Spinal surgery, as a branch of orthopedic surgery, is considered one of the highest-risk and most challenging surgical procedures due to the unique anatomical structure of the spine and its crucial protective role for the central nervous system. In recent years, various surgical robot systems designed for spinal surgery have become a research hotspot in the field of medical robotics. During bone manipulation in spinal surgery, higher demands are placed on the end effector's load-bearing capacity, positioning accuracy, and operational stability. This paper will focus on pedicle screw fixation as an application, examining key technologies such as robot mechanism design, image navigation, and navigation and positioning control to prepare for the clinical application of spinal surgery robots.
Spinal surgery is primarily used to treat spinal deformities, compression fractures, intervertebral disc lesions, and spinal cord compression. Common procedures include laminectomy, spinal fusion, and disc replacement. Typical spinal surgeries often involve the removal or replacement of tissue at the lesion site, which inevitably disrupts the original biomechanical structure of the spine to some extent, reducing its stability in supporting the trunk. Therefore, pedicle screw fixation is often used concurrently in spinal surgery to reconstruct spinal stability. Typical vertebral structures include the vertebral body, spinous process, transverse process, superior/inferior articular processes, mastoid process, and vertebral foramen. Since the transverse processes, laminae, and superior/inferior articular processes all converge at the same point of the pedicle, this location offers the best biomechanical properties. Clinically, surgeons often insert bone screws into the vertebral body through the narrow pedicle to achieve the best fixation strength and stability reconstruction effect; therefore, stability reconstruction surgery is also known as pedicle screw fixation.
Clinically, doctors use image-guided navigation systems to complete surgical diagnoses and planning, determining the path of the surgical pins. During the surgery, the doctor first uses a auger or bone drill to drill the pin path; then, the bone pins are inserted into the vertebrae along the path; next, the fixation rod is bent to an appropriate shape to accommodate the position between the implanted bone pins, and the rod is placed into the U-shaped groove at the rear of the bone pin; finally, the rod is fixed with a nut to complete the reconstruction of spinal stability.
Safe and stable reconstruction places strict requirements on the drilling point, angle, and depth of the pedicle screw. Incorrect selection of the entry point or angle can lead to lateral or medial deviation of the screw, thereby damaging spinal nerves or blood vessels; a screw that is too deep, even penetrating the contralateral cortex, may also damage important tissues, while a screw that is too shallow can lead to problems such as poor stability of subsequently implanted bone screws. Due to its special location, pedicle screw drilling is considered one of the most critical and dangerous procedures in spinal surgery.
Integrated design of spinal surgery robot mechanism
In unstructured clinical surgical environments, surgical robots developed based on industrial robots can no longer meet surgical needs, and research on surgical robots has gradually transitioned to the research of specialized surgical robots. Israel has developed the SpineAssist spinal surgery robot based on the Stewart parallel mechanism. This six-degree-of-freedom robot adjusts the pose of the moving platform to position the guide sleeve at the predetermined pin path, providing accurate operational guidance for the surgeon. The German WISARoMed spinal surgery robot also employs a similar parallel configuration. Its parallel mechanism has advantages such as high rigidity, compact structure, and relatively small size, but it also suffers from limited workspace and poor flexibility.
For spinal surgeries requiring large workspaces, many researchers are opting for serial configurations to build spinal surgery assistance systems. These include the SPINEBOT-I system from Hanyang University in South Korea, the CoRASS system from Pohang University of Science and Technology, and the spinal surgery robot system from Nankai University, all employing a Cartesian coordinate configuration. While Cartesian configurations offer a large and simple workspace, the robots themselves occupy significant space, making them impractical in operating rooms equipped with various instruments. Other systems utilize vertical articulated robots, such as the VectorBot system from Germany and the PA-10-based spinal surgery assistance system from Navarra University. Vertical articulated robots offer a larger workspace-to-volume ratio and a more compact structure, making them well-suited for spinal surgery.
Serial robotic arm design
To meet the clinical needs of spinal surgery, and considering the specific safety requirements of the procedure, the configuration and structural design of the robotic arm must first satisfy the freedom requirements for pin placement and manipulation during spinal surgery. Pin insertion point positioning requires adjustments in three directions and angles in two directions. In addition, to ensure effective tracking of the instrument's end-effector markers by the navigation system, the robotic end-effector's bone-drilling device needs one rotational degree of freedom to adjust the marker's orientation. Therefore, the robotic arm requires at least six independent degrees of freedom: three for position adjustment and three for posture adjustment.
