AI-based magnetic pressure coordinated multi-mode broken needle taking-out system and AI-based magnetic pressure coordinated multi-mode broken needle taking-out method

Abstract

The invention relates to the technical field of artificial intelligence and wisdom medical treatment, and discloses an AI-based magnetic pressure cooperative multi-mode broken needle extraction system and method, and the system comprises an AI auxiliary diagnosis module which generates an obstacle avoidance strategy through three-dimensional image reconstruction and path planning; the intelligent needle taking tool module adopts a broken needle taking device and is combined with a deformation self-adaptive suction head to realize minimally invasive operation; the real-time monitoring module integrates multi-source sensing data and dynamically adjusts the track and adsorption parameters of the mechanical arm; and the data learning module optimizes an operation strategy through a case library, and the method comprises image intelligent planning, magnetic pressure collaborative adsorption, multi-modal monitoring and data closed-loop optimization. By means of multi-source data fusion, the synergistic effect of magnetic attraction and negative pressure, dynamic adjustment of paths and force feedback, accurate grabbing, safe removal and minimally invasive operation of broken needles are achieved, a multi-dimensional sensor monitoring and real-time feedback mechanism is adopted, the surgical risk is effectively reduced, and the operation precision and safety are improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence smart medical technology, and specifically to an AI-based magnetic-pressure collaborative multimodal broken needle removal system and method. Background Art

[0002] During dental root canal treatment and the removal of accidentally broken needles, needle removal remains a significant technical challenge in clinical practice. With advances in microendodontic technology, dentists are now utilizing various auxiliary methods to remove broken needles during treatment. However, due to shortcomings in existing technologies regarding needle location, path planning, and tool adaptation, the reliability and safety of the procedure face numerous challenges.

[0003] Currently, common methods of removing broken needles rely on direct operations under a microscope, usually using clamps, ultrasonic equipment or drilling tools. These methods not only have greater operational risks during the operation, but also easily cause secondary damage to the tooth. In addition, the existing path planning method is still based on human experience. During the operation, the doctor needs to rely on personal experience to select the path, resulting in a high degree of subjectivity in the path planning. Especially when dealing with complex root canal structures, there is a conflict between the rigidity of path planning and the flexibility of operation, which can easily cause unnecessary damage to surrounding tissues.

[0004] Traditional broken needle extraction tools are mostly rigid mechanical devices. These tools struggle to adapt to the tortuous structures within the root canal and often cannot accurately penetrate deep enough to successfully remove broken needles. Especially when faced with complex root canal curves, existing tools can easily cause collisions or scratches on the tooth wall, even leading to complications such as microcracks in the root canal. Furthermore, existing tools lack real-time feedback during operation, preventing the doctor from timely sensing the force and status of the operation. This can easily lead to excessive or uneven force, increasing the risk of complications.

[0005] Based on the above problems, the present invention proposes an AI-based magnetic-pressure collaborative multi-modal broken needle removal system and method. Summary of the Invention

[0006] In response to the shortcomings of the existing technology, the present invention provides an AI-based magnetic-pressure collaborative multi-modal broken needle removal system and method, which solves the problems of broken needle positioning, path planning, tool adaptation and insufficient operation feedback in the existing technology.

[0007] To achieve the above objectives, the present invention is implemented through the following technical solutions: an AI-based magnetic pressure collaborative multi-modal broken needle removal system, comprising:

[0008] The AI-assisted diagnosis module is used to collect and analyze three-dimensional images of the patient's oral alveolar bone and tooth root canals, locate the position of the diagnostic needle, and plan the needle removal path;

[0009] The intelligent needle removal tool module is used to accurately grasp and remove diagnostic needles. It uses a broken needle removal device combined with a deformable adaptive suction tip to achieve minimally invasive surgery.

[0010] Real-time monitoring and feedback module, through real-time image monitoring and patient physiological data feedback, dynamically adjusts the operation path to ensure surgical safety;

[0011] The data learning and case library module is used to store, analyze and optimize case data of diagnostic needle removal and provide personalized recommendations to doctors.

[0012] Preferably, the AI-assisted diagnosis module includes:

[0013] An image acquisition unit, used to obtain three-dimensional image data of the patient's oral cavity;

[0014] An image processing unit, used to eliminate image artifacts and reconstruct a three-dimensional model of the root canal;

[0015] Broken needle positioning unit, used to accurately calibrate the three-dimensional coordinates and spatial posture of the broken needle;

[0016] A path planning unit, used to generate a needle removal path that avoids critical tissues based on image analysis;

[0017] Risk assessment unit for quantifying the risk of root canal curvature and pulp access.

[0018] Preferably, the intelligent needle removal tool module includes:

[0019] A robotic arm control unit for multi-axis coordinated motion control with sub-millimeter precision;

[0020] Force feedback adjustment unit, used to monitor and dynamically limit the operating contact force in real time;

[0021] A broken needle removal device comprises a mounting frame (1), a mechanical arm and an operating panel (18), and is integrated with minimally invasive tools to perform the operation of separating and removing the broken needle.

[0022] Preferably, one end of the mounting frame (1) is connected to the operation panel (18), and the other end thereof is connected to one end of the robotic arm; a negative pressure pump group (2) is installed on the upper surface of the mounting frame (1); the output end of the negative pressure pump group (2) is connected to a hose 1 (10); the other end of the robotic arm is installed with a mounting seat (7); a connector (8) is installed on the outer wall of the mounting seat (7); one end of a rod tube (12) is installed at the bottom of the connector (8); a hose 2 (11) is provided inside the rod tube (12); and the hose 2 (11) is externally penetrated. A connector (8) is provided to be connected to an external water delivery device, a magnetic field generator (9) is provided in the connector (8), the outside of the hose (10) is provided in a rod tube (12) through the connector (8), a suction head (13) is threadedly connected to the other end of the rod tube (12), a magnetic base (14) is installed in the rod tube (12) through a plurality of mounting rods (15), a sensor group (16) is provided on the top of the magnetic base (14), and a filter plate (17) is installed on the inner wall of the rod tube (12) at a position above the sensor group (16).

[0023] Preferably, the robotic arm includes a movable arm 1, one end of the movable arm 1 is connected to the other end of the mounting frame through a driving member, the other end of the movable arm 1 is connected to the movable arm 4 through a driving member, the other end of which is connected to one end of the movable arm 2 through a driving member, the other end of the movable arm 2 is connected to the adjustment frame through a driving member, the inside of the adjustment frame is connected to one end of the movable arm 3 through a driving member, and the other end is connected to the mounting base through a driving member.

