A humanoid embodied intelligent neurosurgical robot system and method

Abstract

The application provides a humanoid embodied intelligent neurosurgery robot system and a control method thereof. The system is constructed based on a perception-cognition-decision-execution-feedback embodied intelligent closed loop architecture, comprising: an intelligent assessment module for disease assessment and neurosurgery indication determination of a patient; a surgery planning module for constructing a three-dimensional fusion model of a brain containing a lesion, a brain function area and a white matter fiber bundle and generating a personalized surgery planning scheme; a humanoid neurosurgery execution module for accurately guiding a surgical instrument to an intracranial target position through stereoscopic directional positioning and performing neurosurgery operation; and an auxiliary rehabilitation module for generating a personalized postoperative rehabilitation scheme. The system realizes intelligentization, autonomy and whole-process closed loop management of neurosurgery.

Description

Technical Field

[0001] This application relates to the field of medical device technology, and more specifically, to a humanoid intelligent neurosurgical robot system and method. Background Technology

[0002] Neurosurgery is an important clinical means of surgical intervention in the human central nervous system. It includes various procedures such as deep brain stimulation (DBS) electrode implantation, stereotactic biopsy of brain tumors, epileptic focus ablation, and cerebral hemorrhage drainage. Its core operation is to precisely guide surgical instruments to the target location in the brain and perform the corresponding surgical operation.

[0003] However, neurosurgery faces extremely stringent technical challenges. Existing systems lack the ability to actively detect and compensate for micro-vibrations at the end of the surgical arm, as well as the real-time monitoring and graded protection mechanism for the spatial distance between surgical instruments and brain functional areas. They have also failed to introduce embodied intelligent decision-making capabilities based on intracranial multimodal perception information into neurosurgical robot systems. There is an urgent need for a new type of surgical robot system that can deeply integrate humanoid embodied intelligence technology with neurosurgery. Summary of the Invention

[0004] To address the aforementioned problems, the first aspect of this application provides a humanoid embodied intelligent neurosurgical robot system. This system is constructed based on an embodied intelligent closed-loop architecture of perception-cognition-decision-execution-feedback, comprising: The intelligent assessment module is used to conduct targeted disease assessments and determine the indications for neurosurgery based on the patient's medical data and vital signs, generate personalized auxiliary assessment results for the patient, and display them. The surgical planning module is used to construct a three-dimensional fusion model of the cranium and brain, including lesions, brain functional areas, and white matter fiber tracts, based on the patient's multimodal neuroimaging data. Based on the three-dimensional fusion model of the cranium and brain, targeted neurosurgical planning is carried out for the patient, generating a personalized surgical plan and displaying it. The humanoid neurosurgical execution module is used to precisely guide surgical instruments to the target intracranial location through stereotactic positioning according to the surgical plan, and to perform neurosurgical operations based on intraoperative multimodal perception. The auxiliary rehabilitation module is used to generate and display personalized postoperative rehabilitation plans for patients based on their medical data and postoperative assessment results.

[0005] The second aspect of this application provides a control method for a humanoid intelligent neurosurgical robot system, applied to the aforementioned humanoid intelligent neurosurgical robot system, comprising: Based on the patient's medical data and physical signs, targeted disease assessments and neurosurgical indications are determined, generating and displaying personalized auxiliary assessment results. Based on the patient's multimodal neuroimaging data, a three-dimensional fusion model of the cranium and brain, including lesions, brain functional areas, and white matter fiber tracts, is constructed. Based on the three-dimensional fusion model of the cranium and brain, targeted neurosurgical planning is carried out for the patient, a personalized surgical plan is generated, and the plan is displayed. According to the surgical plan, the surgical instruments are precisely guided to the target intracranial location through stereotactic positioning, and the neurosurgical operation is performed based on intraoperative multimodal perception. Based on the patient's medical data and postoperative assessment results, a personalized postoperative rehabilitation plan is generated and displayed.

[0006] A third aspect of this application provides an electronic device comprising: a memory and a processor; the memory for storing a program; and the processor, coupled to the memory, for executing the program to implement the control method of the humanoid intelligent neurosurgical robot system described above.

[0007] The fourth aspect of this application provides a computer-readable storage medium having a computer program stored thereon, the program being executed by a processor to implement the control method of the humanoid intelligent neurosurgical robot system described above.

[0008] The humanoid intelligent neurosurgical robot system provided in this application has the following beneficial effects: By adopting an embodied intelligent closed-loop architecture of perception-cognition-decision-execution-feedback, neurosurgical robots are upgraded from passive execution tools to intelligent surgical systems with autonomous decision-making capabilities. They can autonomously adjust surgical strategies according to changes in the intracranial environment during surgery, reducing reliance on the surgeon's personal surgical skills.

[0009] By constructing a three-dimensional fusion model of the cranium that includes lesions, brain functional areas, and white matter fiber tracts through the surgical planning module, a comprehensive visualization of the complex intracranial anatomy and personalized surgical planning can be achieved.

[0010] Through the standardized mechanical interface design and orientation operation module of the humanoid robot body, surgical instruments can be accurately delivered to the intracranial target location along the planned path according to stereotactic coordinates, and various neurosurgical operations such as electrode implantation, tissue sampling or lesion ablation can be performed.

[0011] By integrating a vibration isolation structure with a humanoid torso to suppress environmental vibrations, using a dexterous hand at the end of the surgical arm to perform intracranial positioning and surgical operations with sub-millimeter precision, and using a control module to generate control commands based on navigation images and electrophysiological monitoring data, precise control of neurosurgical operations is achieved.

[0012] By using a displacement compensation module to detect brain tissue displacement and calculate the displacement vector field during critical stages of surgery, the positional information of each anatomical structure in the navigation model is updated, effectively solving the problem of decreased navigation accuracy caused by intraoperative brain tissue displacement.

[0013] Through microelectrode recording and evoked potential monitoring by the electrophysiological monitoring module, the electrophysiological verification of target nuclei and real-time monitoring of the integrity of brain functional pathways were realized, improving the precision and safety of neurosurgery.

[0014] The functional area protection module monitors the spatial distance between surgical instruments and brain functional areas in real time, and executes graded protection responses when the monitored indicators exceed the safety threshold, effectively protecting important brain functional areas such as the motor area, language area, and sensory area.

[0015] By using an end-effector module to detect micro-vibrations at the tip of the surgical arm in real time and generating reverse micro-motion commands through an active compensation algorithm, the tip of the surgical instrument achieves sub-millimeter-level positioning stability during intracranial operations.

[0016] By autonomously performing cognitive understanding and decision-making reasoning based on intracranial multimodal sensory information through the embodied intelligent decision-making module, positioning strategies and operational strategies are generated and iteratively corrected based on execution feedback, thus realizing intelligent autonomous decision-making for neurosurgery. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below.

[0018] Figure 1 A schematic diagram of the overall architecture of the humanoid intelligent neurosurgical robot system provided in the embodiments of this application; Figure 2 A schematic diagram of the humanoid neurosurgical execution module provided in this application embodiment; Figure 3 This is a schematic diagram of the humanoid torso configuration provided in an embodiment of this application; Figure 4 This is a schematic diagram of the shift compensation module architecture provided in an embodiment of this application; Figure 5 This is a schematic diagram of the electrophysiological monitoring module provided in an embodiment of this application; Figure 6 This is a schematic diagram of the functional area protection module provided in an embodiment of this application; Figure 7 This is a schematic diagram of the internal closed-loop architecture of the embodied intelligent decision-making module provided in the embodiments of this application; Figure 8 A flowchart illustrating the control method for a humanoid intelligent neurosurgical robot system provided in this application embodiment; Figure 9 This is a schematic diagram of an electronic device architecture provided in an embodiment of this application. Detailed Implementation

[0019] To make the above-mentioned objects, features, and advantages of this application more apparent and understandable, specific embodiments of this application will be described in detail below with reference to the accompanying drawings. Although exemplary embodiments of this application are shown in the drawings, it should be understood that this application can be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of this application and to fully convey the scope of this application to those skilled in the art.

[0020] It should be noted that, unless otherwise stated, the technical or scientific terms used in this application shall have the ordinary meaning as understood by one of ordinary skill in the art to which this application pertains.

[0021] like Figure 1 As shown, this application provides a humanoid embodied intelligent neurosurgical robot system. The system is built upon an embodied intelligent closed-loop architecture of perception-cognition-decision-execution-feedback, and includes: The intelligent assessment module is used to conduct targeted disease assessments and determine the indications for neurosurgery based on the patient's medical data and vital signs, generate personalized auxiliary assessment results for the patient, and display them. The surgical planning module is used to construct a three-dimensional fusion model of the cranium and brain, including lesions, brain functional areas, and white matter fiber tracts, based on the patient's multimodal neuroimaging data. Based on the three-dimensional fusion model of the cranium and brain, targeted neurosurgical planning is carried out for the patient, generating a personalized surgical plan and displaying it. The humanoid neurosurgical execution module is used to precisely guide surgical instruments to the target intracranial location through stereotactic positioning according to the surgical plan, and to perform neurosurgical operations based on intraoperative multimodal perception. The auxiliary rehabilitation module is used to generate and display personalized postoperative rehabilitation plans for patients based on their medical data and postoperative assessment results.

[0022] Specifically, the embodied intelligent closed-loop architecture is the core architecture that upgrades neurosurgical robots from passive tools that execute doctors' control commands to intelligent agents with the ability to perceive the environment, understand the mind, make autonomous decisions, execute precisely, and provide feedback on results.

[0023] The design philosophy of this architecture is based on the theory of embodied intelligence, which states that intelligent agents must acquire and utilize knowledge through direct interaction with the physical environment.

