Endoscopic brain surgery robot control method

Abstract

The present application relates to the technical field of surgical robots, in particular to a control method for endoscopic brain surgery robot, comprising: constructing a movement vector of an arbitrary channel end; planning a working track of a surgical target area based on a binary voxel map of the movement vector, and converting the working track into a set of depth distribution points; mapping the depth distribution points to a preoperative image, extracting target points corresponding to the surgical target area, constructing a collision detection interval of the channel layout according to the gradient value of the target points and the marked points in the surgical target area, and updating the collision detection interval in real time using intraoperative real-time image; then, according to the feedback position of the intraoperative real-time data, performing distribution detection on the collision detection interval of each channel to determine a safe depth area; and associating the execution order of each channel according to the distribution position of the safe depth area to determine a full-axis control instruction sequence. The control precision and safety of the surgical robot are improved.

Description

Technical Field

[0001] This invention relates to the field of surgical robot technology, specifically a control method for an endoscopic cranial surgery robot. Background Technology

[0002] With the continuous development of surgical robot technology, the existing technology uses master-slave control robotic arms to achieve independent control of the scope and operating instruments. However, due to its split multi-arm structure, it is easily restricted by the narrow cavities of craniocerebral surgery, resulting in poor uniformity of multi-channel motion and a disconnect between collaborative control and spatial constraints, which increases the risk of surgical path planning and further limits the precision of the surgery.

[0003] For example, Chinese Patent Publication No. CN120788733A discloses a natural orifice surgery system, surgical method, and computer equipment. The system includes: a preoperative planning unit for generating preoperative planning information, including a planned surgical path; an intraoperative execution unit for constructing a dynamic surgical environment model and a deformation prediction model, adjusting the planned surgical path displayed in the dynamic surgical environment model in real time based on the real-time intraoperative data and the deformation prediction data output by the deformation prediction model, and performing natural orifice surgery according to the adjusted actual surgical path; and a postoperative evaluation unit for generating postoperative evaluation information for the natural orifice surgery based on the operation data output by the intraoperative execution unit during the execution of the natural orifice surgery.

[0004] For example, Chinese Patent Publication No. CN120053077A discloses an endoscopic control method for a neurosurgical robot based on a hierarchical quadratic programming framework, comprising: determining the current operating mode of the neurosurgical robot endoscope, wherein the operating mode includes: pivot motion mode, path navigation mode, and dynamic tracking mode; determining the corresponding robotic arm motion constraint model based on the operating mode using a hierarchical quadratic programming method; solving the robotic arm motion constraint model based on the output information of the position sensor and force sensor to obtain the control speed of the robot joint; and controlling the robot joint to move based on the control speed, thereby completing the endoscopic control of the neurosurgical robot.

[0005] In existing technologies, the evaluation information under surgical execution is set by the tissue deformation law of the cavity; or the control method of the surgical robot is constrained by the movement mode of the endoscope. However, existing technologies tend to solve data in a single channel and lack collaborative control under the simultaneous operation of multiple channels. This can easily lead to insufficient anatomical protection during the operation, which in turn reduces the accuracy and efficiency of the surgical robot. Summary of the Invention

[0006] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is: an endoscopic craniotomy robot control method, comprising: S1, constructing a movement vector at the end of any channel based on the operating components installed at the end of each channel of the robot.

[0007] S2, based on the binary voxel map of the movement vector, plans the working trajectory of the surgical target area and transforms the working trajectory into a set of depth distribution points containing depth distribution and position distribution.

[0008] S3 maps the depth distribution points to the preoperative image, extracts the target points corresponding to the surgical target area, and constructs the collision detection interval of the channel layout according to the gradient values ​​of the target points and the marked points in the surgical target area.

[0009] S4 acquires real-time intraoperative images, performs non-linear alignment between the real-time intraoperative images and preoperative images for each time period, determines the mapping relationship between intraoperative tissue deformation and marker points, and updates the collision detection range in real time.

[0010] S5, based on the feedback position of real-time data during the operation, performs distributed detection of collision detection zones in each channel to clarify the safe depth area.

[0011] S6 associates the execution order of each channel based on the distribution of the safety depth region to determine the full-axis control command sequence.