[0024] Preferably, the real-time monitoring and feedback module includes:

[0025] Physiological monitoring unit, used to track patients' heart rate and blood pressure vital signs in real time and warn of abnormal fluctuations;

[0026] An image monitoring unit, which provides real-time visual feedback of operations through microscopic imaging and ultrasonic testing;

[0027] A dynamic adjustment unit for automatically optimizing the robot's motion trajectory and adsorption parameters based on multimodal sensing data;

[0028] The safety protection unit is used to trigger the emergency stop mechanism to prevent the risks of overload, temperature rise and pressure exceeding the limit.

[0029] Preferably, the data learning and case library module includes:

[0030] Case storage unit, used for structured archiving of surgical images, operating parameters and postoperative efficacy data;

[0031] A machine learning unit, used to continuously optimize broken needle location and removal strategies through reinforcement learning;

[0032] Personalized recommendation unit, used to generate customized surgical plans based on the patient's anatomical characteristics and historical cases.

[0033] Preferably, the sensor group integrates a magnetic field sensor, a pressure sensor, a temperature sensor, a displacement sensor, and an image monitoring sensor.

[0034] Preferably, a bioimpedance sensor is embedded in the outer wall of the suction tip and is made of medical-grade nickel-titanium alloy.

[0035] The present invention also provides an artificial intelligence-based broken needle removal method, which is applied to the above-mentioned AI-based magnetic pressure coordinated multi-modal broken needle removal system, comprising the following steps:

[0036] S1. Preoperative 3D image acquisition and intelligent planning: 3D image data of the patient's oral cavity and root canals is acquired through a CBCT scanner. The AI-assisted diagnosis module removes artifacts and performs 3D reconstruction on the images. It calibrates the 3D coordinates of the broken needle and generates at least three needle extraction paths that avoid the pulp and curved root canal structures. The path with the lowest risk is selected based on finite element stress analysis.

[0037] S2. Initialize the magnetic-pressure synergistic adsorption system: Install the adapter tip to the end of the rod tube, calibrate the target field strength of the magnetic field generator to 0.5T±0.05T, start the negative pressure pump group to pre-test the negative pressure stability, and set the robot arm motion parameters through the operation panel;

[0038] S3. Dynamic positioning of the robotic arm and activation of adsorption: The robotic arm control unit drives the coordinated movement of movable arms 1, 2, 3, and the adjustment frame, so that the suction head approaches the broken needle along the planned path. The contact force is monitored in real time by the force feedback adjustment unit. After the suction head lightly touches the surface of the broken needle, the magnetic field generator is activated to apply a gradient magnetic field to magnetize the broken needle, and the negative pressure pump group is simultaneously activated to generate a negative pressure of -35kPa to -50kPa for adsorption;

[0039] S4. Multimodal real-time monitoring and dynamic adjustment: The bioimpedance sensor monitors the adhesion state of the broken needle to the dentin. Magnetic field sensors, temperature sensors, and displacement sensors provide real-time feedback on the magnetic field intensity, magnetic base temperature rise, and broken needle displacement. The dynamic adjustment module optimizes the adsorption parameters based on multi-source data.

[0040] S5. Safety protection and emergency treatment: When the patient's heart rate is greater than 100 bpm or the systolic blood pressure is greater than 140 mmHg, the physiological monitoring unit triggers an audible and visual alarm and automatically slows down the robotic arm to 0.1 mm / s. At the same time, the adsorption operation is suspended until vital signs return to normal. If microscopic imaging and ultrasound detect microcracks in the root canal or residual broken needle fragments, the dynamic adjustment unit immediately generates an obstacle avoidance path, and the robotic arm performs avoidance at a speed of 0.2 mm / s. Based on the magnetic field sensor, pressure sensor and displacement feedback data, the negative pressure is adjusted from -35 kPa to -50 kPa and the magnetic field gradient is 5-12 T / m every 100 ms to stabilize the displacement speed of the broken needle at 0.3 mm / s±0.05 mm. When the temperature rise of the magnetic base is greater than 1.5°C, the liquid cooling system automatically increases the pressure. When the dentin stress is greater than 8 MPa, the magnetic field and negative pressure are immediately cut off, triggering an emergency stop. If the broken needle stagnates for more than 10 seconds, the negative pressure automatically switches to pulse mode.

[0041] S6. Broken needle recovery and postoperative verification: After the broken needle is adsorbed, the robotic arm retreats at a speed of 0.2-0.3 mm / s, and the debris is filtered through the filter plate;

[0042] S7. Data upload and model optimization: Upload surgical images, operation parameters and postoperative efficacy data to the case storage unit. The machine learning unit optimizes the broken needle positioning and adsorption strategy through the reinforcement learning algorithm. The personalized recommendation unit generates a postoperative care plan based on the patient's anatomical characteristics.

[0043] The present invention provides an AI-based magnetic-pressure coordinated multi-modal broken needle removal system and method. It has the following beneficial effects:

[0044] 1. The present invention achieves stable grasping and safe removal of broken needles through the synergistic effect of magnetic attraction and negative pressure technology. Compared with the defect of traditional mechanical clamping that is prone to secondary damage, the dual forces of gradient magnetic field directional magnetization and dynamic negative pressure adsorption are combined to solve the problem of clamping slippage caused by the complex surface morphology of broken needles. At the same time, the deformation of the suction head adapts to the curved structure of the root canal, significantly improving the reliability of operation.

[0045] 2. The present invention adopts AI-driven multi-constraint path planning and dynamic adjustment mechanism to effectively avoid key anatomical structures. By integrating three-dimensional image reconstruction and real-time mechanical feedback, it generates a smooth obstacle avoidance path and dynamically corrects the robot arm motion parameters. It overcomes the risk of tissue collision caused by traditional manual operation relying on experience and rigid path planning, ensuring the safety of minimally invasive operations.

