WO2024222402A1 - Catheter robot and registration method thereof - Google Patents

Abstract

A catheter robot and a registration method thereof. The method comprises: acquiring a model point set comprising a plurality of simulated feature points; by means of a first sensor, acquiring a spatial point set comprising a plurality of sampling point subsets; obtaining a plurality of denoising weight matrices corresponding to the plurality of sampling point subsets; based on the plurality of denoising weight matrices, respectively denoising the sampling points in the corresponding plurality of sampling point subsets; based on the denoised sampling points of the plurality of sampling point subsets, respectively determining mean points corresponding to the plurality of sampling point subsets; and based on the plurality of simulated feature points and the plurality of mean points, determining a coarse registration relationship between the model point set and the spatial point set. The invention improves registration precision between a model point set and a spatial point set.

Description

Catheter robot and registration method thereof

This application claims the priority of the Chinese patent application filed with the China Patent Office on April 27, 2023, with application number 202310481670.9 and application name “Catheter Robot and Its Alignment Method”, the entire contents of which are incorporated by reference into this application.

Technical Field

The present application relates to the field of surgical robots, and in particular to a catheter robot and a registration method thereof.

Background Art

Minimally invasive medical techniques are intended to reduce the amount of tissue damaged during a medical procedure, thereby reducing patient recovery time, discomfort, and harmful side effects. Such minimally invasive techniques can be performed through natural orifices in the anatomical structure or through one or more surgical incisions. Doctors can insert minimally invasive medical instruments through these natural orifices or incisions to reach the target tissue location. In order to assist in reaching the target tissue location, the position of the medical instrument can be correlated with a preoperative image or an intraoperative image of the anatomical structure. In order to achieve the correlation of the position of the medical instrument with the preoperative image or an intraoperative image of the anatomical structure, image registration of the medical instrument and the image is required.

The image registration of the prior art is carried out by selecting several obvious feature points on the image, and the coordinate values of these feature points in the image coordinate system constitute a first point set; and controlling the medical instrument to move in the anatomical structure to reach the corresponding feature points, and obtaining the coordinate values of these feature points in the surgical environment coordinate system to constitute a second point set; and then determining the transformation matrix of the surgical environment coordinate system relative to the image coordinate system based on the first point set and the second point set to complete the registration.

However, the existing technology does not consider the intrinsic motion effect of the anatomical structure when performing image registration. For example, when the anatomical structure is a bronchus, the effect of respiratory motion is not considered. This effect causes unexpected deviations in the coordinate values of the acquired feature points in the surgical environment coordinate system, which in turn causes a large error in the determined transformation matrix.

Summary of the invention

Based on this, it is necessary to provide a catheter robot and a registration method thereof that are conducive to improving the registration accuracy.

In one aspect, the present application provides a catheter robot, comprising:

A catheter, wherein the catheter and/or an instrument carried by the catheter is provided with a first sensor for sensing the position of the catheter end and/or the instrument end; and

A control device, coupled to the first sensor and configured to:

Acquire a model point set, the model point set comprising a plurality of simulated feature points, the simulated feature points being acquired from an anatomical model of an anatomical structure of a first patient, the plurality of simulated feature points being respectively associated with a plurality of feature points of the anatomical structure;

Acquire a spatial point set, wherein the spatial point set includes a plurality of sampling point subsets, wherein the sampling point subsets include a plurality of sampling points, wherein the sampling points are acquired from the anatomical structure by the first sensor, and the plurality of sampling point subsets are respectively associated with a plurality of the feature points;

Acquire a plurality of denoising weight matrices corresponding to a plurality of the sampling point subsets, wherein the plurality of denoising weight matrices are respectively associated with a plurality of the feature points;

Based on the plurality of denoising weight matrices, denoising the sampling points in the corresponding plurality of sampling point subsets respectively;

Based on the denoised sampling points of the plurality of sampling point subsets, respectively determining mean points corresponding to the plurality of sampling point subsets;

Based on the plurality of simulated feature points and the plurality of mean points, a rough registration relationship between the model point set and the space point set is determined.

On the other hand, the present application provides a catheter robot registration method, comprising:

Acquire a model point set including a plurality of simulated feature points, wherein the simulated feature points are acquired from an anatomical model of an anatomical structure of a first patient, and the plurality of simulated feature points are respectively associated with a plurality of feature points of the anatomical structure;

Acquire, by a first sensor, a spatial point set including a plurality of sampling point subsets, wherein the sampling point subsets include a plurality of sampling points, the sampling points are acquired from the anatomical structure, and the plurality of sampling point subsets are respectively associated with a plurality of the feature points;

Acquire a plurality of denoising weight matrices corresponding to a plurality of the sampling point subsets, wherein the plurality of denoising weight matrices are respectively associated with a plurality of the feature points;

Based on the plurality of denoising weight matrices, denoising the sampling points in the corresponding plurality of sampling point subsets respectively;

Based on the denoised sampling points of the plurality of sampling point subsets, respectively determining mean points corresponding to the plurality of sampling point subsets;

Based on the plurality of simulated feature points and the plurality of mean points, a rough registration relationship between the model point set and the space point set is determined.

On the other hand, the present application provides a computer-readable storage medium storing a computer program, wherein the computer program is configured to be loaded and executed by a processor to implement the steps of the method described in any one of the above embodiments.

The present application also provides a computer program product, comprising computer instructions, which, when executed on a computer, enable the computer to execute the method described in any one of the above embodiments.

The catheter robot and the registration method thereof of the present application have the following beneficial effects:

By obtaining denoising weight matrices corresponding to each sampling point subset, denoising the sampling points in the corresponding sampling point subsets respectively through multiple denoising weight matrices, and obtaining the mean point of the denoised sampling points in each sampling point subset, and then determining the coarse registration relationship between the image coordinate system and the surgical environment coordinate system based on multiple simulated feature points and multiple mean points, the adverse effect of the intrinsic movement of the patient's anatomical structure on the coarse registration accuracy can be reduced or eliminated.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1 is a schematic diagram of the structure of a catheter robot provided in one embodiment of the present application;

FIG2 is a schematic diagram of the structure of a catheter device and a power unit provided in one embodiment of the present application;

FIG3 is a schematic flow chart of a registration method for a catheter robot according to an embodiment of the present invention;

FIG4 is a schematic diagram of simulated feature points in an anatomical structure provided by an embodiment of the present application;

FIG5 is a schematic flow chart of a registration method for a catheter robot according to an embodiment of the present invention;

FIG6 is a schematic flow chart of a catheter robot registration method provided in yet another embodiment of the application;

FIG7 is a schematic flow chart of a registration method for a catheter robot provided in yet another embodiment of the application;

FIG8 is a schematic diagram of the effect of converting the coordinate system of the movement path of the catheter provided by an embodiment of the present application;

FIG9 is a schematic flow chart of a registration method for a catheter robot according to an embodiment of the present invention;

FIG10 is a schematic diagram of the effect of the catheter motion path before and after denoising according to an embodiment of the present application;

FIG11 is a schematic diagram showing the effect of the registration accuracy before the catheter motion path denoising provided by an embodiment of the present application;

FIG12 is a schematic diagram showing the effect of the registration accuracy after the catheter motion path denoising provided by an embodiment of the present application;

FIG. 13 is a schematic diagram showing the principle of the control device of the telemedicine system of the present application.