[0012] The beneficial effects of this invention are as follows: First, by synchronizing all channels of the robot to the same coordinate system, this invention defines the movement vector at the end of the channel according to the components of linear velocity and angular velocity in three-dimensional space; it quantifies the full-degree-of-freedom motion of axial extension and lateral deflection at the end of the channel; and at the same time, based on the binary voxel map and the method of fitting the working trajectory with the shortest path, it clarifies the reachable and unreachable scenarios of trajectory planning; thus further improving the safety, rationality and feasibility of trajectory planning.

[0013] II. This invention determines the usage sequence of each channel based on the surgical stage and completes time registration. Collision detection is performed based on the minimum Euclidean distance between any two channel depth distribution points, and the depth distribution points are adjusted accordingly. This avoids temporal conflicts and spatial collisions that occur when multiple channels move simultaneously. Then, points along the target path are used as target points, and labels mapped from preoperative images are used to spatially associate the target points with the marked points, thus setting the collision detection interval. This achieves the quantification of the spatial distribution of intracranial blood vessels, functional areas, and other tissue structures, further improving the accuracy of collision detection.

[0014] Third, this invention verifies the depth values ​​of target points near the surgical target area, adjusts the depth values ​​based on gradient values, and determines the corresponding allowable deflection range according to the depth values, ultimately forming a collision detection interval. This limits the relative range of depth and deviation angle values, enabling range calibration of real-time intraoperative data according to the assigned region and offset, thereby achieving dynamic calibration of collision detection and improving the real-time performance and accuracy of intraoperative safety protection. Attached Figure Description

[0015] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0016] Figure 1 This is a flowchart illustrating a control method for an endoscopic craniotomy robot. Figure 2 This is a flowchart illustrating step S2 of a robotic control method for endoscopic cranial surgery. Figure 3 This is a flowchart illustrating step S3 of a robotic control method for endoscopic cranial surgery. Figure 4 This is a flowchart illustrating step S4 of a robotic control method for endoscopic cranial surgery. Figure 5 This is a flowchart illustrating step S5 of a robotic control method for endoscopic cranial surgery. Figure 6 This is a schematic diagram of a control method for an endoscopic brain surgery robot.

[0017] In the diagram: 1. Endoscopic channel; 2. Irrigation / suspension channel; 3. Bipolar electrocoagulation channel; 4. Radiofrequency / plasma channel. Detailed Implementation

[0018] The embodiments of the present invention are described in detail below. The embodiments described below are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention. Where specific techniques or conditions are not specified in the embodiments, they shall be performed in accordance with the techniques or conditions described in the literature in the art or in accordance with the product manual.

[0019] See Figure 1 A control method for an endoscopic craniotomy robot includes: S1, constructing a movement vector at the end of any channel based on the operating components installed at the end of each channel of the robot.

[0020] S2, based on the binary voxel map of the movement vector, plans the working trajectory of the surgical target area and transforms the working trajectory into a set of depth distribution points containing depth distribution and position distribution.

[0021] S3 maps the depth distribution points to the preoperative image, extracts the target points corresponding to the surgical target area, and constructs the collision detection interval of the channel layout according to the gradient values ​​of the target points and the marked points in the surgical target area.

[0022] S4 acquires real-time intraoperative images, performs non-linear alignment between the real-time intraoperative images and preoperative images for each time period, determines the mapping relationship between intraoperative tissue deformation and marker points, and updates the collision detection range in real time.

[0023] S5, based on the feedback position of real-time data during the operation, performs distributed detection of collision detection zones in each channel to clarify the safe depth area.

[0024] S6 associates the execution order of each channel based on the distribution of the safety depth region to determine the full-axis control command sequence.

[0025] Generally, surgical robots have multiple independent drive channels, such as endoscope channel 1, irrigation / suction channel 2, bipolar electrocoagulation channel 3, radiofrequency / plasma channel 4, etc.; each channel is equipped with corresponding operating components at its end.

[0026] like Figure 6 As shown, the endoscope channel 1 is equipped with a built-in 3D binocular camera module and a coaxial ring illumination source for real-time intraoperative illumination, lesion observation and high-definition image acquisition.