[0046] 3. The multi-source data closed-loop feedback system constructed by the present invention realizes real-time early warning and fuse-breaking of intraoperative risks, integrates physiological monitoring, microscopic imaging and multi-dimensional sensor data, and triggers a hierarchical protection strategy through millisecond-level fusion analysis, solving the pain points of delayed monitoring and slow response in traditional surgery, and significantly reducing thermal damage or mechanical complications caused by overload operations. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 is a perspective view of the device of the present invention;

[0048] Figure 2 This is a schematic diagram of the interior of the connector of the present invention;

[0049] Figure 3 for Figure 2 Enlarged view of point A in the middle;

[0050] Figure 4 It is a structural schematic diagram of the magnetic base of the present invention;

[0051] Figure 5 It is a framework diagram of the system of the present invention;

[0052] Figure 6 Schematic diagram of the AI-assisted diagnosis module of the present invention;

[0053] Figure 7 Schematic diagram of the intelligent needle removal tool module of the present invention;

[0054] Figure 8 is a schematic diagram of the real-time monitoring and feedback module of the present invention;

[0055] Figure 9 This is a framework diagram of the data learning and case library module of the present invention;

[0056] Figure 10 Flowchart of the method of the present invention.

[0057] Among them, 1. Mounting frame; 2. Negative pressure pump group; 3. Movable arm 1; 4. Movable arm 2; 5. Adjustment frame; 6. Movable arm 3; 7. Mounting base; 8. Connector; 9. Magnetic field generator; 10. Hose 1; 11. Hose 2; 12. Rod tube; 13. Suction head; 14. Magnetic base; 15. Mounting rod; 16. Sensor group; 17. Filter plate; 18. Operation panel; 19. Movable arm 4. DETAILED DESCRIPTION

[0058] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the present specification. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0059] Please see the attached Figure 1 and attached Figure 5 The embodiment of the present invention provides an AI-based magnetic-pressure collaborative multi-modal broken needle removal system, comprising:

[0060] Please see the attached Figure 6 , AI-assisted diagnosis module, used to collect and analyze three-dimensional images of the patient's oral alveolar bone and tooth root canals, locate the position of the diagnostic needle and plan the needle removal path;

[0061] AI-assisted diagnosis modules include:

[0062] An image acquisition unit, used to obtain three-dimensional image data of the patient's oral cavity;

[0063] An image processing unit, used to eliminate image artifacts and reconstruct a three-dimensional model of the root canal;

[0064] Broken needle positioning unit, used to accurately calibrate the three-dimensional coordinates and spatial posture of the broken needle;

[0065] A path planning unit, used to generate a needle removal path that avoids critical tissues based on image analysis;

[0066] Risk assessment unit for quantifying the risk of root canal curvature and pulp access.

[0067] Specifically, in this embodiment, the image acquisition unit acquires three-dimensional image data of the patient's oral region using a cone-beam CT device. The CT device's scanning parameters are optimized, with the tube voltage set to 85 kV, the tube current to 8 mA, and the exposure time controlled to within 9 seconds to ensure that the radiation dose complies with the ALARA (As Low As Reasonably Achievable) principle.

[0068] In some embodiments, for the fine structure of the root canal system, the scanning layer thickness is set to 0.1 mm, the isotropic voxel size is 0.2 mm × 0.2 mm × 0.2 mm, and the spatial resolution reaches 200 μm. The image reconstruction algorithm uses the Feldkamp-Davis-Kress (FDK) back-projection algorithm, whose mathematical expression is:

[0069]

[0070] Where D is the source-detector distance (typical value 600 mm), P(β,y′,z′) is the projection data, β is the rotation angle, and I recon (x, y, z) is the grayscale value (CT value) of the reconstructed 3D image at position (x, y, z) in the Cartesian coordinate system, y′ and z′ are the local coordinates on the detector plane, corresponding to the projection position of the current voxel (x, y, z) under angle β. is a geometric correction factor used to compensate for the ray divergence effect of cone-beam projection.

[0071] Specifically, the raw data is output in DICOM format and DICOMTag (0018,9345) is attached to record the scanning parameters for easy access by subsequent processing units.

[0072] In this embodiment, the image processing unit uses an improved 3DU-Net network to remove metal artifacts. The network input is a 512×512×512 voxel block, and the output is a three-dimensional root canal model after artifact removal.

[0073] In some embodiments, the network structure includes 16 layers of convolutional modules, each with a convolution kernel size of 3×3×3, a stride of 1, and an activation function of LeakyReLU (α=0.01). Skip connections use channel concatenation to preserve low-frequency anatomical structure information.

[0074] Specifically, for the high-density artifacts in the broken needle area, a hybrid strategy of frequency domain filtering and spatial domain morphological processing is introduced. The frequency domain filter transfer function is designed as:

[0075] in, is the radial spatial frequency, f Nyquist is the Nyquist frequency, which is determined by the image resolution (Δx=0.2mm), f Nyquist =1 / (2Δx)=2.5cycles / mm.

[0076] The morphological opening operation uses a 3×3×3 spherical structure element to eliminate residual noise points while maintaining the continuity of the root canal wall.

[0077] In this embodiment, the broken needle positioning unit achieves sub-millimeter positioning accuracy through a multi-scale feature fusion strategy. The specific process includes:

[0078] CT value threshold segmentation: extract high-density areas with CT values ​​greater than 3000HU and generate a binary mask;

[0079] Morphological refinement: skeleton extraction is performed on the mask to obtain the center line of the broken needle;

[0080] Principal component analysis (PCA): Calculate the eigenvectors of the centerline voxel covariance matrix to determine the principal axis direction of the broken needle.

[0081] Mathematically, the eigendecomposition of the covariance matrix C is expressed as:

[0082]

[0083] Among them, μ is the voxel coordinate mean, V is the eigenvector matrix, and the eigenvector corresponding to the maximum eigenvalue is the direction of the broken needle axis v axis , xi is the three-dimensional coordinate (x, y, z) of the broken needle voxel, in mm, Λ is the eigenvalue diagonal matrix, reflecting the extension degree of the broken needle along each main axis direction, N is the total number of samples in the data set, (x i -μ)(x i -μ) T is the deviation outer product matrix of a single sample, which represents the covariation relationship of the sample in each dimension.

[0084] In some embodiments, the positioning error compensation model is:

[0085] ΔP=k1·||v axis -v gt ||+k2·σ CT ;

[0086] Among them, k1=0.1, k2=0.05 are empirical coefficients, σ CT is the standard deviation of local CT values.

[0087] In this embodiment, the path planning unit adopts the A* algorithm under multiple constraints, the search space is discretized into 0.1mm grids, and the cost function integrates path length, safety and operational feasibility.