DETAILED DESCRIPTION

In order to facilitate the understanding of the present application, the present application will be described more comprehensively with reference to the relevant drawings below. The preferred embodiments of the present application are given in the drawings. However, the present application can be implemented in many different forms and is not limited to the embodiments described in the present application. On the contrary, the purpose of providing these embodiments is to make the disclosure of the present application more thoroughly and comprehensively understood.

It should be noted that when an element is referred to as being "disposed on" another element, it may be directly on the other element or there may be a central element. When an element is considered to be "connected" to another element, it may be directly connected to the other element or there may be a central element at the same time. When an element is considered to be "coupled" to another element, it may be directly coupled to the other element or there may be a central element at the same time. The terms "vertical", "horizontal", "left", "right" and similar expressions used in this application are for illustrative purposes only and do not represent the only implementation method. The terms "distal end" and "proximal end" used in this application are used as directional words, which are commonly used terms in the field of interventional medical devices, where "distal end" refers to the end close to the patient during surgery, and "proximal end" refers to the end away from the patient during surgery. The terms "first/second" and the like used in this application represent a component and a class of more than two components with common characteristics.

Unless otherwise defined, all technical and scientific terms used in this application have the same meaning as commonly understood by those skilled in the art to which this application belongs. The purpose of describing specific implementation methods is not intended to limit this application. The term "and/or" used in this application includes any and all combinations of one or more related listed items. The term "each" used in this application includes one or more than two. The term "plurality" used in this application refers to two or more.

The following describes the catheter robot registration method, the catheter robot, and the computer-readable storage medium proposed according to the embodiments of the present application in conjunction with the accompanying drawings.

FIG1 shows a catheter system 1000 provided in an embodiment of the present application. The catheter system 1000 includes an imaging vehicle 100, a trolley 200 and a main controller 300 respectively connected to the imaging vehicle 100, a catheter instrument 400 that can be coupled to the trolley 200, a sensor system 500 connected to the trolley 200, and a control system 600 for realizing control between the catheter instrument 400, the main controller 300, the sensor system 500 and the imaging vehicle 100. Among them, the main controller 300 can be connected to the trolley 200 by wire or wirelessly. When the operator performs various procedures on the patient next to the trolley 200, the control instruction can be triggered by operating the main controller 300, and the catheter instrument 400 can be controlled to advance, retract, bend and turn, etc. through the drive of the trolley 200.

The trolley 200 can usually be moved to the side of the operating bed to engage the catheter instrument 400, and control the catheter instrument 400 to move up and down in the vertical direction, or translate in the horizontal direction, or move in non-vertical and non-horizontal directions under the control command, so as to provide a better preoperative preparation angle for the operation of the catheter instrument 400. The control command can be a command triggered by the operator by operating the main controller 300, or a command triggered by the operator directly clicking or pressing a button set on the trolley 200. Of course, in other embodiments, the control command can also be a command triggered by voice control or a force feedback mechanism.

As shown in FIG1 , further, the trolley 200 may include a base 210, a sliding seat body 220 that can be lifted and moved along the base 210, and two mechanical arms 230 fixedly connected to the sliding seat body 220. The mechanical arm 230 may include a plurality of arm segments connected at a joint, and the plurality of arm segments provide the mechanical arm 230 with a plurality of degrees of freedom, for example, seven degrees of freedom corresponding to seven arm segments. A power unit (not shown in the figure) is installed at the end of the mechanical arm 230, and the power unit of the mechanical arm 230 is used to engage the catheter instrument 400 and to move the catheter instrument 400 under the driving force of the power unit. Under the action, the end of the catheter instrument 400 is controlled to bend and turn accordingly. The two mechanical arms 230 can be structures that are completely the same or partially the same, one mechanical arm 230 is used to engage the inner catheter instrument 410, and the other mechanical arm 230 is used to engage the outer catheter instrument 420. During installation, the outer catheter instrument 420 can be installed first, and when the outer catheter instrument 420 is installed, the catheter of the inner catheter instrument 410 is inserted into the catheter of the outer catheter instrument 420.

The sensor system 500 has one or more subsystems for receiving information about the catheter device 400. The subsystems may include: a position sensor system; a shape sensor system for determining the position, orientation, speed, velocity, pose, and/or shape of the tip of the catheter device 400 and/or along one or more segments of the catheter that may constitute the catheter device 400; and/or a visualization system for capturing images from the tip of the catheter device 400.

The imaging vehicle 100 may be provided with a display system 110 and a flushing system (not shown in the figure), etc. The display system 110 is used to display images or representations of the surgical site and the catheter instrument 400 generated by the subsystem of the sensor system 500. Real-time images of the surgical site and the catheter instrument 400 captured by the visualization system may also be displayed. Images of the surgical site recorded before or during surgery may also be presented using image data from imaging techniques such as computed tomography (CT), magnetic resonance imaging (MRI), optical coherence tomography (OCT), and ultrasound, etc. The preoperative or intraoperative image data may be presented as a two-dimensional, three-dimensional, or four-dimensional (such as time-based or rate-based information) image and/or presented as an image from a model created based on a preoperative or intraoperative image data set, and a virtual navigation image may also be displayed. In the virtual navigation image, the actual position of the catheter instrument 400 is registered with the preoperative image to present a virtual image of the catheter instrument 400 in the surgical site to the operator from the outside.

The control system 600 includes at least one memory and at least one processor. It is understood that the control system 600 can be integrated into the trolley 200 or the imaging trolley 100, or can be independently provided. The control system 600 can support wireless communication protocols such as IEEE 802.11, IrDA, Bluetooth, HomeRF, DECT, and wireless telemetry. The control system 600 can transmit one or more signals indicating the movement of the catheter instrument 400 by the power unit to move the catheter instrument 400. The catheter instrument 400 can extend to a surgical location in the body via an opening of a natural cavity or a surgical incision of the patient.

Further, the control system 600 may include a mechanical control system (not shown in the figure) and an image processing system (not shown in the figure), wherein the mechanical control system is used to control the movement of the catheter instrument 400, and therefore, can be integrated into the trolley 200. The image processing system is used for virtual navigation path planning, and therefore, can be integrated into the imaging vehicle 100. Of course, the various subsystems of the control system 600 are not limited to the specific situations listed above, and can also be reasonably set according to actual conditions. Among them, the image processing system can image the surgical site based on the image of the surgical site recorded before or during the operation, using the above imaging technology. The software used in combination with manual input can also convert the recorded image into a two-dimensional or three-dimensional synthetic image of a part or the entire anatomical organ or segment. During the virtual navigation procedure, the sensor system 500 can be used to calculate the position of the catheter instrument 400 relative to the patient's anatomical structure, which can be used to generate an external tracking image and an internal virtual image of the patient's anatomical structure, so as to realize the actual position of the catheter instrument 400 and the preoperative image registration, so that the virtual image of the catheter instrument 400 in the surgical site can be presented to the operator from the outside.

