Interaction stiffness coefficient-based robot-assisted nail placement guide needle puncture state identification method

Abstract

The invention discloses a robot-assisted screw placement guide needle puncture state recognition method based on an interaction stiffness coefficient. The recognition method comprises the steps that force feedback data marked by an expert is collected and utilized; based on the force feedback data and a pre-constructed expert evaluation standard, training a neural network to learn time sequence and frequency spectrum features of force signals in different interaction states, and establishing a state recognition model; and performing real-time state recognition according to real-time data acquired in the operation by utilizing the state recognition model. According to the recognition method, the problems that in the prior art, preoperative planning is seriously depended in an operation, the state of the tail end of the flexible instrument cannot be accurately sensed, and reliability is poor when visual information is limited are solved.

Description

Technical Field

[0001] This invention belongs to the field of surgical robot technology, specifically relating to a method for identifying the puncture status of a robot-assisted pin placement guide needle based on the interaction stiffness coefficient. Background Technology

[0002] With the continuous advancement of technology, robots, due to their advantages such as high precision and high stability, have been widely used in the medical field, especially in orthopedics.

[0003] Typically, surgical robot state perception technology relies primarily on multimodal data fusion. A common approach is to integrate a six-dimensional force / torque sensor into the end effector of the robotic arm. This sensor directly measures the force feedback signal generated when the instrument comes into contact with tissue to determine the interaction state, and then combines this information with CT images, preoperative planning, and other data for navigation and safety control.

[0004] However, in orthopedics, the application of robots in cervical spine and spinal deformity fixation surgery faces many serious challenges. Patients with cervical spine and spinal deformities have complex physiological structures and a certain degree of mobility during surgery. Factors such as the patient's breathing and anesthesia response can cause deviations between the vertebral body position and the preoperative CT planning model. Screw placement surgery in cervical spine and spinal deformity patients often uses flexible guide needles or drills, whose tips undergo slight deflection under force. This deformation is a key basis for fine-tuning during surgery based on rich experience, but traditional force sensors can only measure the force at the end of the robotic arm and cannot directly reflect the actual interaction between the flexible instrument tip and bone tissue, leading to the loss of crucial information. Furthermore, there is limited research on determining the interaction state between instruments and tissue during screw placement in cervical spine and spinal deformity patients. Existing technologies disclose a surgical robot state encoding method based on a cross-attention mechanism. Although its model can control the robot based on current visual-force feedback, it fails to effectively solve the problem of state determination when image information is unreliable, and it also fails to effectively solve the problem of state determination when the flexible instrument tip deflects.

[0005] In summary, existing technologies for robots used in orthopedic surgery have significant limitations, low accuracy, and poor reliability. Summary of the Invention

[0006] This invention provides a robot-assisted pin placement and guide needle puncture status recognition method based on interactive stiffness coefficient, which solves the problems of existing technologies that rely heavily on preoperative planning during surgery, cannot accurately perceive the status of flexible instrument ends, and have poor reliability when visual information is limited.

[0007] To achieve the above objectives, the present invention adopts the following specific technical solution: First, this invention provides a method for identifying the puncture state of a robot-assisted pin placement guide needle based on interactive stiffness coefficient, including: Collect and utilize force feedback data annotated by experts; Based on the force feedback data and the pre-built expert evaluation criteria, a neural network is trained to learn the temporal and spectral characteristics of force signals under different interactive states, and a state recognition model is established. Using the aforementioned status recognition model, real-time status recognition is performed based on real-time data collected during the operation; The expert evaluation criteria are constructed as follows: Based on expert evaluation, an expert evaluation criterion based on the interaction stiffness coefficient between surgical instruments and bone tissue during screw placement is constructed; wherein the interaction stiffness coefficient is used to characterize the interaction state between the robot and the surgical process, and its expression is: ; In the above formula, K TB For interactive stiffness coefficients, F TB The real-time interactive force is measured by an end force sensor. v T This refers to the real-time feed speed of the robot's gripper.

