Training data processing method for neurosurgery MR teaching system

Abstract

The application provides a kind of training data processing method, system, equipment and medium for neurosurgery MR teaching system.The method determines target neurosurgery training task by responding to training scene selection instruction in MR space, and presents target surgical area corresponding to target neurosurgery training task in MR space, then, based on medical image data, anatomical structure segmentation and spatial registration are performed on target surgical area, to obtain three-dimensional voxel data of at least one key neural structure object, then, based on three-dimensional voxel data and user operation behavior data in MR space, real-time interactive simulation is carried out on target neurosurgery training task to generate simulation process data, finally, according to simulation process data, training evaluation information is obtained, and training evaluation information is displayed in MR space in a visual manner, so that the authenticity and safety of neurosurgery training are effectively improved through visual display and interaction in MR space.

Description

Technical Field

[0001] This application relates to data processing technology, and more particularly to a method for processing training data for a neurosurgical MR teaching system. Background Technology

[0002] Neurosurgery, as a highly specialized discipline in the medical field, has long faced numerous challenges in its teaching and surgical practice. In terms of teaching, the traditional model mainly relies on cadaver dissection and two-dimensional imaging data.

[0003] Among these challenges, cadaver dissection resources are scarce and costly, making it difficult to meet the needs of large-scale teaching and resulting in limited practical opportunities for students. While two-dimensional imaging can provide some anatomical information, it cannot intuitively present the complex three-dimensional structure of the brain, such as the spatial relationships between blood vessels, nerves, and lesions, which students find difficult to understand, especially those with weak spatial imagination. In surgical practice, real surgical operations are extremely risky, and students are prone to making mistakes due to lack of experience, which not only endangers patient safety but also affects their learning motivation.

[0004] Therefore, there is an urgent need for a neurosurgical data processing method that can achieve intuitive spatial cognition, realistic interactive simulation, and safe operation. Summary of the Invention

[0005] This application provides a training data processing method for a neurosurgical MR teaching system, which enables visualization and real interaction in MR (Mixed Reality) space, effectively improving the realism and safety of neurosurgical training.

[0006] Firstly, this application provides a method for processing training data for a neurosurgical MR teaching system, including: In the MR space used for neurosurgical teaching, in response to the training scene selection command in the MR space, the target neurosurgical training task is determined, and the target surgical area corresponding to the target neurosurgical training task is presented in the MR space. Based on medical imaging data, anatomical structure segmentation and spatial registration are performed on the target surgical area to obtain three-dimensional voxel data of at least one key neural structure object. Based on the three-dimensional voxel data and the user operation behavior data in the MR space, the target neurosurgical training task is subjected to real-time interactive simulation to generate simulation process data. Based on the simulation process data, training evaluation information is obtained and displayed in a visual manner in the MR space.

[0007] Optionally, the step of performing anatomical segmentation and spatial registration on the target surgical region based on medical imaging data includes: Determine reference volume data for the target surgical region; In a preset object coordinate system, fine-grained segmentation is performed on the target surgical area to obtain one or more neural key structure objects, including lesion objects, functional area objects, and dangerous blood vessel objects. The preset object coordinate system is transformed with the world coordinate system of the MR space to obtain the three-dimensional voxel data of the neural key structure object in the MR space.

[0008] Optionally, the step of performing real-time interactive simulation of the target neurosurgical training task based on the three-dimensional voxel data and the user operation behavior data in the MR space to generate simulation process data includes: A tissue biomechanical model of the target surgical area is constructed based on the three-dimensional voxel data; Based on the user's operational behavior data in the MR space, the contact area and contact depth between the virtual surgical instrument object and the tissue mechanics model are determined; The amount of tissue deformation and / or resection is determined based on the contact area and the contact depth to update the local deformation of the three-dimensional neuroanatomical model; The updated 3D neuroanatomical model and the virtual surgical instrument object are rendered in real time in the MR space to generate continuous interactive simulation screens.

[0009] Optionally, the tissue mechanics model includes one or more mechanical models of brain tissue, blood vessel walls, and tumor tissue constructed based on the finite element method, mass spring model, and / or position-based dynamics model.

[0010] Optionally, the generation of simulation process data includes: Based on the sequence of trajectory points of the virtual surgical instrument object in the MR space, an instrument trajectory curve is formed; Based on the tissue mechanics model and the instrument trajectory curve, the contact force, cutting force and tensile force applied to the virtual surgical instrument object are determined to form a sequence of operational mechanics parameters. Based on the spatial relationship between the instrument trajectory curve and the key neural structure, contact events of the key structure at each moment are detected and recorded.