The structures of the inner catheter device 410 and the outer catheter device 420 are substantially the same, and they respectively have a slender and flexible inner catheter 41 and an outer catheter 42, wherein the diameter of the outer catheter 42 is slightly larger than the inner catheter 41, so that the inner catheter 41 can pass through the outer catheter 42 and provide a certain support for the inner catheter 41, so that the inner catheter 41 can reach the target position in the patient's body, so as to facilitate operations such as tissue or cell sampling from the target position.

Certain movements of the main controller 300 may cause corresponding movements of the catheter device 400. For example, when the operator operates the direction lever of the main controller 300 to move upward or downward, the movement of the direction lever of the main controller 300 may be mapped to the corresponding pitch movement of the end of the catheter device 400; when the operator operates the direction lever of the main controller 300 to move left or right, the movement of the direction lever of the main controller 300 may be mapped to the corresponding yaw movement of the end of the catheter device 400. In this embodiment, the main controller 300 may control the end of the catheter device 400 to move within a 360° spatial range.

2 shows a catheter instrument 400 provided in an embodiment of the present application. The catheter instrument 400 is configured to engage with the power unit 240 of the mechanical arm 230 , and includes an instrument box 45 configured to engage with the power unit 240 and a catheter 48 connected to the instrument box 45 . The “engagement” refers to a state in which the driving force of the power unit 240 can be transmitted to the inside of the instrument box 45 and the catheter 48 can move normally when the instrument box 45 is installed in the power unit 240. For example, under the driving force of the power unit 240, the end of the catheter 48 can be bent and turned.

The end in this application may also be referred to as the distal end or the head, which refers to the end away from the instrument box 45 ; the front end may also be referred to as the proximal end or the tail, which refers to the end close to the instrument box 45 .

The processor of the control system 600 is configured to perform the following steps to implement the catheter robot registration method provided in an embodiment of the present application. As shown in FIG3 , the method includes:

Step S11, obtaining a model point set including a plurality of simulated feature points from the anatomical model of the anatomical structure.

Each simulated feature point may correspond (match) to a feature point of the anatomical structure, and different simulated feature points generally correspond to different feature points of the anatomical structure. Each simulated feature point has coordinates in the image coordinate system of the anatomical model. The anatomical structure may be one of the natural or surgically created connecting channels such as bronchus, urinary tract, (cardio) blood vessels and intestines. This application is described by taking the anatomical structure as a bronchus as an example.

Feature points can be selected from points in the anatomical structure that have obvious features and are easy to identify. Simulated feature points are points in the anatomical model, which are the same as the points referred to by the feature points in the anatomical structure. For example, when the anatomical structure is a bronchus (usually referring to the bronchus of the lung), the feature points can be selected from various levels of carina, for example, the main carina, the first level and second level carina of the left lobe and the right lobe can be selected respectively.

When selecting feature points, it is possible to cover as many areas of the anatomical structure as possible. The wider the distribution range of the feature points in the anatomical structure and the greater the number of distribution, the more accurate the subsequent registration will be, and it is possible to avoid falling into local registration errors, that is, one part of the area is aligned and the rest of the area is not aligned. For example, when the anatomical structure is a pulmonary bronchus, multiple feature points can be selected to cover as many pulmonary bronchus as possible. For example, one or more feature points can be selected from one or more areas of the main carina, right upper lobe, right middle lobe, right lower lobe, left upper lobe, left middle lobe and left lower lobe. Exemplarily, one feature point can be selected from each of the above-mentioned areas to cover the entire bronchus. As shown in FIG4 , feature point 1 representing the main carina, feature point 2 representing the carina of the left upper lobe, and feature point 3 representing the carina of the left upper lobe are selected in the anatomical model. There are seven feature points in total, namely, feature point 2, feature point 3 representing the carina of the left middle lobe, feature point 4 representing the carina of the left lower lobe, feature point 5 representing the carina of the right upper lobe, feature point 6 representing the carina of the right middle lobe, and feature point 7 representing the carina of the right lower lobe.

When obtaining the simulated feature points in the model point set, these simulated feature points can be manually selected, or these simulated feature points can be automatically identified and selected.

Step S12: acquiring a spatial point set including a plurality of sampling point subsets from the anatomical structure, and acquiring a denoising weight matrix corresponding to the sampling point subsets in the spatial point set.

Each sampling point subset may correspond to (match) a feature point of the anatomical structure, different sampling point subsets may correspond to different feature points of the anatomical structure, and the sampling point subsets and the simulated feature points are associated through the feature points. Each sampling point subset includes multiple sampling points, and the sampling points are generally described by the coordinates of the catheter end or the instrument end in the world coordinate system (surgical environment coordinate system or physical space coordinate system). The sampling points can be acquired by a tracking sensor, and the tracking sensor can include a position sensor and/or a shape sensor.

For example, the visualization system can be used to guide the end of the catheter inserted into the anatomical structure and/or the end of the instrument to move to the feature point, and the position sensor set at the end of the catheter and/or the end of the instrument can be used to obtain the position of the end of the catheter and/or the end of the instrument. Among them, a position sensor can be a component in the electromagnetic positioning system. The electromagnetic positioning system can further include a magnetic field generating component and a magnetic field detecting component. The magnetic field generating component is used to generate a magnetic field, the position sensor is an EM sensor (i.e., an electromagnetic sensor), the position sensor will cause a change in the magnetic field in the magnetic field, and the magnetic field detection component can detect the change in the magnetic field to detect the position of the position sensor relative to the magnetic field/magnetic field generating component. Combined with the coordinate conversion relationship between the position sensor and the catheter end and/or the instrument end obtained by presetting or calibration, the position of the catheter end and/or the instrument end relative to the magnetic field/magnetic field generating component can be calculated, and then combined with the coordinate conversion relationship between the magnetic field/magnetic field generating component and the world coordinate system, the position of the catheter end and/or the instrument end in the world coordinate system, i.e., the sampling point, can be calculated. The position sensor can sense at least three translational degrees of freedom in the magnetic field.

For another example, a shape sensor system can be used to obtain the position of a catheter tip and/or an instrument tip inserted into an anatomical structure. For example, the shape sensor can include an optical fiber aligned with the catheter, and the optical fiber shape sensor can be used to obtain the position of a catheter tip and/or an instrument tip inserted into an anatomical structure. The fiber optic bending sensor can feedback the shape of the catheter, based on which the position of the catheter end and/or the instrument end relative to the base of the shape sensor can be calculated. Combined with the position of the base of the shape sensor in the world coordinate system, the position of the catheter end and/or the instrument end in the world coordinate system, i.e., the sampling point, can be calculated.

Some types of anatomical structures, such as bronchi, (cardio) blood vessels, etc., have intrinsic movements such as respiratory movement and cardiac movement with high frequency and large amplitude, which can easily cause large fluctuations in the position of the catheter end and/or the instrument end at the feature point, and thus easily lead to the problem of inaccurate subsequent registration. In some embodiments, at least for this type of anatomical structure, the catheter end and/or the instrument end can be manipulated to stay at the feature point in the anatomical structure for a certain period of time to continuously obtain its position. In some embodiments, the duration may include at least one movement cycle of the intrinsic movement of the anatomical structure. For example, for bronchi, the duration may include at least one respiratory cycle, such as one, two or more, and a normal respiratory cycle is usually 3 to 5 seconds. For another example, for (cardio) blood vessels, the duration may include at least one cardiac cycle, such as one, two or more, and a normal cardiac cycle is usually 0.5 to 1 second. The longer the sampling duration or the more movement cycles included, the more sampling points will be obtained during such a duration, and individual differences caused by different movement cycles can be avoided.