[0027] The irrigation / suction channel 2 is a shared tubing channel with an inner diameter of 2-3 mm. It connects to either the irrigation fluid source or a negative pressure source via an external valve assembly. In irrigation mode, it is used for pulsed / continuous saline irrigation of the surgical field, blood dilution, and removal of debris and adhesions. In suction mode, it is used to aspirate accumulated blood, cerebrospinal fluid, irrigation waste fluid, electrocoagulation / plasma fume, and tissue debris. Irrigation and suction can be performed separately, or they can be used alternately or in combination.

[0028] The bipolar electrocoagulation channel 3 is equipped with bipolar electrocoagulation forceps / electrode assembly for intraoperative hemostasis and tissue coagulation, and after the treatment of brain tumors, it is equipped with tumor retrieval forceps for removal.

[0029] Radiofrequency / plasma channel 4 is equipped with radiofrequency electrodes or plasma blades for fragmentation, cutting, and vaporization ablation of lesions.

[0030] In craniocerebral surgery, surgical robots present working scenarios with access constraints, target area offset, and multi-channel collaboration. Among them, access constraints mean that the robot is limited by the straight channel axis formed by the skull drilling. The instrument can only perform axial extension and limited lateral deflection. Generally, the deflection angle is less than 30 degrees, and the axial extension stroke is greater than 30 mm. The outer diameter of the lumen body set in the corresponding channel is set to 15 mm.

[0031] Target area displacement refers to the significant drift of brain tissue during surgery, after incising the dura mater, releasing cerebrospinal fluid, or removing lesions. Real-time synchronization of the trajectories of each channel's endpoints is necessary to determine the control and processing of each channel. Simultaneously, multi-channel coordination requires determining the deflection direction of multiple independent channels, such as those for the inserted endoscope, suction device, and bipolar electrocoagulation device, to avoid collisions caused by deflection and movement, which could affect the surgical procedure. The robot's current operational process needs to be checked in real-time based on the movement trajectory of each independent channel, and the surgical procedure needs to be controlled in real-time according to the depth distribution of the corresponding components at the endpoints of each channel.

[0032] One implementation of step S1 includes: synchronizing all channels of the robot to the same coordinate system, and defining the movement vector at the end of the channel according to the components of linear velocity and angular velocity in three-dimensional space.

[0033] Specifically, first, a fixed point is selected on the robot body as the origin, and three orthogonal axes XYZ are defined to form a right-handed coordinate system, which is the common reference system for all channel movements.

[0034] Secondly, in the scenario where four channels are introduced, the movement vector of each channel is defined based on its constituent components.

[0035] The tool coordinate system for the four channels is represented as follows: For endoscope channel 1, the center of the lens is taken as the origin, the Z-axis is the line pointing into the patient's body (line of sight) along the endoscope axis, the X-axis is the line pointing to the right of the image, and the Y-axis is the line pointing to the bottom of the image.

[0036] For the irrigation / suction channel 2, with the center of the tube opening as the origin, the Z-axis points along the tube axis into the patient's body (in the direction of advance), and the XY axis is arranged in the same way as the endoscope channel 1.

[0037] For bipolar electrocoagulation channel 3, the midpoint between the two tweezer tips is selected as the origin, the Z-axis points to the tip along the axis of the tweezer bar, the X-axis is along the direction of the line connecting the two tweezer tips (opening direction), and the Y-axis is perpendicular to the XZ plane.

[0038] For the radio frequency / plasma channel 4, the origin is the electrode tip, the Z-axis points along the electrode rod towards the tip, and the XY axis is arranged in the same way as the endoscope channel 1.

[0039] Next, for each channel, a mapping from a common reference system to the tool coordinate system is established. Through a homogeneous transformation matrix, the relative coordinate system of the fixed tool offset is determined under robot joint changes.

[0040] Then, the robot joint velocities are mapped to the spatial velocities of the end effector using the Jacobian matrix.

[0041] Specifically, the position change of the joint encoder in adjacent control cycles is read, and the approximate derivative of the value is taken as the joint velocity vector at this point; then the product of the joint velocity vector and the Jacobian matrix is ​​taken as the movement vector at this point.

[0042] It should be noted that the Jacobian matrix is ​​a 6×n matrix. In the scenario of axial scaling + deflection angle + orientation angle, the value of n is 3. Each column of the Jacobian matrix corresponds to a joint, and it can be decomposed into two 3×n forms according to the linear velocity and angular velocity components of the movement vector.