[0011] Optionally, obtaining training evaluation information based on the simulation process data includes: Based on the standard surgical path and the instrument trajectory curve, a path deviation metric is determined to generate operational path deviation information. Based on the key structural contact events, the number of times, depth of invasion, and area of ​​invasion of preset functional area objects or dangerous blood vessels are determined to form key structural invasion information. Based on the simulation start time and simulation end time, the operation time information for the user to complete the target neurosurgical training task is determined.

[0012] Optionally, displaying the training evaluation information in the MR space in a visual manner includes: Based on the operation path deviation information, the critical structure infringement information, and the operation time information, a comprehensive score result is determined, and the comprehensive score result is presented in the MR space in the form of a score interface and / or text prompts.

[0013] Secondly, this application provides an MR teaching system for neurosurgery, comprising: The selection module is used to determine a target neurosurgical training task in response to a training scene selection command in the MR space used for neurosurgical teaching, and to present the target surgical area corresponding to the target neurosurgical training task in the MR space. The processing module is used to perform anatomical structure segmentation and spatial registration on the target surgical area based on medical imaging data to obtain three-dimensional voxel data of at least one key neural structure object. The processing module is also used to perform real-time interactive simulation of the target neurosurgical training task based on the three-dimensional voxel data and the user operation behavior data in the MR space, so as to generate simulation process data. The display module is used to obtain training evaluation information based on the simulation process data, and to display the training evaluation information in a visual manner in the MR space.

[0014] Thirdly, this application provides an electronic device, comprising: Processor; and, Memory for storing the executable instructions of the processor; The processor is configured to perform any of the possible methods described in the first aspect by executing the executable instructions.

[0015] Fourthly, this application provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement any of the possible methods described in the first aspect.

[0016] The training data processing method for a neurosurgical MR teaching system provided in this application determines the target neurosurgical training task in response to a training scene selection command in the MR space used for neurosurgical teaching. It then presents the target surgical area corresponding to the target neurosurgical training task in the MR space. Next, based on medical imaging data, it performs anatomical structure segmentation and spatial registration on the target surgical area to obtain three-dimensional voxel data of at least one key neural structure. Then, based on the three-dimensional voxel data and user operation behavior data in the MR space, it performs real-time interactive simulation of the target neurosurgical training task to generate simulation process data. Finally, based on the simulation process data, it obtains training evaluation information and displays it visually in the MR space. This visualization and interactive approach in the MR space effectively improves the realism and safety of neurosurgical training. Attached Figure Description

[0017] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0018] Figure 1 This is a schematic flowchart illustrating a training data processing method for a neurosurgical MR teaching system according to an exemplary embodiment of this application; Figure 2 This is a schematic flowchart illustrating a specific implementation of S130 according to an example embodiment of this application; Figure 3 This is a schematic diagram of the structure of an MR teaching system for neurosurgery, according to an example embodiment of this application; Figure 4 This is a schematic diagram of the structure of an electronic device according to an example embodiment of this application.

[0019] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0020] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0021] Figure 1 This is a flowchart illustrating a training data processing method for a neurosurgical MR teaching system according to an example embodiment of this application. Figure 1 As shown, the method provided in this embodiment includes: S110, in response to the training scene selection instruction in MR space, determine the target neurosurgical training task.

[0022] In this step, in the MR space used for neurosurgical teaching, in response to the training scene selection instruction in the MR space, the target neurosurgical training task is determined, and the target surgical area corresponding to the target neurosurgical training task is presented in the MR space.

[0023] Specifically, a list of training scenarios can be displayed in the MR space. This list may include one or more of the following: craniotomy approach, ventricular puncture, tumor resection, and / or aneurysm clipping. In response to a user's selection of a training scenario from the MR space, the identifier of the selected training scenario is retrieved, and this identifier is used as the target of the training scenario selection command.

[0024] Then, based on the training scenario identifier, case data entries associated with the training scenario identifier are obtained from the preset case dataset. The case data entries include at least case type information, target anatomical region information, and recommended surgical approach information.

[0025] Next, based on the case type information and the target anatomical region information, the target neurosurgical training task corresponding to the case data entry is determined.