Exemplarily, when the anatomical structure is a bronchus, the end of the catheter and/or the end of the instrument are affected by respiratory motion at different amplitudes in different lung regions. For example, the fluctuation amplitude caused by the respiratory effect in the upper lobe region may be 10 mm, while the fluctuation amplitude caused by the respiratory effect in the lower lobe region may reach 20 mm. It can be seen that corresponding to the different positions of the end of the catheter and/or the end of the instrument in the anatomical structure, the influence of the intrinsic motion of the anatomical structure on the end of the catheter and/or the end of the instrument is usually different. Since the feature points are associated with the sampling point subsets, different denoising weight matrices can be constructed for different feature points or for different sampling point subsets, so that the sampling point subset can be denoised later by the denoising weight matrix corresponding to the sampling point subset to eliminate the influence of the intrinsic motion of the anatomical structure on the sampling points as much as possible.

In the present application, each selected feature point can be considered to represent an area of the anatomical structure. At different positions within the area, it can be simply considered that the intrinsic movement of the area has basically the same effect on the position of the catheter tip and/or the instrument tip.

In some embodiments, the number of simulated feature points in the model point set may be the same as or different from the number of the sampling point subset in the spatial point set. Typically, the simulated feature points in the model point set and the sampling point subset in the spatial point set may be associated with more than three feature points, such as four, five, six, seven or more, and the more associated feature points, the more accurate the registration.

Step S13, denoising the sampling points in the sampling point subset based on the denoising weight matrix, and determining the mean point of the sampling point subset based on the denoised sampling points.

Among them, all sampling points in the sampling point subset can be denoised separately through the denoising weight matrix corresponding to the sampling point subset. The mean point of the sampling point subset is a point representing the average position of all sampling points in the sampling point subset, which is actually a coordinate value described in the surgical environment coordinate system. Each sampling point subset corresponds to a mean point.

Step S14: determining a first transformation matrix between the model point set and the space point set based on the multiple simulated feature points of the model point set and the multiple mean points of the space point set.

Among them, the simulated feature points have a corresponding relationship with the feature points, and the mean point also has a corresponding relationship with the feature points. Then, the first transformation matrix between the model point set and the space point set can be determined through the one-to-one correspondence between multiple simulated feature points and multiple mean points, that is, the first registration relationship between the image coordinate system of the anatomical structure and the surgical environment coordinate system can be determined. The first registration relationship is a rough registration relationship. Exemplarily, the first transformation matrix can be determined by a landmark transform algorithm.

Since the selected feature points (i.e., landmark points) are relatively few and only the simulated feature points and mean points associated with the feature points are registered, the accuracy is relatively low, and the first transformation matrix determined in step S14 can be considered as a coarse registration matrix, or the first registration relationship determined can be considered as a coarse registration relationship. Among them, the coarse registration is a rigid registration, that is, the position (i.e., coordinate value) of the feature point only undergoes translation and rotation changes, and there is no scale change. The distance between any feature points remains unchanged in theory after the coordinate system is converted.

In the above steps S11 to S14, each sampling point in the sampling point subset associated with the feature point is denoised by using the denoising weight matrix associated with the feature point in the anatomical structure, and the mean point of the denoised sampling points in the sampling point subset is obtained, and then the simulated feature points based on the model point set and the mean of the spatial point set are obtained. The one-to-one correspondence between the points is used to determine the first transformation matrix, which can reduce or eliminate the influence of the intrinsic movement of the anatomical structure on the registration accuracy.

The anatomical model of the anatomical structure can generally be a three-dimensional or four-dimensional model. In some embodiments, taking the anatomical structure as a bronchus and the anatomical model as a three-dimensional model as an example, before step S11, a step of obtaining the anatomical model of the anatomical structure may also be included, and the implementation method of this step is as follows.

Images of anatomical structures are obtained by scanning (or photographing) such as CT, MRI, OCT or ultrasound. During the scan, the patient usually needs to take a deep breath and hold it until the scan is completed. The bronchi are largest and the fine branches are easier to image when the human body is in the inhalation state.

Segment and reconstruct the image of the anatomical structure. For example, the patient's image can be segmented by segmentation algorithms such as region growing and convolutional neural networks to segment the bronchi of the lungs; and then the segmented image can be reconstructed into a three-dimensional model by, for example, a marching cube algorithm.

In some embodiments, the denoising weight matrix corresponding to the sampling point subset obtained may include two sources.

The first one can come from the patient himself.

The second type can be derived from other patients with the same or similar physical characteristics as the patient. A certain deviation is allowed. For example, the allowable deviation is defined as within 5%. For patients with the same or similar physical characteristics, the amplitude of the intrinsic movement of the anatomical structure is basically the same for the same feature point in the anatomical structure, and the influence on the position of the catheter end and/or the instrument end is basically the same. Among them, the physical characteristics include, for example, body shape characteristics and/or physiological characteristics. When the anatomical structure is a bronchus or a (cardio)vascular vessel, the body shape characteristics include at least the chest circumference; when the anatomical structure is a bronchus, the physiological characteristics include at least the amplitude of the respiratory movement; when the anatomical structure is a (cardio)vascular vessel, the physiological characteristics include at least the amplitude of the heartbeat movement.

In some embodiments, for the case where the denoising weight matrix is derived from the second type, as shown in FIG5 , the above step S12 may include:

Step S1201, obtaining the patient's physical characteristics.

Physical characteristics include, for example, body shape characteristics and/or physiological characteristics.

Step S1202, determining the type of the patient's anatomical structure.

The types of anatomical structures include, for example, bronchi or (cardio)vascular tubes, and of course, may also include Other organs. The type can be determined by manual input or automatic identification.

Step S1203, based on the patient's physical features and anatomical structure type, target patients with the same anatomical structure type and the same or similar physical features are matched.

Among them, one or more databases can be constructed, which store information of multiple patients who have been treated in this hospital or other hospitals. This information includes but is not limited to the patient's name, gender, age, physical characteristics, physiological characteristics of anatomical structures, and the same or different denoising weight matrices of different feature points of one or more anatomical structures. Therefore, it is advantageous to achieve accurate matching through the same or different denoising weight matrices of different feature points of physical characteristics, physiological characteristics, and one or more anatomical structures.

Step S1204: determining feature points associated with the patient's sampling point subset.

Step S1205: Based on the feature points, a denoising weight matrix of the target patient associated with the feature points is matched.

Among them, the feature points of the same anatomical structure of different patients are usually selected to have the same or great similarity. Taking the bronchus as an example, these feature points are usually selected from the most easily recognizable bifurcation points of the bronchus at all levels, that is, the carina at all levels. Exemplarily, when the feature points of the sampling point subset correspond to the main carina, the denoising weight matrix matched in step S1204 should also be the denoising weight matrix of the main carina of the target patient; when the feature points of the sampling point subset correspond to the first-level carina of the left lung lobe, the denoising weight matrix matched in step S1204 should also be the denoising weight matrix of the first-level carina of the left lung lobe of the target patient.