[0008] Furthermore, including: Specimen processing: Obtain at least one embalmed vertebral body specimen, wherein each specimen of the at least one embalmed vertebral body specimen is a T12-L4 vertebral body, and determine the properties of each vertebral body specimen, wherein the properties include at least one of the following: congenital malformation, fracture, tumor, and tuberculosis; remove the paravertebral muscles of each specimen, remove all soft tissues outside the ligaments and intervertebral discs, divide from the L2 / L3 intervertebral space, and fix the anterior column 1 / 2 of the vertebral body with acrylic bone cement; Experimental equipment preparation: The experimental equipment includes the Tianji orthopedic surgical robot system, C-arm, flexible guide needle, and six-dimensional force sensor; Preparation before bone guide needle puncture: After the bone cement is fixed, the vertebral specimen is fixed on an X-ray-permeable operating table. The robot control panel and optical camera are placed at the head end of the operating table, and the robotic arm is placed at the tail end of the operating table. The patient tracker is fixed to the spinous process / bone cement on the head side of the specimen. The scale is mounted on the robotic arm and placed in the surgical area. Fluoroscopy confirms that the scale is in good position. Real-time three-dimensional images are acquired using a C-arm and transmitted to the robot control panel. Path planning and implementation: Simulate the pedicle screw puncture path of the spine, using the normal pedicle screw insertion point as the test puncture point, and adjust the guide needle angle to invade the medial cortical bone of the pedicle by 2mm; replace the end of the robotic arm with a flexible guide needle puncture tool equipped with a six-dimensional force sensor; control the robotic arm to guide the flexible guide needle to determine the planned screw puncture point, grind away the cortical bone within a 2mm radius centered on the insertion point, and use the flexible guide needle to confirm that the cortical bone grinding meets the puncture requirements; control the robotic arm to puncture the cancellous bone, record the force sensor data, the actual displacement and speed of the robotic arm end, until the insertion depth reaches more than 45mm or the robotic arm stops due to excessive resistance, and record whether the robotic arm has completed the puncture normally; Postoperative CT: After the puncture is completed, CT data of specimens with and without guide needles are obtained by C-arm scanning; Postoperative evaluation: Based on the DICOM data from the postoperative CT scan, three-dimensional reconstruction was performed to obtain the puncture path of the flexible guide needle, and the puncture status of the guide needle was evaluated in conjunction with whether the robotic arm successfully completed the puncture.

[0009] Furthermore, the postoperative evaluation results include: puncture via the cancellous bone pathway; puncture completed in the cancellous bone after the flexible guide needle deflects upon encountering the cortex; and failure to compensate after the flexible guide needle encounters the cortex, resulting in an error in the robot's robotic arm and inability to complete the puncture.

[0010] Furthermore, the vertebral specimen was fixed to the operating table after the bone cement was fixed with hot melt adhesive; the robotic arm was controlled to puncture the cancellous bone at a speed of 0.2 mm / s.

[0011] Furthermore, the C-arm is a 3D-C-arm; The flexible guide needle uses a flexible probe commonly used in spinal surgery.

[0012] Furthermore, based on the force feedback data and pre-built expert evaluation criteria, a neural network is trained to learn the temporal and spectral characteristics of force signals under different interaction states, and a state recognition model is established, including: Acquire force sensor data, actual displacement and velocity of the robotic arm end effector from the force feedback data; The time series data of the six-dimensional interaction stiffness coefficients in each CSV file and the corresponding state labels are sliced ​​through a sliding window of length L and encapsulated into training samples; each sample Xi contains a segment of time series data and corresponds to a state label Yi. A multi-layer temporal network structure based on TimesNet, in which the input is 6-channel temporal data, the network contains E coding layers, and each coding layer uses multi-scale convolution to effectively extract complex features in the time domain and frequency domain. Using the Adam optimizer, set the initial learning rate to... After each training cycle, the optimal model parameters are dynamically saved based on the model's accuracy on the validation set to establish a state recognition model.

[0013] Furthermore, using the aforementioned state recognition model, real-time state recognition is performed based on real-time data collected during the operation, including: During the surgical procedure, the robot's force, displacement, and velocity signals are collected in real time and sent to a timing data buffer with a length of L. Whenever a new frame of filtered time-series data is received, the input to the state recognition model is constructed, and forward inference is performed based on the input to obtain the current puncture state based on the inference result.