[0043] Then take the coordinate value Z1 of its axis direction in the common reference system, the coordinate value P1 of its origin, and the coordinate value P2 of the origin of a certain end tool; if the joint is a rotary joint, the linear velocity column is Z1×(P2-P1), and the angular velocity column is Z1; for a translating joint, the linear velocity column is Z1, and the angular velocity column is 0.

[0044] Finally, based on the joint encoder, the joint angle / displacement is read and decomposed into linear velocity and angular velocity corresponding to the three-dimensional coordinates to obtain the output translation vector.

[0045] In one embodiment of the present invention, the movement vector is used as a standard to define the feasible motion space. The movement vectors under all possible positions and orientations are binarized. If there is a direction that allows the instrument to reach safely, it is marked as 1, otherwise it is marked as 0, forming a binary voxel map of the reachable area at the surgical target area. On this binary map, a path planning algorithm is used to calculate a working trajectory from the entry point to the target area. The working trajectory is a three-dimensional spatial curve. By combining the depth value of the surgical target area with the three-dimensional coordinates, the depth distribution points under the independent depth and deflection of each channel are determined.

[0046] It should be noted that the current embodiment is based on preoperative planning, and converts the various collected data into relative depth and coordinates to enhance the robot's control process during surgery.

[0047] Specifically, such as Figure 2 As shown, one implementation of step S2 includes: S21. Based on the reachable area of ​​the movement vector at the surgical target area, construct a three-dimensional voxel mesh to form a binary voxel map of the movement vector.

[0048] S22, in the binary voxel map, the shortest path from the skull inlet coordinates to the surgical target area is fitted as the working trajectory of the surgical target area.

[0049] The shortest path represents the movement based on preoperative image planning. In the current surgical scenario that relies on axial extension and angle adjustment, the shortest path will serve as the basis for setting the depth distribution points to plan the relative position of the movement to the surgical target area.

[0050] S23, for any working trajectory at any time, calculate the depth value, deflection angle and orientation angle of axial extension to obtain the depth distribution points; where the lateral deflection angle corresponds to the deflection angle of the channel around a plane perpendicular to the axis, and the value range is ±30 degrees; the azimuth angle corresponds to the rotation angle of the channel around the axis, and the value range is 0-360 degrees.

[0051] In this step, the relative velocity vector of the movement trajectory, combined with its relative three-dimensional coordinates, depth, and angle values, is used to perform a constraint check on the depth distribution for each channel.

[0052] S24. For each channel's depth distribution point, perform geometric verification and combine the verified data into an output set of depth distribution points.

[0053] Specifically, the implementation methods for performing geometric verification include: First, determine the order of use of each channel according to the surgical stage, and then perform time registration for each channel.

[0054] Secondly, based on the minimum Euclidean distance between the depth distribution points of any two channels, channel collision detection is performed. If the Euclidean distance is less than the channel safety threshold, the depth distribution points of each channel are adjusted.

[0055] Specifically, the channel safety threshold is set to 5mm to prevent multiple channels from being too close together during operation, thus causing mutual interference. When a value less than the channel safety threshold is detected, the voxel value corresponding to the depth distribution point is marked as 0, and the depth distribution point is regenerated to avoid collisions between channels.

[0056] The channel safety threshold is set based on the average distance between channels during normal surgery to prevent channels from getting too close together.

[0057] In one embodiment of the present invention, by spatially mapping depth distribution points to preoperative image features, anatomical labels near the point, such as brain tissue structures like gray matter, white matter, blood vessels, cerebrospinal fluid, and tumors, are examined. A detection interval containing anatomical labels and gradient values ​​is defined along the structures at each point in the surgical target region. The gradient value is used to describe the intensity of feature value changes for a certain structure type along the possible movement direction of the instrument.

[0058] Specifically, such as Figure 3 As shown, one implementation of step S3 includes: S31, select multiple target points from the surgical target area, and use the depth distribution points along the path from the end of each channel to the target point as the target points for treatment.

[0059] S32, spatially map the target location to the preoperative image to determine the anatomical labels near the point.