[0024] In neurosurgical scenarios, the physical environment is the complex intracranial brain tissue structure (including brain parenchyma, brain functional areas, white matter fiber tracts, ventricular system, cerebral vascular network, etc.). The intelligent agent directly interacts with the intracranial tissue environment through surgical instruments, perceives navigation imaging information, electrophysiological monitoring information, surgical instrument position information, and intracranial tissue state information. After cognitive reasoning and decision planning, it controls the surgical instruments to perform corresponding positioning and surgical operations, and continuously corrects and optimizes the surgical strategy through feedback mechanisms.

[0025] The effect of this closed-loop architecture is that it enables neurosurgical robots to autonomously locate, operate, and respond to unexpected situations such as brain tissue displacement and functional area proximity in complex intracranial tissue environments.

[0026] Specifically, the intelligent assessment module is the system's preoperative assessment entry point, responsible for comprehensively analyzing the patient's multi-dimensional medical data to determine whether the patient is suitable for neurosurgery.

[0027] The medical data includes, but is not limited to, the patient's medical history, neurological examination results, laboratory test results (complete blood count, coagulation function, electrolytes, etc.), neuroimaging data (MRI plain scan and enhanced scan, CT plain scan and enhanced scan, MRA / CTA cerebral angiography, fMRI brain functional localization, DTI white matter fiber tract tracking, PET-CT brain metabolic imaging, etc.), electroencephalogram (EEG) data, cardiopulmonary function test data, and previous surgical records.

[0028] The vital signs data include the patient's vital signs, Glasgow Coma Scale (GCS) score, Karnofsky Performance Status (KPS) score, neurological deficit assessment data (muscle strength grading, language function assessment, sensory function assessment, etc.), intracranial pressure monitoring data, etc.

[0029] The intelligent assessment module uses a multimodal basic model to fuse and analyze the above-mentioned multi-source data and perform cognitive reasoning to generate personalized auxiliary assessment results for neurosurgery. These assessment results include not only disease diagnosis and lesion nature determination (such as brain tumor classification and grading, Parkinson's disease staging, epileptic focus localization, and cerebral hemorrhage volume assessment), but also surgical feasibility assessment conclusions (target location accessibility, surgical approach feasibility, expected surgical risk level), surgical method recommendations (craniotomy, stereotactic surgery, endoscopic surgery, etc.), and surgical procedure recommendations (DBS electrode implantation, stereotactic biopsy, tumor resection, lesion ablation, hematoma drainage, etc.).

[0030] The personalization of the personalized auxiliary assessment results is essentially generated by the multimodal basic model based on the patient's individual medical history, imaging characteristics and neurological functional status to generate assessment conclusions, rather than using a preset general assessment template.

[0031] The effect of this feature is that it provides a scientific and comprehensive preoperative assessment basis for subsequent surgical planning, avoiding intraoperative accidents caused by insufficient assessment.

[0032] Specifically, the surgical planning module is responsible for constructing a three-dimensional fusion model of the brain based on the patient's multimodal neuroimaging data and generating a personalized surgical planning scheme.

[0033] This module first acquires the patient's multimodal neuroimaging data, including MRI T1-weighted images (for displaying brain structure anatomy), MRI T2-weighted images (for displaying ventricles and edema), MRI contrast-enhanced scans (for displaying the extent of tumor enhancement and cerebral blood vessels), fMRI brain functional imaging (for precise localization of brain functional areas such as motor, language, and sensory areas), DTI diffusion tensor imaging (for three-dimensional tracking and reconstruction of white matter fiber tracts, including pyramidal tract, arcuate tract, optic radiation, corpus callosum fibers, etc.), CT thin-slice scans (for displaying skull and calcification), and cerebral angiography data (for displaying the distribution of cerebral arteries and veins). Through a multimodal image fusion algorithm, the image data of different modalities are precisely aligned to a unified stereotactic coordinate system to construct a three-dimensional fusion model of the cranium that includes the three-dimensional boundaries of lesions, the location of brain functional areas, and the course of white matter fiber tracts.

[0034] Neurosurgical planning is based on the aforementioned three-dimensional fusion model of the brain: For DBS electrode implantation surgery, the stereotactic coordinates of the target nucleus (STN, GPi, or VIM) are precisely located in the three-dimensional model to determine the optimal surgical approach and electrode implantation trajectory, ensuring that the trajectory avoids the ventricles, cerebral blood vessels, and brain functional areas; for stereotactic biopsy, the stereotactic coordinates of the biopsy target point and the optimal entry path are determined to ensure that the sampling location is diagnostically representative and the path avoids important structures; for lesion ablation, the ablation range and the implantation position and angle of the ablation probe are planned to ensure that the ablation covers the lesion without damaging the surrounding functional tissues.

[0035] The system employs a multi-objective optimization algorithm to select the surgical path with the optimal comprehensive score while meeting safety constraints such as avoiding brain functional areas and white matter fiber tracts. It also generates a personalized surgical plan, which includes the coordinates of the craniotomy point or burr hole, the instrument entry angle, the instrument advancement depth, the stereotactic coordinates of the target point, and the operational parameters for the specific surgical procedure.

[0036] The effect of this feature is that it provides precise path guidance and operation parameter guidance for intraoperative surgical procedures, significantly improving the predictability and safety of neurosurgical procedures.

[0037] Specifically, the humanoid neurosurgical execution module is responsible for precisely guiding surgical instruments to the target intracranial location through stereotactic positioning according to the surgical plan, and performing neurosurgical operations based on intraoperative multimodal perception.

[0038] This module includes a humanoid robot body and a directional operation module mounted on it.

[0039] The humanoid robot can autonomously move to the surgical position and dock with the operating table. Its end effector is equipped with a standardized mechanical interface to adapt to different types of surgical execution modules.

[0040] The orientation module is installed on the humanoid robot body through a standardized mechanical interface. Based on the stereoscopic coordinates, it accurately delivers surgical instruments to the target location in the brain along the planned path, and performs surgical operations such as electrode implantation, tissue sampling or lesion ablation after reaching the target location.

[0041] Intraoperative multimodal perception includes navigation imaging data, electrophysiological monitoring data, instrument position data, and brain tissue displacement data, providing real-time environmental perception support for surgical procedures.

[0042] The effect of this feature is that it enables the precise delivery of surgical instruments to the target intracranial location and the execution of surgical procedures under stereotactic guidance and multimodal perception.

[0043] Specifically, the auxiliary rehabilitation module is responsible for generating personalized postoperative rehabilitation plans based on the patient's medical data and postoperative assessment results.

[0044] This module comprehensively considers factors such as surgical procedure type (DBS electrode implantation, stereotactic biopsy, tumor resection, lesion ablation, etc.), surgical site and depth, immediate postoperative imaging assessment results (electrode position verification, biopsy needle tract assessment, ablation range assessment, examination for complications such as intracranial hemorrhage, etc.), intraoperative operation records, and individual patient differences. It uses a multimodal basic model for reasoning and analysis to generate personalized postoperative rehabilitation plans.

[0045] For patients with DBS electrode implantation, rehabilitation programs include postoperative programming (initial settings of stimulation parameters: recommended values ​​for voltage, frequency, and pulse width, as well as follow-up plans for subsequent parameter optimization), medication adjustment programs, and exercise rehabilitation training programs.

[0046] For patients after brain tumor surgery, the rehabilitation program includes a postoperative imaging follow-up plan, recommendations for subsequent radiotherapy and chemotherapy, assessment of neurological function recovery, and rehabilitation training programs (including motor rehabilitation, speech rehabilitation, and cognitive rehabilitation).

[0047] For patients who have undergone epileptic focus ablation, the rehabilitation plan includes adjusting the antiepileptic drug regimen and an EEG follow-up plan.

[0048] The effect of this feature is that it enables closed-loop management of the entire process from preoperative assessment to postoperative rehabilitation, avoiding information gaps between postoperative rehabilitation and intraoperative treatment.

[0049] The beneficial effects of the aforementioned humanoid intelligent neurosurgical robot system are as follows: Through an embodied intelligent closed-loop architecture of perception-cognition-decision-execution-feedback, the neurosurgical robot possesses autonomous decision-making and adaptive adjustment capabilities; the intelligent assessment module enables comprehensive preoperative disease assessment and determination of neurosurgical indications; the surgical planning module enables the construction of a three-dimensional fusion model of the cranium based on multimodal neuroimaging and personalized surgical plan planning; the humanoid neurosurgical execution module enables precise delivery of surgical instruments and execution of surgical procedures under stereotactic guidance and multimodal perception; and the auxiliary rehabilitation module enables personalized postoperative rehabilitation management.

[0050] like Figure 2 As shown, in a preferred embodiment, the humanoid neurosurgical execution module includes: a humanoid robot body for autonomously moving to the surgical position and docking with the operating table; the end of the humanoid robot body is provided with a standardized mechanical interface for adapting to different types of surgical execution modules; and a directional operation module, which is installed on the humanoid robot body through the standardized mechanical interface, for accurately delivering surgical instruments to the intracranial target location along a planned path according to stereotactic coordinates, and performing surgical operations after reaching the target location, wherein the surgical operations include at least one of electrode implantation, tissue sampling, or lesion ablation.

[0051] Specifically, the humanoid robot body adopts a humanoid robot structure design, including an autonomous mobile chassis, a humanoid torso, and a surgical execution module mounted on the end effector.

[0052] The autonomous mobile chassis integrates omnidirectional wheels and a lidar navigation system, enabling the humanoid robot to navigate and move autonomously within the operating room to reach the target parking position next to the operating table.

