A percutaneous spinal puncture positioning system based on robotic ultrasound scan imaging

Abstract

This invention discloses a percutaneous spinal puncture positioning system based on robotic ultrasound imaging, comprising: a collaborative robot, a medical ultrasound machine, a motion control module, and an image processing system. The image processing system further includes a CT image-based three-dimensional spinal reconstruction module, an ultrasound data processing module, and a registration and fusion module. This invention can reconstruct a CT three-dimensional spinal model from thin-slice CT images, obtain an ultrasound three-dimensional spinal model through ultrasound image processing, and register and fuse the two models into a unified coordinate system. Using this invention, a "three-dimensional positioning + two-dimensional monitoring" navigation puncture positioning mode can be achieved, fully leveraging the advantages of robotic ultrasound imaging, minimizing labor costs and reliance on experience, and safely and reliably completing percutaneous spinal puncture positioning. This invention requires minimal consumables, provides intuitive results, and is radiation-free, making it suitable for widespread use.

Description

Technical Field

[0001] This invention relates to the fields of medicine and computers, and specifically to a percutaneous spinal puncture positioning system based on robotic ultrasound scanning imaging. Background Technology

[0002] The spine, with its irregular bone structure and multiple interconnected segments, serves as the body's weight-bearing "backbone." It houses the spinal cord, which processes and transmits nerve impulses, and is covered by a complex layer of back muscles, creating a complex musculoskeletal environment. Degenerative diseases occurring here, such as herniated discs, have seen an increasing incidence in recent years due to changes in lifestyle, and are now common clinical conditions. Tumors, such as intramedullary tumors, present extremely high surgical challenges due to the need for complete tumor removal, preservation of nerve function, and maintenance of spinal stability—a long-standing and difficult clinical problem.

[0003] In recent years, with the growing acceptance of the concept of "minimally invasive surgery," both doctors and patients hope to achieve surgical goals to the greatest extent possible through small incisions, short pathways, and precise exposure, such as complete tumor resection or adequate nerve decompression. Therefore, for spinal cord surgery, the planning of incisions and approaches is paramount, and one of the main reasons for the high technical threshold and steep learning curve of minimally invasive surgery. This clinical "pain point" presents both a strong demand and a significant challenge for the innovation and development of intraoperative online precision navigation puncture positioning technology. Taking percutaneous spinal endoscopy, a representative technique in minimally invasive spinal surgery, as an example, segmental localization and precise puncture target location are essential preoperative steps, requiring a positioning error of <1cm. To meet surgical positioning needs, there are clinically recognized and widely used positioning methods, as well as various emerging technologies and equipment.

[0004] Firstly, intraoperative fluoroscopy or CT-assisted localization using a C-arm X-ray machine or an O-arm CT scanner is currently the mainstream clinical method for intraoperative localization. The process can be summarized as: "anteroposterior and lateral fluoroscopy (CT imaging) – insertion of the puncture needle – repeated fluoroscopy (CT imaging) to determine the puncture needle position (– adjustment of the puncture needle – re-verification) – localization completed." This method of localization has the advantages of relatively simple operation and relatively high accuracy. However, the non-real-time nature of the imaging is a major drawback. It also relies heavily on experience and has high equipment requirements. Furthermore, due to the radiation characteristics of X-rays and CT scans, it causes significant radiation damage to patients and doctors and nurses who frequently perform such surgeries, thus shortening the professional lifespan of the surgical team to some extent.

[0005] Secondly, navigation and positioning technologies, represented by various navigation devices and robots, generally integrate positioning systems (optical or magnetic positioning devices), robots, navigation systems, image registration, and approach planning systems. They are primarily designed for fusion surgeries involving spinal degeneration and instability requiring screw placement. These technologies can effectively improve the accuracy of pedicle screw placement, achieving an astonishing sub-millimeter level (≤1mm). However, the invasive bone marker linking required by these technologies does not align with the minimally invasive concept, and they also suffer from long registration times, cumbersome preliminary preparations, and expensive equipment.