[0060] S33: Extract markers corresponding to anatomical labels from the surgical target area, calculate the approach rate between the target point and the markers in any direction, and use it as the gradient value corresponding to the target point; specifically, extract tissues such as blood vessels and nerve bundles from the surgical target area. These tissues represent tissues that affect surgical safety and do not include targets that need to be removed / stopped, such as tumors, hematomas, and malformed vascular clusters. Markers are set for these tissues; calculate the distance between the target point and the corresponding marker in the surgical target area according to the spatial gradient, which is used to help verify the tissue structure targeted in the current surgery.

[0061] It should be noted that when determining the approach rate between the target point and the marker point, a distance field is constructed based on the distance between the target point and the nearest marker point, so that each location has a distance value that can point towards or away from the target point. Secondly, the magnitude of the distance gradient in this distance field is always 1. The relative rate of approach to the surgical target marker is determined by the angle between the instrument's movement direction and the gradient direction. Using the movement direction corresponding to the current target point as the basis for verification, and combining it with the gradient of the distance field, the approach rate between the target point and the marker point is calculated.

[0062] Specifically, when the instrument is moving towards the marked point, its direction of motion is opposite to the gradient direction, and the approach rate is expressed as follows: ;in, This indicates the approach rate, meaning whether the instrument is rapidly approaching the marked point; The gradient of the distance field is a vector pointing in the direction of the fastest increase in the current distance value; the distance to the marker point is the current target point, indicating the signed distance from the current target point to the nearest marker point. This represents the magnitude of the distance gradient, which is 1 everywhere in an ideal distance field; A unit vector representing the direction of movement along the trajectory, used to describe the direction of movement along the target point; This represents the angle between the direction of motion and the gradient direction, used to explain the relative angle when approaching each other.

[0063] Since the direction of motion and the gradient direction are opposite, when the included angle is 180 degrees, the cosine of this angle is represented as -1, making the approach speed -1. Since we are primarily concerned with the relative decrease in distance, a negative sign is introduced to correct this value and explain the gradient's value. A positive approach speed is then represented as follows: ;in, This represents the accumulated distance the instrument travels along the trajectory to the target point from its starting point. In other words, the approach rate is represented by the accumulated distance plus the distance between the current target point and the marker. When the instrument is directly facing the marker, the approach rate is 1, indicating maximum risk. When the instrument moves along an equidistant line (tangentially), the approach rate is 0. When the instrument moves away from the structure, the approach rate is negative, indicating no risk. This current approach rate is then used as the gradient value for subsequent settings, thus illustrating the relative speed of the instrument's overall movement.

[0064] S34. Based on the gradient values ​​of each target point, combined with the depth range and deflection angle range, the collision detection range is defined. The collision detection area is different from the aforementioned end verification of the recognition channel. This collision detection area introduces anatomical structural features and uses the gradient values ​​of these features to adjust the safe range of collision detection, so as to avoid each channel directly touching additional brain tissue.

[0065] Specifically, when combining depth range and deflection angle range, the implementation methods include: Based on the minimum distance from the target point to the surgical target area, depth values ​​are checked for target points near the surgical target area.

[0066] For the target points after depth value verification, the depth values ​​are adjusted according to the gradient values ​​of each target point.

[0067] Based on the given depth value, determine the allowable deflection range corresponding to each depth value, and form the output collision detection interval according to the value of the allowable deflection range.

[0068] Specifically, the distance from the target point to the target area will be used separately for proximity to the surgical target area. When the distance is less than the safe distance threshold (e.g., 2mm) for proximity to the surgical target area, it is considered to have entered the vicinity of the target area. Within this area, regardless of the gradient value, a depth value check is performed. The expected depth is compared with the currently simulated depth, the currently simulated depth is corrected, and synchronized to the depth distribution point set corresponding to the target point. The safe distance threshold is set according to the average distance of the ends of various channels to the marked blood vessels, nerve bundles, and other tissues during normal surgery. That is, if it is less than the safe distance, the current target point is regressed to the previous target point to meet the safe distance. Then, the depth value and the allowable deflection range corresponding to the depth value are checked again to achieve bypass processing for specific tissue mutations.