[0053] The docking mechanism is located at the connection between the chassis and the operating table. It adopts a combination design of precision guide pins and locking buckles. After the robot reaches the target position, the guide pin of the docking mechanism is precisely inserted into and locked with the pre-set docking hole on the side of the operating table, realizing a rigid connection between the humanoid robot body and the operating table.

[0054] This rigid connection ensures that the robot body will not shift or vibrate relative to the operating table during the operation, providing a stable physical support platform for precise intracranial surgical procedures.

[0055] The effect of this feature is that it enables the humanoid robot to autonomously reach the surgical site and precisely dock with the operating table, providing a stable support foundation for subsequent precise intracranial operations.

[0056] Specifically, the standardized mechanical interface is located at the end of the humanoid robot body (arm end or torso end), and adopts a unified mechanical connection standard and electrical communication protocol, so that different types of surgical execution modules (such as directional operation modules, endoscope control modules, micromanipulation modules, etc.) can be quickly installed and disassembled through the same interface.

[0057] The standardized mechanical interface includes a mechanical locking part and an electrical connection part: the mechanical locking part adopts a quick-release bayonet design with a positioning pin. The surgical execution module is aligned with the bayonet and then rotated to lock, which can complete the loading and unloading operation within 10 seconds; the electrical connection part includes a power supply interface, a control signal interface and a sensor data interface, which automatically establishes an electrical connection while mechanically locking.

[0058] The effect of this feature is that it makes the humanoid robot body a universal surgical execution platform. By changing the end effector module, it can be adapted to different types of neurosurgical operations, which greatly improves the system's flexibility and scalability.

[0059] Specifically, the orientation operation module is installed on the humanoid robot body through a standardized mechanical interface.

[0060] The orientation operation module integrates a stereotactic motion platform, including a three-axis translation stage (X / Y / Z directions, with a travel range covering the intracranial surgical operation space) and a two-axis rotation stage (pitch and yaw angles, used to set the instrument entry angle), for a total of five degrees of freedom (5DOF), which can accurately position surgical instruments to the planned intracranial target stereotactic coordinates and entry angle.

[0061] The instrument delivery mechanism is driven by a high-precision linear motor, achieving a delivery accuracy of 0.1 mm and a delivery speed that can be continuously adjusted from 0.05 mm / s to 10 mm / s.

[0062] The directional manipulation module can be adapted to different types of surgical instruments to perform different types of surgical procedures: For electrode implantation, it is adapted to DBS electrodes or deep brain electrodes, and the electrodes are precisely implanted into the target nuclei (STN, GPi, VIM, etc.) or deep brain target locations along the planned path. During delivery, the microelectrode signals are recorded in real time for electrophysiological target area verification; For tissue sampling, it is adapted to stereotactic biopsy needles, and after the biopsy needles are delivered to the target location, the biopsy sampling mechanism is triggered to obtain tissue specimens. Multi-point sampling at different depths can be performed to improve diagnostic accuracy; For lesion ablation, it is adapted to laser interstitial thermotherapy (LITT) fiber optic probes or radiofrequency ablation (RF) probes, and after the ablation probes are delivered into the lesion, the ablation energy is activated. The ablation range is controlled by intraoperative MRI thermal imaging or impedance monitoring.

[0063] The effect of this feature is that it enables precise positioning of surgical instruments in a stereotactic coordinate system and the execution of various types of neurosurgical procedures.

[0064] The beneficial effects of the aforementioned humanoid neurosurgical execution module are as follows: rapid preoperative deployment is achieved through the autonomous movement and docking of the humanoid robot body; modular replacement of the surgical execution module and expansion of system capabilities are achieved through standardized mechanical interfaces; and precise positioning of surgical instruments and execution of various neurosurgical operations such as electrode implantation, tissue sampling, and lesion ablation are achieved through the stereotactic motion platform and high-precision instrument delivery mechanism of the directional operation module.

[0065] like Figure 3 As shown, in a preferred embodiment, the humanoid neurosurgical execution module includes a control module, a humanoid torso, and a surgical arm; the humanoid torso integrates a vibration isolation structure to suppress the influence of environmental vibration on surgical accuracy; the surgical arm is equipped with a dexterous hand at its end for holding surgical instruments and performing intracranial positioning and surgical operations with sub-millimeter precision; the control module generates control commands based on intraoperative navigation images and electrophysiological monitoring data to control the surgical arm to perform surgical actions.

[0066] Specifically, neurosurgery requires surgical instruments to be positioned in the brain with an accuracy better than 1 millimeter. There are various environmental vibration sources in the operating room (vibration of air conditioning system, vibration of surgical microscope gantry, micro-vibration of floor, vibration of adjacent equipment, vibration of neuronavigation equipment, etc.). These vibrations are transmitted to the end effector of the robot and will cause tiny displacement jitters. In intracranial surgery, even an additional 0.5 millimeters of jitter may cause accidental damage to brain functional areas or important blood vessels.

[0067] The humanoid torso of this system adopts a high-rigidity frame structure design. The frame material is a hybrid frame of carbon fiber composite material and aerospace-grade aluminum alloy, which maximizes the frame stiffness while ensuring the structure is lightweight.

[0068] The humanoid torso integrated vibration isolation structure adopts a combination of passive vibration isolation and active vibration suppression: the passive vibration isolation layer is set between the bottom of the torso and the mobile chassis, and is made of high-damping elastomer material to isolate the low-frequency vibration (environmental vibration with a frequency below 50Hz) transmitted by the chassis; the active vibration suppression system sets high-precision acceleration sensors and piezoelectric actuators at key locations on the torso to detect residual vibration in real time and actively cancel it by generating reverse vibration through piezoelectric actuators, suppressing the residual vibration amplitude to below 0.01 mm.

[0069] The effect of this feature is that it ensures that the surgical instruments at the end of the humanoid robot are not affected by environmental vibrations when operating inside the skull, thus guaranteeing the stability of intracranial positioning.

[0070] Specifically, the surgical arm is mounted on the humanoid torso.

[0071] The surgical arm adopts a human-like upper limb joint structure design (3 degrees of freedom for the shoulder joint, 1 degree of freedom for the elbow joint, and 3 degrees of freedom for the wrist joint, for a total of 7 degrees of freedom), which has a wide range of spatial movement capabilities. It can quickly move surgical instruments from their initial position to above the intracranial surgical operation area and adjust them to the planned entry angle and direction.

[0072] The end of the surgical arm is equipped with a dexterous hand, which uses a precision clamping mechanism to reliably hold various neurosurgical instruments (DBS electrodes, stereotactic biopsy needles, LITT fiber optic probes, RF ablation probes, microelectrode recording probes, ventricular drainage tubes, etc.). The clamping force can be precisely adjusted within the range of 0.1N to 5N to ensure that the surgical instruments will not slip or rotate undesirably during intracranial operations.

[0073] The dexterous hand also has the ability to fine-tune its posture (pitch and yaw each by ±3 degrees), which is used to make precise corrections to the final entry direction of surgical instruments.

[0074] The dexterous hand performs intracranial localization and surgical procedures with sub-millimeter precision, and the localization repeatability is better than 0.2 mm.

[0075] The dexterous hand integrates a miniature six-axis force / torque sensor to sense the changes in three-dimensional force and three-dimensional torque of surgical instruments during intracranial manipulation in real time, providing high-precision force perception data for closed-loop force control of surgical operations.

[0076] The effect of this feature is that it enables the end effector of humanoid robots to perform precise clamping of surgical instruments and intracranial manipulation with sub-millimeter precision.

[0077] Specifically, the control module is a central control unit with a human torso-like configuration, which generates control commands based on intraoperative navigation images and electrophysiological monitoring data to control the surgical arm to perform surgical actions.

[0078] The control module receives real-time navigation image data (real-time position of surgical instruments in the skull, deviation information from the target point, and distance information from brain functional areas) and electrophysiological monitoring data (neuronal firing characteristics and evoked potential amplitude changes of microelectrode recording signals) provided by the navigation system. It extracts the current position coordinates of the surgical instruments, the target point coordinates, and the three-dimensional deviation vector between the two through data analysis algorithms. At the same time, it analyzes electrophysiological signals to determine the brain region characteristics where the instrument tip is located. Then, it combines the surgical planning scheme to generate motion control commands for each joint of the surgical arm and operation control commands for the dexterous hand.

[0079] The control module employs a dual-modal fusion control approach combining navigation imaging and electrophysiology: the navigation imaging modality corrects instrument position deviations in real time using intraoperative images, while the electrophysiology modality verifies whether the instrument tip has reached the target brain region by recording neuronal discharge characteristics of signals from microelectrodes (e.g., in DBS surgery, STN characteristic discharge signals are used to determine whether the electrode has entered the subthalamic nucleus). The two modalities mutually verify and complement each other, achieving dual precision in both anatomical localization and electrophysiological functional localization.

[0080] When an abnormal decrease in the amplitude of evoked potentials is detected (such as a decrease in the amplitude of motion evoked potentials exceeding 50%), the control module automatically reduces the operating speed or suspends the operation.

[0081] The effect of this feature is that it enables intelligent surgical control based on both navigation imaging and electrophysiological feedback, making neurosurgical procedures safer and more precise.

[0082] The beneficial effects of the aforementioned humanoid torso configuration are as follows: the high rigidity frame and vibration isolation structure of the humanoid torso solve the problem of the impact of environmental vibration on the precision of intracranial surgery; the multi-degree-of-freedom joint structure of the surgical arm enables intracranial instrument positioning in a large workspace; the dexterous hand enables precise control of various neurosurgical instruments through fine gripping and posture micro-adjustment; and the control module's navigation image and electrophysiological dual-modal fusion control enables intelligent execution of surgical strategies.