[0006] Third, navigation and positioning technology based on ultrasound scanning imaging. Ultrasound has the advantages of real-time imaging and safety without radiation, and has been widely used in clinical diagnosis and guided puncture positioning. However, due to sound wave attenuation, refraction, and distortion in ultrasound imaging, bone identification in the images is difficult. Continuous ultrasound images cannot directly reconstruct an effective three-dimensional spinal bone model for clinical guidance through threshold selection and voxel reconstruction, as is the case with CT. To solve this problem, the emerging multimodal image matching and fusion technology has emerged in recent years. This technology matches preoperative CT or MRI images in two-dimensional or three-dimensional form with intraoperative ultrasound images to help fill in missing information in the ultrasound images, thereby improving the visualization and positioning accuracy of navigation guidance. However, in practical applications, this strictly limits the ultrasound scanning path, sacrificing the original advantage of flexible imaging. Furthermore, due to the similarity of spinal segment images and the varying quality of ultrasound images, there are problems such as high registration difficulty and large positioning errors, resulting in limited universality and reliability of this technology. In addition, there are methods to extract features from three-dimensional ultrasound spinal body data, such as extracting the sagittal profile of the spine, the three-dimensional curvature of the spinous process and transverse process to identify segments or assess the scoliosis angle. However, the measurement error of the imaging surface deformation is large and cannot be accurately estimated. The degree of visualization is limited and there is a large difference from the conventional images used by doctors.

[0007] In summary, intraoperative navigation and localization techniques for spinal cord and spinal cord-related diseases have developed rapidly. Ultrasound technology, due to its unique characteristics of flexible imaging, real-time operation, and radiation-free operation, shows great research and application potential. However, ultrasound imaging also has drawbacks, including poor bone tissue imaging, difficulty in identifying signs, and reliance on experience in scanning and image interpretation. Organically combining ultrasound imaging technology, robotics, and image processing technology brings new ideas to its application, thereby leveraging the characteristics and advantages of ultrasound scanning imaging, reducing labor costs and reliance on experience, and achieving high-precision, safe, and reliable percutaneous spinal puncture localization, replacing traditional X-ray localization methods and achieving the goal of intraoperative puncture localization. Summary of the Invention

[0008] To address the shortcomings of existing systems, the technical problem this invention aims to solve is to provide a percutaneous spinal puncture positioning system based on robotic ultrasound scanning imaging, innovatively proposing a "three-dimensional positioning + two-dimensional monitoring" puncture positioning mode. "Three-dimensional positioning" refers to establishing a surgical coordinate system adapted to the intraoperative position through automated bone surface image segmentation and three-dimensional reconstruction during the scanning process, registered with a CT spinal three-dimensional model, thus providing intuitive and precise navigation guidance for the operator. "Two-dimensional monitoring" involves the robot controlling the movement of the ultrasound probe, making the puncture target point visible on the ultrasound image, with real-time monitoring of the entire puncture process using two-dimensional ultrasound images. This puncture positioning mode offers excellent accuracy while maximizing safety and reliability during the puncture process, significantly reducing reliance on experience.

[0009] To address the technical problem, this invention proposes a percutaneous spinal puncture localization system based on robotic ultrasound scanning imaging, comprising:

[0010] A collaborative robot having at least one robotic arm with several joints for moving an integrated support at the end of the robotic arm in three-dimensional space.

[0011] A medical ultrasound machine, wherein the ultrasound probe of the medical ultrasound machine is mounted on an integrated bracket at the end of a robotic arm, for use in adaptive ultrasound scanning during surgery to acquire ultrasound data adapted to the patient's position during surgery; and the ultrasound probe is also used to provide real-time ultrasound images during procedures such as puncture.

[0012] The motion control module is used to control the movement of the integrated support on the end of the robotic arm, so that the ultrasound probe can perform adaptive ultrasound scanning according to the scanning path and probe posture preset by the motion control module during the operation, and control the ultrasound probe to reach the vicinity of the target puncture point and keep it pressed against the skin during the operation.

