Interventional surgery robot puncture path intelligent planning method

Abstract

This invention discloses an intelligent puncture path planning method for interventional surgical robots, belonging to the field of interventional robot path planning technology. The method acquires multimodal medical images of the patient's puncture site, completes organ tissue segmentation and three-dimensional reconstruction, generates a digital twin anatomical model, and calibrates basic puncture parameters; constructs a respiratory phase tracking model, combines the basic puncture parameters to lock the respiratory cycle and real-time respiratory phase, and outputs the three-dimensional predicted coordinates of the puncture target point for a pre-set duration; acquires needle insertion force data through a six-dimensional force sensor, calculates tissue compression deformation and needle body flexible bending offset in real time, and generates a feedforward compensation vector; fuses the three-dimensional predicted coordinates and the feedforward compensation vector to generate a dynamic puncture path, performing continuous needle insertion without interruption throughout the entire process. This invention improves upon the shortcomings of traditional static path planning, which cannot adapt to respiratory displacement and tissue deformation, effectively improving puncture positioning accuracy and is suitable for minimally invasive interventional puncture surgery scenarios.

Description

Technical Field

[0001] This invention relates to the field of interventional robot path planning technology, specifically to an intelligent planning method for puncture paths in interventional surgical robots. Background Technology

[0002] Interventional puncture surgery is a core means of minimally invasive diagnosis and treatment such as organ biopsy and tumor ablation. Clinically, the requirements for puncture positioning accuracy are very strict. The widespread use of interventional surgical robots has provided a hardware foundation for the precise implementation of puncture operations. Puncture path planning technology is the core key to ensuring the accuracy of robot puncture operations.

[0003] Existing interventional puncture path planning methods are all based on preoperative static medical images to complete the path setting. During the operation, they cannot synchronously and collaboratively compensate for the dynamic displacement of the target point caused by respiratory movements and the soft tissue deformation caused by the puncture needle. The coupling of these two factors directly causes excessive puncture positioning deviation, which cannot meet the clinical needs for high-precision puncture. Summary of the Invention

[0004] The purpose of this invention is to provide an intelligent planning method for puncture paths in interventional surgical robots, so as to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: an intelligent planning method for puncture paths in interventional surgical robots, comprising the following steps: Step 100: Acquire multimodal medical images of the patient's puncture site, and based on the multimodal medical images, complete the segmentation and three-dimensional reconstruction of the lesion area, danger area and organ tissues to generate a digital twin anatomical model, and simultaneously complete the calibration of the basic puncture parameters; Step 200: Construct a respiratory phase tracking model based on the phase-locked loop algorithm, collect the patient's respiratory signal in real time, and lock the patient's respiratory cycle and real-time respiratory phase in combination with the puncture basic parameters, and output the three-dimensional predicted coordinates of the puncture target point after a preset duration. Step 300: The six-dimensional force sensor installed on the end effector of the interventional surgical robot collects the needle insertion force data in real time during the puncture process. Based on the basic puncture parameters, the tissue compression deformation and needle body flexible bending offset are calculated in real time according to the needle insertion force data to generate a feedforward compensation vector. Step 400: The three-dimensional predicted coordinates and the feedforward compensation vector are fused in three-dimensional space to generate a dynamic puncture path that is updated in real time. The interventional surgical robot is then controlled to perform continuous needle insertion along the dynamic puncture path without any interruption. Step 500: Real-time acquisition of the actual position of the puncture needle tip and the real-time needle insertion depth; deviation verification between the actual position and the dynamic puncture path; and execution of deviation correction operation or safe shutdown operation based on a preset threshold.

[0006] Preferably, step 100 includes: Step 110: Acquire four-dimensional computed tomography (CT) images and ultrasound images of the patient's puncture site using medical imaging equipment as multimodal medical images. Step 120: Using a three-dimensional convolutional neural network model, the lesion area, danger area and organ tissue in the multimodal medical image are automatically segmented, and three-dimensional reconstruction is completed based on the segmentation results to generate a digital twin anatomical model. Step 130: Complete the calibration of basic puncture parameters, which include the puncture needle stiffness coefficient, the preset correspondence between the needle insertion force and tissue and needle deformation, and the patient's respiratory characteristic parameters.