In some embodiments, for the case where the denoising weight matrix is derived from the second type, as shown in FIG6 , the above step S12 also includes:

Step S1201', obtaining the patient's physical characteristics.

Step S1203', based on the patient's physical characteristics, match target patients with the same or similar physical characteristics.

Step S1204', determining the characteristic points associated with the patient's sampling point subset.

Step S1205 ′: based on the feature points, a denoising weight matrix of the target patient associated with the feature points is matched.

Furthermore, in the above step S13, the sampling points in the sampling point subset may be denoised using the matched denoising weight matrix.

Through the above steps S1201 to S1205, there is no need to reconstruct the denoising weight matrix of different feature points of the anatomical structure for the patient himself, and the denoising weight matrix of the corresponding feature points of the existing anatomical structure of other patients with the same or similar body features as the patient can be used to greatly reduce the preoperative preparation time, which at least includes the time for arranging sensors for detecting the intrinsic movement of the anatomical structure. Of course, in a scenario with more time, it is also possible to take into account the individual differences between patients and construct the denoising weight matrix of different feature points of the anatomical structure for the patient alone.

In some embodiments, as shown in FIG. 7 , whether based on the patient itself or on other patients, for each of the plurality of feature points, a method for initially constructing a denoising weight matrix for each feature point may include:

Step S1211: acquiring a sampling point subset corresponding to a feature point from the anatomical structure.

The sampling point subset includes multiple sampling points, and the multiple sampling points are obtained by sampling the same tracking sensor at different times. In this embodiment, the multiple sampling points generally include sampling points in the order of ten, one hundred or even more.

Step S1212: acquiring a test point set from the body surface corresponding to the anatomical structure.

Each test point set includes multiple test point subsets, and each test point subset includes at least one test point. The test point is sampled by a surface sensor arranged on the patient's body surface corresponding to the anatomical structure. The test point is generally described by the coordinates of the surface sensor in the surgical environment coordinate system. A corresponding test point of different test point subsets is usually sampled by a corresponding surface sensor at different times.

Corresponding to the expected number of test points in each test point subset, the number of body surface sensors is configured in the same manner. For example, when the expected number of test points in each test point subset is one, the number of body surface sensors is one; for another example, when the expected number of test points in each test point subset is two or three, the number of body surface sensors is correspondingly two or three.

In this embodiment, the plurality of test point subsets generally include test point sets in the order of ten, one hundred or even more.

In some embodiments, the number of sampling points in the sampling point subset is usually required to be the same as that in the test point set. The number of point subsets is the same for subsequent calculations. However, this does not necessarily require that the same number must be obtained by sampling, and the same number can also be maintained by data processing such as interpolation, filtering, etc.

In some embodiments, for the same feature point, sampling can be performed separately and successively to obtain a sampling point subset and a test point set respectively; however, from the perspective of saving preoperative time, sampling can also be performed simultaneously and at the same frequency to obtain a sampling point subset and a test point set.

The body surface sensor used to sense the intrinsic movement of the anatomical structure can generally be a position sensor or a posture sensor. The body surface sensor is usually non-invasively and stably set on the patient's body surface corresponding to the anatomical structure, that is, the body surface sensor is generally exposed on the patient's body surface. Therefore, compared with the tracking sensor used to sense the position of the catheter end and/or the instrument end mentioned above, there are relatively more types of body surface sensors to choose from. For example, the body surface sensor can not only use the EM sensor as mentioned above, but also can use optical positioning sensors.

In some embodiments, when the anatomical structure is a bronchus or a (cardio)vascular vessel, at least one body surface sensor is arranged on the patient's chest. Exemplarily, the body surface sensors may include three, the first body surface sensor may be arranged in the middle of the patient's chest, the second body surface sensor may be arranged in the area of the patient's chest corresponding to the seventh rib on the left, and the third body surface sensor may be arranged in the area of the patient's chest corresponding to the seventh rib on the right. The more body surface sensors there are, the more areas of the patient's body surface are covered, and the more accurately the intrinsic movement of the anatomical structure is sensed.

Step S1213, based on the test point set and the sampling point subset in the same surgical environment coordinate system, determine a denoising weight matrix corresponding to the sampling point subset.

In some embodiments, the body surface sensor and the tracking sensor are EM sensors, and the two may belong to the same electromagnetic positioning system or different electromagnetic positioning systems. When the two belong to the same electromagnetic positioning system, the positions sensed by the two are in the same surgical environment coordinate system; when the two belong to different electromagnetic positioning systems, the positions sensed by the two may also be transformed to the same surgical environment coordinate system by coordinate conversion. In this case, the surgical environment coordinate system may also be referred to as an electromagnetic coordinate system.

In some embodiments, the above step S1213 can be more specifically implemented by the following steps:

For the convenience of description, the position data of the test point set obtained by the body surface sensor (i.e., the test point) is defined as The body surface sensor data Ref is defined, and the position data of a subset of sampling points acquired by the tracking sensor (ie, sampling points) is defined as the in-vivo catheter data Tube.

(1) De-average the body surface sensor data Ref to obtain the body surface sensor data And de-average the in vivo catheter data Tube to obtain the in vivo catheter data

(2) Obtaining body surface sensor data The inverse solution of . For example, the inverse solution can be obtained by SVD (Singular Value Decomposition) method. The obtained inverse solution can usually be expressed as:

in, is the inverse solution of the body surface sensor data, V is the right singular matrix, S is the singular value matrix, and U is the left singular matrix.

(3) Inverse solution based on body surface sensor data and in vivo catheter data Determine the denoising weight matrix of the corresponding feature point. The denoising weight matrix associated with the corresponding feature point can be expressed as:

Among them, Weight is the denoising weight matrix.

In some embodiments, still taking the anatomical structure of the bronchus as an example, three body surface sensors can be used to obtain a test point set in more than two respiratory cycles (usually 6 to 10 seconds), and a subset of sampling points associated with the feature points can be simultaneously obtained through the tracking sensor. Among them, the body surface sensor data Ref is a matrix of N rows and 9 columns, and the format of each row can be expressed as {X1, Y1, Z1, X2, Y2, Z2, X3, Y3, Z3}, for example, and each 3 columns correspond to a position data sensed by a body surface sensor. The in vivo catheter data Tube is a matrix of N rows and 3 columns, and the format of each row can be expressed as {X0, Y0, Z0}, for example, and the 3 columns correspond to the position data of the catheter end and/or the instrument end sensed by the tracking sensor.

Furthermore, each feature point corresponds to a certain denoising weight matrix. Exemplarily, the data structure ST = {PosTube, PosRef, Weight} can be used for storage. Among them, PosTube is the mean value of the sampling points in the sampling point set corresponding to each feature point. For example, when an electromagnetic positioning system is used, the mean value is the mean value of the electromagnetic coordinate points at the end of the catheter and/or the end of the instrument, as shown in ; PosRef is the mean value of the test points in the test point set corresponding to each feature point. If an electromagnetic positioning system is used, the mean The value is the average of the electromagnetic coordinate points of the three body surface sensors, as shown in Weight is a matrix with 9 rows and 3 columns.