[0013] The image processing system is capable of acquiring ultrasound image data obtained from a medical ultrasound machine. The image processing system includes a CT image spinal three-dimensional reconstruction module, an ultrasound data processing module, and a registration and fusion module.

[0014] Among them, the CT image spine 3D reconstruction module is used to acquire the patient's preoperative thin-slice CT image data, set the CT threshold range to separate and extract human bone signals, and output the reconstructed bone model based on voxel drawing 3D reconstruction output after extraction. After removing impurity signals, the CT spine model is obtained and saved.

[0015] The ultrasound data processing module uses a neural network-based image processing algorithm to identify and segment bone surface image regions in real time. It uses the VTK open-source library to perform real-time voxel-based volume rendering on the segmented bone surface ultrasound image sequence at known locations and render the images to obtain an ultrasound three-dimensional model of the spine that is adapted to the patient's intraoperative position.

[0016] The registration and fusion module acquires the CT spine model from the CT image spine 3D reconstruction module and the ultrasound spine 3D model from the ultrasound data processing module. Using the ultrasound spine 3D model as the base coordinate, the two models are registered to establish a unified surgical coordinate system.

[0017] Furthermore, the slice thickness of the thin-slice CT obtained by the CT image spinal three-dimensional reconstruction module should be less than 1.0 mm; the CT threshold should be selected to display only bone density and not other tissue densities; the CT spinal model should be saved at a resolution of 1*1*1 mm.

[0018] Furthermore, the CT threshold range should be >400 HU.

[0019] Furthermore, the motion control module, robotic arm, and ultrasonic probe need to be calibrated before acquiring ultrasonic images to eliminate the time difference between sampling and transmission. The calibration is divided into time calibration and spatial calibration. The time calibration accuracy is less than 0.01s, and the spatial calibration accuracy error must be <0.5mm.

[0020] Furthermore, when registering two models, the registration and fusion module performs manual point selection under the guidance of anatomical segment landmarks, with ≥3 manually selected points. Registration is performed based on point features and the ICP algorithm, and rigid transformation is performed with the ultrasound three-dimensional spinal model as the reference, with a registration error of <1mm.

[0021] Furthermore, the ultrasonic probe is capable of imaging at a depth of 50-90 mm.

[0022] Furthermore, the integrated stent is also equipped with an adaptive clamping follow-up device and an auxiliary puncture device;

[0023] The adaptive clamping follow-up device is used to maintain a good relative position and posture between the ultrasound probe and the patient's body during the scanning process to ensure scanning safety and obtain high-quality ultrasound images.

[0024] The auxiliary puncture stent is used to assist doctors in inserting the puncture needle along a prescribed path and to display the insertion depth of the puncture needle.

[0025] Compared with existing technologies, this invention has the following advantages and beneficial effects:

[0026] This invention can replace traditional intraoperative X-ray fluoroscopy or CT imaging-assisted positioning systems. It innovatively proposes a "three-dimensional positioning + two-dimensional monitoring" navigation puncture positioning mode, fully leveraging the advantages of robotic ultrasound scanning imaging to minimize labor costs and reliance on experience, ensuring safe and reliable percutaneous spinal puncture positioning. The robotic arm's end effector holds the ultrasound probe, giving ultrasound scanning powerful mobility and real-time image pose information. Combined with the U-Net convolutional neural network image segmentation algorithm, it achieves automated segmentation of the bone surface region in ultrasound images, making three-dimensional bone imaging via ultrasound possible, initially achieving clinical guidance effects comparable to CT. The unified surgical coordinate system obtained through registration and fusion with the CT-reconstructed spinal model, adapted to the patient's intraoperative position, is like giving the surgeon "X-ray vision glasses," freeing them from complex spatial imagination and information integration, greatly reducing reliance on clinical experience. The image processing system's registration of CT-reconstructed and ultrasound-reconstructed 3D models based on image feature points greatly avoids segmental registration errors and helps surgeons develop a global anatomical view based on intraoperative positioning, thus improving positioning accuracy. During the puncture process, puncture is performed under the assistance of a puncture frame and continuous ultrasound image monitoring, maximizing the safety and reliability of the puncture process without the need for secondary verification. This invention's system can be applied to positioning in open and minimally invasive spinal cord surgeries, and is particularly suitable for surgical positioning and pathway planning in minimally invasive spinal cord surgeries, such as percutaneous spinal endoscopy. It is also suitable for clinical scenarios such as spinal anesthesia assistance, bedside lumbar puncture positioning, and clinical teaching demonstrations. The technology and equipment used in this invention require low economic investment, no expensive consumables, provide intuitive results, have a high degree of automation, pose no radiation hazards, and ensure safe and accurate puncture positioning, making it suitable for widespread use. Attached Figure Description