[0014] In addition, the present invention provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the robot-assisted pin placement guide needle puncture state recognition method based on interactive stiffness coefficient.

[0015] Compared with the prior art, the technical solution of the present invention has the following beneficial effects: 1. Significantly improved the accuracy of sensing the actual state of the end effector of flexible devices.

[0016] In existing technologies, force sensors are typically mounted at the end effector of a robot, measuring only the force at rigid connections. They cannot directly reflect the subtle deflections and force transmission changes that occur when the flexible guide needle tip interacts with bone tissue, leading to the loss of crucial information. This invention, however, does not rely on absolute force values. Instead, it calculates the interaction stiffness coefficient and analyzes this coefficient as a time-series signal. When the flexible guide needle tip encounters hard cortical bone and deflects, even if the robot continues to advance at a constant speed, the end effector force sensor readings exhibit a non-linear, continuously increasing trend. This unique dynamic change pattern is fully recorded in the time series of the interaction stiffness coefficient. The temporal deep learning model employed in this invention excels at capturing such complex patterns in the time dimension, accurately distinguishing between states such as "normal puncture," "flexible deflection," and "rigid collision." This is equivalent to indirectly decoding the microscopic instrument tip state from macroscopic force signals, solving the problem of existing technologies being unable to perceive the true interaction state of flexible instruments.

[0017] 2. It reduces reliance on preoperative planning and visual information, enhancing the robustness and safety of the system.

[0018] Existing technologies heavily rely on preoperative CT image-planned paths and are susceptible to interference from intraoperative tissue displacement (such as patient breathing), smoke, or blood obstruction when fusing visual information, leading to incorrect state judgments. The state recognition of this invention is entirely based on real-time physical interaction data (force and motion), independent of any preoperative planning model or external visual information. Regardless of whether the vertebral body undergoes slight displacement due to physiological activity, as long as the guide needle tip interacts with bone tissue, corresponding force feedback is generated. By analyzing these real-time feedback signals, the current physical contact state is directly determined. Therefore, even if the actual puncture point deviates from the preoperative plan, or the surgical field is poor, this invention can still reliably identify potential risks (such as contact with hard cortical bone), providing the robot with a basis for timely safe shutdown or adjustment strategies, thereby greatly improving the safety and robustness of surgery in complex and dynamic environments.

[0019] 3. By transforming the implicit experience of experts into objective mathematical models, the standardization and popularization of high-level surgical techniques have been achieved.

[0020] In traditional surgery, experienced surgeons rely on "tactile sensation" to determine the puncture status—a form of tacit knowledge that is difficult to quantify and teach. This invention, through a standardized data acquisition and expert annotation process, correlates the expert's "tactile" judgment with objective postoperative CT imaging results, and solidifies this correlation into time-series data of interaction stiffness coefficients. A deep learning model then learns from this high-quality annotated data, successfully transforming the expert's diagnostic logic into a stable and reproducible mathematical model. This not only endows the robotic system with perceptual capabilities similar to those of a seasoned expert but also provides objective evidence for training young surgeons and establishing standardized surgical procedure evaluation systems, possessing significant clinical application and educational value. Attached Figure Description

[0021] Figure 1 This is a flowchart of the robot-assisted pin placement and guide needle puncture state recognition method based on interactive stiffness coefficient of the present invention; Figure 2 This is a flowchart illustrating the processing of an embodiment of the present invention. Detailed Implementation

[0022] 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.

[0023] The main objective of this invention is to: construct an evaluation standard for the interaction stiffness between surgical instruments and bone tissue during robotic screw placement based on expert evaluation; train a neural network to directly learn the temporal and spectral characteristics of force signals under different interaction states by collecting and utilizing force feedback data annotated by experts, thereby establishing an accurate state recognition model; and utilize the constructed state recognition model to identify the interaction state between instruments and tissues in real time and robustly based on real-time interaction stiffness coefficient signals even when vertebral displacement occurs during surgery, providing a decision-making basis for surgical safety control.

[0024] The innovative aspects of this invention, taking into account the characteristics of robot-assisted screw placement surgery, are as follows: 1. A new evaluation standard definition and calculation method are proposed, namely, the "interaction stiffness coefficient" is calculated based on the real-time interaction force and real-time feed speed of the robot end effector, which is used to characterize the interaction state between surgical instruments and human tissue.