[0083] like Figure 4 As shown, in a preferred embodiment, the system further includes a displacement compensation module, which is used to acquire the current state of the brain tissue through image acquisition during key stages of the operation, calculate the displacement, deformation and displacement vector field of the brain tissue, update the position information of each anatomical structure in the navigation model, and compensate for the corresponding spatial deviation.

[0084] Specifically, the intraoperative key stage image acquisition is the basic data acquisition function of the displacement compensation module.

[0085] In neurosurgery, brain shift is a key factor affecting the accuracy of intraoperative navigation.

[0086] The mechanisms of brain tissue displacement include: loss of cerebrospinal fluid (CSF) from the subarachnoid space after craniotomy, leading to a decrease in intracranial pressure and brain tissue collapse; bulging or collapse of brain tissue towards the incision due to gravity after dural incision; displacement of surrounding brain tissue towards the incision due to the formation of the resection cavity during tumor resection; and brain tissue deformation caused by factors such as intraoperative traction and cerebral edema.

[0087] The brain tissue displacement can reach 10 to 20 millimeters, causing a serious deviation between the preoperative image navigation model and the actual position of the brain tissue during surgery.

[0088] The displacement compensation module triggers image acquisition at key stages during surgery (after craniotomy / dural incision, key time points during tumor resection, after tumor resection cavity formation, etc.) to obtain the current state of brain tissue.

[0089] Intraoperative imaging methods include: intraoperative ultrasound (iUS, which acquires real-time ultrasound images of brain tissue through a sterile ultrasound probe at the craniotomy window), intraoperative magnetic resonance imaging (iMRI, which acquires high-contrast tomographic images of brain tissue through a dedicated intraoperative MRI scanner), intraoperative CT scan (iCT, which acquires tomographic images of brain tissue through a dedicated intraoperative CT scanner), and intraoperative surface stereoscopic vision (acquiring three-dimensional morphological data of the exposed cerebral cortex surface through a binocular stereo camera).

[0090] The effect of this feature is that it provides real-time intraoperative brain tissue state imaging data for subsequent brain tissue displacement detection and vector field calculation.

[0091] Specifically, the displacement compensation module registers and compares the intraoperative imaging data with the preoperative imaging data to detect displacement and deformation of brain tissue.

[0092] The detection algorithm employs a non-rigid registration method: First, feature points are extracted and matched from preoperative and intraoperative images, and anatomical landmarks such as sulci, gyri, ventricular margins, and vascular bifurcation points are extracted as registration reference points; then, a non-rigid registration algorithm based on B-spline transformation is used to calculate the spatial transformation field from preoperative to intraoperative images. This transformation field describes the displacement vector of each voxel position from preoperative to intraoperative.

[0093] For scenarios using intraoperative ultrasound, ultrasound-MRI cross-modal registration is also required to resolve grayscale differences between different image modalities.

[0094] The displacement compensation module also integrates a biomechanical finite element model (FEM). When intraoperative images only cover part of the brain region (such as ultrasound only covering a limited area below the craniotomy window), the displacement amount of brain tissue in the uncovered area is estimated by using the FEM model based on known mechanical parameters of brain tissue materials (elastic modulus, Poisson's ratio, viscoelastic coefficient, etc.) and boundary conditions (CSF loss, gravity direction, resection cavity shape, etc.), thus achieving displacement field estimation across the entire brain.

[0095] The effect of this feature is that it enables comprehensive and quantitative detection of intraoperative brain tissue displacement and deformation, providing accurate displacement data for displacement vector field calculation.

[0096] Specifically, the displacement compensation module constructs a three-dimensional displacement vector field based on the calculation results of non-rigid registration and biomechanical FEM model. This vector field assigns a three-dimensional displacement vector to each location point in three-dimensional space, describing the displacement direction and amount of brain tissue at that location from the preoperative position to the current intraoperative position.

[0097] Based on the three-dimensional displacement vector field, the displacement compensation module updates the positional information of all anatomical structures in the navigation model: the preoperative target coordinates (such as the center coordinates of the DBS target nucleus, the biopsy target coordinates, and the boundary coordinates of the ablation target area), the boundary coordinates of brain functional areas (motor area, language area, sensory area), the course of white matter fiber tracts (pyramidal tract, arcuate tract, etc.), the location of cerebral blood vessels, and the location of the ventricular margins are translated and updated according to the displacement vectors of the corresponding positions, generating a displacement-compensated intraoperative navigation model.

[0098] The updated navigation model can accurately reflect the actual location of brain tissue during surgery, providing a reliable spatial reference for the precise navigation of surgical instruments.

[0099] When image acquisition is triggered again during surgery (such as when tumor resection reaches a new stage), the displacement compensation module re-executes the complete process of displacement detection, vector field calculation and navigation model update to achieve multiple iterative compensation for brain tissue displacement.

[0100] The effect of this feature is that, through the calculation of the displacement vector field and the updating of the navigation model, it effectively compensates for the spatial deviation caused by the displacement of brain tissue during surgery, so that the navigation model always keeps in line with the actual position of brain tissue during surgery, thus ensuring the accuracy of surgical instrument navigation.

[0101] The beneficial effects of the aforementioned displacement compensation module are as follows: real-time state data of brain tissue are obtained through various imaging acquisition methods during key stages of the operation; comprehensive quantitative detection of brain tissue displacement and deformation is achieved through non-rigid registration and biomechanical FEM model; and precise compensation for intraoperative brain tissue displacement is achieved through three-dimensional displacement vector field calculation and navigation model update.

[0102] The displacement compensation module effectively solves the problem of decreased navigation accuracy caused by brain tissue displacement during neurosurgery, and significantly improves the accuracy of surgical instrument positioning during surgery.

[0103] like Figure 5 As shown, in a preferred embodiment, the system further includes an electrophysiological monitoring module, which is used to record and acquire neuronal discharge signals in the brain region where the tip of the surgical instrument is located through microelectrodes to determine whether the instrument has reached the target brain region, and to monitor the integrity of brain functional pathways in real time through evoked potentials, and to trigger an early warning when the monitoring signal is abnormal.

[0104] Specifically, the microelectrode recording (MER) is the target verification function of the electrophysiological monitoring module.

[0105] In functional neurosurgical procedures such as DBS electrode implantation, it is difficult to accurately determine whether the tip of the surgical instrument has reached the functionally active area of ​​the target nucleus by relying solely on image-guided positioning. Microelectrode recording is the gold standard technology for verifying electrophysiological target areas.

[0106] The electrophysiological monitoring module performs single-unit recording by integrating high-impedance microelectrodes (with impedance typically ranging from 0.5 to 1.5 megohms and a tip diameter of approximately 15 to 25 micrometers) into the tip of the surgical instrument or mounted in parallel, acquiring neuronal firing signals in the brain region where the surgical instrument tip is located.

[0107] The collected neuronal firing signals are pre-amplified, band-pass filtered (300Hz to 3000Hz) and converted from analog to digital before being sent to a signal analysis algorithm for real-time processing.

[0108] Signal analysis algorithms extract features from neuronal firing signals, including firing frequency, firing mode (regular firing, cluster firing, irregular firing, etc.), firing amplitude, and signal-to-noise ratio.

[0109] Neurons in different brain regions exhibit characteristic firing patterns: neurons in the subthalamic nucleus (STN) show high-frequency irregular firing (firing frequency 25 to 45 Hz), accompanied by obvious burst firing and oscillatory activity; neurons in the medial part of the globus pallidus (GPi) show high-frequency continuous firing (firing frequency 60 to 80 Hz), with a relatively regular firing pattern; neurons in the ventral intermediate nucleus of the thalamus (VIM) show rhythmic firing activity synchronized with tremors.

[0110] The electrophysiological monitoring module compares the real-time collected neuronal firing characteristics with the standard firing pattern feature library of the target nucleus. When the matching degree exceeds the preset threshold, it is determined that the surgical instruments have reached the target brain region, providing electrophysiological evidence for the final confirmation of the electrode implantation location.

[0111] The effect of this feature is that it provides electrophysiological verification capabilities for the target area localization of surgical instruments that surpass the accuracy of image localization, ensuring that instruments such as DBS electrodes are accurately implanted into the functionally active areas of the target nucleus.

[0112] Specifically, the evoked potential monitoring is a brain functional pathway protection function of the electrophysiological monitoring module.

[0113] There is a risk of damaging brain functional pathways during the delivery of surgical instruments to the target intracranial location and the execution of surgical procedures.

[0114] Evoked potential monitoring assesses the integrity of brain functional pathways in real time by continuously monitoring the electrophysiological responses of specific brain functional pathways during surgery.

[0115] The electrophysiological monitoring module supports multiple evoked potential monitoring types: Motor evoked potentials (MEP) are generated by transcranial electrical stimulation of the motor cortex or direct cortical electrical stimulation. Motor evoked potentials are recorded on the muscles of the contralateral limb to monitor the integrity of the pyramidal tract (corticospinal tract). When the MEP amplitude decreases by more than 50% or disappears completely, it indicates that the pyramidal tract may be damaged. Somatosensory evoked potentials (SEP) are generated by stimulating the contralateral median nerve or posterior tibial nerve and recorded on the scalp or cortex to monitor the integrity of the sensory conduction pathway. When the SEP amplitude decreases by more than 50% or the latency prolongs by more than 10%, it indicates that the sensory pathway may be damaged. Brainstem auditory evoked potentials (BAEP) are generated by intraocular sound stimulation to monitor the integrity of the auditory pathway and brainstem function during posterior fossa surgery.