In order to facilitate the implementation of the above step S13, when acquiring the data of the sampling points corresponding to a certain feature point, the control device can mark these sampling points or the sampling point subsets corresponding to these sampling points, for example, assign serial numbers to establish the association between the sampling points or the sampling point subsets and the corresponding feature points. Furthermore, when determining the corresponding denoising weight matrix based on the association between the feature points and the sampling points (or sampling point subsets) and the denoising weight matrix, the sampling points (or sampling point subsets) and the denoising weight matrix can be matched extremely conveniently and quickly.

In some embodiments, when denoising the sampling points in the sampling point subset based on the denoising weight matrix, for the same feature point, the sampling point can be denoised by combining a corresponding pair of sampling points and a test point, the denoising weight matrix weight, and the mean value PosRef of the test points in the test point set. Exemplary examples include:

(1) Determine the offset between the current test point associated with the same feature point and the mean of the test points in the test point set. For example, the offset can be determined by the following formula:

in, is the offset, and Ro is the current test point. Exemplarily, when the number of body surface sensors is 3, the offset is a row vector containing 9 elements.

(2) De-noising the current sampling point based on the feature vector associated with the same feature point, the current sampling point, and the denoising weight matrix. For example, denoising can be performed using the following formula:

Wherein, Tf is the current sampling point after denoising; To is the current sampling point.

In this embodiment, a corresponding pair of sampling points and test points may refer to sampling points and test points that are sampled at the same moment in the same motion cycle of the anatomical structure, for example, the pair of points are collected at the same moment in the same motion cycle; or may refer to sampling points and test points that are sampled at the same moment in different motion cycles of the anatomical structure.

In some embodiments, referring to FIG. 3 , after step S14, the method further includes:

Step S15, obtaining a path point cloud from the anatomical structure.

The path point cloud includes a plurality of actual path points sensed in the anatomical structure by the first sensor.

A path point cloud may also be referred to as a path point set, and includes multiple actual path points that the catheter end and/or the instrument end passes through in the anatomical structure. An actual path point reflects the position of the catheter end and/or the instrument end when the sensor feedbacks. The actual path points obtained by multiple feedbacks from the same sensor are arranged according to the feedback time to obtain the movement path of the catheter end and/or the instrument end. For example, the first path point cloud may include not only the sampling point, but also other points different from the sampling point. For another example, the first path point cloud may only include other points different from the sampling point.

The simulated feature points and the actual path points may be obtained in one process, or may be obtained in different processes.

Step S16, obtaining a model point cloud from the anatomical model.

The model point cloud includes multiple skeleton points of the pipeline centerline of the anatomical model and/or multiple vertices of the pipeline wall. The skeleton points can be obtained, for example, by first extracting the pipeline centerline from the anatomical model and then discretizing the pipeline centerline to obtain these skeleton points. The vertices of the pipeline wall can be obtained, for example, by first extracting the pipeline wall from the anatomical model and then discretizing the pipeline wall to obtain these pipeline wall vertices.

Step S17, determining a second transformation matrix of the path point cloud and the model point cloud based on the first transformation matrix, a plurality of actual path points of the path point cloud, and a plurality of skeleton points of the model point cloud.

In step S17, by performing point cloud registration on the path point cloud and the model point cloud, the second transformation matrix between the surgical environment coordinate system and the anatomical model, i.e., the second registration relationship, can be determined. The second registration relationship is a precise registration relationship. Since the first transformation matrix obtained in the coarse registration stage has a high precision, the precision of the second transformation matrix obtained in the precise registration stage is also relatively high.

Since the registration uses relatively more points and has relatively higher precision, the second transformation matrix determined in step S17 can be considered as a precise registration matrix, or the determined second registration relationship can be considered as a precise registration relationship.

The representative algorithm for point cloud registration is the Iterative Closest Point (ICP) algorithm. The core idea is to minimize the distance between two point sets and make the two point sets close to each other through iteration. The following is a brief introduction to the principle of ICP.

The basic method of ICP includes two steps: 1. Match point clouds Q and P to find the corresponding point pairs between them; 2. Calculate the transformation matrix between point clouds Q and P based on the corresponding point pairs. A corresponding point pair consists of two points, one from point cloud P and the other from point cloud Q. These two points are each other's matching points, and these two points are considered to be corresponding, that is, the two points are essentially equivalent.

If the real and accurate corresponding point pairs can be found directly, then the above process only needs to be performed once to directly calculate a sufficiently accurate transformation matrix. However, it is difficult to directly find accurate corresponding point pairs in practical applications, so ICP will match according to the principle of the closest distance (generally Euclidean distance), that is, for each point in the point cloud P, find the point closest to it in the point cloud Q as its matching point. Then calculate a transformation matrix based on these point pairs. After completing a round of calculations, ICP will determine whether the iteration stop condition is met. If not, the transformation matrix obtained by this round of calculations is used to transform and update the point cloud P, and then the point cloud Q and the transformed point cloud P are used to repeat the above process until the iteration stops. Exemplarily, the default condition for stopping the iteration is that the difference in error between the previous and next two iterations is less than 0.01, but sometimes the difference between the point clouds is so large that the convergence condition cannot be met, so it is necessary to set an upper limit on the number of iterations, for example, it can be set to 500. For example, the following formula can be used for iteration or applied to virtual navigation:
PosTubeCT=MatrixFine*MatrixCoarse*PosTubeEM formula (5)

Among them, PosTubeEM is the position of the catheter end and/or the instrument end in the surgical environment coordinate system, PosTubeCT is the position of the catheter end and/or the instrument end in the image coordinate system of the anatomical model, MatrixCoarse is the first transformation matrix, and MatrixFine is the second transformation matrix.

Referring to Figure 8, the motion path of the catheter tip and/or the instrument tip in the surgical environment coordinate system is located in the upper part of Figure 8. After each actual path point in the motion path is first multiplied by the rough registration matrix and then by the fine registration matrix, the motion path is basically located inside the planned bronchial anatomical model. The more iterations, the higher the registration accuracy will be.

Since ICP is sensitive to initial values, if the initial difference between point clouds P and Q is large, directly using Convergence may not be guaranteed using ICP. Therefore, point cloud registration can be divided into two parts: coarse registration from step S11 to step S14 and fine registration from step S15 to step S18. First, coarse registration is used to obtain an initial transformation matrix (i.e., the first transformation matrix) to roughly align the two sets of point clouds. Then, ICP is used for fine registration, using more points for refined alignment. At this time, the first round of ICP is to match the point cloud Q with the point cloud P transformed by the initial transformation matrix.

In some embodiments, as shown in FIG. 9 , step S15 may include:

Step S151, obtaining the current actual path point.

The first sensor acquires the actual path point in real time, and the actual path point acquired at the current moment can be defined as the current actual path point. As time changes, the current actual path point will be continuously updated.

Step S152, determining the feature point closest to the current actual path point.

That is, it is necessary to determine the feature point closest to each actual path point. For example, the Euclidean distance between each actual path point and the mean PosTube of the sampling points in the sampling point set corresponding to each feature point can be compared to obtain the feature point closest to it.

Step S153, determining a denoising weight matrix associated with the feature point.