[0025] 2. A standardized experimental procedure based on specimens, combined with robots, six-dimensional force / torque sensors and postoperative CT three-dimensional reconstruction was established, and the puncture status of the guide needle was classified and labeled accordingly, serving as the gold standard for model training and validation.

[0026] This invention provides a method for identifying the puncture state of a robot-assisted pin placement guide needle based on interactive stiffness coefficient, such as... Figure 1 and Figure 2 As shown, the identification method specifically includes the following: To quantitatively describe the interaction characteristics between surgical instruments and bone tissue during robot-assisted screw placement surgery, the "interaction stiffness coefficient" is introduced as an evaluation criterion. During surgical operations, both the robot's feed speed and end effector force are readily obtainable parameters, but the interaction force values ​​vary significantly at different feed speeds. The interaction stiffness coefficient aims to eliminate the influence of speed variations and more stably reflect the ease of instrument feeding. Its definition is as follows: ; In the above formula, K TB For interactive stiffness coefficients, F TB The real-time interactive force is measured by an end force sensor. v T This refers to the real-time feed speed of the robot's gripper.

[0027] Step 1: Collect and utilize force feedback data annotated by experts; Step two: Based on the force feedback data and pre-built expert evaluation criteria, train the neural network to learn the temporal and spectral characteristics of force signals under different interaction states, and establish a state recognition model; wherein, the expert evaluation criteria are constructed in the following manner: Based on expert evaluation, an expert evaluation standard is constructed for the interaction stiffness coefficient between surgical instruments and bone tissue during screw placement; wherein the interaction stiffness coefficient is used to characterize the interaction state between the robot and the surgical process, and its expression is: ; In the above formula, K TB For interactive stiffness coefficients, F TB The real-time interactive force is measured by an end force sensor. v T This refers to the real-time feed speed of the robot's gripper.

[0028] The expert evaluation process includes the following specific steps: 1.1 Specimen Processing: Obtain at least one embalmed vertebral body specimen, wherein each specimen comprises a T12-L4 vertebra, and determine the properties of each vertebral body specimen, wherein the properties include at least one of the following: congenital malformation, fracture, tumor, and tuberculosis; remove the paravertebral muscles of each specimen, remove all soft tissues except ligaments and intervertebral discs, divide the specimen from the L2 / L3 intervertebral space, and fix the anterior column 1 / 2 of the vertebral body with acrylic bone cement to facilitate subsequent processing. At least one embalmed vertebral body specimen may be used, consisting of one, two, three, or more specimens. The properties of the vertebral body specimens can be determined by X-ray examination.

[0029] 1.2 Experimental Equipment Preparation: The experimental equipment includes the TiRobot orthopedic surgical robot system, a C-arm, a flexible guide needle, and a six-dimensional force sensor. The TiRobot orthopedic surgical robot system (National Medical Device Registration Certificate 20163542280) is used for robot-assisted vertebral screw placement. The C-arm is a Siemens 3D-C-arm. The flexible guide needle uses a standard flexible probe for spinal surgery, with a spherical tip diameter of 2mm; the flexible portion of the flexible guide needle is 50mm long and 1.5mm in diameter; the rigid portion of the flexible guide needle is 20cm long and consists of a combination of conical and cylindrical shapes with a diameter of 1.5-5mm. The six-dimensional force sensor used is an ATI-MINI45-E-5.

[0030] 1.3 Preparation before bone guide needle puncture: After fixing the bone cement with hot melt adhesive, the vertebral specimen is fixed on an X-ray-permeable operating table. The robot main control panel and optical camera are placed at the head end of the operating table, and the robot arm is placed at the tail end of the operating table. The patient tracker is fixed to the spinous process / bone cement on the head side of the specimen. The scale is installed on the robot arm and placed in the surgical area. Fluoroscopy confirms that the scale is in good position. Real-time three-dimensional images are acquired using a C-arm and transmitted to the robot main control panel.