[0026] Finally, based on the target anatomical region information, the anatomical region range corresponding to the target neurosurgical training task is determined in the three-dimensional neuroanatomical model. The anatomical region range is marked as the target surgical region in MR space and presented by highlighting the outline, semi-transparent coloring and / or overlaying text annotations.

[0027] It is worth noting that most existing preoperative planning systems focus on single-case, single-perspective observation, which makes it impossible to overlay the planning results onto the operation interface of the virtual surgical instrument object in the form of intuitive spatial path markings. As a result, trainees are still prone to path deviation or accidentally entering high-risk areas such as functional areas or dangerous blood vessels during actual drills, thus limiting the safety and teachability of approach planning.

[0028] Therefore, in MR space, a three-dimensional neuroanatomical model corresponding to the patient's anatomical structure can be generated by loading the patient's medical imaging data based on a preset case dataset. Then, in response to the user's path planning request in MR space, surgical approach planning is performed on the three-dimensional neuroanatomical model to determine at least one recommended surgical path that meets safety constraints; these safety constraints include one or more of the following: avoiding functional area objects, avoiding dangerous blood vessels, and / or limiting the range of puncture angles. Finally, path visualization markers are generated along the recommended surgical path in MR space, and path deviation prompts are displayed to the user in the guidance interface of the virtual surgical instrument object.

[0029] Furthermore, the specific implementation method for surgical approach planning on a three-dimensional neuroanatomical model can be to determine a set of candidate approach points on the extracranial surface of the three-dimensional neuroanatomical model and a set of candidate target points on the lesion objects of key neural structures.

[0030] Then, based on the set of candidate entry points and the set of candidate target points, multiple candidate spatial curves are constructed as candidate surgical paths.

[0031] Next, for each candidate surgical path, the volume of normal brain tissue traversed, the minimum distance to dangerous blood vessels, and the spatial relationship with functional areas are determined to obtain a safety cost function value. Specifically, spatial sampling can be performed along the candidate surgical path in a three-dimensional neuroanatomical model to obtain multiple sets of path sampling points. Then, based on the voxel label information of the multiple sets of path sampling points in the three-dimensional neuroanatomical model, the number of voxels belonging to lesion objects, functional areas, and dangerous blood vessels is determined, and the volume of normal brain tissue traversed by the candidate surgical path is calculated based on the number of voxels and the volume of each voxel. Next, based on the spatial distance between the multiple sets of path sampling points and the surface of dangerous blood vessels, the minimum distance between the candidate surgical path and dangerous blood vessels is determined. Finally, based on the spatial positional relationship between the multiple sets of path sampling points and functional areas, a relative positional relationship index between the candidate surgical path and functional areas is determined. The relative positional relationship index includes whether the path traverses a functional area, the minimum distance to the functional area, and / or the proportion of the path length within a preset safety buffer zone. Then, based on the indicators of the volume of normal brain tissue traversed, the minimum distance, and the relative positional relationship, a safety cost function value is constructed according to preset weights and / or penalty coefficients, which is used to quantitatively evaluate the safety of different candidate surgical paths.

[0032] Finally, based on the safety cost function value, recommended surgical paths that meet the safety constraints are selected.

[0033] Through the aforementioned path planning in MR space, on the one hand, trainees' three-dimensional intuitive understanding of individualized anatomical structures and potential safe access routes is significantly enhanced, reducing the risk of misjudgment of access routes caused by relying solely on two-dimensional images and experience. On the other hand, through path visualization marking and real-time deviation prompts, trainees can be effectively constrained to avoid functional areas and dangerous blood vessels during training, forming a repeatable and comparable standardized access operation mode.

[0034] S120. Based on medical imaging data, perform anatomical structure segmentation and spatial registration on the target surgical area to obtain three-dimensional voxel data of at least one key neural structure.

[0035] First, reference volume data for the target surgical region can be determined. Specifically, this can be done by identifying an initial volume of interest (ROI) encompassing the general area of ​​the lesion based on anatomical landmarks in medical imaging data corresponding to the target surgical type. Then, boundary expansion processing is performed on the initial ROI, expanding it in three dimensions by a preset safety margin to obtain an expanded volume including surrounding normal tissue and / or important anatomical structures. Next, spatial matching is performed between the expanded volume and a preset anatomical structure template to remove anatomical regions irrelevant to the target surgical type, thus obtaining reference volume data for the target surgical region.