[0088] Specifically, the path node expansion rules are:

[0089] Allowed movement directions: 26 neighborhoods (including diagonal movement)

[0090] Curvature constraint: The angle of change in the direction of adjacent nodes is ≤ 15°

[0091] Collision detection: Euclidean distance from the grid to the root canal wall ≥ 0.3mm

[0092] In some embodiments, a dynamic weight adjustment mechanism is introduced. When approaching a high-risk area (less than 0.5 mm from the pulp), the safety weight w2 is increased from 0.3 to 0.7. The mathematical expression is:

[0093]

[0094] Where d is the minimum distance from the current path point to the dental pulp.

[0095] The path smoothing process uses B-spline curve fitting with a control point spacing of 0.5 mm to ensure the continuity of the robot arm's motion trajectory.

[0096] In this embodiment, the risk assessment unit quantifies the operational risk through multi-physics field coupling simulation, including:

[0097] Biomechanical risk: Dentin stress distribution was calculated by finite element analysis, and the material model adopted the linear elastic assumption (elastic modulus 18 GPa, Poisson's ratio 0.3);

[0098] Thermal damage risk: To simulate the eddy current thermal effect of the magnetic field generator, the heat conduction equation is expressed as:

[0099]

[0100] Among them, Q eddy =σ|E| 2 is the eddy current heat source term, σ=1.5×10 6 S / m is tissue conductivity, E is the induced electric field strength, and ρ is the dentin density, which is 2.1 g / cm 3 , c p is the thermal conductivity, which takes a value of 0.8W / (m·K), T is the temperature field function, which represents the temperature of a point in space at time t, and k is the thermal conductivity of the material, which represents the heat conduction ability;

[0101] Nerve injury risk: Nerve excitability was predicted based on the impedance phase angle change rate Δφ / Δt, with a threshold set at 10° / s.

[0102] In some embodiments, the risk report generation module integrates a visualization tool to overlay the risk heat map onto the three-dimensional model, and the color mapping rule is:

[0103] Red area (risk value > 0.8): Operation prohibited

[0104] Yellow area (0.4≤Risk value≤0.8): Manual confirmation is required

[0105] Green area (risk value < 0.4): safe operation area

[0106] Please see the attached Figure 7 , intelligent needle removal tool module, used to accurately grasp and remove diagnostic needles;

[0107] The intelligent needle removal tool module includes:

[0108] A robotic arm control unit for multi-axis coordinated motion control with sub-millimeter precision;

[0109] Force feedback adjustment unit, used to monitor and dynamically limit the operating contact force in real time;

[0110] Broken needle removal device, integrated with minimally invasive tools to perform broken needle separation and removal operations.

[0111] Specifically, in this embodiment, each joint of the robotic arm is equipped with a high-precision harmonic reducer (transmission error ≤ 0.01°) and an absolute encoder (resolution 19 bit) to achieve an end-end repeat positioning accuracy of ±0.05mm.

[0112] Specifically, the kinematic model was established using an improved DH parameter method, and the inverse kinematics solution was implemented using a numerical iteration method, with a convergence condition of end position deviation less than 0.05 mm. Path tracking control employed a PID algorithm, with the proportional coefficient, integral time, and derivative time optimized through calibration using 500 clinical data sets, and a response bandwidth ≥100 Hz.

[0113] In some embodiments, the robot arm motion trajectory data is provided by the path planning unit of the AI-assisted diagnosis module, and the path curvature constraint is ≤ 2.0 mm. -1 The end linear velocity is limited to 0.5 mm / s and the angular velocity is limited to 1.2 rad / s to avoid root canal wall damage caused by excessive movement.

[0114] In this embodiment, the force feedback adjustment unit monitors the contact force between the suction tip 13 and the broken needle in real time through the pressure sensor integrated in the sensor group 16. The pressure sensor has a range of 0-2N and an accuracy of ±0.01N. The dynamic force limitation strategy is based on the coordinated control of the defined pulp approach distance (d) and the phase angle (φ) fed back by the bioimpedance sensor:

[0115] When d≥0.5mm and φ≤45°: the maximum allowable contact force is 0.3N;

[0116] When 0.3mm≤d<0.5mm or 45°<φ≤60°: the contact force is limited to 0.1N-0.3N according to the linear relationship;

[0117] When d<0.3mm or φ>60°: the adsorption operation is forcibly cut off and an emergency stop is triggered.

[0118] Specifically, the pressure sensor data is processed in real time by the dynamic adjustment unit, and multi-parameter fusion is performed by combining the temperature sensor (monitoring the temperature rise ΔT of the magnetic base) and the displacement sensor (detecting the axial displacement Δx of the suction head) to dynamically correct the force threshold:

[0119]

[0120] Wherein, ΔT is in °C, and Δx is in mm. When the limit is exceeded, the safety protection unit directly cuts off the power supply of the negative pressure pump group (2) and the magnetic field generator (9) without relying on external damping devices.

[0121] In some embodiments, the pressure data is synchronized to the risk assessment unit and cross-validated with the dentin stress finite element model (maximum allowable stress 8 MPa). If the model predicts a stress value σ FEM and the measured pressure F meas If the deviation is greater than 20%, the manual review mechanism will be triggered.

[0122] Please see the attached Figure 1 - Attachment Figure 4The broken needle removal device includes a mounting frame 1, a robotic arm and an operation panel 18. One end of the mounting frame 1 is connected to the operation panel 18, and the other end thereof is connected to one end of the robotic arm. A negative pressure pump group 2 is installed on the upper surface of the mounting frame 1, and the output end of the negative pressure pump group 2 is connected to a hose 10. The other end of the robotic arm is installed with a mounting seat 7, a connector 8 is installed on the outer wall of the mounting seat 7, and one end of a rod tube 12 is installed at the bottom of the connector 8. A hose 2 11 is arranged inside the rod tube 12, and a connector 8 is passed through the outside of the hose 2 11 and is connected to an external water supply device. A magnetic field generator 9 is arranged in the connector 8. The outside of the hose 10 is arranged in the rod tube 12 through the connector 8. The other end of the rod tube 12 is threadedly connected to a suction head 13. A magnetic base 14 is installed in the rod tube 12 through a plurality of mounting rods 15. A sensor group 16 is arranged on the top of the magnetic base 14, and a filter plate 17 is installed on the inner wall of the rod tube 12 above the sensor group 16.