[0116] The electrophysiological monitoring module continuously acquires and analyzes various evoked potential signals in real time, compares key parameters such as amplitude and latency of the real-time monitored waveform with baseline values, and triggers an early warning when abnormal changes occur in the monitored signal (significant decrease in amplitude or prolongation of latency), reminding the surgical system and surgeon that the current operation may affect brain functional pathways.

[0117] The effect of this feature is that it provides real-time electrophysiological monitoring and early warning capabilities for the integrity of brain functional pathways during surgical procedures, enabling the surgical system to detect risk signals in time before functional damage occurs.

[0118] The beneficial effects of the above-mentioned electrophysiological monitoring module are as follows: by recording the discharge signals of neurons in the target area through microelectrodes and comparing them with the standard discharge pattern feature library, the electrophysiological verification of the target area localization of surgical instruments is realized; through multi-channel continuous monitoring of motor evoked potentials, somatosensory evoked potentials and brainstem auditory evoked potentials, the real-time assessment of the integrity of brain functional pathways and abnormal early warning are realized.

[0119] The electrophysiological monitoring module provides functional electrophysiological support for the precision and safety of neurosurgical procedures.

[0120] like Figure 6 As shown, in a preferred embodiment, the system further includes a functional area protection module for real-time monitoring of the spatial distance between the surgical instruments and the brain functional areas, as well as the degree of impact of the surgical operation on the motor area, language area, and sensory area during the surgical operation. When the monitored indicators exceed the safety threshold, a graded protection response is executed, which includes at least one of early warning prompts, operation speed limits, and automatic stop.

[0121] Specifically, the real-time spatial distance monitoring is a basic monitoring function of the functional area protection module.

[0122] The functional area protection module acquires the three-dimensional coordinates of the surgical instrument tip in real time through the neuronavigation system, and calculates the spatial distance between the location of the brain functional areas (motor area, i.e., precentral gyrus; language area, i.e., Broca's area and Wernicke's area; sensory area, i.e., postcentral gyrus, etc.) located by fMRI before surgery and the white matter fiber tract pathways (pyramidal tract, arcuate tract, optic radiation, etc.) reconstructed by DTI before surgery.

[0123] Spatial distance calculation takes into account the actual intraoperative location of brain functional areas and white matter fiber tracts after being updated by the displacement compensation module, ensuring the accuracy of distance calculation.

[0124] The functional area protection module also integrates DTI fiber tract tracking data, which not only calculates the distance between the device and the brain functional area cortex, but also the closest distance between the device and the subcortical white matter fiber tracts (especially key motor fiber tracts such as the pyramidal tract), because damage to the subcortical white matter fiber tracts can also lead to severe neurological deficits.

[0125] In addition, the functional area protection module supports intraoperative cortical stimulation mapping, which records the patient's motor responses (such as contralateral limb muscle contraction) or language responses (such as speech interruption) by applying low-intensity electrical stimulation to the exposed cerebral cortex, directly verifying the precise boundary location of the functional area, and is used to calibrate and supplement the functional area location located by preoperative fMRI.

[0126] The effect of this feature is that it provides real-time and accurate distance monitoring data for the spatial relationship between surgical instruments and brain functional areas.

[0127] Specifically, the functional area protection module integrates spatial distance monitoring data, evoked potential monitoring data, and cortical electrical stimulation mapping results to assess in real time the impact of surgical procedures on the motor, language, and sensory areas.

[0128] Evaluation indicators include: the closest distance between the device and each functional area, the closest distance between the device and each key white matter fiber bundle, and the percentage change in evoked potential amplitude of the corresponding functional pathway.

[0129] The graded protection response is divided into three levels based on the degree of impact.

[0130] The first level is a warning prompt: when the surgical instrument is less than 10 mm away from the nearest brain functional area or key white matter fiber bundle, the system will issue a warning to the surgical system and the surgeon through an audible and visual alarm and a colored distance ring on the navigation screen, reminding that the instrument is approaching the brain functional area, but without interrupting the surgical operation.

[0131] The second level is operation speed limit: when the distance between the surgical instrument and the nearest brain functional area or key white matter fiber bundle is less than 5 mm, the system automatically limits the movement speed of the surgical instrument to the minimum safe speed (usually less than 20% of the normal operation speed), while increasing the navigation update frequency and evoked potential monitoring sensitivity, so that the surgical operation can be performed in a slow, high-precision mode.

[0132] The third level is automatic stop: when the distance between the surgical instrument and the nearest brain functional area or key white matter fiber bundle is less than 2 mm, or when the amplitude of the evoked potential decreases by more than 50%, the system automatically stops all movement of the surgical instrument, issues an emergency alarm, and waits for the surgeon's confirmation before deciding whether to continue the operation or adjust the surgical strategy.

[0133] The effect of this feature is that, through a graded protection response mechanism, progressively escalating safety measures are taken for functional area approach risks of different degrees. This avoids excessive sensitivity leading to frequent operation interruptions, while ensuring immediate protection of brain functional areas in the event of severe risks.

[0134] The beneficial effects of the aforementioned functional area protection module are as follows: it enables continuous quantitative monitoring of the distance relationship between surgical instruments and brain functional areas and white matter fiber tracts through real-time spatial distance monitoring; it enables multimodal precise positioning of functional areas by integrating DTI fiber tract tracking and cortical electrical stimulation mapping; and it adopts progressively upgraded safety measures such as early warning prompts, operation speed limits, and automatic stops for functional area approach risks of different degrees through a graded protection response mechanism.

[0135] The functional area protection module provides comprehensive and multi-layered intelligent protection for the safety of important brain functional areas such as the motor area, language area, and sensory area during neurosurgery.

[0136] In a preferred embodiment, the system further includes an end-effector module for real-time detection of micro-vibration signals at the end of the surgical arm and for generating reverse micro-motion commands through an active compensation algorithm to counteract end-effector jitter, thereby achieving sub-millimeter-level positioning stability of the surgical instrument tip during intracranial manipulation.

[0137] Specifically, the real-time detection of micro-vibrations is a fundamental sensing function of the end effector anti-vibration module.

[0138] In neurosurgery, the micro-vibrations and tremors of surgical instruments during intracranial manipulation directly affect the precision and safety of the surgery. Even with the suppression of vibration by the humanoid trunk vibration isolation structure, micro-vibrations may still remain at the end of the surgical arm caused by factors such as control tremors of the surgical arm joint servo system, backlash of joint gears, and resonance of the surgical arm's own structure.

[0139] The end effector module is equipped with a high-precision triaxial accelerometer and a high-precision triaxial angular velocity sensor (gyroscope) at the end of the surgical arm (at the dexterous hand base position) to collect linear acceleration and angular velocity data of the end of the surgical arm in real time at a sampling frequency of not less than 1000Hz.

[0140] The micro-vibration detection algorithm performs frequency domain analysis (Fast Fourier Transform FFT and Short-Time Fourier Transform STFT) on the acquired acceleration and angular velocity signals to extract the frequency components, amplitude, and phase information of the micro-vibrations at the end of the surgical arm.

[0141] Frequency domain decomposition can separate vibration components from different sources: joint servo jitter usually manifests as high-frequency narrowband vibration (frequency range of 50Hz to 200Hz) related to the servo control frequency, vibration caused by gear backlash manifests as low-frequency pulse vibration related to the reversal of joint movement direction, and structural resonance manifests as specific frequency vibration corresponding to the natural frequency of the surgical arm.

[0142] The effect of this feature is that it provides precise frequency, amplitude, and phase data of micro-vibrations at the end of the surgical arm for the active compensation algorithm.

[0143] Specifically, the active compensation algorithm adopts a control strategy that combines feedforward compensation and feedback compensation.

[0144] Feedforward compensation is based on the dynamic model of the surgical arm and joint motion commands. It predicts the micro-vibration components (including joint servo jitter and vibration caused by gear backlash) that the surgical arm end will generate when executing the current motion command, and pre-adds the reverse compensation component to the motion command to cancel the predicted vibration.

[0145] Feedback compensation is based on the micro-vibration detection results. The residual micro-vibration signal detected is analyzed in real time. The vibration component that needs to be compensated is extracted by an adaptive filtering algorithm (such as an adaptive notch filter or a least mean square LMS adaptive filter) and a reverse micro-motion command is generated.

[0146] The reverse micro-motion commands are executed by a high-bandwidth micro-motion actuator mounted at the end of the surgical arm. The micro-motion actuator employs a piezoelectric ceramic actuator (PZT) with nanometer-level displacement resolution and kilohertz-level response bandwidth, enabling it to generate precise reverse micro-motions in three directions to counteract end jitter.

[0147] The piezoelectric actuator's maximum stroke covers the maximum amplitude range that the surgical arm tip may generate (typically ±0.5 mm), with a response time of less than 1 millisecond, ensuring that the reverse compensation motion is precisely synchronized in time with the compensated vibration.

[0148] After active compensation, the residual vibration amplitude of the surgical instrument tip during intracranial manipulation is suppressed to below 0.05 mm, and the positioning stability reaches the sub-millimeter level (better than 0.1 mm).

[0149] The effect of this feature is that, through the coordinated work of the active compensation algorithm and the piezoelectric micro-motion actuator, it achieves efficient cancellation of micro-vibrations at the end of the surgical arm, enabling the tip of the surgical instrument to achieve sub-millimeter-level positioning stability during intracranial operations.

[0150] The beneficial effects of the aforementioned end-effector module are as follows: it enables precise frequency domain analysis of micro-vibration signals at the end of the surgical arm through real-time detection of micro-vibrations by high-precision sensors; it achieves accurate compensation for micro-vibrations from multiple sources through an active compensation algorithm that combines feedforward compensation and feedback compensation; and it achieves vibration cancellation with nanometer-level precision through reverse micro-motion execution by piezoelectric ceramic actuators.