[0036] Secondly, within a pre-defined object coordinate system, fine-grained segmentation is performed on the target surgical region to obtain one or more neural key structure objects, including lesion objects, functional area objects, and dangerous blood vessel objects. This pre-defined object coordinate system can be based on the patient's location. Specifically, within the pre-defined object coordinate system, the reference volume data corresponding to the target surgical region is preprocessed, including one or more of image denoising, intensity normalization, and resampling. Based on a pre-trained segmentation network model, multi-class voxel-level segmentation is performed on the preprocessed reference volume data to obtain candidate lesion regions, candidate functional area regions, and candidate blood vessel regions. Then, lesion candidate regions are filtered based on morphological operations and connected component analysis to identify lesion objects that meet the volume range and shape constraints. Next, based on functional magnetic resonance imaging data and / or brain regional atlases, candidate functional area regions are mapped to pre-defined functional region labels to identify functional area objects associated with the target surgical type. Furthermore, based on angiography data and / or vascular enhancement algorithms, centerline extraction and diameter estimation can be performed on candidate vascular regions to screen out dangerous vascular objects that meet diameter and / or location thresholds. Finally, lesion objects, functional area objects, and dangerous vascular objects are stored in a preset object coordinate system in the form of object labels, forming a fine-grained segmentation result that includes one or more key neural structures.

[0037] Next, a coordinate transformation is performed between the preset object coordinate system and the world coordinate system of MR space to obtain three-dimensional voxel data of the key neural structures in MR space. Specifically, a first transformation matrix between the preset object coordinate system and the patient's physical coordinate system can be determined based on imaging parameters and / or registration markers from medical imaging data. Then, a second transformation matrix between the patient's physical coordinate system and the world coordinate system of MR space is determined based on the spatial positioning results of the MR terminal device and / or calibration data from the external optical tracking system. Next, the first and second transformation matrices are combined to obtain the coordinate transformation matrix from the preset object coordinate system to the world coordinate system of MR space. Based on the coordinate transformation matrix, batch coordinate mapping is performed on the voxel coordinates corresponding to lesion objects, functional area objects, and dangerous blood vessel objects stored in the preset object coordinate system to generate three-dimensional voxel coordinates corresponding to the world coordinate system of MR space. Finally, the mapped three-dimensional voxel coordinates are associated and stored with their respective object labels to form three-dimensional voxel data of the key neural structures in MR space.

[0038] S130. Based on three-dimensional voxel data and user operation behavior data in MR space, a real-time interactive simulation of the target neurosurgical training task is performed to generate simulation process data.

[0039] Figure 2 This is a schematic flowchart illustrating a specific implementation of S130 according to an example embodiment of this application. For example... Figure 2 As shown, the above-mentioned S130 includes: S131. Construct a tissue biomechanical model of the target surgical area based on three-dimensional voxel data.

[0040] In this step, the aforementioned tissue mechanics model includes one or more of the mechanical models of brain tissue, blood vessel walls, and tumor tissue constructed based on the finite element method, mass spring model, and / or position basis dynamics model.

[0041] In one specific implementation, tissue type identification can be performed on each voxel in the target surgical area based on three-dimensional voxel data, classifying the voxels into one or more of the following: brain parenchyma voxels, cerebrospinal fluid voxels, vascular wall voxels, and / or tumor voxels. Secondly, based on the tissue type identification results, adjacent voxels belonging to the same tissue type are clustered to form multiple tissue sub-regions, and each tissue sub-region corresponds to a candidate mechanical sub-model.

[0042] Next, for the tissue sub-regions corresponding to the brain parenchyma voxels, finite element meshes and / or mass are constructed based on the adjacency relationships of each voxel in three-dimensional space. A discrete spring structure is used, and preset elastic modulus, Poisson's ratio, and damping coefficient are assigned to the grid cells and / or mass points to form a sub-model of brain tissue mechanics.

[0043] Then, for the tissue sub-region corresponding to the blood vessel wall voxel, a tubular mesh model and / or a location-based dynamic model are constructed based on the blood vessel centerline and blood vessel radius parameters. Preset anisotropic elastic parameters and wall thickness parameters are assigned to the tubular mesh model to form a blood vessel wall mechanical sub-model.