[0007] Preferably, the preset correspondence between the needle insertion force and the tissue and needle deformation is a preset mapping relationship between the needle insertion force value and the corresponding tissue compression deformation and needle body flexible bending offset when the puncture needle enters different types of soft tissue. The different types of soft tissue include skin, fat, muscle and organ tissue.

[0008] Preferably, step 200 includes: Step 210: Construct a breathing phase tracking model based on a first-order phase-locked loop algorithm coupled with a recurrent neural network; Step 220: The patient's respiratory signal is collected in real time through a respiratory acquisition device, and the respiratory signal is input into the respiratory phase tracking model. Combined with the respiratory characteristic parameters in the puncture baseline parameters, the patient's respiratory cycle and real-time respiratory phase are locked. Step 230: Using the respiratory phase tracking model, output the predicted coordinates of the puncture target point in three-dimensional space according to the preset lead time, as the three-dimensional predicted coordinates.

[0009] Preferably, the preset duration ranges from 80 milliseconds to 120 milliseconds, the prediction error of the three-dimensional predicted coordinates is no greater than 0.2 millimeters, the recurrent neural network adopts a lightweight long short-term memory network, and the inference delay of the lightweight long short-term memory network is no greater than 3 milliseconds.

[0010] Preferably, step 300 includes: Step 310: Collect the axial insertion force and radial deflection force of the puncture needle in real time during the puncture process at a fixed sampling frequency, and use them as insertion force data. Step 320: Retrieve the preset correspondence between the needle insertion force and the tissue and needle deformation in the basic puncture parameters, and match the tissue compression deformation and needle flexible bending offset corresponding to the needle insertion force data. Step 330: Based on the tissue compression deformation and the needle body flexible bending offset, a feedforward compensation vector is generated to correct the puncture path.

[0011] Preferably, the sampling frequency of the six-dimensional force sensor is not less than 1000 Hz, and the calculation delay of the feedforward compensation vector is not greater than 5 milliseconds.

[0012] Preferably, step 400 includes: Step 410: Perform spatial coordinate weighted fusion on the three-dimensional predicted coordinates and the feedforward compensation vector to generate the corrected real-time puncture target coordinates. Step 420: Based on the real-time puncture target coordinates, generate a real-time updated dynamic puncture path at a fixed update frequency. Step 430: The dynamic puncture path is synchronously input into the motion control unit of the interventional surgical robot to control the interventional surgical robot to perform continuous needle insertion operations without interruption along the dynamic puncture path.

[0013] Preferably, the update frequency of the dynamic puncture path is not less than 100 Hz, and in the spatial coordinate weighted fusion, the weight of the three-dimensional predicted coordinates accounts for 70% to 90%, and the weight of the feedforward compensation vector accounts for 10% to 30%.

[0014] Preferably, step 500 includes: Step 510: The actual position of the puncture needle tip is collected in real time by intraoperative ultrasound equipment, and the real-time needle insertion depth is collected in real time by the encoder of the interventional surgical robot. Step 520: Calculate the positional deviation value between the actual position and the corresponding position of the dynamic puncture path, and compare the positional deviation value with a preset positional deviation threshold. Step 530: If the position deviation value is greater than the preset position deviation threshold, perform a deviation correction operation for the dynamic puncture path. The preset position deviation threshold is not greater than 0.5 mm. Step 540: The real-time collected needle insertion force data and real-time needle insertion depth are compared with the preset force threshold and preset depth threshold, respectively. If either parameter exceeds the corresponding threshold, a safety shutdown operation is immediately performed.

[0015] Compared with the prior art, the beneficial effects of the present invention are: By optimizing puncture path planning through coupled respiratory phase tracking and puncture force feedforward compensation, puncture deviations caused by respiratory motion and soft tissue deformation can be reduced, thus improving puncture positioning accuracy. By adopting a continuous needle insertion method without interruption, traction damage to body tissues during puncture can be reduced, thereby improving the stability of surgical operations. This solution is compatible with existing interventional surgical robot platforms, is easy to implement, and can provide certain technical support for precise interventional puncture surgery. Attached Figure Description

[0016] Figure 1 The main flowchart of an intelligent puncture path planning method for an interventional surgical robot provided in an embodiment of the present invention is shown. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Please see Figure 1 This invention provides an intelligent planning method for puncture paths in interventional surgical robots, comprising: Step 100: Acquire multimodal medical images of the patient's puncture site, and based on the multimodal medical images, segment and reconstruct the lesion area, danger area and organ tissues in three dimensions to generate a digital twin anatomical model, and simultaneously complete the calibration of the basic puncture parameters.