[0151] The end-effector module enables sub-millimeter-level positioning stability of surgical instrument tips during intracranial operations, providing end-effector-level assurance for the precision and safety of neurosurgical procedures.

[0152] like Figure 7 As shown, in a preferred embodiment, the system further includes an embodied intelligent decision-making module, which is connected to the intelligent assessment module, surgical planning module, humanoid neurosurgical execution module, and auxiliary rehabilitation module. This module is used to autonomously perform cognitive understanding and decision-making reasoning of intracranial multimodal sensory information based on the neurosurgical task objectives, generate positioning and operation strategies for surgical instruments and distribute them to the humanoid neurosurgical execution module, and autonomously iterate and correct the strategies based on execution feedback until the surgical task objectives are achieved.

[0153] Specifically, the embodied intelligent decision-making module adopts a multi-cognitive intelligent agent collaborative architecture, including multiple cognitive intelligent agents and a surgical situation sharing space, wherein the multiple cognitive intelligent agents include: The system comprises three cognitive agents: a perception agent, a planning agent, and a safety agent. The former is independent of the perception, planning, and safety agents and possesses an independent perception path. The former performs security audits on the safety commands and rejects them when security risks are detected. The safety agent serves as the communication medium between the multiple cognitive agents, employing an asynchronous read-write mechanism for information exchange and collaboration.

[0154] The perceptual and cognitive agent is responsible for the acquisition, fusion, and semantic processing of intraoperative multimodal perceptual data. This agent independently connects to the intraoperative multimodal sensing system, including surgical field visual sensors, end-effector force / torque sensors, intraoperative navigation imaging equipment, electrophysiological monitoring devices, and a spatial navigation and positioning system.

[0155] The perceptual and cognitive intelligent agent, based on a multimodal perceptual data fusion algorithm, transforms heterogeneous sensor data into a unified structured scene representation. This structured scene representation encompasses the following semantic dimensions: tissue state description (including tissue type labeling, elastic modulus estimation, blood perfusion status, and damage severity grading), instrument pose information (including the six-degree-of-freedom pose and motion state vectors of each surgical instrument in the surgical area coordinate system), anatomical structure mapping (including the location of key anatomical landmarks, delineation of safety boundaries, and marking of surgical restricted areas), and abnormal event labeling (including spatiotemporal identifiers of abnormal states such as bleeding detection, brain tissue damage identification, and instrument displacement determination).

[0156] Scene representations are continuously refreshed at a fixed frequency and written into the scene state domain of the surgical situation sharing space. The other three cognitive agents obtain a unified understanding of the current surgical situation by reading the scene representations in the shared space, without having to process the raw sensor data independently, thus eliminating redundant computation and data consistency risks among multiple agents.

[0157] The planning and reasoning agent undertakes the hierarchical decomposition of surgical tasks and the dynamic generation of operational strategies. This agent reads the structured scene representation provided by the perceptual and cognitive agent, as well as the preset surgical task objectives, from the shared surgical situation space, and executes hierarchical surgical planning and reasoning.

[0158] At the task level, the planning and reasoning agent decomposes the overall surgical objective into an ordered sequence of sub-tasks (e.g., localization → surgical approach establishment → intracranial target manipulation → electrophysiological verification → instrument withdrawal). Explicit state transition conditions are defined between each sub-task, and the agent autonomously determines the completion status of the current sub-task and whether the triggering conditions for subsequent sub-tasks are met based on real-time changes in the scenario representation.

[0159] At the policy level, the planning and reasoning agent generates specific operational policies for the currently executed subtask. These operational policies include positioning policies (precise spatial coordinates of the target point and optimal entry angle), entry policies (path selection, velocity curve planning, obstacle avoidance constraints), operation policies (operation type, force range, duration parameters), and force control policies (force threshold setting, compliance control parameters, adaptive gain coefficient). The generated candidate policies are written into the policy queue field of the shared space for the motion control agent to consume and execute.

[0160] When the scene representation deviates significantly, the planning and reasoning agent automatically triggers a policy revision mechanism to reassess the applicability of the current policy and generate an alternative policy. Leveraging the characteristics of an asynchronous architecture, the planning and reasoning agent can pre-generate operational policies for subsequent subtasks while the current subtask is being executed, achieving pipelined parallel processing of planning and execution.

[0161] The motion control agent is responsible for converting and issuing real-time motion commands from the operational strategy to the surgical execution mechanism. This agent reads the operational strategy from the strategy queue domain of the surgical posture sharing space and transforms the abstract strategy into a sequence of real-time motion control commands for the surgical arm and dexterous hand.

[0162] The motion control agent comprises three functional sub-modules: a trajectory planning module, which maps high-level operational strategies to time-series motion trajectories in joint space and Cartesian space, generating directly executable trajectories based on kinematic constraints, collision detection, and motion smoothness optimization criteria; a force-position hybrid control module, which dynamically switches between position control and force control modes according to the operation type—switching to force control mode for electrode delivery to precisely constrain propulsion force, and switching to position control mode for alignment and positioning to ensure spatial positioning accuracy—with force-position control parameters adaptively adjusted based on real-time force feedback; and a real-time anomaly response module, which detects abnormal conditions such as sudden changes in contact force, instrument slippage, and accidental collisions with millisecond-level latency, and immediately executes local protective actions, including emergency braking, safe retraction, and joint locking, without waiting for intervention from other cognitive agents, thus endowing the motion control agent with independent local safety decision-making capabilities.

[0163] Before being sent to the surgical execution mechanism, the control commands generated by the motion control agent must be reviewed and approved by the safety guardian agent. Execution feedback data (including actual motion trajectory, end-effector contact force, execution status, and trajectory deviation) is written back to the execution feedback domain in the shared space for the perception and cognition agent to use for closed-loop updates of the scene state.

[0164] The safety-guarding intelligent agent is responsible for independent safety monitoring and risk veto throughout the entire surgical procedure. This agent is architecturally independent of the decision-making chain of the other three cognitive agents and possesses two key mechanisms: Independent Perception Path: The safety guardian directly accesses the raw sensor data stream, without relying on the scene representation generated by the perception and cognition agent in the surgical situation sharing space. This design ensures that even if the perception and cognition agent malfunctions, experiences latency, or data anomalies, the safety guardian can still independently complete the safety assessment of the surgical situation based on the raw data, forming a perception redundancy guarantee at the architectural level.

[0165] A veto mechanism: The safety-guarding intelligent agent continuously monitors three types of information sources in parallel: the content written by each cognitive agent in the shared space, the raw sensor data acquired by its own independent sensing channels, and the consistency deviation between the two. A veto is triggered when any of the following risk conditions are detected: the operation strategy violates anatomical safety constraints (such as the instrument trajectory intruding into the safety envelope of critical blood vessels or nerves), the motion command exceeds the force / velocity / acceleration safety threshold, there is a significant consistency deviation between its own sensing results and the representation of the shared space scene (indicating that the sensing link may have failed), or the system's cumulative risk rating exceeds the preset safety threshold.

[0166] Upon triggering a veto, the safety guardian agent executes a tiered response strategy based on the risk level: Mild risk: Freezes current motion control commands, sends a strategy regeneration request to the planning and reasoning agent, pauses the surgical procedure but maintains the system's ready state, and automatically resumes execution after the new strategy passes safety review. Moderate risk: Downgrades the system control mode to physician-assisted mode, suspends the autonomous decision-making link, and allows the surgeon to determine the subsequent operation plan. Severe risk: Immediately triggers an emergency stop command, resets all motion control commands to zero, automatically retracts the surgical execution mechanism along a preset safe trajectory to a safe position, and the system enters a safety lock state, which can only be restarted after being manually unlocked by the physician.

[0167] The security guardian agent continuously writes real-time risk rating information into the risk rating domain of the shared space, and each cognitive agent has reading permission to ensure that the entire multi-agent system maintains a global consensus on the current security situation.

[0168] The surgical situation sharing space serves as the unified communication medium among the four cognitive agents (with the sole exception of the independent perception pathway of the security guardian agent). Essentially, it is a structured shared memory space with strict access control, internally divided into five functional data domains: Scene state domain: Written at a fixed frequency by the perceptual and cognitive agents, and accessed only by other agents. It stores a unified structured representation of the current surgical scene, covering tissue state, instrument pose, anatomical structure mapping, and abnormal event marking.

[0169] Task progress domain: Written, maintained, and updated by the planning and reasoning agent. It records the hierarchical execution progress of the surgical task, including the set of completed subtasks, the current execution status of subtasks, and the ordered queue of subtasks to be executed.

[0170] Policy queue field: written by the planning and reasoning agent and read by the motion control agent. The sequence of operation policies to be executed is managed using a first-in, first-out (FIFO) queue.

[0171] Risk rating domain: Written by the security guardian agent, all agents have read access. It records the system's overall risk rating index and the sub-assessment of specific risk factors in real time.

[0172] Execution feedback domain: written by the motion control agent and read by the perception and cognition agent for closed-loop updates of the scene state. It records the actual execution results of motion commands, including the actual motion trajectory, end-effector contact force, execution status flag, and deviation from the planned trajectory.

[0173] The shared space employs an asynchronous read-write mechanism, where each cognitive agent independently executes read and write operations according to its own operating cycle, without blocking, waiting, or synchronization lock dependencies. This mechanism enables the four cognitive agents to operate in true parallelism—the perception agent continuously refreshes the scene state, the planning and reasoning agent continuously generates operation strategies, the motion control agent continuously issues motion commands, and the security guardian agent continuously performs security monitoring—each working independently at its optimal operating frequency, achieving loosely coupled asynchronous information exchange and multi-agent collaboration through the shared space.