[0123] The robotic arm includes a movable arm 3, one end of which is connected to the other end of the mounting frame 1 through a driving member, the other end of which is connected to the movable arm 4 19 through a driving member, the other end of which is connected to one end of the movable arm 2 4 through a driving member, the other end of which is connected to the adjusting frame 5 through a driving member, and the inside of the adjusting frame 5 is connected to one end of the movable arm 3 6 through a driving member, and the other end of which is connected to the mounting base 7 through a driving member.

[0124] The sensor group 16 integrates a magnetic field sensor, a pressure sensor, a temperature sensor, a displacement sensor, and an image monitoring sensor.

[0125] The outer wall of the suction tip 13 is embedded with a bioimpedance sensor and is made of medical-grade nickel-titanium alloy.

[0126] Specifically, after determining the needle removal path, the adapter suction head 13 is installed to the end of the rod tube 12, the target field strength of the magnetic field generator 9 is calibrated to 0.5T±0.05T, the negative pressure pump group 2 is started to pre-test the negative pressure stability, and the robot arm movement parameters are set through the operation panel 18 to enable the robot arm to operate, thereby controlling the suction head 13 to enter the tooth root canal, so that the suction head 13 approaches the broken needle along the planned path, and the contact force is monitored in real time through the force feedback adjustment unit. After the suction head 13 lightly touches the surface of the broken needle, the magnetic field generator 9 is started to apply a gradient magnetic field to magnetize the broken needle, and the negative pressure pump group 2 is simultaneously activated to generate -35kPa to -50kPa Negative pressure adsorption monitors the adhesion state of the broken needle and dentin through the bioimpedance sensor, and uses the magnetic field sensor, temperature sensor and displacement sensor to provide real-time feedback of the magnetic field strength, temperature rise of the magnetic base 14 and displacement of the broken needle. The dynamic adjustment module optimizes the adsorption parameters based on multi-source data. When the broken needle is sucked into the suction head 13, the robotic arm runs to control the suction head 13 to withdraw. During the needle removal process, the flushing water enters the rod tube 12 through the hose 11 and flushes the root canal in real time through the suction head 13, thereby avoiding drying out the root canal in the tooth. Under the influence of suction, the debris is filtered through the filter plate 17 to avoid entering the space between the filter plate 17 and the rod tube 12.

[0127] The suction head 13 is made of medical grade nickel-titanium alloy and can bend and deform during the needle removal process to adapt to the environment inside the root canal and facilitate the removal of broken needles.

[0128] Please see the attached Figure 8 ,Real-time monitoring and feedback module, through real-time image monitoring and patient physiological data feedback, dynamically adjusts the operation path to ensure surgical safety;

[0129] Real-time monitoring and feedback modules include:

[0130] Physiological monitoring unit, used to track patients' heart rate and blood pressure vital signs in real time and warn of abnormal fluctuations;

[0131] An image monitoring unit, which provides real-time visual feedback of operations through microscopic imaging and ultrasonic testing;

[0132] A dynamic adjustment unit for automatically optimizing the robot's motion trajectory and adsorption parameters based on multimodal sensing data;

[0133] The safety protection unit is used to trigger the emergency stop mechanism to prevent the risks of overload, temperature rise and pressure exceeding the limit.

[0134] Specifically, in this embodiment, the physiological monitoring unit collects the patient's electrocardiogram (sampling rate 1kHz) and photoplethysmography (PPG) through a wireless body surface patch, and extracts heart rate (HR) and systolic blood pressure (SBP) data based on an adaptive filtering algorithm. The calculation formula is:

[0135]

[0136] Among them, K1=12.5 and K2=90 are clinical calibration coefficients, A PPG is the pulse wave amplitude, A 基线 is the resting state baseline value, N R波 is the number of R waves of the ECG signal detected during the sampling period, t 采样周期 is the length of the time window for data collection.

[0137] Specifically, according to ADA and JADA clinical guidelines, when a heart rate >100 bpm (a significant increase in intraoperative stress response threshold) or systolic blood pressure >140 mmHg (a critical value for hypertension risk) is detected, a three-level audible and visual alarm (green, yellow, and red) is triggered, and the robotic arm's movement speed is reduced to a safe mode (0.1 mm / s). If the limit is exceeded for >30 seconds (covering the time boundary between instantaneous interference and real risk), an emergency stop request signal is sent to the safety protection unit.

[0138] In some embodiments, data anomaly determination introduces sliding window variance analysis (window length 5 seconds, step length 1 second) to eliminate the influence of short-term interference signals.

[0139] In this embodiment, the image monitoring unit integrates a fiber optic microendoscope (0.8mm diameter, 8K resolution) and a 50MHz high-frequency ultrasonic probe, and realizes multimodal data fusion through feature matching algorithm. The depth of field compensation model of microscopic imaging is:

[0140]

[0141] Among them, z 原始 is the original depth data, Δz = 0.05mm is the depth of field correction, I 灰度 is the current pixel grayscale value (0-255).

[0142] Specifically, when ultrasound detects root canal microcracks (length > 50 μm) or broken needle fragments, the dynamic adjustment unit immediately generates an obstacle avoidance path. The path offset is calculated according to the following rules:

[0143]

[0144] At the same time, the speed of the robot arm end is limited to 0.2 mm / s to avoid secondary damage. The fused image data is transmitted to the operation panel 18 via Gigabit Ethernet, where the safety boundary line (red) and the real-time path line (green) are superimposed and displayed.

[0145] In this embodiment, the dynamic adjustment unit receives the multimodal data stream (magnetic field strength B, contact force F, temperature T, displacement x, bioimpedance phase angle φ) from the sensor group 16 and performs the following optimization in a 100 ms cycle:

[0146] Compensation for path deviation based on PID control algorithm:

[0147]

[0148] Where, e(t)=x 目标 -x 实际 , K p =2.5, K i =0.8, K d =0.2, the output u(t) is converted into joint angle increment, K p is the proportional gain coefficient, which determines the response strength of the system to the current error. e(t) is the error signal, which indicates the instantaneous deviation between the set value and the actual value. K i is the integral gain coefficient, used to eliminate steady-state error, is the time integral of the error, reflecting the cumulative amount of historical errors, K d is the differential gain coefficient, which suppresses system oscillation and predicts future error trends. is the error change rate, which reflects the dynamic characteristics of the error.

[0149] Adjust the negative pressure and magnetic field gradient according to the impedance phase angle φ:

[0150]

[0151] Specifically, when φ>60°, it was determined that the broken needle was severely adhered to the dentin, and the system was directly switched to the pulsed negative pressure mode.