[0019] Specifically, step 100 includes: Step 110: Acquire four-dimensional computed tomography (CT) images and ultrasound images of the patient's puncture site using medical imaging equipment as multimodal medical images. Step 120: Using a three-dimensional convolutional neural network model, the lesion area, danger area and organ tissue in the multimodal medical image are automatically segmented, and three-dimensional reconstruction is completed based on the segmentation results to generate a digital twin anatomical model. Step 130: Complete the calibration of basic puncture parameters, which include the puncture needle stiffness coefficient, the preset correspondence between the needle insertion force and tissue and needle deformation, and the patient's respiratory characteristic parameters.

[0020] Specifically, in this embodiment, step 100 serves as a preoperative foundational construction step, providing data and model support for subsequent intelligent planning of the puncture path; the multimodal medical imaging is used to comprehensively acquire anatomical structural information of the puncture site; the three-dimensional convolutional neural network model is a deep learning model suitable for three-dimensional medical image processing, which can achieve precise and automatic segmentation of lesion areas, danger areas, and organ tissues; the digital twin anatomical model is used to restore the true three-dimensional anatomical morphology of the patient's puncture site; and the calibration of the basic puncture parameters provides standardized data support for subsequent respiratory phase tracking and deformation compensation, avoiding planning deviations caused by individual differences and instrument differences.

[0021] In addition, this step establishes a foundation for personalized preoperative planning through an integrated process of image acquisition, 3D reconstruction, and parameter calibration, thereby improving the adaptability and accuracy of subsequent dynamic planning.

[0022] In one optional implementation, the three-dimensional convolutional neural network model adopts a three-dimensional lightweight medical image segmentation network structure, which balances segmentation accuracy and inference efficiency.

[0023] As an example, in one feasible implementation, for the patient's target puncture site, corresponding four-dimensional computed tomography (CT) images and ultrasound images are acquired through medical imaging equipment. A three-dimensional convolutional neural network model is used to complete the automatic segmentation and three-dimensional reconstruction of the lesion area, danger area and organ tissue, generating a digital twin anatomical model. At the same time, the calibration of the puncture needle stiffness coefficient, the needle insertion force and the preset correspondence between tissue and needle deformation, and the patient's respiratory characteristic parameters are completed.

[0024] The preset correspondence between the needle insertion force and the tissue and needle deformation is a preset mapping relationship between the needle insertion force value and the corresponding tissue compression deformation and needle body flexible bending offset when the puncture needle enters different types of soft tissue. The different types of soft tissue include skin, fat, muscle and organ tissue.

[0025] Specifically, in this embodiment, the preset correspondence between the needle insertion force and the tissue and needle deformation is a data correspondence established in advance based on the biomechanical characteristics of soft tissue and the mechanical characteristics of the puncture needle, which is used to realize the rapid calculation of puncture deformation; the different types of soft tissue are classified according to the tissue classification standards of routine clinical interventional puncture, and the deformation response of different soft tissues is different. Clarifying the type can improve the accuracy of deformation compensation.

[0026] In addition, this step, by clarifying the deformation correspondence and soft tissue type, can reduce the computational complexity during surgery and ensure the real-time and reliable nature of deformation compensation.

[0027] As an example, in one feasible implementation, the relationship between needle insertion force and deformation data corresponding to skin, fat, muscle, and organ tissues is established in advance, and the corresponding tissue compression deformation and needle body flexible bending offset are quickly matched according to the real-time needle insertion force during the operation.

[0028] Step 200: Construct a respiratory phase tracking model based on the phase-locked loop algorithm, collect the patient's respiratory signal in real time, and lock the patient's respiratory cycle and real-time respiratory phase in combination with the puncture baseline parameters, and output the three-dimensional predicted coordinates of the puncture target point after a preset duration.