[0031] 1.4 Path Planning and Implementation: Simulate the pedicle screw puncture path of the spine, using the normal pedicle screw insertion point as the test puncture point, and adjust the guide needle angle to penetrate 2mm of the medial cortical bone of the pedicle; replace the end of the robotic arm with a flexible guide needle puncture tool equipped with a six-dimensional force sensor; control the robotic arm to guide the flexible guide needle to determine the planned screw puncture point, grind away the cortical bone within a 2mm radius centered on the insertion point, and use the flexible guide needle to confirm that the cortical bone grinding meets the puncture requirements; control the robotic arm to puncture the cancellous bone at a speed of 0.2mm / s, record the force sensor data, the actual displacement and speed of the robotic arm end, until the penetration depth reaches more than 45mm or the robotic arm stops due to excessive resistance (theoretical load of the robotic arm is 5kg), and record whether the robotic arm has completed the puncture normally.

[0032] 1.5 Postoperative CT: After the puncture is completed, CT data of specimens with and without guide needles are obtained by C-arm scanning.

[0033] 1.6 Postoperative Evaluation: Based on the DICOM data from postoperative CT scans, three-dimensional reconstruction was performed to obtain the puncture path of the flexible guide needle. The puncture status of the guide needle was evaluated in conjunction with whether the robotic arm successfully completed the puncture. The postoperative evaluation results include: puncture through the cancellous bone path, i.e., "normal puncture"; the flexible guide needle deflects flexibly after encountering the cortical bone and completes the puncture in the cancellous bone, i.e., "flexible deflection"; the flexible guide needle encountering the cortical bone and being unable to compensate, resulting in an error from the robotic arm and failure to complete the puncture, i.e., "rigid collision".

[0034] Based on the force feedback data and pre-built expert evaluation criteria, the specific process of training a neural network to learn the temporal and spectral characteristics of force signals under different interaction states and establishing a state recognition model is as follows: 2.1 Acquire the force sensor data, actual displacement and velocity of the robotic arm end effector from the force feedback data, that is, the data collected in step 1.4 above (force sensor data, actual displacement and velocity of the robotic arm end effector).

[0035] 2.2. The time-series data of the six-dimensional interaction stiffness coefficients in each CSV file and the corresponding state labels are sliced ​​using a sliding window of length L and encapsulated into training samples; each sample X i It contains a time series data and corresponds to a state label. Y i ; 2.3 A multi-layer temporal network structure based on TimesNet, wherein the input is 6-channel temporal data, the network contains E coding layers, and each coding layer uses multi-scale convolution to effectively extract complex features in the time domain and frequency domain; 2.4. Use the Adam optimizer and set the initial learning rate. After each training cycle, the optimal model parameters are dynamically saved based on the model's accuracy on the validation set to establish a state recognition model.

[0036] Step 3: Using a status recognition model, real-time status recognition is performed based on real-time data collected during the operation. The specific process includes: During the surgical procedure, the robot's force, displacement, and velocity signals are collected in real time and sent to a time-series data buffer with a length of L. Whenever a new frame of filtered time-series data is received, an input to the state recognition model is constructed, and forward inference is performed based on the input to obtain the current puncture state based on the inference result.

[0037] The above-mentioned robot-assisted pin placement and guide needle puncture state recognition method based on interactive stiffness coefficient has the following advantages: 1. Significantly improved the accuracy of sensing the actual state of the end effector of flexible devices.

[0038] In existing technologies, force sensors are typically mounted at the end effector of a robot, measuring only the force at rigid connections. They cannot directly reflect the subtle deflections and force transmission changes that occur when the flexible guide needle tip interacts with bone tissue, leading to the loss of crucial information. This invention, however, does not rely on absolute force values. Instead, it calculates the interaction stiffness coefficient and analyzes this coefficient as a time-series signal. When the flexible guide needle tip encounters hard cortical bone and deflects, even if the robot continues to advance at a constant speed, the end effector force sensor readings exhibit a non-linear, continuously increasing trend. This unique dynamic change pattern is fully recorded in the time series of the interaction stiffness coefficient. The temporal deep learning model employed in this invention excels at capturing such complex patterns in the time dimension, accurately distinguishing between states such as "normal puncture," "flexible deflection," and "rigid collision." This is equivalent to indirectly decoding the microscopic instrument tip state from macroscopic force signals, solving the problem of existing technologies being unable to perceive the true interaction state of flexible instruments.