[0027] Figure 1 This is a flowchart illustrating the application of the system of the present invention;

[0028] Figure 2 A schematic diagram of a CT spinal model acquired before surgery; A in the diagram represents the posterior view, and B represents the lateral view.

[0029] Figure 3 Schematic diagram of the ultrasound spinal model acquired during surgery; A in the figure represents the rear view and B represents the side view;

[0030] Figure 4 Schematic diagram of registration and fusion of CT spinal model and ultrasound spinal model; A in the figure represents rear view and B represents side view;

[0031] Figure 5 This is a schematic diagram of the robotic arm of the system of the present invention and the process of puncturing the articular processes of the target segment.

[0032] In the figure, 1. Costotransverse joint, 2. Lamine, 3. Vertebral body, 4. Spinous process, 5. Transverse process, 6. Rib, 7. Interlaminar space, 8. Costovertebral angle, 9. Robotic arm, 10. Puncture assist device, 11. Ultrasound image, 12. Predicted puncture needle path. Detailed Implementation

[0033] The present invention will be further described and illustrated below with reference to specific embodiments. The embodiments described are merely examples of the content of this disclosure and do not limit the scope of the invention. The technical features of each embodiment in the present invention can be combined accordingly, provided that there is no mutual conflict.

[0034] This embodiment presents a percutaneous spinal puncture positioning system based on robotic ultrasound scanning imaging, which comprises four main components: a collaborative robot, a medical ultrasound machine, a motion control module, and an image processing system.

[0035] Among them, such as Figure 5 As shown, the collaborative robot has at least one robotic arm with multiple joints, which possesses high-precision end-effector position acquisition capabilities. It is connected to a motion control module to receive control and provides real-time feedback of the end-effector position to the motion control module. Through the coordination of the joints, the integrated support on the end of the robotic arm can move in three-dimensional space. In one embodiment of the invention, the collaborative robot is a 6DPF UR collaborative robot.

[0036] A medical ultrasound machine is used to acquire ultrasound images and is an existing device in medical systems. The ultrasound probe of the medical ultrasound machine is mounted on an integrated support at the end of a robotic arm for adaptive ultrasound scanning during surgery, acquiring ultrasound data adapted to the patient's intraoperative position. The ultrasound probe also provides real-time ultrasound images during surgery.

[0037] The motion control module is mounted on the computer and is the main control module of this system. It is used to control the movement of the integrated support on the end of the robotic arm, so that the ultrasound probe can perform adaptive ultrasound scanning according to the scanning path and probe posture preset by the robotic arm navigation module during the operation, and to control the ultrasound probe to reach the vicinity of the target puncture point and press the skin during the operation.

[0038] The image processing system, mounted on a computer, includes a CT image spinal 3D reconstruction module, an ultrasound data processing module, and a registration and fusion module. The CT image spinal 3D reconstruction module is used to acquire preoperative thin-slice CT images of the patient, set a CT threshold range to separate and extract human bone signals, output a reconstructed bone model based on voxel-based 3D volume drawing, remove impurity signals from the reconstructed bone model, obtain a CT spinal model, and save it.

[0039] The ultrasound data processing module uses a neural network-based image segmentation algorithm to identify and segment bone surface image regions in real time. It uses the VTK open-source library to perform real-time voxel-based volume rendering on the segmented bone surface ultrasound image sequence at known locations and render the images to obtain an ultrasound spine model that adapts to the patient's intraoperative position.