[0174] In one specific implementation, the multimodal perception data fusion algorithm of the perceptual cognitive agent is implemented using a cross-modal attention mechanism based on Vision Transformer. Specifically, intraoperative navigation images (intraoperative MRI or CT), microscopic / endoscopic visual images, electrophysiological monitoring signals, and microelectrode recording signals are respectively input into the encoders of the corresponding modalities to extract multi-scale feature maps. The visual images are processed by a weighted visual encoder to extract spatial features, while the electrophysiological signals are processed by a one-dimensional convolutional encoder to extract temporal features. Then, a cross-modal cross-attention module enables bidirectional information interaction between features from different modalities. Navigation image features are used as a query and a key-value pair with microscopic visual features for attention calculation to achieve spatial alignment between anatomical structures and the surgical field. Electrophysiological signal features are used as a query and a key-value pair with navigation image features for attention calculation to achieve mapping association between functional area localization and anatomical structures. Finally, the fused multimodal features are output in parallel through a semantic segmentation head and a target detection head to generate a unified scene representation that includes the three-dimensional distribution of intracranial anatomical structures, lesion boundaries, locations of brain functional areas, white matter fiber tract routing, six-DOF device pose, brain tissue displacement status, and neuronal firing activity characteristics. This cross-modal attention mechanism enables deep fusion of imaging anatomical information and electrophysiological functional information at the feature level, and compared with traditional multimodal data stitching methods, it can more effectively capture the spatial correspondence between anatomical structures and functional areas.

[0175] In one specific implementation, the decision engine of the planning and reasoning agent is implemented using a visual-language large model fine-tuned for the surgical domain, specifically a variant of the LLaVA architecture. The visual projection layer maps the scene representation output by the perceptual cognitive agent to the token embedding space of the language model, concatenates it with textualized neurosurgical system prompts, and then feeds it into the Transformer decoder to autoregressively generate operational strategies. Fine-tuning employs a LoRA low-rank adaptation strategy, training only the adapter weights and visual projection layer parameters, while freezing the base language model weights, to achieve precise adaptation to the neurosurgical domain while maintaining the general reasoning capabilities of the large model. The agent is configured with a stage-specific adaptive meta-query mechanism for neurosurgical procedures: defining a neurosurgical stage state machine (stereotactic localization → surgical approach establishment → dural incision → intracranial target exposure → precise target manipulation → electrophysiological verification → hemostasis and cranial closure → instrument withdrawal), with each stage configured with a dedicated set of learnable query vectors as additional key-value pairs injected into the Transformer decoder's cross-attention layer. For example, the meta-query-guided model in the "intracranial target exposure" phase focuses on anatomical landmarks of brain sulci and gyri, distances between functional areas, and brain tissue displacement compensation status, while the "precise target manipulation" phase focuses on changes in electrophysiological signals, safe distances between instruments and functional areas, and manipulative force feedback. During phase transitions, the meta-query vector undergoes smooth linear interpolation over several frames to prevent abrupt changes in attention patterns.

[0176] In one specific implementation, the motion control agent employs a conditional flow matching framework to train a diffusion strategy model. This method defines a linear interpolation path from random noise to the target motion trajectory, and trains a vector field network to learn the instantaneous direction at any position along this path. The vector field network uses a one-dimensional temporal UNet architecture, with three-way conditional information (policy conditions, scene conditions, and safety conditions) injected layer by layer through a feature modulation mechanism. During inference, the algorithm solves along the vector field from random noise using multi-step Euler integration, rereading the latest conditions from the shared space at each integration step to achieve dynamic coupling. Each step's output is projected through three levels of safety constraints: functional area potential field projection (a safety envelope to prevent instruments from intruding into the motor, speech, and sensory areas), force constraint truncation (limiting intracranial manipulation forces within a safe range), and velocity constraint truncation (limiting the speed of intracranial instrument movement). During training, frequency domain regularization is applied to suppress high-frequency jitter, ensuring sub-millimeter-level stability of intracranial manipulation. To address the delicate nature of neurosurgical procedures, the size of motion blocks is dynamically adjusted according to the stage: large blocks are used during the surgical approach establishment stage to ensure motion continuity, while the smallest blocks are used during the precise manipulation of the intracranial target area to provide frame-by-frame fine control. Transitions are achieved through crossfading when blocks change. Vibration compensation signals from the end-effector anti-flicker module are injected as an additional safety channel, ensuring that the generated motion trajectory automatically includes anti-jitter compensation components.

[0177] In one specific implementation, the independent safety assessment of the safety guardian agent is achieved through adversarial diffusion training combined with dynamic time warping distance. The adversarial diffusion module shares the same network architecture as the motion control agent but has the opposite training objective: it learns to generate the action patterns most likely to cause accidents in the current scenario, using neurosurgical safety accident trajectories (such as instrument intrusion into functional areas leading to motor or language function impairment, electrode implantation deviating from the target nucleus, excessive traction of brain tissue causing hemorrhage, surgical instruments damaging important blood vessels or white matter fiber bundles, etc.) as the generation target. During inference, multiple predicted failed trajectories are generated in parallel to form a danger envelope. The optimal trajectory output by the motion control agent is compared with the danger envelope using dynamic time warping (DTW) distance measurement (using a fast approximation algorithm). A safety response is triggered when the safety distance is lower than the dynamic threshold. The dynamic threshold is obtained by multiplying the base threshold by a stage correction coefficient (more stringent for intracranial target area operation stage) and an anatomical correction coefficient (more stringent for areas closer to functional areas and important blood vessels). Compared with Euclidean distance, DTW distance measurement can handle the nonlinear changes in the time axis caused by brain tissue displacement and cerebrospinal fluid loss during neurosurgery, and more accurately captures the morphological similarity of the surgical trajectory.

[0178] In one specific implementation, the perceptual consistency verification of the safety guardian agent is achieved using a contrastive learning dual-path verification mechanism. Specifically, the safety guardian agent reads the scene representation generated by the perceptual cognitive agent from the shared space, and the safety assessment representation generated by independently processing the original sensor data (including intraoperative navigation images, electrophysiological monitoring signals, microelectrode recording signals, and instrument force feedback) through its own independent perceptual pathway. These two representations are then input into a similarity evaluation network trained based on contrastive learning. During the training phase, this network learns to map semantically consistent representation pairs to nearby locations in the embedding space and semantically inconsistent representation pairs to distant locations using a contrastive learning loss function. During the inference phase, it determines perceptual consistency by calculating the cosine similarity of the two representations in the embedding space. When the cosine similarity is below a preset threshold for several consecutive frames, the safety guardian agent determines that there may be an anomaly in the perceptual link (such as navigation registration drift, uncompensated brain tissue displacement, electrophysiological signal artifact interference, etc.), triggers a safety response, and switches to a safety degradation mode based on its own independent perception. During training, hidden risk scenarios such as "brain tissue displacement causing inconsistencies between the navigation model and the actual anatomy, but not obvious in navigation images" are used as negative samples to enable the dual pathways to actively generate divergent signals under such hidden risks.

[0179] In one specific implementation, the surgical situation sharing space is further configured with a forward scene prediction engine, employing a lightweight causal Transformer architecture. It takes the recent frame scene state embeddings and current policy encoding as input, and regresses the scene state increments at future time steps through multiple prediction heads, with a prediction window of approximately 0.5 seconds. In neurosurgery, the prediction engine incorporates brain tissue displacement trends and cerebrospinal fluid dynamic changes into the prediction variables, enabling the prediction to reflect the impact of intracranial pressure changes on target area coordinate drift. The prediction results are used by the motion control agent as diffusion conditions, by the safety guardian agent for predictive safety pre-screening, and by the planning and reasoning agent for evaluating the applicability of the strategy. The prediction engine is configured with an online calibration mechanism, continuously updating parameters using an exponential moving average (EMA) based on execution feedback bias to adapt to the individual patient's brain tissue biomechanical characteristics and intraoperative brain tissue displacement evolution. When the prediction bias continuously exceeds the normal range, the system is marked as invalid, and automatically switches to reactive control and conservative safety threshold modes.

[0180] The data flow and collaborative relationships between the embodied intelligent decision-making module and other modules in the system are as follows: With the intelligent assessment module, the perceptual and cognitive agent obtains the patient's preoperative assessment results (including disease diagnosis, surgical indication determination, intracranial lesion localization results, functional area risk assessment results, etc.) from the intelligent assessment module, encodes them as surgical task constraints, and writes them into the task progress domain of the surgical situation sharing space, providing the planning and reasoning agent with constraint inputs when formulating surgical strategies; With the surgical planning module, the planning and reasoning agent obtains a three-dimensional cranial model, surgical approach planning, instrument selection scheme, and target area coordinate parameters from the surgical planning module, serving as a reference benchmark for task hierarchy decomposition and operational strategy generation. Simultaneously, when intraoperative scene representation shows brain tissue displacement or target area coordinate drift due to cerebrospinal fluid loss, the planning and reasoning agent feeds back the corrected target area coordinates and approach parameters to the module. The surgical planning module updates the surgical plan. Between the surgical planning module and the humanoid neurosurgery execution module, the motion control agent sends real-time motion control commands to the humanoid neurosurgery execution module (including the orientation operation module, surgical arm, and dexterous hand). The perception and cognition agent receives intraoperative navigation image information, electrophysiological monitoring data, microelectrode recording signals, and instrument position data from the humanoid neurosurgery execution module as multimodal perception inputs. The safety protection agent directly accesses the raw sensor data stream of the execution module through an independent perception pathway for parallel safety monitoring. Between the surgical planning module and the rehabilitation module, after the surgery is completed, the planning and reasoning agent transmits the intraoperative operation records (including the actual surgical approach, target area operation parameters, electrophysiological verification results, intraoperative brain tissue displacement compensation records, etc.) to the rehabilitation module through the task progress domain of the surgical situation sharing space, providing intraoperative data support for the formulation of personalized rehabilitation plans.