[0069] Secondly, the gradient value is checked against the given gradient value to determine if the rate of change of the gradient value is too fast. If the gradient value is greater than the preset gradient value, the allowable depth endpoint of the current target point is modified, and the previous target point before the surgical target area is selected as the allowable endpoint to avoid damage to other tissues caused by rapid changes. The preset gradient value is set based on the average value of each channel in historical data. At this time, when the previous target point is selected as the allowable endpoint, its depth value and related deflection angle are rechecked. After the check is completed, the spatial range formed by the depth and deflection angle is used as the passable area after facing tissue distortion or relative deformation. These areas are stored in the collision detection area, and the robot's working process at each moment is continuously checked.

[0070] Then, based on the verified depth value, the discretization process is performed according to the allowed deflection angle. For each deflection angle, all direction angles are traversed to form the points of direction angle and deflection angle.

[0071] Then, based on the points where the orientation angle and deflection angle are formed, the distance between their locations and the marked points is calculated. If the distance exceeds the aforementioned safe distance threshold, the corresponding points are marked as the allowable deflection range. After gradually merging related points, the setting of the allowable deflection range is completed. All verified depth values ​​and allowable deflection ranges are used as the current output collision detection interval. At this time, the collision detection interval represents the relative spatial range obtained by combining depth values ​​and angle values. This spatial range will be constructed based on the preoperative analysis data and then updated using real-time intraoperative data to address the impact of brain tissue deformation on the robot's operation.

[0072] Furthermore, in one embodiment of the present invention, the preoperatively planned target points and related marker points are used to perform multi-dimensional alignment processing on the intraoperative real-time images for each time period, and on the basis of the alignment processing, the content to be updated in the collision detection interval is marked.

[0073] like Figure 4 As shown, step S4 is implemented in the following ways: S41 compares intraoperative real-time images with preoperative images at each time point, and constructs a three-dimensional deformation field of tissue deformation using the vascular tree as a spatial marker.

[0074] Specifically, the intraoperative real-time images will be matched with the preoperative images in real time through the tree structure of blood vessel distribution. The images will be recorded in three-dimensional coordinates, and the areas where the three-dimensional coordinates change will be formed by linear interpolation.

[0075] It should be noted that when processing intraoperative real-time images and preoperative images, the vessel centerline and vessel bifurcation feature points are extracted by segmenting the vessels to construct a feature point set; then the three-dimensional offset vectors of the intraoperative images and preoperative images at each time period are compared to serve as the three-dimensional deformation field at that time.

[0076] Since the brain is rich in blood vessels, and when blood vessels change, the markers of related tissues will change accordingly, the differences in feature points in the image can be quickly identified based on the three-dimensional deformation of the blood vessels. Then, according to the three-dimensional offset vector of the difference points, the position of the channel blood vessel is used to form a three-dimensional deformation field covering the surgical target area by three-dimensional interpolation.

[0077] In addition, in the current intraoperative real-time imaging, it is necessary to record the relative distance between brain tissues such as nerve bundles and spatial landmarks, so as to determine the relative position of these brain landmarks at each update, thereby making the three-dimensional deformation field closer to the relative position of tissue deformation.

[0078] S42, from the three-dimensional deformation field of tissue deformation, determine the common feature points and differential feature points corresponding to the marker points at each time step; among them, the common feature points represent the consistent content of the blood vessel position before and during the operation, and the differential feature points represent the positions where there are three-dimensional offset vectors. These positions where three-dimensional offsets occur will be used as the basis for the current data update to determine the depth distribution points to be updated, as well as the corresponding marker points.

[0079] S43, based on the displacement of the differential feature points, updates the marker points and depth distribution points in real time, and updates the collision detection interval in real time according to the gradient changes of the marker points and target points. In the current update process, the depth distribution points are essentially non-tissue movement content, and this position is less affected by tissue deformation. When adjusting them, first adjust the position of the marker points according to the displacement corresponding to the differential feature points, then determine the distance between the marker point position and the original depth distribution points, and check whether it meets the safe distance threshold. If it does, the depth distribution points are not modified; if it does not meet the safe distance threshold, the position of the depth distribution points is adjusted synchronously according to the displacement of the marker points, ultimately achieving a global update of each point and the real-time surgical status.

[0080] In one embodiment of the present invention, the actual operation position and the collision detection range simulated before surgery are limited by a real-time feedback processing method to obtain a safe depth range under normal surgical procedure, thereby avoiding the relative risk caused by target area deviation during surgery.