[0040] The registration and fusion module acquires the CT spine model from the CT image spine 3D reconstruction module and the ultrasound spine model from the ultrasound data processing module. Using the ultrasound spine model as the base coordinate, the two models are registered to establish a unified surgical coordinate system.

[0041] In a preferred embodiment, the integrated stent is further provided with an adaptive clamping follow-up device and an auxiliary puncture device; the adaptive clamping follow-up device is used to maintain a good relative position and posture between the ultrasound probe and the patient's body during the scanning process to ensure scanning safety and obtain high-quality ultrasound images; the auxiliary puncture stent is used to assist the doctor in inserting the puncture needle along the prescribed path and to display the insertion depth of the puncture needle.

[0042] like Figure 1 As shown, the procedure for radiation-free percutaneous spinal puncture localization using the system of the present invention based on robotic ultrasound scanning imaging can be as follows:

[0043] S1. Preoperative CT Image-Based Three-Dimensional Spinal Reconstruction Process

[0044] 1.1: Obtain the patient's preoperative thin-section CT image data, set the CT threshold range to separate and extract human bone signals, and output the reconstructed bone model based on voxel-based rendering.

[0045] 1.2: Remove unwanted signals other than the spine and some ribs, and save as a CT spine model at an appropriate resolution, such as... Figure 2 As shown, A and B are the dorsal and right lateral views of the CT spinal model, respectively. By identifying anatomical markers on the model (usually the spinous process 4, transverse process 5, and interlaminar space 7), the spinal segment localization, incision marking, or puncture localization can be guided.

[0046] S2. Robotic Ultrasonic Scanning Imaging and 3D Spinal Reconstruction Process

[0047] 2.1: Determine the patient's intraoperative position according to the surgical requirements and properly fix the body to prevent the long axis of the body from moving during the positioning process;

[0048] 2.1: After the instrument is assembled, it needs to be calibrated. Only after the calibration accuracy is satisfactory can the subsequent steps be carried out. The ultrasonic probe and the adaptive ultrasonic scanning auxiliary device are mounted on the integrated bracket at the end of the robotic arm. The robotic arm drives the scanning device to perform the scanning. The adaptive clamping follow-up device can be the adaptive clamping follow-up device disclosed in CN201910277733.2.

[0049] 2.2: Based on surgical needs, the back contour, and the location of the lesion target, plan the scanning path and probe position. The scanning path should include at least one anatomical segment landmark (select the nearest one, such as the costovertebral angle or lumbosacral junction). Adjust the probe position according to the intraoperative position. After scanning, the robot moves the probe to naturally lift it off the skin surface, and then hovers and locks it in the set safe area.

[0050] 2.3: Real-time force control monitoring of the robot-assisted dynamic adaptive ultrasound scanning, during which a neural network-based image segmentation algorithm is used to identify and segment bone surface image regions in real time;

[0051] 2.4: The VTK (Visualization Toolkit) open-source library was used to perform real-time voxel-based volume rendering on segmented ultrasound image sequences of known locations on the bone surface. Appropriate rendering modes and indices were selected to obtain an ultrasound spinal model adapted to the patient's intraoperative position, such as... Figure 3 As shown, clearly identifiable anatomical landmarks that correspond to the CT spinal model can be used as point features in 2.5, including but not limited to costotransverse joint 1, transverse process 5, costovertebral angle 8, etc.

[0052] 2.5: Import the CT spinal model obtained from S1. Guided by anatomical segment landmarks, manually select points and then perform registration based on point features, supplemented by the Iterative Closest Point (ICP) algorithm. Establish a unified surgical coordinate system using the ultrasound spinal model as the base coordinate. Figure 4 The diagram shows the registration and fusion of the CT spinal model and the ultrasound spinal model. The costovertebral angle 8 is a point feature selection.

[0053] S3. Surgical coordinate system-guided puncture positioning procedure

[0054] 3.1: The surgeon identifies the lesion segment on the surgical coordinate system and further confirms the target puncture point within the same system. Once selected, the exact location of this point on the surgical coordinate system is obtained. After the robot moves the probe to the vicinity of the target point, it slowly descends and presses against the skin until the puncture target point falls within a suitable area on the real-time ultrasound image plane.