[0029] Specifically, step 200 includes: Step 210: Construct a breathing phase tracking model based on a first-order phase-locked loop algorithm coupled with a recurrent neural network; Step 220: The patient's respiratory signal is collected in real time through a respiratory acquisition device, and the respiratory signal is input into the respiratory phase tracking model. Combined with the respiratory characteristic parameters in the puncture baseline parameters, the patient's respiratory cycle and real-time respiratory phase are locked. Step 230: Using the respiratory phase tracking model, output the predicted coordinates of the puncture target point in three-dimensional space according to the preset lead time, as the three-dimensional predicted coordinates.

[0030] Specifically, in this embodiment, step 200 is used to realize the dynamic prediction of breathing at the puncture target. The breathing phase tracking model is constructed by coupling a first-order phase-locked loop algorithm and a recurrent neural network. The phase-locked loop algorithm is used to stably lock the breathing cycle and phase, and the recurrent neural network is used to mine the temporal change characteristics of breathing motion. The combination of the two can solve the problem of target displacement tracking lag caused by breathing motion. According to the preset advance time, the three-dimensional prediction coordinates are output in advance, which can make the puncture path adapt to the dynamic position of the target in advance and avoid the lag defects of static planning.

[0031] In addition, this step, combined with the respiratory characteristic parameters calibrated before surgery, can quickly adapt to the breathing patterns of different patients and simplify the intraoperative calibration process.

[0032] As an example, in one feasible implementation, a respiratory phase tracking model is constructed based on a first-order phase-locked loop algorithm and a recurrent neural network. The patient's real-time respiratory signal is acquired through a respiratory acquisition device, and the respiratory cycle and real-time respiratory phase are locked in combination with the respiratory characteristic parameters calibrated before the operation, and the three-dimensional predicted coordinates of the puncture target point are output in advance.

[0033] The preset duration ranges from 80 milliseconds to 120 milliseconds, the prediction error of the three-dimensional predicted coordinates is no greater than 0.2 millimeters, the recurrent neural network adopts a lightweight long short-term memory network, and the inference delay of the lightweight long short-term memory network is no greater than 3 milliseconds.

[0034] Specifically, in this embodiment, the preset duration is the advance prediction duration of the respiratory phase tracking model. This duration is determined based on the rated motion response delay of the interventional surgical robot to ensure that the predicted coordinates are synchronized with the robot's actions. The error index of the three-dimensional predicted coordinates is used to ensure the puncture positioning accuracy. The lightweight long short-term memory network is a recurrent neural network structure that optimizes computing power usage, which can reduce inference latency and ensure the real-time performance of the prediction process.

[0035] In addition, this step balances prediction accuracy and computational efficiency through parameter constraints to ensure that dynamic target prediction meets the needs of real-time use during surgery.

[0036] As an example, in one feasible implementation, the preset duration is set to a value between 80 milliseconds and 120 milliseconds to control the prediction error of the three-dimensional prediction coordinates to be no greater than 0.2 millimeters, and a lightweight long short-term memory network is used to ensure that the inference latency is no greater than 3 milliseconds.

[0037] Step 300: The six-dimensional force sensor installed on the end effector of the interventional surgical robot collects needle insertion force data in real time during the puncture process. Based on the basic puncture parameters, the tissue compression deformation and needle body flexible bending offset are calculated in real time according to the needle insertion force data to generate a feedforward compensation vector.

[0038] Specifically, step 300 includes: Step 310: Collect the axial insertion force and radial deflection force of the puncture needle in real time during the puncture process at a fixed sampling frequency, and use them as insertion force data. Step 320: Retrieve the preset correspondence between the needle insertion force and the tissue and needle deformation in the basic puncture parameters, and match the tissue compression deformation and needle flexible bending offset corresponding to the needle insertion force data. Step 330: Based on the tissue compression deformation and the needle body flexible bending offset, a feedforward compensation vector is generated to correct the puncture path.

[0039] Specifically, in this embodiment, step 300 is used to realize deformation feedforward compensation during the puncture process. The needle insertion force data is used to reflect the interaction state between the puncture needle and the tissue. Based on the preset correspondence, the tissue compression deformation and the needle body flexible bending offset can be quickly calculated. The feedforward compensation vector is used to directly correct the puncture path and realize deformation advance compensation.