[0044] Furthermore, for the tissue sub-regions corresponding to tumor voxels, a closed surface mesh is constructed based on the voxel set of the tumor boundary, and solid mesh units are generated inside the closed surface mesh. Nonlinear elastic parameters and / or viscoelastic parameters corresponding to the tumor type are assigned to form a tumor tissue mechanical sub-model. Finally, based on the spatial contact relationships and / or topological constraints between the brain tissue mechanical sub-model, the blood vessel wall mechanical sub-model, and the tumor tissue mechanical sub-model, boundary coupling conditions between each mechanical sub-model are established to form an overall tissue mechanical model of the target surgical area. S132. Based on the user's operational behavior data in MR space, determine the contact area and contact depth between the virtual surgical instrument object and the tissue mechanics model.

[0045] First, based on gesture tracking data, handle pose data, and / or head-mounted display pose data in MR space, the pose information of the virtual surgical instrument object in the world coordinate system of MR space is calculated in real time.

[0046] Secondly, based on the pose information of the virtual surgical instrument object, the spatial bounding body of the working end of the virtual surgical instrument object in MR space is determined. The spatial bounding body includes one or more of the following: spherical bounding body, cylindrical bounding body and / or polyhedral bounding body.

[0047] Next, collision detection is performed between the spatial enclosing body and the three-dimensional voxel coordinates corresponding to each mechanical unit in the tissue mechanics model to determine the target voxel set that intersects with the spatial enclosing body, and a contact area is formed based on the spatial distribution of the target voxel set.

[0048] Then, based on the relative overlap distance between the spatial enclosing volume and the target voxel set in the normal direction, the local contact depth of each target voxel in the contact area is calculated, and the overall contact depth corresponding to the virtual surgical instrument object is determined based on the contact depth of each target voxel.

[0049] Finally, the contact area and overall contact depth are input as external boundary conditions into the tissue mechanics model to drive the determination of subsequent tissue deformation and / or resection amount.

[0050] S133. Determine the amount of tissue deformation and / or resection based on the contact area and contact depth to update the local deformation of the three-dimensional neuroanatomical model.

[0051] In this step, based on the contact depth of each target voxel in the contact area, the contact depth is mapped to an external force load applied to the corresponding node or mass point of the tissue mechanics model. The external force load includes normal compressive load and / or tangential shear load.

[0052] Secondly, under a preset time step, dynamic and / or quasi-static solutions are performed on the tissue mechanics model after the application of external force load to obtain the displacement field of each node in the tissue mechanics model, thereby determining the amount of tissue deformation corresponding to the target surgical area.

[0053] Next, based on the amount of tissue deformation, the three-dimensional voxel coordinates in MR space are updated for voxels belonging to deformable but unresectable tissues, such as normal brain tissue and / or vascular wall voxels, to obtain a deformed local three-dimensional neuroanatomical model.

[0054] Then, when the user's operation behavior data is detected to meet the preset resection judgment conditions, the target voxels belonging to the resectable tissue are marked as resectable voxels based on the contact area and contact depth, such as tumor voxels and / or partial brain tissue voxels, and the resectable voxels are deleted or hidden from the three-dimensional neuroanatomy model to determine the corresponding resection amount.

[0055] Finally, based on the amount of tissue deformation and resection, the topology and / or surface mesh of the three-dimensional neuroanatomical model are reconstructed to generate updated local deformation results. The local deformation results are then fused with the anatomical regions that have not undergone deformation to form an overall updated three-dimensional neuroanatomical model.

[0056] S134. Render the updated 3D neuroanatomical model and virtual surgical instrument objects in real time in MR space to generate continuous interactive simulation screens.

[0057] First, based on the updated 3D neuroanatomy model, rendering material parameters corresponding to their respective tissue types are assigned to brain tissue, blood vessel walls, and tumor tissue. The rendering material parameters include color, transparency, and / or surface texture parameters.

[0058] Secondly, based on the virtual light source configuration in MR space, the illumination intensity and shadow area of ​​the updated 3D neuroanatomical model surface at the current viewpoint are calculated to generate a lighting rendering effect that meets the requirements of stereoscopic perception.

[0059] Next, based on the pose and contact area information of the virtual surgical instrument object in MR space, contact highlights, cutting marks and / or blood vessel rupture warning marks are superimposed on the working end of the virtual surgical instrument object to enhance the interactive feedback effect.

[0060] Then, under the preset rendering frame rate constraint, the continuously updated three-dimensional neuroanatomical model and virtual surgical instrument objects are rendered in chronological order to generate multiple frames of simulation images. These multiple frames of simulation images are then presented in a stereoscopic display in MR space to form a continuous interactive simulation image.