[0152] In this embodiment, the safety protection unit achieves millisecond-level emergency stop response through a hard-wired connection. The triggering logic includes:

[0153] Overload temperature rise protection: When the temperature sensor of the magnetic base 14 detects a temperature rise ΔT>1.5°C (reference temperature 37°C), the liquid cooling system is immediately activated (flow rate increases from 10mL / min to 15mL / min) and the magnetic field strength is reduced by 20%. If the temperature rise is not suppressed within 30 seconds, the power supply of the magnetic field generator 9 is cut off;

[0154] Stress overlimit protection: The stress distribution of dentin is calculated in real time by the finite element model. max When the pressure is higher than 8MPa, the three-level response is triggered:

[0155] Cut off the power supply of negative pressure pump group 2;

[0156] The robotic arm moves back 0.5mm;

[0157] Activate the bioimpedance sensor to recheck the adhesion status;

[0158] Displacement stagnation protection: If the needle breaks and the displacement stagnation lasts for more than 10 seconds, the system will automatically switch to the negative pressure pulse mode (-35kPa / +5kPa alternating, frequency 2Hz) and increase the magnetic field gradient to 15T / m.

[0159] Stagnation determination logic:

[0160] Axial displacement change threshold:

[0161] Determination formula:

[0162] (n=100 points, corresponding to 10 seconds of data);

[0163] Among them, Δz is the average change of the axial displacement of the broken needle, and the threshold value of 0.01mm is used to judge stagnation; z i is the axial coordinate value of the displacement sensor at the i-th sampling; n is the number of sampling points in the calculation window (10 seconds of data corresponds to n = 100).

[0164] Standard Deviation Limits:

[0165] Among them, σ z is the standard deviation of displacement.

[0166] Verification method: Use sliding window mean filtering (window length 1 second, step length 0.1 second) to eliminate instantaneous noise, and determine stagnation when 10 consecutive windows trigger the threshold.

[0167] Radial displacement compensation: If the X / Y axis displacement change is greater than 0.02mm (50% of the root canal wall safety distance), it is determined to be a lateral deviation of the broken needle, triggering path correction rather than stagnation.

[0168] Trigger condition: Axial displacement stagnation (Δz<0.01mm) lasts for 10 seconds and there is no radial offset alarm (Δx / Δy<0.02mm).

[0169] Exit conditions:

[0170] Successful release of stagnation: Δz ≥ 0.01 mm for 3 seconds (30 consecutive sampling points meet the standard);

[0171] Timeout protection: If the stagnation is not resolved within 60 seconds, the system switches to safe mode and triggers a manual intervention signal.

[0172] The data learning and case library module is used to store, analyze and optimize case data of diagnostic needle removal and provide personalized recommendations to doctors.

[0173] Please see the attached Figure 9 , data learning and case library modules include:

[0174] Case storage unit, used for structured archiving of surgical images, operating parameters and postoperative efficacy data;

[0175] A machine learning unit, used to continuously optimize broken needle location and removal strategies through reinforcement learning;

[0176] Personalized recommendation unit, used to generate customized surgical plans based on the patient's anatomical characteristics and historical cases.

[0177] Specifically, in this embodiment, the case storage unit adopts a distributed database architecture to structure and archive the following data types:

[0178] Preoperative data: patient's oral 3D images (DICOM format), root canal curvature (Schneider classification), broken needle material (CT value > 3000HU area volume);

[0179] Intraoperative data: robotic arm motion trajectory (0.1mm accuracy), adsorption parameters (negative pressure -35kPa to -50kPa, magnetic field gradient 5-12T / m), temperature rise record (ΔT≤1.5°C);

[0180] Postoperative data: duration of surgery, landmarks for complete removal of broken needles, and types of complications (microcracks / pulp damage, etc.).

[0181] Specifically, the data is stored in a timestamp-aligned manner with a time resolution of 10ms, supporting multimodal retrieval (such as "root canal curvature grade III + broken titanium alloy needle").

[0182] In some embodiments, data anonymization uses a k-anonymity model (k=5) to ensure patient privacy compliance. Encrypted storage uses the AES-256 algorithm and key separation management

[0183] In this embodiment, the machine learning unit deploys a reinforcement learning framework, and the agent action space includes:

[0184] Path planning parameters: robot end speed (0.1-0.5mm / s), path curvature upper limit (1.0-2.0mm -1 );

[0185] Adsorption parameters: negative pressure adjustment step (±5kPa), magnetic field gradient change (±2T / m).

[0186] The reward function is designed as:

[0187] R=w1·S 成功 -w2·T 手术 -w3·σ max -w4·ΔT;

[0188] Among them, S success is the mark of successful removal (1 / 0), T operation is the standardized operation time, σ max is the maximum stress in dentin (MPa), ΔT is the temperature rise of the magnetic base (°C), and the weight coefficients are w1=10, w2=0.5, w3=2, and w4=1.

[0189] Specifically, the policy network uses the Deep Deterministic Policy Gradient (DDPG) algorithm. The Actor network outputs continuous action values, and the Critic network evaluates Q values. The learning rate is set to 0.001 and the batch size is 256. The model is updated offline every 24 hours, and the incremental learning data comes from the latest 50 surgical cases.

[0190] In this embodiment, the personalized recommendation unit generates a customized solution through feature embedding and similarity matching. The process includes:

[0191] Patient feature extraction:

[0192] Anatomical characteristics: root canal curvature, dentin thickness, and broken needle depth;

[0193] Dynamic characteristics: real-time impedance phase angle (φ) and contact force fluctuation variance during surgery

[0194] The eigenvector is represented as:

[0195]

[0196] Case similarity calculation:

[0197] Use cosine similarity to match historical cases:

[0198]

[0199] Where v is the feature vector of the current patient, which contains multidimensional clinical parameters. v′ is the feature vector of a case in the historical case library, and its data structure is exactly the same as v. v·v′ is the dot product operation of the vectors, which calculates the sum of the products of each dimension of the two vectors. ||v|| is the Euclidean norm (modulus) of vector v.

[0200] The similarity threshold is set to 0.85, and after a successful match, the top three strategies with the highest reward value R are recommended first.

[0201] Parameter optimization:

[0202] Based on the adsorption parameters and path planning data of the matching case, the particle swarm algorithm (PSO) is further optimized, and the objective function is:

[0203] f 优化 =α·T 预测 +β·σ 预测 ;

[0204] Among them, α=0.6, β=0.4, the upper limit of iteration number is 100 times, and the population size is 50.