[0039] 2. It reduces reliance on preoperative planning and visual information, enhancing the robustness and safety of the system.

[0040] Existing technologies heavily rely on preoperative CT image-planned paths and are susceptible to interference from intraoperative tissue displacement (such as patient breathing), smoke, or blood obstruction when fusing visual information, leading to incorrect state judgments. The state recognition of this invention is entirely based on real-time physical interaction data (force and motion), independent of any preoperative planning model or external visual information. Regardless of whether the vertebral body undergoes slight displacement due to physiological activity, as long as the guide needle tip interacts with bone tissue, corresponding force feedback is generated. By analyzing these real-time feedback signals, the current physical contact state is directly determined. Therefore, even if the actual puncture point deviates from the preoperative plan, or the surgical field is poor, this invention can still reliably identify potential risks (such as contact with hard cortical bone), providing the robot with a basis for timely safe shutdown or adjustment strategies, thereby greatly improving the safety and robustness of surgery in complex and dynamic environments.

[0041] 3. By transforming the implicit experience of experts into objective mathematical models, the standardization and popularization of high-level surgical techniques have been achieved.

[0042] In traditional surgery, experienced surgeons rely on "tactile sensation" to determine the puncture status—a form of tacit knowledge that is difficult to quantify and teach. This invention, through a standardized data acquisition and expert annotation process, correlates the expert's "tactile" judgment with objective postoperative CT imaging results, and solidifies this correlation into time-series data of interaction stiffness coefficients. A deep learning model then learns from this high-quality annotated data, successfully transforming the expert's diagnostic logic into a stable and reproducible mathematical model. This not only endows the robotic system with perceptual capabilities similar to those of a seasoned expert but also provides objective evidence for training young surgeons and establishing standardized surgical procedure evaluation systems, possessing significant clinical application and educational value.

[0043] This invention also provides a computer-readable storage medium storing a computer program, which is executed by a processor to implement the robot-assisted pin placement guide needle puncture state recognition method based on interactive stiffness coefficient in the above embodiments.

[0044] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of the present invention can be implemented using various computer languages, such as the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0045] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0046] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0047] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0048] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0049] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

[0050] Obviously, those skilled in the art can make various modifications and variations to the embodiments of the present invention without departing from the spirit and scope of the invention. Therefore, if these modifications and variations fall within the scope of the claims of the present invention and their equivalents, the present invention also intends to include these modifications and variations.

[0051] In summary, 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 identifying the state of a robot-assisted drill guide needle puncture based on an interaction stiffness coefficient, characterized by, The method comprises the following steps: Collect and use expert-labeled force feedback data; Based on the force feedback data and pre-constructed expert evaluation criteria, train a neural network to learn the time sequence and frequency spectrum characteristics of force signals under different interaction states, and establish a state recognition model; Using the state recognition model, real-time state recognition is performed according to real-time data collected during the operation; Wherein the expert evaluation criteria are constructed as follows: based on expert evaluation, the expert evaluation criteria of the interaction stiffness coefficient between the surgical instrument and the bone tissue during the screwing process are constructed; wherein the interaction stiffness coefficient is used to represent the interaction state between the robot and the surgical process, and its expression is: ； In the above formula, K TB is the interaction stiffness coefficient, F TB is the real-time interaction force measured by the end force sensor, v T is the real-time feed speed of the robot to clamp the instrument.