[0061] Finally, during the rendering process, the display ratio, viewpoint position, and occlusion culling strategy of the target surgical area are adjusted in real time based on the user's viewpoint changes and / or interaction commands in MR space to ensure the visibility and interactive coherence of key neuroanatomical structures and virtual surgical instruments.

[0062] S135. Based on the sequence of trajectory points of the virtual surgical instrument object in MR space, form the instrument trajectory curve.

[0063] In this step, pose sampling data corresponding to virtual surgical instrument objects can be obtained based on user operation behavior data in MR space. The pose sampling data includes spatial position coordinates and / or posture angle information collected within a preset time interval.

[0064] Then, the spatial coordinates of the virtual surgical instrument objects are sorted on the time axis according to the sampling time order to form a discrete sequence of trajectory points.

[0065] Then, denoising and smoothing are performed on adjacent trajectory points in the trajectory point sequence. The smoothing process includes one or more of the following: sliding window-based averaging filtering, spline interpolation-based curve fitting, and / or Kalman filter-based trajectory estimation, in order to reduce high-frequency noise caused by jitter.

[0066] Furthermore, after smoothing, the trajectory point sequence is interpolated or fitted based on one or more of spline functions, polyline segments, and / or higher-order polynomials to obtain a continuously differentiable instrument trajectory curve in three-dimensional space.

[0067] Finally, the instrument trajectory curve is associated with and stored with the corresponding timestamp information to facilitate subsequent time-based trajectory playback and / or segmented analysis.

[0068] S136. Based on the tissue mechanics model and instrument trajectory curve, determine the sequence of operational mechanics parameters.

[0069] In this step, based on the tissue mechanics model and instrument trajectory curve, the contact force, cutting force, and traction force applied to the virtual surgical instrument object are determined, forming a sequence of operational mechanics parameters.

[0070] Specifically, it can be based on the spatial positional relationship between the instrument trajectory curve and the tissue mechanics model of the target surgical area, and determine the contact state between the working end of the virtual surgical instrument and the tissue mechanics model at each time step. The contact state includes one of non-contact, compression contact and / or cutting contact.

[0071] Then, when it is detected that the virtual surgical instrument object and the tissue mechanics model are in a state of compressive or cutting contact, the normal contact force applied to the virtual surgical instrument object is calculated based on the contact area and contact depth, according to the mechanical parameters corresponding to different tissue types in the tissue mechanics model, including: elastic modulus, yield stress and / or shear strength. Next, when relative sliding is detected at the working end of the virtual surgical instrument object within the contact area, the cutting force and / or friction force applied to the virtual surgical instrument object are determined based on the friction coefficient and relative sliding speed of the tissue mechanics model.

[0072] Furthermore, when the virtual surgical instrument object is detected to move away from the lesion or along a preset traction direction and cause displacement to the surrounding tissue, the magnitude and direction of the corresponding traction force are calculated based on the amount of tissue displacement, the equivalent stiffness of the tissue, and the traction direction.

[0073] Finally, at each time step, the calculated contact force, cutting force, and tensile force are recorded as operational mechanical parameters in chronological order, and the operational mechanical parameters corresponding to multiple time steps are combined to form an operational mechanical parameter sequence. Furthermore, within the operational mechanical parameter sequence, abnormal peak force values ​​can be optionally identified and suppressed to eliminate non-physiological mechanical anomalies caused by tracking errors.

[0074] S137. Based on the spatial relationship between the instrument trajectory curve and the key neural structures, detect and record the key structural contact events at each moment.

[0075] In this step, critical neural structures can be categorized into lesion objects, functional area objects, and dangerous blood vessel objects, and corresponding safe distance thresholds and / or contact determination thresholds can be pre-set for each type of critical neural structure.

[0076] Then, at each time step, based on the instrument position and orientation corresponding to the current moment of the instrument trajectory curve, a spatial detection region corresponding to the working end of the virtual surgical instrument object is constructed. The spatial detection region includes one or more of the following: spherical neighborhood, cylindrical neighborhood, and / or bounding box neighborhood.

[0077] Next, spatial intersection calculations are performed on the three-dimensional voxel data corresponding to the spatial detection region and the neural key structure object. When an overlap between the spatial detection region and any neural key structure object is detected, it is determined that there is potential key structure contact at the current time step. Based on the nearest distance and / or overlap depth between the spatial detection region and the surface of the neural key structure object, the potential key structure contact is further classified into one of the following: proximity event, surface contact event, and / or depth intrusion event.