[0205] The artificial intelligence-based broken needle removal method described below and the AI-based magnetic-pressure collaborative multimodal broken needle removal system described above can be referenced to each other.

[0206] Please see the attached Figure 10 A method for removing broken needles by combining magnetic pressure and AI, which is applied to the above-mentioned system for removing broken needles by combining magnetic pressure and AI, comprises the following steps:

[0207] S1. Preoperative 3D image acquisition and intelligent planning: 3D image data of the patient's oral cavity and root canals is acquired through a CBCT scanner. The AI-assisted diagnosis module removes artifacts and performs 3D reconstruction on the images. It calibrates the 3D coordinates of the broken needle and generates at least three needle extraction paths that avoid the pulp and curved root canal structures. The path with the lowest risk is selected based on finite element stress analysis.

[0208] S2. Initialize the magnetic-pressure synergistic adsorption system: Install the adapter suction head 13 to the end of the rod tube 12, calibrate the target field strength of the magnetic field generator 9 to 0.5T±0.05T, start the negative pressure pump group 2 to pre-test the negative pressure stability, and set the robot arm movement parameters through the operation panel 18;

[0209] S3. Dynamic positioning of the robotic arm and activation of adsorption: The robotic arm control unit drives the movable arms 1 3, 2, 3, and 6, as well as the adjustment frame 5, to move in coordination, causing the suction head 13 to approach the broken needle along the planned path. The force feedback adjustment unit monitors the contact force in real time. After the suction head 13 lightly touches the surface of the broken needle, the magnetic field generator 9 is activated to apply a gradient magnetic field to magnetize the broken needle, and the negative pressure pump group 2 is simultaneously activated to generate a negative pressure of -35 kPa to -50 kPa for adsorption.

[0210] S4. Multimodal real-time monitoring and dynamic adjustment: The bioimpedance sensor monitors the adhesion state between the broken needle and the dentin. The magnetic field sensor, temperature sensor, and displacement sensor provide real-time feedback on the magnetic field intensity, temperature rise of the magnetic base 14, and displacement of the broken needle. The dynamic adjustment module optimizes the adsorption parameters based on multi-source data.

[0211] S5. Safety protection and emergency treatment: When the patient's heart rate is greater than 100bpm or the systolic blood pressure is greater than 140mmHg, the physiological monitoring unit triggers an audible and visual alarm and automatically slows down the robotic arm to 0.1mm / s. At the same time, the adsorption operation is suspended until vital signs return to normal. If microscopic imaging and ultrasound detect microcracks in the root canal or residual broken needle fragments, the dynamic adjustment unit immediately generates an obstacle avoidance path, and the robotic arm performs avoidance at a speed of 0.2mm / s. Based on the magnetic field sensor, pressure sensor and displacement feedback data, the negative pressure is adjusted from -35kPa to -50kPa and the magnetic field gradient is 5-12T / m every 100ms to stabilize the broken needle displacement speed at 0.3mm / s±0.05mm. When the temperature rise of the magnetic base 14 is greater than 1.5℃, the liquid cooling system automatically increases the pressure. When the dentin stress is greater than 8MPa, the magnetic field and negative pressure are immediately cut off, triggering an emergency stop. If the broken needle stagnates for more than 10 seconds, the negative pressure automatically switches to pulse mode.

[0212] S6, Broken needle recovery and postoperative verification: After the broken needle is adsorbed, the robotic arm retreats at a speed of 0.2-0.3 mm / s, and the debris is filtered through the filter plate 17;

[0213] S7. Data upload and model optimization: Upload surgical images, operation parameters and postoperative efficacy data to the case storage unit. The machine learning unit optimizes the broken needle positioning and adsorption strategy through the reinforcement learning algorithm. The personalized recommendation unit generates a postoperative care plan based on the patient's anatomical characteristics.

[0214] The method of this embodiment can be used to execute the above system embodiment. Its principles and technical effects are similar and will not be described in detail here.

[0215] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. An AI-based magnetic-pressure collaborative multi-modal broken needle removal system, characterized in that: include: The AI-assisted diagnosis module is used to collect and analyze three-dimensional images of the patient's oral alveolar bone and tooth root canals, locate the position of the diagnostic needle, and plan the needle removal path; The intelligent needle removal tool module is used to accurately grasp and remove diagnostic needles. It uses a broken needle removal device combined with a deformable adaptive suction tip to achieve minimally invasive surgery. Real-time monitoring and feedback module, through real-time image monitoring and patient physiological data feedback, dynamically adjusts the operation path to ensure surgical safety; The data learning and case library module is used to store, analyze and optimize case data of diagnostic needle removal and provide personalized recommendations to doctors.

2. The AI-based magnetic-pressure coordinated multi-modal broken needle removal system according to claim 1, characterized in that: The AI-assisted diagnosis module includes: An image acquisition unit, used to obtain three-dimensional image data of the patient's oral cavity; An image processing unit, used to eliminate image artifacts and reconstruct a three-dimensional model of the root canal; Broken needle positioning unit, used to accurately calibrate the three-dimensional coordinates and spatial posture of the broken needle; A path planning unit, used to generate a needle removal path that avoids critical tissues based on image analysis; Risk assessment unit for quantifying the risk of root canal curvature and pulp access.

3. The AI-based magnetic-pressure coordinated multi-modal broken needle removal system according to claim 1, characterized in that: The intelligent needle removal tool module includes: A robotic arm control unit for multi-axis coordinated motion control with sub-millimeter precision; Force feedback adjustment unit, used to monitor and dynamically limit the operating contact force in real time; A broken needle removal device comprises a mounting frame (1), a mechanical arm and an operating panel (18), and is integrated with minimally invasive tools to perform the operation of separating and removing the broken needle.