In some embodiments, a KD search tree (KD-tree) can be constructed with the mean point PosTube of the sampling points in the sampling point subset corresponding to all feature points as a node for subsequent retrieval. The KD tree constructed here is a key-value search tree in a three-dimensional Euclidean space, which is used for range search and nearest neighbor search. By inputting the real-time position of the catheter end and/or the instrument end (i.e., the real-time actual path point), the feature point corresponding to the nearest node can be quickly queried through the KD search tree, and then the denoising weight matrix corresponding to the nearest feature point can be obtained. By constructing a KD search tree, the process of obtaining the denoising weight matrix is greatly accelerated.

Step S154, denoising the current actual path point based on the denoising weight matrix, obtaining the denoised current actual path point and adding it to the path point cloud.

Wherein, each actual path point can be denoised by using the formula (4) and the formula (5).

Furthermore, since the actual path points in the path point cloud are all subjected to denoising, the second transformation matrix determined by calculating the path point cloud including the denoised actual path points and the model point cloud is The array has a higher registration accuracy.

In Figures 10 to 12, the point cloud composed of hollow circle-shaped points represents the catheter movement path before denoising (i.e., before compensation), the point cloud composed of five-pointed star-shaped points represents the catheter movement path after denoising, and the continuous path in the middle is the path spliced together from multiple bronchial centerlines.

As shown in FIG. 10 , it can be seen that the denoised point cloud has a smaller deviation (ie, better convergence) relative to the bronchial centerline, which means that the influence of respiratory motion is greatly weakened.

As shown in Figures 11 and 12, the comparison of the registration accuracy before and after denoising is shown. In Figure 11, the registration accuracy before compensation is approximately 6.5 mm. In Figure 12, the registration accuracy after compensation is approximately 3.2 mm. It can be seen that the denoising process greatly improves the registration accuracy, and the fluctuation of the catheter position is greatly reduced, making it easier to observe the position of the catheter in the anatomical model.

In some embodiments, when performing virtual navigation, the actual path points of the catheter tip in the anatomical structure sensed by the first sensor can be obtained; based on the first transformation matrix and/or the second transformation matrix, the simulated path points of the actual path points in the anatomical model can be determined; a marker graphic representing the catheter tip and/or the instrument tip is generated; and the marker graphic is displayed at the simulated path points. Exemplarily, after the fine registration iteration ends, the second transformation matrix can be used to determine the simulated path points of the actual path points in the anatomical model; after the fine registration starts and before it ends, the first transformation matrix and the second transformation matrix can be used to determine the simulated path points of the actual path points in the anatomical model; before the fine registration starts, the first transformation matrix can be used to determine the simulated path points of the actual path points in the anatomical model.

The marking graphic may be, for example, a sphere or any other shape.

The method of the present application has the following beneficial effects:

It can improve the accuracy of the registration algorithm. The position of the catheter end and/or the instrument end displayed on the anatomical model is closer to the actual position. Higher accuracy is helpful to determine whether the catheter end and/or the instrument end (such as a biopsy needle) has reached the lesion, reduce the number of X-rays taken during surgery, and reduce the radiation intake of doctors and patients.

It can improve the stability of the registration algorithm. The combination of coarse registration and fine registration greatly improves the robustness of the algorithm and reduces the possibility of prolonged surgery or surgical failure due to registration failure.

The stability of the data obtained by the tracking sensor can be improved. The data obtained by the tracking sensor after denoising filters out the jitter caused by the intrinsic movement of the anatomical structure. After being converted into the image coordinate system of the anatomical model, it can be displayed in a certain position of the anatomical model in a stable form without fluctuating and drifting everywhere, making it more convenient for users to observe the current position of the catheter tip and/or the instrument tip.

The present application also provides a control device. As shown in FIG13 , the control device may include: a processor 501, a communications interface 502, a memory 503, and a communication bus 504.

The processor 501 , the communication interface 502 , and the memory 503 communicate with each other via the communication bus 504 .

The communication interface 502 is used to communicate with other devices such as various sensors, rotating motors, solenoid valves, or network elements of other clients or servers.

The processor 501 is used to execute the program 505, and specifically can execute the relevant steps in the above method embodiment.

Specifically, the program 505 may include program codes, which include computer operation instructions.

The processor 505 may be a central processing unit CPU, or an application specific integrated circuit ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of the present application, or an FPGA, or a graphics processing unit GPU (Graphics Processing Unit). The one or more processors included in the control device may be processors of the same type, such as one or more CPUs, or one or more GPUs; or they may be processors of different types, such as one or more CPUs and one or more GPUs.

The memory 503 is used to store the program 505. The memory 503 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk storage.

The program 505 may be specifically used to enable the processor 501 to perform the following operations:

Acquire a model point set including a plurality of simulated feature points, wherein the simulated feature points are acquired from an anatomical model of an anatomical structure of a first patient, and the plurality of simulated feature points are respectively associated with a plurality of feature points of the anatomical structure;

Acquire, by a first sensor, a spatial point set including a plurality of sampling point subsets, wherein the sampling point subsets include a plurality of sampling points, the sampling points are acquired from the anatomical structure, and the plurality of sampling point subsets are respectively associated with a plurality of the feature points;

Acquire a plurality of denoising weight matrices corresponding to a plurality of the sampling point subsets, wherein the plurality of denoising weight matrices are respectively associated with a plurality of the feature points;

Based on the plurality of denoising weight matrices, denoising the sampling points in the corresponding plurality of sampling point subsets respectively;

Based on the denoised sampling points of the plurality of sampling point subsets, respectively determining mean points corresponding to the plurality of sampling point subsets;

Based on the plurality of simulated feature points and the plurality of mean value points, a first transformation matrix between the model point set and the space point set is determined.

The present application also provides a computer-readable storage medium storing a computer program, wherein the computer program is configured to be loaded by a processor and executed to implement the steps of the method described in any of the above embodiments.

The present application also provides a computer program product, including computer instructions, which, when executed on a computer, enable the computer to execute the method described in any one of the above embodiments.

The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims (18)

