made of incompressible material and is filled in the closed shell (1) without air bubbles. The material of the medium liquid includes, but is not limited to, silicone oil, glycerin, and water. The closed shell (1) is made of high-transparency acrylic material.

3. The method for detecting collision force of a vascular model for interventional surgery training as described in claim 1, characterized in that: The enclosed shell (1) is a sealable cuboid structure equipped with a top cover (4). Pressure sensors (5) are evenly distributed on each inner wall surface of the enclosed shell (1). The medium liquid fills the inner wall space of the enclosed shell (1), and the pressure changes of the medium liquid inside the enclosed shell (1) are monitored in real time by the pressure sensors (5) on each inner wall surface.

4. The method for detecting collision force of a vascular model for interventional surgery training as described in claim 1, characterized in that: The closed shell (1) has an inlet (2) and an outlet (3) on its side that are connected to the end of the vascular model (6). The vascular model (6) is circulated with an external flow pump through the inlet (2) and the outlet (3) to simulate the flow of blood in the blood vessels. During training, the guide wire is inserted into the vascular model (6) through the inlet (2) and / or the outlet (3) to train the interventional surgery.

5. The method for detecting collision force of a vascular model for interventional surgery training as described in claim 1, characterized in that: When the guidewire is in normal operation and the intervention is not colliding with the wall of the vascular model (6), the fluid in the vascular model (6) simulates the flow of blood. The vascular model (6) acts on the medium liquid to form a basic pressure field, which is continuously monitored by each pressure sensor (5) to obtain the dynamic fluctuation baseline of the basic pressure field. When a collision occurs, the pressure field of the medium liquid changes, and the monitoring value of the corresponding pressure sensor (5) will change instantaneously. The data processing unit takes the difference of the instantaneous change as the pressure value.

6. The method for detecting collision force of a vascular model for interventional surgery training as described in claim 1, characterized in that: The data processing unit includes a database that maps pressure values ​​to impact forces, and the impact force is calculated from the pressure values ​​based on the database.

7. The method for detecting collision force of a vascular model for interventional surgery training as described in claim 6, characterized in that: The database is configured with n pressure sensors (5). During training, the pressure values ​​monitored by each pressure sensor (5) constitute n. column vectors , ； in, The pressure value obtained by the nth pressure sensor (5) is represented. The collision force is defined as a three-dimensional force vector. , constitute 3 Column vectors ； in, , , These represent the three components of the force in a three-dimensional Cartesian coordinate system; Database consists of column vectors and column vectors Constructing a mapping function , 。 8. The method for detecting collision force of a vascular model for interventional surgery training as described in claim 6, characterized in that: The database is generated as follows: In the simulation model, multiple collision sections are set at intervals along the central axis of the blood vessel model (6), and multiple collision points are set at intervals along the circumference on each collision section. Forces are applied at each collision point along the normal and tangential directions, with the force ranging from 0.1N to 2.0N in steps of 0.1N to form discrete forces. The pressure values ​​corresponding to each collision point on each collision section under each discrete force are obtained, thereby obtaining the relationship between force and pressure value, and generating a database corresponding to the mapping between collision force and pressure value.

9. The method for detecting collision force of a vascular model for interventional surgery training as described in claim 8, characterized in that: Before the database is generated, the simulation model is physically verified and calibrated. The physical verification and calibration method is as follows: apply a known force to the preset key position of the real prototype and the pressure sensor (5) feeds back the measured pressure value; construct the simulation model and apply force under the same conditions to obtain the pressure simulation value; compare the pressure simulation value with the pressure measurement value, and repeatedly verify and iterate by adjusting the parameters in the simulation model until the difference between the pressure simulation value and the pressure measurement value is within the threshold.

10. The method for detecting collision force of a vascular model for interventional surgery training as described in claim 6, characterized in that: The data processing unit includes a collision force calculation model, which is a neural network model trained based on a database. Set the number of pressure sensors (5) to n, construct an input layer with n neurons, and receive the pressure values ​​monitored by the n pressure sensors (5). Set 1-3 hidden layers, each with 16-128 neurons; use ReLU as the activation function; set an output layer with 3 neurons to output collision force. The three components; input the dataset from the database into the neural network model, and train the model by adjusting the internal weights and biases through the backpropagation algorithm and gradient descent optimizer; After training, the neural network model is deployed to the data processing unit, and the predicted collision force is obtained by inputting the real-time collected pressure data.

Images