4. The AI-based magnetic-pressure coordinated multi-modal broken needle removal system according to claim 3, characterized in that: One end of the mounting frame (1) is connected to the operation panel (18), and the other end thereof is connected to one end of the robotic arm. A negative pressure pump group (2) is installed on the upper surface of the mounting frame (1), and the output end of the negative pressure pump group (2) is connected to a hose 1 (10). The other end of the robotic arm is installed with a mounting seat (7), and a connector (8) is installed on the outer wall of the mounting seat (7). One end of a rod tube (12) is installed at the bottom of the connector (8), and a hose 2 (11) is provided inside the rod tube (12). A connector is provided on the outside of the hose 2 (11). The connector (8) is connected to an external water delivery device, a magnetic field generator (9) is arranged in the connector (8), the outside of the hose (10) is arranged in a rod tube (12) through the connector (8), the other end of the rod tube (12) is threadedly connected to a suction head (13), a magnetic base (14) is installed in the rod tube (12) through a plurality of mounting rods (15), a sensor group (16) is arranged on the top of the magnetic base (14), and a filter plate (17) is installed on the inner wall of the rod tube (12) at a position above the sensor group (16).

5. The AI-based magnetic-pressure coordinated multi-modal broken needle removal system according to claim 4, characterized in that: The mechanical arm comprises a movable arm (3), one end of the movable arm (3) is connected to the other end of the mounting frame (1) through a driving member, the other end of the movable arm (3) is connected to the movable arm (4) through a driving member, the other end of the movable arm (3) is connected to one end of the movable arm (4) through a driving member, the other end of the movable arm (4) is connected to the adjustment frame (5) through a driving member, the inside of the adjustment frame (5) is connected to one end of the movable arm (3) through a driving member, the other end of the movable arm (6) is connected to the mounting seat (7) through a driving member.

6. The AI-based magnetic-pressure coordinated multi-modal broken needle removal system according to claim 1, characterized in that: The real-time monitoring and feedback module includes: Physiological monitoring unit, used to track patients' heart rate and blood pressure vital signs in real time and warn of abnormal fluctuations; An image monitoring unit, which provides real-time visual feedback of operations through microscopic imaging and ultrasonic testing; A dynamic adjustment unit for automatically optimizing the robot's motion trajectory and adsorption parameters based on multimodal sensing data; The safety protection unit is used to trigger the emergency stop mechanism to prevent the risks of overload, temperature rise and pressure exceeding the limit.

7. The AI-based magnetic-pressure coordinated multi-modal broken needle removal system according to claim 1, characterized in that: The data learning and case library module includes: Case storage unit, used for structured archiving of surgical images, operating parameters and postoperative efficacy data; A machine learning unit, used to continuously optimize broken needle location and removal strategies through reinforcement learning; Personalized recommendation unit, used to generate customized surgical plans based on the patient's anatomical characteristics and historical cases.

8. The AI-based magnetic-pressure coordinated multi-modal broken needle removal system according to claim 4, characterized in that: The sensor group (16) integrates a magnetic field sensor, a pressure sensor, a temperature sensor, a displacement sensor, and an image monitoring sensor.

9. The AI-based magnetic-pressure coordinated multi-modal broken needle removal system according to claim 4, characterized in that: The outer wall of the suction head (13) is embedded with a bioimpedance sensor and is made of medical-grade nickel-titanium alloy.

10. A magnetic-pressure coordinated multi-modal broken needle removal method based on AI, characterized in that: The AI-based magnetic-pressure coordinated multi-modal broken needle removal system according to any one of claims 1 to 9 is used, comprising the following steps: S1. Preoperative 3D image acquisition and intelligent planning: 3D image data of the patient's oral cavity and root canals is acquired through a CBCT scanner. The AI-assisted diagnosis module removes artifacts and performs 3D reconstruction on the images. It calibrates the 3D coordinates of the broken needle and generates at least three needle extraction paths that avoid the pulp and curved root canal structures. The path with the lowest risk is selected based on finite element stress analysis. S2. Initialization of the magnetic-pressure coordinated adsorption system: Install the adapter suction head (13) to the end of the rod tube (12), calibrate the target field strength of the magnetic field generator (9) to 0.5T±0.05T, start the negative pressure pump group (2) to pre-test the negative pressure stability, and set the robot arm movement parameters through the operation panel (18); S3. Dynamic positioning and adsorption start of the robotic arm: the robotic arm control unit drives the movable arm 1 (3), movable arm 2 (4), movable arm 3 (6) and the adjustment frame (5) to move in coordination, so that the suction head (13) approaches the broken needle along the planned path, and the contact force is monitored in real time by the force feedback adjustment unit. After the suction head (13) lightly touches the surface of the broken needle, the magnetic field generator (9) is started to apply a gradient magnetic field to magnetize the broken needle, and the negative pressure pump group (2) is simultaneously activated to generate a negative pressure of -35kPa to -50kPa for adsorption; S4. Multimodal real-time monitoring and dynamic adjustment: The adhesion state of the broken needle and dentin is monitored by a bioimpedance sensor, and the magnetic field intensity, temperature rise of the magnetic base (14) and displacement of the broken needle are fed back in real time using a magnetic field sensor, a temperature sensor and a displacement sensor. The dynamic adjustment module optimizes the adsorption parameters based on multi-source data; S5. Safety protection and emergency treatment: When the patient's heart rate is greater than 100 bpm or the systolic blood pressure is greater than 140 mmHg, the physiological monitoring unit triggers an audible and visual alarm and automatically slows down the robotic arm to 0.1 mm / s, while suspending the adsorption operation until vital signs return to normal. If microscopic imaging and ultrasound detect microcracks in the root canal or residual broken needle fragments, the dynamic adjustment unit immediately generates an obstacle avoidance path, and the robotic arm performs avoidance at a speed of 0.2 mm / s. Based on the magnetic field sensor, pressure sensor and displacement feedback data, the negative pressure is adjusted from -35 kPa to -50 kPa and the magnetic field gradient is 5-12 T / m every 100 ms to stabilize the displacement speed of the broken needle at 0.3 mm / s ± 0.05 mm. When the temperature rise of the magnetic base (14) is greater than 1.5 ° C, the liquid cooling system automatically increases the pressure. When the dentin stress is greater than 8 MPa, the magnetic field and negative pressure are immediately cut off, triggering an emergency stop. If the broken needle stagnates for more than 10 seconds, the negative pressure automatically switches to the pulse mode. S6. Broken needle recovery and postoperative verification: After the broken needle is adsorbed, the robotic arm retreats at a speed of 0.2-0.3 mm / s, and the debris is filtered through the filter plate (17); S7. Data upload and model optimization: Upload surgical images, operation parameters and postoperative efficacy data to the case storage unit. The machine learning unit optimizes the broken needle positioning and adsorption strategy through the reinforcement learning algorithm. The personalized recommendation unit generates a postoperative care plan based on the patient's anatomical characteristics.

Images