[0055] 3.2: For example Figure 5As shown, the puncture assist device 10, which is also mounted on the integrated bracket at the end of the robotic arm 9, plans the puncture angle and skin incision. It is expected that after the tip of the puncture needle coincides with the puncture target point on the real-time ultrasound image 11, the surgeon will insert the puncture needle percutaneously under the full guidance of the real-time ultrasound image 11 and perform puncture along the predicted puncture needle path 12.

[0056] 3.3: Fix the puncture needle, remove the robot, and perform enlargement or skin incision as needed for the surgery, followed by subsequent surgical procedures.

[0057] In a preferred embodiment, to obtain a more accurate CT-reconstructed three-dimensional spinal model, the slice thickness of the thin-slice CT in step 1.1 is required to be 1.0 mm or less; the CT threshold range should be sufficient to display only bone density and not other tissue densities. More preferably, the CT threshold range in step 1.1 should be >400 HU; some bone that affects the field of vision can be removed according to surgical needs.

[0058] In a preferred embodiment, the storage resolution in step 1.2 is selected as 1*1*1mm.

[0059] In a preferred embodiment, step 2.1 involves eliminating the time difference between sampling and transmission, and converting the ultrasonic image coordinate system to the robot base coordinate system. A strict calibration process must be completed before scanning, which can be divided into time calibration and spatial calibration. The time calibration accuracy is less than 0.01s, and the spatial calibration accuracy error must be <0.5mm.

[0060] In step 2.2, the intelligent planning of the scanning path includes at least one anatomical segment landmark. Based on the location of the target point, the nearest costovertebral angle or lumbosacral junction is selected, and this guides the subsequent manual selection of points for registration.

[0061] During the scanning process in step 2.3, the recognition and segmentation of bone surface image regions are automatically performed based on the U-Net convolutional neural network. The recognition and segmentation frame rate is greater than the sampling frame rate, generally greater than 20fps.

[0062] In step 2.4 of the three-dimensional reconstruction process, the VTK open-source program library is used to perform real-time voxel-based volume rendering on the segmented ultrasound image sequence of the bone surface at known locations, and output a three-dimensional ultrasound spine model.

[0063] In step 2.5, during the registration and fusion process, points are manually selected under the guidance of anatomical segment landmarks (costovertebral angle or lumbosacral junction), with a number of ≥3 points. Registration is performed based on point features and the ICP algorithm, and rigid transformation is performed based on the ultrasound three-dimensional model of the spine. The registration error is <1mm.

[0064] The medical ultrasound machine can acquire ultrasound images and send them to the image processing system. The equipped ultrasound probe is capable of imaging at a depth of 50-90mm. The image processing system can import thin-slice CT image data (DICOM format data), reconstruct and output a CT-reconstructed three-dimensional spinal model (.stl or .obj format), and can edit the three-dimensional spinal model to remove other impurity signals besides bone. It can import ultrasound images from the medical ultrasound machine and robotic arm end-effector position information from the robotic arm, and convert them according to the calibrated 4*4 rotation matrix. It can achieve registration and fusion of the CT three-dimensional spinal model and the ultrasound three-dimensional spinal model based on point features and the ICP algorithm, using the ultrasound three-dimensional spinal model as the base coordinate to unify the coordinate system.

[0065] This invention can replace traditional intraoperative X-ray fluoroscopy or CT imaging-assisted positioning systems. It can be applied to the positioning of open and minimally invasive spinal cord surgeries, and is particularly suitable for surgical positioning and pathway planning in minimally invasive spinal cord surgeries represented by percutaneous spinal endoscopy. It is also suitable for clinical scenarios such as spinal anesthesia assistance, bedside lumbar puncture positioning, and clinical teaching demonstrations.

[0066] The above-described embodiments are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. Those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention.