[0078] For cases determined to be surface contact events or deep intrusion events, the corresponding key structure type, timestamp of the event, contact location coordinates, contact normal direction, contact area and / or intrusion depth are recorded to form a single key structure contact event entry.

[0079] Finally, the multiple key structural contact event entries detected along the time axis are stored in chronological order as a key structural contact event sequence, and optionally classified and archived according to lesion objects, functional area objects, and dangerous blood vessel objects, for subsequent generation of key structural invasion information and training evaluation information.

[0080] S140. Based on the simulation process data, obtain the training evaluation information and display the training evaluation information in a visual manner in the MR space.

[0081] In this step, path deviation metrics can be determined based on standard surgical paths and instrument trajectory curves to generate operational path deviation information. Based on critical structure contact events, the number of invasions, depth of invasion, and area of ​​invasion of preset functional areas or dangerous blood vessels can be determined to generate critical structure invasion information. Based on the simulation start and end times, the operation time information for the user to complete the target neurosurgical training task can be determined.

[0082] Next, based on the operation path deviation information, critical structure violation information, and operation time information, the comprehensive scoring result is determined, and the comprehensive scoring result is presented in the MR space in the form of a scoring interface and / or text prompts.

[0083] In this embodiment, in the MR space used for neurosurgical teaching, in response to the training scene selection command in the MR space, the target neurosurgical training task is determined, and the target surgical area corresponding to the target neurosurgical training task is presented in the MR space. Then, based on medical imaging data, anatomical structure segmentation and spatial registration are performed on the target surgical area to obtain three-dimensional voxel data of at least one key neural structure. Next, based on the three-dimensional voxel data and user operation behavior data in the MR space, the target neurosurgical training task is simulated in real time to generate simulation process data. Finally, training evaluation information is obtained based on the simulation process data, and the training evaluation information is displayed in a visual manner in the MR space. Thus, through visualization and interaction in the MR space, the realism and safety of neurosurgical training are effectively improved.

[0084] Figure 3 This is a schematic diagram illustrating the structure of an MR teaching system for neurosurgery according to an example embodiment of this application. Figure 3 As shown, the MR teaching system 300 for neurosurgery provided in this embodiment includes: Selection module 310 is used to determine a target neurosurgical training task in response to a training scene selection instruction in the MR space used for neurosurgical teaching, and to present a target surgical area corresponding to the target neurosurgical training task in the MR space. Processing module 320 is used to perform anatomical structure segmentation and spatial registration on the target surgical area based on medical imaging data to obtain three-dimensional voxel data of at least one key neural structure object; The processing module 320 is also used to perform real-time interactive simulation of the target neurosurgical training task based on the three-dimensional voxel data and the user operation behavior data in the MR space, so as to generate simulation process data. Display module 330 is used to obtain training evaluation information based on the simulation process data, and to display the training evaluation information in a visual manner in the MR space.

[0085] Figure 4 This is a schematic diagram of the structure of an electronic device according to an example embodiment of this application. For example... Figure 4 As shown, the electronic device 400 provided in this embodiment includes: a processor 401 and a memory 402; wherein: Memory 402 is used to store computer programs, and the memory may also be flash memory.

[0086] Processor 401 is used to execute the execution instructions stored in the memory to implement the various steps in the above method. For details, please refer to the relevant descriptions in the preceding method embodiments.

[0087] Alternatively, the memory 402 can be either standalone or integrated with the processor 401.

[0088] When the memory 402 is a device independent of the processor 401, the electronic device 400 may further include: Bus 403 is used to connect the memory 402 and the processor 401.

[0089] This embodiment also provides a readable storage medium storing a computer program, which, when executed by at least one processor of an electronic device, enables the electronic device to perform the methods provided in the various embodiments described above.

[0090] This embodiment also provides a program product including a computer program stored in a readable storage medium. At least one processor of an electronic device can read the computer program from the readable storage medium, and the at least one processor executes the computer program to cause the electronic device to perform the methods provided in the various embodiments described above.

[0091] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this application are indicated by the claims.

[0092] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.