Commonly used position adjustment structures include Cartesian coordinate, cylindrical coordinate, and spherical coordinate types. During spinal surgery, the patient is in a prone position, with their spine approximately parallel to the horizontal direction. Simultaneously, the operating room conditions require a compact and lightweight robot structure. Comparing Cartesian and spherical coordinate robots reveals that the latter has a larger workspace/volume ratio and a more compact structure. In rotary joint robots, compared to spherical coordinate robots, cylindrical coordinate robots have fewer joints affected by gravity, which allows for further reduction in the size of the robotic arm and also lowers the possibility of the robotic arm falling and injuring the patient in case of joint failure, better meeting the higher safety requirements of surgical robots. Considering several serial robot configurations, this paper selects the cylindrical coordinate type for position adjustment, plus three attitude adjustment degrees of freedom at the robot's front end.
In this configuration, all joints except the first joint, which is a linear joint with vertical movement, are rotary joints. The geometric model of the spinal surgery robot is shown in Figure 1. The axes of joints 2 and 3 are perpendicular to the horizontal plane, allowing the robot to move within the horizontal plane; the axes of joints 4, 5, and 6 intersect at a point, forming the robot's "wrist joint." The end effector is mounted along the axis of joint 6 and rotates to adjust its direction.
The workspace of a spinal surgery robot needs to accommodate all positional and orientation requirements within a specific surgical area. Clinically, a Cartesian coordinate system composed of three orthogonal planes (transverse, sagittal, and coronal planes) is used to describe the patient. The selection of the screw entry point during spinal surgery is primarily performed on the coronal plane. During spinal surgery, the patient is in a prone position, with the coronal plane approximately horizontal. The robot is placed on the side of the operating table, with its x-axis approximately perpendicular to the sagittal plane and its y-axis approximately perpendicular to the transverse plane, as shown in Figure 2. Therefore, the robot's end effector primarily locates the screw entry point on the xy-plane.
Force feedback bone drilling device design
For the drilling of the screw path during spinal surgery, this paper designs a dedicated end-effector bone drilling device. As shown in Figure 3, it includes dedicated surgical instruments (dedicated medical bone drill motor, quick-release connector, dedicated bone drill bit), linear transmission unit, drive motor, connecting flange, 6-axis force/torque sensor, and infrared positioning target.
The bone drill bit is connected to the output shaft of the bone drill motor via a dedicated quick-release connector, and the high-speed rotational motion required for drilling the screw path is provided by the medical bone drill motor. The linear feed motion required for the drilling process is provided by the tail drive motor and linear transmission unit. An infrared target is fixed to the moving end of the linear transmission unit to track the feed motion of the bone drill in real time. To ensure that the surgical instruments and the moving end of the linear transmission components do not slip in the event of a power outage, a self-locking trapezoidal lead screw is selected as the linear transmission component. The fixed end of the linear transmission unit is connected to a 6-axis force/torque sensor via a connecting flange. This structure ensures that the feedback force during the drilling process can be transmitted to the force/torque sensor through the transmission unit, connecting flange, and other components, realizing real-time force sensing at the robot's end effector.
Image 3D reconstruction technology
Image navigation technology is widely used in orthopedic surgical robot systems, and can be mainly divided into two categories: two-dimensional image navigation and three-dimensional image navigation. Among them, two-dimensional image navigation is constructed by taking two perspective images of the patient's surgical site from the front and side, and registering them based on the method of picking corresponding points.
3D image navigation allows surgeons to intuitively observe the spatial location of the planned path through a 3D reconstructed anatomical model of the surgical area and directly complete surgical planning in 3D space. Israel's SpineAssist system uses a combination of 3D images acquired through preoperative computed tomography (CT) and intraoperative 2D image registration to achieve 3D image navigation. South Korea's SPINEBOT series robots also use a similar method. Another 3D image-based navigation method uses geometric feature-based registration through bony markers or artificially implanted markers. This method is used in the intervertebral disc replacement surgical robot system at Harbin Institute of Technology; however, intraoperative bony markers are often not obvious, thus limiting the accuracy of point selection and registration.