[0040] In addition, this step does not require the real-time construction of complex biomechanical models, effectively reducing the computational load and ensuring the real-time performance of compensation actions.

[0041] As an example, in one feasible implementation, the axial insertion force and radial deflection force during the puncture process are collected at a fixed sampling frequency, a preset deformation correspondence is retrieved to match the corresponding deformation and offset, and a feedforward compensation vector for path correction is generated.

[0042] The sampling frequency of the six-dimensional force sensor is not less than 1000 Hz, and the calculation delay of the feedforward compensation vector is not greater than 5 milliseconds.

[0043] Specifically, in this embodiment, the sampling frequency of the six-dimensional force sensor determines the precision of force data acquisition, ensuring that high-frequency sampling can capture minute force changes and avoid compensation errors; the calculation delay of the feedforward compensation vector is the core indicator of real-time compensation, and low delay can ensure that the compensation action and the puncture action are executed synchronously.

[0044] In addition, this step strengthens the accuracy foundation of deformation compensation by limiting the sensor and calculation parameters, making it suitable for surgical scenarios involving continuous needle insertion.

[0045] As an example, in one feasible implementation, a six-dimensional force sensor with a sampling frequency of not less than 1000 Hz is used, and the calculation delay of the feedforward compensation vector is controlled to be no more than 5 milliseconds.

[0046] Step 400: The three-dimensional predicted coordinates and the feedforward compensation vector are fused in three-dimensional space to generate a dynamic puncture path that is updated in real time. The interventional surgical robot is then controlled to perform continuous needle insertion operations along the dynamic puncture path without any interruption.

[0047] Specifically, step 400 includes: Step 410: Perform spatial coordinate weighted fusion on the three-dimensional predicted coordinates and the feedforward compensation vector to generate the corrected real-time puncture target coordinates. Step 420: Based on the real-time puncture target coordinates, generate a real-time updated dynamic puncture path at a fixed update frequency. Step 430: The dynamic puncture path is synchronously input into the motion control unit of the interventional surgical robot to control the interventional surgical robot to perform continuous needle insertion operations without interruption along the dynamic puncture path.

[0048] Specifically, in this embodiment, step 400 is used to realize dynamic path fusion and puncture execution. The spatial coordinate weighted fusion is used to integrate target prediction data and deformation compensation data. The dynamic puncture path is used to adapt to the positional changes of tissue and target during the operation in real time. Continuous needle insertion without interruption can reduce tissue traction and reduce surgical trauma.

[0049] In addition, this step will transmit the fusion path synchronously to the motion control unit to achieve seamless integration between path planning and robot execution.

[0050] As an example, in one feasible implementation, the three-dimensional predicted coordinates and the feedforward compensation vector are fused using spatial coordinate weighting. Based on the fusion result, a dynamic puncture path is generated and updated in real time. The path is then transmitted to the robot motion control unit to execute continuous needle insertion.

[0051] The update frequency of the dynamic puncture path is no less than 100 Hz, and in the weighted fusion of spatial coordinates, the weight of the three-dimensional predicted coordinates is 70% to 90%, and the weight of the feedforward compensation vector is 10% to 30%.

[0052] Specifically, in this embodiment, the update frequency of the dynamic puncture path determines the path's responsiveness to dynamic changes, and high-frequency updates can match the target point and tissue displacement in real time; the weight allocation of the spatial coordinate weighted fusion is determined based on the preoperatively calibrated respiratory displacement amplitude and puncture deformation level, taking into account the compensation effect of both factors.

[0053] In addition, this step improves the stability and accuracy of the dynamic puncture path by reasonably setting the update frequency and weight.

[0054] As an example, in one feasible implementation, the update frequency of the dynamic puncture path is set to no less than 100 Hz, the three-dimensional predicted coordinates are weighted at 70% to 90% during spatial coordinate weighted fusion, and the feedforward compensation vector is weighted at 10% to 30%.

[0055] Step 500: Real-time acquisition of the actual position of the puncture needle tip and the real-time needle insertion depth; deviation verification between the actual position and the dynamic puncture path; and execution of deviation correction operation or safe shutdown operation based on a preset threshold.