[0081] Specifically, such as Figure 5 As shown, one implementation of step S5 includes: S51: Based on the feedback position of real-time data during surgery, obtain the real-time recorded depth value, match the depth value with the collision detection interval, and determine the area to which the feedback position belongs.

[0082] One way to determine the region to which the feedback location belongs includes: When the feedback position is an axial extension movement, a depth correlation is established between the feedback position and the collision detection area.

[0083] When the feedback position is a deflection motion, establish a deflection correlation between the feedback position and the collision detection area; Based on deep correlation and deflection correlation, the region to which the feedback location belongs is quantified.

[0084] Specifically, based on the real-time depth value corresponding to the current feedback position, check whether the feedback position belongs to axial extension or deflection motion in adjacent time periods; if it belongs to axial extension motion, determine the lowest and highest depths of the current motion as the depth correlation; if it belongs to deflection motion, mark the deflection angle and direction angle of the current motion, and select the maximum and minimum values ​​of the angle as the deflection correlation here.

[0085] As for the region to which the feedback location belongs, the current region is identified by mapping the real-time depth value to the distance field simulated before surgery and then querying the region corresponding to that location.

[0086] S52, calculate the offset of each assigned region based on the offset vector of the centroid of the surgical target area; wherein, the offset vector is set according to the vector difference between the preoperative simulation and the intraoperative image about the centroid of the surgical target area, and the offset of the real-time operation is determined according to the location of the assigned region.

[0087] S53 remaps the offset of the home region to the collision detection interval to form the output safe depth region.

[0088] After knowing the mapping relationship between the home region and the collision detection interval, the offset is synchronized to the coordinates corresponding to the collision detection interval according to the distribution of each home region. The simulated collision detection interval is then corrected according to the offset, and the corrected data is used as the safe depth region.

[0089] As for the association relationship bound to the current feedback location, it will be synchronously recorded as relevant data in the safety depth area to determine whether the real-time operation exceeds the coverage of the collision detection area.

[0090] It should be noted that the current surgical target area is set with multiple assigned regions, each with a corresponding offset to interpret the relative spatial coordinates of each position. This offset will be used as the verification mapping content of the collision detection interval. After the offset is known, it can be further updated and remapped into the collision detection interval to re-trigger the collision detection interval update for the current time period, thereby obtaining the corresponding safe depth area.

[0091] In one embodiment of the present invention, the order of the current safe depth region on the depth axis is analyzed according to the distribution corresponding to the safe depth region, and sorted according to the depth value to obtain the segmented coverage of each depth value range; for example, 0-15mm, only endoscope; 15-30mm, endoscope + suction device; 30-42mm, endoscope + suction device + bipolar; 42-45mm, endoscope + suction device; these distributions are synchronized to different surgical stages to obtain the full-axis mapping control process.

[0092] One implementation of step S6 includes: S61 projects the safe depth region onto the depth axis to obtain channel combinations corresponding to different depth ranges.

[0093] S62 uses the time period executed by each channel as an index to associate the execution order with the channel combination of each time period, forming a full-axis control command sequence. The execution order refers to the order in which each channel is used under a preset surgical stage.

[0094] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention, which are still covered within the protection scope of the present invention.

Claims

1. A control method for an endoscopic craniotomy robot, characterized in that, include: S1, construct the movement vector at the end of any channel based on the operating components installed on the robot at the end of each channel; S2, based on the binary voxel map of the movement vector, plans the working trajectory of the surgical target area and transforms the working trajectory into a set of depth distribution points containing depth distribution and position distribution; S3 maps the depth distribution points to the preoperative image, extracts the target points corresponding to the surgical target area, and constructs the collision detection interval of the channel layout according to the gradient values ​​of the target points and the marked points in the surgical target area. S4: Acquire real-time intraoperative images, perform non-linear alignment between real-time intraoperative images and preoperative images for each time period, determine the mapping relationship between intraoperative tissue deformation and marker points, and update the collision detection range in real time. S5, based on the feedback position of real-time data during the operation, performs distributed detection on the collision detection range of each channel to clarify the safe depth area; S6 associates the execution order of each channel based on the distribution of the safety depth region to determine the full-axis control command sequence.