Claims

1. A method for processing training data for a neurosurgical MR teaching system, characterized in that, include: In the MR space used for neurosurgical teaching, in response to the training scene selection command in the MR space, the target neurosurgical training task is determined, and the target surgical area corresponding to the target neurosurgical training task is presented in the MR space. Based on medical imaging data, anatomical structure segmentation and spatial registration are performed on the target surgical area to obtain three-dimensional voxel data of at least one key neural structure object. Based on the three-dimensional voxel data and the user operation behavior data in the MR space, the target neurosurgical training task is subjected to real-time interactive simulation to generate simulation process data. Based on the simulation process data, training evaluation information is obtained and displayed in a visual manner in the MR space.

2. The training data processing method for a neurosurgical MR teaching system according to claim 1, characterized in that, The step of performing anatomical segmentation and spatial registration on the target surgical area based on medical imaging data includes: Determine reference volume data for the target surgical region; In a preset object coordinate system, fine-grained segmentation is performed on the target surgical area to obtain one or more neural key structure objects, including lesion objects, functional area objects, and dangerous blood vessel objects. The preset object coordinate system is transformed with the world coordinate system of the MR space to obtain the three-dimensional voxel data of the neural key structure object in the MR space.

3. The training data processing method for a neurosurgical MR teaching system according to claim 1, characterized in that, The method of performing real-time interactive simulation of the target neurosurgical training task based on the three-dimensional voxel data and user operation behavior data in the MR space to generate simulation process data includes: A tissue biomechanical model of the target surgical area is constructed based on the three-dimensional voxel data; Based on the user's operational behavior data in the MR space, the contact area and contact depth between the virtual surgical instrument object and the tissue mechanics model are determined; The amount of tissue deformation and / or resection is determined based on the contact area and the contact depth to update the local deformation of the three-dimensional neuroanatomical model; The updated 3D neuroanatomical model and the virtual surgical instrument object are rendered in real time in the MR space to generate continuous interactive simulation screens.

4. The training data processing method for a neurosurgical MR teaching system according to claim 3, characterized in that, The tissue mechanics model includes one or more of the mechanical models of brain tissue, blood vessel walls, and tumor tissue constructed based on the finite element method, mass spring model, and / or position basis dynamics model.

5. The training data processing method for a neurosurgical MR teaching system according to claim 3 or 4, characterized in that, The generated simulation process data includes: Based on the sequence of trajectory points of the virtual surgical instrument object in the MR space, an instrument trajectory curve is formed; Based on the tissue mechanics model and the instrument trajectory curve, the contact force, cutting force and tensile force applied to the virtual surgical instrument object are determined to form a sequence of operational mechanics parameters. Based on the spatial relationship between the instrument trajectory curve and the key neural structure, contact events of the key structure at each moment are detected and recorded.

6. The training data processing method for a neurosurgical MR teaching system according to claim 5, characterized in that, The step of obtaining training evaluation information based on the simulation process data includes: Based on the standard surgical path and the instrument trajectory curve, a path deviation metric is determined to generate operational path deviation information. Based on the key structural contact events, the number of times, depth of invasion, and area of ​​invasion of preset functional area objects or dangerous blood vessels are determined to form key structural invasion information. Based on the simulation start time and simulation end time, the operation time information for the user to complete the target neurosurgical training task is determined.

7. The training data processing method for a neurosurgical MR teaching system according to claim 6, characterized in that, The step of visually displaying the training evaluation information in the MR space includes: Based on the operation path deviation information, the critical structure infringement information, and the operation time information, a comprehensive score result is determined, and the comprehensive score result is presented in the MR space in the form of a scoring interface and / or text prompts.

8. An MR teaching system for neurosurgery, characterized in that, include: The selection module is used to determine a target neurosurgical training task in response to a training scene selection command in the MR space used for neurosurgical teaching, and to present the target surgical area corresponding to the target neurosurgical training task in the MR space. The processing module is used to perform anatomical structure segmentation and spatial registration on the target surgical area based on medical imaging data to obtain three-dimensional voxel data of at least one key neural structure object. The processing module is also used to perform real-time interactive simulation of the target neurosurgical training task based on the three-dimensional voxel data and the user operation behavior data in the MR space, so as to generate simulation process data. The display module is used to obtain training evaluation information based on the simulation process data, and to display the training evaluation information in a visual manner in the MR space.

9. An electronic device, characterized in that, include: processor; as well as, Memory for storing the executable instructions of the processor; The processor is configured to execute the method of any one of claims 1 to 7 by executing the executable instructions.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method as described in any one of claims 1 to 7.

Images