[0056] Specifically, step 500 includes: Step 510: The actual position of the puncture needle tip is collected in real time by intraoperative ultrasound equipment, and the real-time needle insertion depth is collected in real time by the encoder of the interventional surgical robot. Step 520: Calculate the positional deviation value between the actual position and the corresponding position of the dynamic puncture path, and compare the positional deviation value with a preset positional deviation threshold. Step 530: If the position deviation value is greater than the preset position deviation threshold, perform a deviation correction operation for the dynamic puncture path. The preset position deviation threshold is not greater than 0.5 mm. Step 540: The real-time collected needle insertion force data and real-time needle insertion depth are compared with the preset force threshold and preset depth threshold, respectively. If either parameter exceeds the corresponding threshold, a safety shutdown operation is immediately performed.

[0057] Specifically, in this embodiment, step 500 is used to achieve closed-loop verification and safety protection of the puncture process. The actual position of the puncture needle tip is obtained by intraoperative ultrasound equipment to complete the deviation verification, and the real-time needle depth is collected by robot encoder to monitor the puncture stroke. Deviation correction is used to ensure puncture accuracy, and safe shutdown is used to avoid surgical risks caused by excessive puncture force or depth.

[0058] Furthermore, this step forms a closed loop encompassing planning, execution, verification, and protection, ensuring surgical safety while improving accuracy.

[0059] As an example, in one feasible implementation, the actual position of the puncture needle tip is acquired by intraoperative ultrasound equipment, the real-time needle insertion depth is acquired by robot encoder, the position deviation value is calculated and compared with a preset threshold, and correction is performed if the value exceeds the threshold; at the same time, the needle insertion force and needle insertion depth are compared with the corresponding preset thresholds, and a safety shutdown is immediately performed if the value exceeds the limit.

[0060] In this embodiment, by optimizing the puncture path planning through coupling respiratory phase tracking and puncture force feedforward compensation, the puncture deviation caused by respiratory motion and soft tissue deformation can be reduced, thereby improving the puncture positioning accuracy. By adopting a continuous needle insertion method without interruption, the traction damage to the body tissues caused by the puncture can be reduced, thereby improving the stability of the surgical operation. This solution is compatible with existing interventional surgical robot platforms, is easy to implement, and can provide certain technical support for precise interventional puncture surgery.

[0061] Furthermore, it should be noted that the combination of the various technical features in this case is not limited to the combination methods described in the claims of this case or the combination methods described in the specific embodiments. All technical features described in this case can be freely combined or combined in any way, unless they contradict each other.

[0062] It should be noted that the above examples are merely specific embodiments of the present invention, and the present invention is obviously not limited to the above embodiments, with many similar variations. All modifications that can be directly derived or conceived by those skilled in the art from the content disclosed in this invention should fall within the protection scope of this invention.

[0063] The above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for intelligent planning of puncture paths in an interventional surgical robot, characterized in that, Includes the following steps: Step 100: Acquire multimodal medical images of the patient's puncture site, and based on the multimodal medical images, complete the segmentation and three-dimensional reconstruction of the lesion area, danger area and organ tissues to generate a digital twin anatomical model, and simultaneously complete the calibration of the basic puncture parameters; Step 200: Construct a respiratory phase tracking model based on the phase-locked loop algorithm, collect the patient's respiratory signal in real time, and lock the patient's respiratory cycle and real-time respiratory phase in combination with the puncture basic parameters, and output the three-dimensional predicted coordinates of the puncture target point after a preset duration. Step 300: The six-dimensional force sensor installed on the end effector of the interventional surgical robot collects the needle insertion force data in real time during the puncture process. Based on the basic puncture parameters, the tissue compression deformation and needle body flexible bending offset are calculated in real time according to the needle insertion force data to generate a feedforward compensation vector. Step 400: The three-dimensional predicted coordinates and the feedforward compensation vector are fused in three-dimensional space to generate a dynamic puncture path that is updated in real time. The interventional surgical robot is then controlled to perform continuous needle insertion along the dynamic puncture path without any interruption. Step 500: Real-time acquisition of the actual position of the puncture needle tip and the real-time needle insertion depth; deviation verification between the actual position and the dynamic puncture path; and execution of deviation correction operation or safe shutdown operation based on a preset threshold.