A catheter robot, characterized by comprising: a catheter, wherein at least one of the catheter and an instrument carried by the catheter is provided with a first sensor for sensing a position of at least one of a distal end of the catheter and a distal end of the instrument; and A control device, coupled to the first sensor and configured to: Acquire a model point set, the model point set comprising a plurality of simulated feature points, the simulated feature points being acquired from an anatomical model of an anatomical structure of a first patient, the plurality of simulated feature points being respectively associated with a plurality of feature points of the anatomical structure; Acquire a spatial point set, wherein the spatial point set includes a plurality of sampling point subsets, wherein the sampling point subsets include a plurality of sampling points, wherein the sampling points are acquired from the anatomical structure by the first sensor, and the plurality of sampling point subsets are respectively associated with a plurality of the feature points; Acquire a plurality of denoising weight matrices corresponding to a plurality of the sampling point subsets, wherein the plurality of denoising weight matrices are respectively associated with a plurality of the feature points; Based on the plurality of denoising weight matrices, denoising the sampling points in the corresponding plurality of sampling point subsets respectively; Based on the denoised sampling points of the plurality of sampling point subsets, respectively determining mean points corresponding to the plurality of sampling point subsets; Based on the plurality of simulated feature points and the plurality of mean points, a rough registration relationship between the model point set and the space point set is determined. The catheter robot according to claim 1, characterized in that the control device is configured to: Acquire a path point cloud, the path point cloud comprising a plurality of actual path points sensed in the anatomical structure by the first sensor; Acquire a model point cloud, wherein the model point cloud includes at least one of a plurality of skeleton points of a pipeline centerline of the anatomical model and a plurality of vertices of a pipeline wall; The path point cloud is determined based on the rough registration relationship, the actual path points in the path point cloud, and at least one of the skeleton points and the vertices of the pipeline wall in the model point cloud. The precise registration relationship between the radius point cloud and the model point cloud. The catheter robot according to claim 2, characterized in that the step of obtaining a path point cloud comprises: Get the current actual path point; Determine the feature point that is closest to the current actual path point; Determining the denoising weight matrix associated with the feature point; The current actual path point is denoised based on the denoising weight matrix to obtain the denoised current actual path point and add it to the path point cloud. The catheter robot according to claim 3, characterized in that the step of determining the feature point closest to the current actual path point comprises: Respectively determine mean points of the sampling points in the plurality of sampling point subsets associated with the plurality of feature points; Using the plurality of mean points as nodes, constructing a KD search tree; Based on the KD search tree, the feature point closest to the current actual path point is determined. The catheter robot according to any one of claims 2 to 4, characterized in that after determining the precise registration relationship between the second path point cloud and the model point cloud, it further comprises: acquiring an actual path point of the catheter tip in the anatomical structure sensed by the first sensor; Determining a simulated path point of the actual path point in the anatomical model based on at least one of the coarse registration relationship and the fine registration relationship; generating a marker pattern representative of at least one of the catheter tip and the instrument tip; The marker graphic is displayed at the simulated path point. The catheter robot according to claim 1, characterized in that the step of obtaining a plurality of denoising weight matrices corresponding to a plurality of the sampling point subsets comprises: Acquiring a physical characteristic of the first patient, wherein the physical characteristic comprises at least one of a body shape characteristic and a physiological characteristic of the anatomical structure; matching a second patient having the same or similar physical characteristics; determining the feature points associated with the subset of sampling points of the first patient; The denoising weight matrix associated with the feature point of the second patient is matched, wherein a plurality of feature points of the anatomical structure of the second patient are respectively associated with a plurality of the denoising weight matrices. The catheter robot according to claim 1, characterized in that the step of obtaining a plurality of denoising weight matrices corresponding to a plurality of the sampling point subsets comprises: Acquiring a physical characteristic of the first patient, wherein the physical characteristic comprises at least one of a body shape characteristic and a physiological characteristic of the anatomical structure; determining a type of anatomy of the first patient; matching a second patient with the same type and the same or similar physical features; determining the feature points associated with the subset of sampling points of the first patient; The denoising weight matrix associated with the feature point of the second patient is matched, wherein a plurality of feature points of different types of anatomical structures of the second patient are respectively associated with a plurality of the denoising weight matrices. The catheter robot according to claim 6 or 7 is characterized in that the physical characteristics include at least one of body shape characteristics and physiological characteristics, the body shape characteristics include chest circumference, and the physiological characteristics include at least one of the amplitude of respiratory movement and the amplitude of heartbeat movement. The catheter robot according to claim 1, characterized in that the catheter robot further comprises at least one second sensor, the second sensor being coupled to the control device and being configured on the patient's body surface for sensing the intrinsic movement of the patient's anatomical structure, and the control device being further configured to: Acquire a sampling point subset of the patient, wherein the sampling point subset of the patient includes a plurality of sampling points, the sampling points are acquired from the anatomical structure of the patient by the first sensor, and the sampling point subset is associated with a feature point of the patient; Acquire a test point set of the patient, wherein the test point set of the patient includes a plurality of test points, wherein the test points are acquired from the body surface of the patient by the second sensor, and the sampling points and the test points are in the same surgical environment coordinate system or are in the same surgical environment coordinate system after conversion; The denoising weight matrix associated with the feature points of the patient is determined based on the sampling point subset and the test point set obtained from the patient, wherein the patient is the first patient or the second patient. The catheter robot according to claim 9, characterized in that the determining the denoising weight matrix associated with the feature points of the patient based on the sampling point subset and the test point set obtained from the patient comprises: Performing a de-averaging process on a plurality of test points in the test point set obtained from the patient, and performing a de-averaging process on a plurality of sampling points in the sampling point subset obtained from the patient; Determining an inverse solution of the test point set after the de-averaging process; Based on the sampling point subset subjected to the de-averaging process and the inverse solution of the test point set, the denoising weight matrix corresponding to the sampling point subset obtained from the patient is determined. The catheter robot according to claim 9, characterized in that the anatomical structure is a bronchus, the intrinsic motion is a respiratory motion, and the acquisition cycle for acquiring the test point set of the patient includes at least one respiratory cycle; or The anatomical structure is a blood vessel, the intrinsic motion is a cardiac motion, and the acquisition cycle for acquiring the test point set of the patient includes at least one cardiac cycle. The catheter robot according to claim 11 is characterized in that when the anatomical structure is a bronchus, the feature points include multiple of the main carina of the bronchus, the first-level carina of the left lobe, the second-level carina of the left lobe, the first-level carina of the right lobe, and the second-level carina of the right lobe. The catheter robot according to claim 11 is characterized in that when the anatomical structure is a bronchus, the feature points include one or more feature points in one or more areas of the main carina, right upper lobe, right middle lobe, right lower lobe, left upper lobe, left middle lobe and left lower lobe. The catheter robot according to claim 9 is characterized in that, when the anatomical structure is a bronchus, the anatomical model is an anatomical model obtained when the human body is in an inhalation state. The catheter robot according to claim 9, characterized in that The first sensor includes an electromagnetic sensor or a shape sensor, and the second sensor includes an electromagnetic sensor or an optical positioning sensor. The catheter robot according to claim 1, characterized in that the anatomical structure includes a urinary tract or an intestine. A catheter robot registration method, characterized by comprising: Acquire a model point set including a plurality of simulated feature points, wherein the simulated feature points are acquired from an anatomical model of an anatomical structure of a first patient, and the plurality of simulated feature points are respectively associated with a plurality of feature points of the anatomical structure; Acquire, by a first sensor, a spatial point set including a plurality of sampling point subsets, wherein the sampling point subsets include a plurality of sampling points, the sampling points are acquired from the anatomical structure, and the plurality of sampling point subsets are respectively associated with a plurality of the feature points; Acquire a plurality of denoising weight matrices corresponding to a plurality of the sampling point subsets, wherein the plurality of denoising weight matrices are respectively associated with a plurality of the feature points; Based on the plurality of denoising weight matrices, denoising the sampling points in the corresponding plurality of sampling point subsets respectively; Based on the denoised sampling points of the plurality of sampling point subsets, respectively determining mean points corresponding to the plurality of sampling point subsets; Based on the plurality of simulated feature points and the plurality of mean points, a rough registration relationship between the model point set and the space point set is determined. A computer-readable storage medium, characterized in that the computer-readable storage medium stores a computer program, and the computer program is configured to be loaded and executed by a processor to implement the steps of the method according to claim 17.