In spinal surgery robot systems, the image navigation system needs to provide surgeons with visual images of the patient's surgical area for diagnosis and planning, while also providing the robot with surgical planning information for monitoring and inducing its movement. Therefore, the development of the navigation system needs to meet the requirements of robot movement, operational accuracy, and precision, while also satisfying the surgeon's needs for user-friendliness during intraoperative diagnosis, planning, and real-time monitoring.
For example, surgical area images acquired by CT or MRI can be reconstructed into a 3D model, allowing surgeons to more intuitively observe the spatial morphology of the patient's spine, the surgical path, and instrument positions. Medical image 3D reconstruction methods are mainly divided into volume rendering and surface rendering. From the perspective of 3D reconstruction image quality, volume rendering is superior to surface rendering; however, in terms of algorithm efficiency and interactivity, surface rendering far surpasses volume rendering. Intraoperative navigation systems focus on software interactivity; therefore, this paper uses the Marching Cubes (MC) surface rendering algorithm for 3D reconstruction of the spine in the surgical area.
The Monte Carlo (MC) algorithm is an algorithm based on isosurface extraction from 3D spatial image data fields. Its basic idea is to construct isosurface triangles within voxels by setting an image grayscale threshold, and then stitch these triangles together according to their vertices and normal vectors to obtain the overall 3D isosurface as the outline of the 3D model. This paper establishes an interactive 3D reconstruction function based on the Observer-Command interactive mode provided by the VisualizationToolkit. This allows for convenient and accurate selection of the 3D image reconstruction threshold, effectively improving reconstruction efficiency.
Image registration technology
One of the core aspects of image-based navigation is establishing a unified coordinate transformation relationship among the image, patient, and robot. The transformation relationship between the robot and patient is calculated through real-time intraoperative positioning and tracking, while the coordinate transformation relationship between the image and patient needs to be established through image registration, which essentially involves obtaining the transformation matrix between the image coordinate system and the patient coordinate system. Registration in spinal surgery is a rigid registration process, and commonly used methods include geometric feature-based registration methods based on point/point, point/plane, contour/plane, and plane/plane relationships. Additionally, there are methods that perform registration at the image pixel level.
This paper employs a registration method combining two algorithms: the point-picking method and the Iterative Closest Point (ICP) method, both based on point-to-point relationships. This improves registration accuracy while avoiding registration failures caused by significant differences in pose between the image coordinate system and the reference coordinate system. During surgical registration, the point set in the patient's reference coordinate system is first registered with the point set in the image coordinate system to obtain an initial coarse registration transformation matrix and a coarsely registered reference point set. Then, the ICP algorithm is used to perform a secondary fine registration between the reference point set and the image points, obtaining a fine registration matrix, thus yielding the final registration matrix. Figure 4 illustrates the coordinate transformation relationships during image registration.
Navigation and positioning control technology
Navigation and positioning utilize infrared binocular cameras and infrared target points fixed to the robot's end effector and the patient's spine for real-time localization and tracking. This enables the spatial transformation coordinate system of the intraoperative navigation-robot-patient system to describe the operation, transforming surgical planning information into robot control parameters, thereby guiding the surgical robot. Therefore, the entire coordinate system can be divided into three sub-coordinate systems: the surgical instrument coordinate system, the patient coordinate system, and the optical locator coordinate system, as shown in Figure 5.
In the surgical instrument coordinate system, the pose matrix of the instrument tip in the instrument marker coordinate system is established through manual instrument calibration. In the patient coordinate system, a spatial transformation matrix is established through image registration to convert the pose information of the surgical area located in the patient target coordinate system in real space to the image coordinate system. In the optical locator coordinate system, the optical locator tracks the pose of the surgical instrument target and the patient target in real time, and forms a spatial transformation closed loop between the locator, surgical instrument, and patient image through coordinate transformation within the instrument and patient sub-coordinate systems. During surgical planning, the surgeon plans the operation in the patient image coordinate system to determine the entry point and target point of the pin path. During navigation and positioning, the navigation information is transformed into the robot end effector coordinate system. This allows for convenient adjustment of the optical locator's position during the operation without changing the values of the navigation information.
For more information, please visit the Industrial Robots channel.