2. The method according to claim 1, characterized in that, Step 100 includes: Step 110: Acquire four-dimensional computed tomography (CT) images and ultrasound images of the patient's puncture site using medical imaging equipment as multimodal medical images. Step 120: Using a three-dimensional convolutional neural network model, the lesion area, danger area and organ tissue in the multimodal medical image are automatically segmented, and three-dimensional reconstruction is completed based on the segmentation results to generate a digital twin anatomical model. Step 130: Complete the calibration of basic puncture parameters, which include the puncture needle stiffness coefficient, the preset correspondence between the needle insertion force and tissue and needle deformation, and the patient's respiratory characteristic parameters.

3. The method according to claim 2, characterized in that, The preset correspondence between the needle insertion force and the tissue and needle deformation is a preset mapping relationship between the needle insertion force value and the corresponding tissue compression deformation and needle body flexible bending offset when the puncture needle enters different types of soft tissue. The different types of soft tissue include skin, fat, muscle and organ tissue.

4. The method according to claim 1, characterized in that, Step 200 includes: Step 210: Construct a breathing phase tracking model based on a first-order phase-locked loop algorithm coupled with a recurrent neural network; Step 220: The patient's respiratory signal is collected in real time through a respiratory acquisition device, and the respiratory signal is input into the respiratory phase tracking model. Combined with the respiratory characteristic parameters in the puncture baseline parameters, the patient's respiratory cycle and real-time respiratory phase are locked. Step 230: Using the respiratory phase tracking model, output the predicted coordinates of the puncture target point in three-dimensional space according to the preset lead time, as the three-dimensional predicted coordinates.

5. The method according to claim 4, characterized in that, The preset duration ranges from 80 milliseconds to 120 milliseconds, the prediction error of the three-dimensional predicted coordinates is no greater than 0.2 millimeters, the recurrent neural network adopts a lightweight long short-term memory network, and the inference delay of the lightweight long short-term memory network is no greater than 3 milliseconds.

6. The method according to claim 1, characterized in that, Step 300 includes: Step 310: Collect the axial insertion force and radial deflection force of the puncture needle in real time during the puncture process at a fixed sampling frequency, and use them as insertion force data. Step 320: Retrieve the preset correspondence between the needle insertion force and the tissue and needle deformation in the basic puncture parameters, and match the tissue compression deformation and needle flexible bending offset corresponding to the needle insertion force data. Step 330: Based on the tissue compression deformation and the needle body flexible bending offset, a feedforward compensation vector is generated to correct the puncture path.

7. The method according to claim 6, characterized in that, The sampling frequency of the six-dimensional force sensor is not less than 1000 Hz, and the calculation delay of the feedforward compensation vector is not greater than 5 milliseconds.

8. The method according to claim 1, characterized in that, Step 400 includes: Step 410: Perform spatial coordinate weighted fusion on the three-dimensional predicted coordinates and the feedforward compensation vector to generate the corrected real-time puncture target coordinates. Step 420: Based on the real-time puncture target coordinates, generate a real-time updated dynamic puncture path at a fixed update frequency. Step 430: The dynamic puncture path is synchronously input into the motion control unit of the interventional surgical robot to control the interventional surgical robot to perform continuous needle insertion operations without interruption along the dynamic puncture path.

9. The method according to claim 8, characterized in that, The update frequency of the dynamic puncture path is no less than 100 Hz. In the weighted fusion of spatial coordinates, the weight of the three-dimensional predicted coordinates accounts for 70% to 90%, and the weight of the feedforward compensation vector accounts for 10% to 30%.

10. The method according to claim 1, characterized in that, Step 500 includes: Step 510: The actual position of the puncture needle tip is collected in real time by intraoperative ultrasound equipment, and the real-time needle insertion depth is collected in real time by the encoder of the interventional surgical robot. Step 520: Calculate the positional deviation value between the actual position and the corresponding position of the dynamic puncture path, and compare the positional deviation value with a preset positional deviation threshold. Step 530: If the position deviation value is greater than the preset position deviation threshold, perform a deviation correction operation for the dynamic puncture path. The preset position deviation threshold is not greater than 0.5 mm. Step 540: The real-time collected needle insertion force data and real-time needle insertion depth are compared with the preset force threshold and preset depth threshold, respectively. If either parameter exceeds the corresponding threshold, a safety shutdown operation is immediately performed.

Images