Claims

1. A percutaneous spinal puncture positioning system based on robotic ultrasound scanning imaging, characterized in that, include: A collaborative robot having at least one robotic arm with several joints for moving an integrated support at the end of the robotic arm in three-dimensional space. A medical ultrasound machine, wherein the ultrasound probe of the medical ultrasound machine is mounted on an integrated bracket at the end of a robotic arm, for use in adaptive ultrasound scanning during surgery to acquire ultrasound data adapted to the patient's position during surgery; and the ultrasound probe is also used to provide real-time ultrasound images during surgery. The motion control module is used to control the movement of the integrated support on the end of the robotic arm, so that the ultrasound probe can perform adaptive ultrasound scanning according to the scanning path and probe posture preset by the motion control module during the operation, and control the ultrasound probe to reach the vicinity of the target puncture point and press the skin during the operation. The image processing system is capable of acquiring ultrasound data obtained from a medical ultrasound machine. The image processing system includes a CT image spinal three-dimensional reconstruction module, an ultrasound data processing module, and a registration and fusion module. Among them, the CT image spine three-dimensional reconstruction module is used to acquire the patient's preoperative thin-slice CT image data, set the CT threshold range to separate and extract human bone signals, and output the reconstructed bone model based on voxel three-dimensional volume drawing after extraction. Impurity signals are removed from the reconstructed bone model to obtain and save the CT spine model. The ultrasound data processing module uses a neural network-based image segmentation algorithm to identify and segment bone surface image regions in real time. It uses the VTK open-source library to perform real-time voxel-based volume rendering on the segmented bone surface ultrasound image sequence at known locations and render the images to obtain an ultrasound three-dimensional model of the spine that is adapted to the patient's intraoperative position. The registration and fusion module acquires the CT spine model from the CT image spine 3D reconstruction module and the ultrasound spine 3D model from the ultrasound data processing module. Using the ultrasound spine 3D model as the base coordinate, the two models are registered to establish a unified surgical coordinate system. The slice thickness of the thin-slice CT obtained by the CT image spinal 3D reconstruction module should be less than 1.0 mm; the CT threshold should be selected to display only bone density and not other tissue densities; the resolution of the CT spinal model should be 1*1*1 mm. The ultrasound data processing module uses a U-Net convolutional neural network to automatically identify and segment bone surface image regions, with the identification and segmentation frame rate being greater than the sampling frame rate. When registering two models, the registration and fusion module performs manual point selection under the guidance of anatomical segment landmarks, with a manual selection of ≥3 points. Registration is performed based on point features and the ICP algorithm, and rigid transformation is performed with the ultrasound three-dimensional spinal model as the reference. The registration error is <1mm.

2. The percutaneous spinal puncture positioning system based on robotic ultrasound scanning imaging according to claim 1, characterized in that, The CT threshold range should be >400 HU.

3. The percutaneous spinal puncture positioning system based on robotic ultrasound scanning imaging according to claim 1, characterized in that, The motion control module, robotic arm, and ultrasonic probe need to be calibrated before acquiring ultrasonic images to eliminate the time difference between sampling and transmission. The calibration is divided into time calibration and spatial calibration. The time calibration accuracy is less than 0.01s, and the spatial calibration accuracy error must be <0.5mm.

4. The percutaneous spinal puncture positioning system based on robotic ultrasound scanning imaging according to claim 1, characterized in that, The motion control module's preset scanning path includes at least one anatomical segment landmark during the scanning process.

5. The percutaneous spinal puncture positioning system based on robotic ultrasound scanning imaging according to claim 1, characterized in that, The ultrasonic probe is capable of imaging at a depth of 50-90 mm.

6. The percutaneous spinal puncture positioning system based on robotic ultrasound scanning imaging according to claim 1, characterized in that, The integrated stent is also equipped with an adaptive compression follow-up device and an auxiliary puncture device; The adaptive clamping follow-up device is used to maintain a good relative position and posture between the ultrasound probe and the patient's body during the scanning process to ensure scanning safety and obtain high-quality ultrasound images. The auxiliary puncture device is used to assist doctors in inserting the puncture needle along a prescribed path and to display the insertion depth of the puncture needle.

Images