2. The identification method of claim 1, wherein, The method comprises the following steps: Specimen processing: obtain at least one preserved vertebral body specimen, wherein each specimen in the at least one preserved vertebral body specimen is a T12-L4 vertebral body, and the properties of each vertebral body specimen are determined, wherein the properties at least include one of the following: congenital deformity, fracture degree, tumor and tuberculosis disease; the paravertebral muscle of each specimen is cut off, all soft tissues outside the ligament and intervertebral disc are removed, the L2 / L3 intervertebral space is segmented, and 1 / 2 of the anterior column of the vertebral body is fixed by using acrylic bone cement; Test equipment preparation: the test equipment includes a Tianji orthopedic surgical robot system, a C-arm, a flexible guide needle and a six-dimensional force sensor; Specimen bone guide needle puncture preparation: the bone cement fixed vertebral body specimen is fixed on a radiolucent operating bed, a robot main console and an optical camera are arranged at the head end of the operating bed, and a robot mechanical arm is arranged at the tail end of the operating bed; a patient tracker is fixed on the head side of the spinous process / bone cement of the specimen, a mechanical arm scale is arranged in the operation area, and the scale position is determined to be good under fluoroscopy; real-time three-dimensional images are obtained by using a C-arm and transmitted to a robot main console; Path planning and implementation: a spinal pedicle screw puncture path is simulated, a normal pedicle screw puncture entry point is used as a test puncture point, the angle of the guide needle is adjusted to be 2mm from the inner cortical bone of the pedicle, the end of the robot mechanical arm is replaced with a flexible guide needle puncture tool provided with a six-dimensional force sensor, the robot mechanical arm is controlled to guide the flexible guide needle to determine the planned screw puncture entry point, the cortical bone within a range of 2mm from the entry point is removed, the removal of the cortical bone is confirmed by using the flexible guide needle to meet the puncture requirements, the robot mechanical arm is controlled to puncture the cancellous bone, and the force sensor data, the actual displacement and speed of the end of the mechanical arm are recorded until the depth reaches 45mm or more or the mechanical arm stops due to excessive resistance, and whether the mechanical arm successfully completes the puncture is recorded; Postoperative CT: after the puncture is completed, CT data of the specimen containing the guide needle and the puncture without the guide needle are obtained by using a C-arm; Postoperative evaluation: based on the dicom data of the postoperative CT, three-dimensional reconstruction is performed to obtain the flexible guide needle puncture path, and the puncture state of the guide needle is evaluated in combination with whether the mechanical arm successfully completes the puncture.

3. The identification method of claim 2, wherein, The evaluation results of the postoperative evaluation include: cancellous bone path puncture; the flexible guide needle meets the cortical bone, flexibly deflects in the cancellous bone and completes the puncture; the flexible guide needle cannot compensate after meeting the cortical bone, the mechanical arm of the robot reports an error and cannot complete the puncture.

4. The identification method of claim 2, wherein, The bone cement fixed vertebral body specimen is fixed on the operating bed by using hot melt adhesive; the mechanical arm is controlled to puncture the cancellous bone at a speed of 0.2mm / s.

5. The identification method of claim 2, wherein, The C-arm is a 3D-C-arm. The flexible needle is a flexible probe used in spine surgery.

6. The identification method of claim 2, wherein, Based on the force feedback data and the pre-constructed expert evaluation standard, the neural network learns the time sequence and spectral characteristics of the force signal under different interaction states, establishes a state recognition model, including: Obtaining force sensor data, actual displacement and speed of the end of the robot arm in the force feedback data; The time series data of the six-dimensional interaction stiffness coefficients in each CSV file and the corresponding state label are sliced by a sliding window with a length of L to encapsulate training samples; each sample X i contains a segment of time series data and corresponds to a state label Y i ; Based on the multi-layer time sequence network structure of TimesNet, the input is 6-channel time sequence data, the network contains E encoding layers, and multi-scale convolution is used to effectively extract complex features in time domain and frequency domain in each encoding layer; Using Adam optimizer, set the initial learning rate to At the end of each training cycle, the optimal model parameters are dynamically saved according to the accuracy of the model on the validation set, and the state recognition model is established.

7. The identification method of claim 6, wherein, Using the state recognition model, real-time state recognition is performed according to real-time data collected during the operation, including: When performing the operation, the force, displacement and speed signals of the robot are collected in real time and sent to a time sequence data buffer, and the length of the time sequence data buffer is L; Whenever a new time sequence data after filtering is received, the input of the state recognition model is constructed, and forward inference is performed according to the input, and the current puncture state is obtained according to the inference result.

8. A computer-readable storage medium having stored thereon a computer program, characterized in that, The program is executed by the processor to implement the steps of the robot-assisted needle placement puncture state recognition method based on the interaction stiffness coefficient in any one of claims 1-7.

Images