2. The control method for an endoscopic craniotomy robot according to claim 1, characterized in that, The methods for implementing a movement vector include: Synchronize all channels of the robot to the same coordinate system, and define the movement vector at the end of the channel according to the components of linear velocity and angular velocity in three-dimensional space.

3. The control method for an endoscopic craniotomy robot according to claim 1, characterized in that, The implementation methods for the depth distribution point set in step S2 include: S21. Based on the reachable area of ​​the movement vector at the surgical target area, construct a three-dimensional voxel mesh to form a binary voxel map of the movement vector. S22, in the binary voxel map, the shortest path from the skull entry point coordinates to the surgical target area is fitted as the working trajectory of the surgical target area. S23, for the working trajectory at any time, calculate the depth value, deflection angle and direction angle of axial extension and contraction to obtain the depth distribution points; S24. For each channel's depth distribution point, perform geometric verification and combine the verified data into an output set of depth distribution points.

4. The control method for an endoscopic craniotomy robot according to claim 3, characterized in that, When performing geometric verification, the implementation methods include: The order of use of each channel is determined according to the surgical stage, and time registration is performed for each channel; Based on the minimum Euclidean distance between the depth distribution points of any two channels, perform channel collision detection. If the Euclidean distance is less than the channel safety threshold, adjust the generated depth distribution points for each channel.

5. The control method for an endoscopic craniotomy robot according to claim 1, characterized in that, The collision detection interval in step S3 is implemented in the following ways: S31, Select multiple target points from the surgical target area, and use the depth distribution points along the path from the end of each channel to the target point as the target points for treatment; S32, Spatial mapping of the target point with preoperative images to determine the anatomical labels near the point; S33, extract the marker points corresponding to the anatomical labels from the surgical target area, calculate the approach rate between the target point and the marker point in any direction, and use it as the gradient value corresponding to the target point; S34. Based on the gradient values ​​of each target point, combined with the depth range and deflection angle range, the collision detection range is defined.

6. The control method for an endoscopic craniotomy robot according to claim 5, characterized in that, When combining depth range and deflection angle range, the implementation methods include: Based on the minimum distance from the target point to the surgical target area, perform depth value verification for target points adjacent to the surgical target area; For the target points after depth value verification, adjust the depth value according to the gradient value of each target point; Based on the given depth value, determine the allowable deflection range corresponding to each depth value, and form the output collision detection interval according to the value of the allowable deflection range.

7. The control method for an endoscopic craniotomy robot according to claim 5, characterized in that, The implementation methods for updating the collision detection range in step S4 include: S41, compare intraoperative real-time images with preoperative images at each time period, and construct a three-dimensional deformation field of tissue deformation using the vascular tree as a spatial marker. S42, from the three-dimensional deformation field of tissue deformation, determine the common feature points and difference feature points corresponding to the marked points at each time step; S43 updates the marker points and depth distribution points in real time based on the displacement of the differential feature points, and updates the collision detection range in real time according to the gradient changes of the marker points and target points.

8. The control method for an endoscopic craniotomy robot according to claim 1, characterized in that, The implementation methods of the safe depth region in step S5 include: S51, based on the feedback position of real-time data during the operation, obtain the real-time recorded depth value, match the depth value with the collision detection range, and determine the area to which the feedback position belongs; S52, calculate the offset of each assigned region based on the offset vector of the centroid of the surgical target area; S53 remaps the offset of the home region to the collision detection interval to form the output safe depth region.

9. The control method for an endoscopic craniotomy robot according to claim 8, characterized in that, The methods for determining the region to which the feedback location belongs include: When the feedback position is an axial telescopic motion, a depth correlation relationship is established between the feedback position and the collision detection area; When the feedback position is a deflection motion, establish a deflection correlation between the feedback position and the collision detection area; Based on deep correlation and deflection correlation, the region to which the feedback location belongs is quantified.

10. The control method for an endoscopic craniotomy robot according to claim 1, characterized in that, The implementation methods of the full-axis control command sequence in step S6 include: S61, Project the safe depth region onto the depth axis to obtain channel combinations corresponding to different depth ranges; S62 uses the time period executed by each channel as an index to associate the execution order with the channel combination of each time period, forming a full-axis control command sequence.

Images