[0181] The beneficial effects of the embodied intelligent decision-making module based on the multi-cognitive agent collaborative architecture are as follows: Through the perceptual cognitive agent, unified cognitive understanding of intracranial multimodal perceptual information is achieved, transforming heterogeneous data such as intraoperative navigation images, electrophysiological signals, microelectrode recordings, and force feedback into standardized scene representations; through the planning and reasoning agent, hierarchical autonomous planning and dynamic adjustment of neurosurgical operations are achieved, enabling real-time correction of target area coordinates and access strategies based on brain tissue displacement; through the motion control agent, the conversion from high-level operational strategies to precise surgical movement commands is realized, ensuring sub-millimeter precision requirements for intracranial operations; through the independent perceptual pathways and veto mechanism of the safety guardian agent, architectural-level safety redundancy is constructed, effectively preventing irreversible damage such as instrument intrusion into functional areas or important blood vessels; through the asynchronous read / write mechanism of the surgical posture sharing space, true parallel operation and loosely coupled collaboration of the four cognitive agents are achieved; and through the real-time fusion of electrophysiological monitoring data, the system can continuously verify treatment effects during target area operations.

[0182] This application provides a control method for the aforementioned humanoid intelligent neurosurgical robot system. The specific solution of this method is as follows: Figure 8 As shown below, the control method of the humanoid intelligent neurosurgical robot system will be described in detail.

[0183] Combination Figure 8 As shown, the control method of the humanoid intelligent neurosurgical robot system includes: S101, based on the patient's medical data and physical signs, conducts targeted disease assessments and determines the indications for neurosurgery, generates personalized auxiliary assessment results for the patient, and displays them. S102. Based on the patient's multimodal neuroimaging data, construct a three-dimensional fusion model of the cranium and brain that includes lesions, brain functional areas, and white matter fiber tracts. Based on the three-dimensional fusion model of the cranium and brain, perform targeted neurosurgical planning for the patient, generate a personalized surgical plan, and display it. S103, according to the surgical plan, the surgical instruments are precisely guided to the target position in the brain through stereotactic positioning, and the neurosurgical operation is performed based on intraoperative multimodal perception; S104 generates and displays a personalized postoperative rehabilitation plan for the patient based on their medical data and postoperative assessment results.

[0184] The control method of the humanoid intelligent neurosurgical robot system provided in the above embodiments of this application corresponds to the humanoid intelligent neurosurgical robot system provided in the embodiments of this application. Therefore, the specific content of the control method corresponds to the humanoid intelligent neurosurgical robot system. The specific content can be referred to the records in the humanoid intelligent neurosurgical robot system, which will not be repeated in this application.

[0185] The control method of the humanoid intelligent neurosurgical robot system provided in the above embodiments of this application is based on the same inventive concept as the humanoid intelligent neurosurgical robot system provided in the embodiments of this application, and has the same beneficial effects as the methods adopted, run or implemented by the application stored therein.

[0186] Based on the same inventive concept, another embodiment of this application provides an electronic device for implementing the control method of the humanoid intelligent neurosurgical robot system described in the above embodiments. Figure 9 As shown, the electronic device includes a memory 301 and a processor 303.

[0187] Memory 301 can be configured to store a program.

[0188] Additionally, memory 301 can also be configured to store various other data to support operation on the electronic device. Examples of this data include instructions for any application or method used to operate on the electronic device, contact data, phonebook data, messages, pictures, videos, etc.

[0189] Memory 301 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. Processor 303, coupled to memory 301, is used to execute programs in memory 301 for: Based on the patient's medical data and physical signs, targeted disease assessments and neurosurgical indications are determined, generating and displaying personalized auxiliary assessment results. Based on the patient's multimodal neuroimaging data, a three-dimensional fusion model of the cranium and brain, including lesions, brain functional areas, and white matter fiber tracts, is constructed. Based on the three-dimensional fusion model of the cranium and brain, targeted neurosurgical planning is carried out for the patient, a personalized surgical plan is generated, and the plan is displayed. According to the surgical plan, the surgical instruments are precisely guided to the target intracranial location through stereotactic positioning, and the neurosurgical operation is performed based on intraoperative multimodal perception. Based on the patient's medical data and postoperative assessment results, a personalized postoperative rehabilitation plan is generated and displayed.

[0190] In this application, Figure 9 The diagram only shows some components and does not mean that the electronic device includes only these components. Figure 9 The components shown.

[0191] The electronic device provided in this embodiment is based on the same inventive concept as the control method of the humanoid intelligent neurosurgical robot system provided in this application embodiment, and has the same beneficial effects as the methods adopted, run or implemented by the application stored therein.

[0192] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-readable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0193] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 The computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0194] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0195] In a typical configuration, a computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory. Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0196] This application also provides a computer-readable storage medium corresponding to the control method of the humanoid intelligent neurosurgical robot system provided in the foregoing embodiments, wherein a computer program (i.e., a program product) is stored thereon, and the computer program, when run by a processor, executes the control method of the humanoid intelligent neurosurgical robot system provided in any of the foregoing embodiments.

[0197] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, optical disc read-only memory (CDROM), digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0198] The computer-readable storage medium provided in the above embodiments of this application and the control method of the humanoid intelligent neurosurgical robot system provided in the embodiments of this application are based on the same inventive concept and have the same beneficial effects as the methods adopted, run or implemented by the applications stored therein.

[0199] It should be noted that numerous specific details are set forth in the specification provided herein. However, it will be understood that embodiments of this application may be practiced without these specific details. In some instances, well-known structures and techniques have not been shown in detail so as not to obscure the understanding of this specification.

[0200] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, system, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, system, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, system, article, or apparatus that includes said element.

[0201] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A humanoid intelligent neurosurgical robot system, characterized in that, The system is built upon an embodied intelligence closed-loop architecture of perception-cognition-decision-execution-feedback, including: The intelligent assessment module is used to conduct targeted disease assessments and determine the indications for neurosurgery based on the patient's medical data and vital signs, generate personalized auxiliary assessment results for the patient, and display them. The surgical planning module is used to construct a three-dimensional fusion model of the cranium and brain, including lesions, brain functional areas, and white matter fiber tracts, based on the patient's multimodal neuroimaging data. Based on the three-dimensional fusion model of the cranium and brain, targeted neurosurgical planning is carried out for the patient, generating a personalized surgical plan and displaying it. The humanoid neurosurgical execution module is used to precisely guide surgical instruments to the target intracranial location through stereotactic positioning according to the surgical plan, and to perform neurosurgical operations based on intraoperative multimodal perception. The auxiliary rehabilitation module is used to generate and display personalized postoperative rehabilitation plans for patients based on their medical data and postoperative assessment results.

2. The humanoid intelligent neurosurgical robot system according to claim 1, characterized in that, The humanoid neurosurgical execution module includes: The humanoid robot body is used to autonomously move to the surgical position and dock with the operating table. The end of the humanoid robot body is equipped with a standardized mechanical interface to adapt to different types of surgical execution modules. The orientation operation module is installed on the humanoid robot body through the standardized mechanical interface. It is used to accurately deliver surgical instruments to the intracranial target location along the planned path according to stereoscopic coordinates, and to perform surgical operations after reaching the target location. The surgical operations include at least one of electrode implantation, tissue sampling, or lesion ablation.

3. The humanoid intelligent neurosurgical robot system according to claim 1, characterized in that, The humanoid neurosurgical execution module includes a control module, a humanoid torso, and a surgical arm. The humanoid torso integrates a vibration isolation structure to suppress the impact of environmental vibrations on surgical precision. The surgical arm is equipped with a dexterous hand at its end for holding surgical instruments and performing intracranial positioning and surgical operations with sub-millimeter precision. The control module generates control commands based on intraoperative navigation images and electrophysiological monitoring data to control the surgical arm to perform surgical actions.

4. The humanoid intelligent neurosurgical robot system according to claim 1, characterized in that, The system also includes a displacement compensation module, which is used to acquire the current state of brain tissue through image acquisition during key stages of surgery, calculate the displacement, deformation and displacement vector field of brain tissue, update the position information of each anatomical structure in the navigation model, and compensate for the corresponding spatial deviations.

5. The humanoid intelligent neurosurgical robot system according to claim 1, characterized in that, The system also includes an electrophysiological monitoring module, which is used to record neuronal discharge signals in the brain region where the tip of the surgical instrument is located through microelectrodes to determine whether the instrument has reached the target brain region, and to monitor the integrity of brain functional pathways in real time through evoked potentials, triggering an early warning when the monitoring signal is abnormal.

6. The humanoid intelligent neurosurgical robot system according to claim 1, characterized in that, The system also includes a functional area protection module, which is used to monitor in real time the spatial distance between surgical instruments and brain functional areas during surgical operations, as well as the degree of impact of surgical operations on motor areas, language areas, and sensory areas. When the monitored indicators exceed the safety threshold, a graded protection response is executed, which includes at least one of early warning prompts, operation speed limits, and automatic stop.

7. The humanoid intelligent neurosurgical robot system according to claim 1, characterized in that, The system also includes an end-effector module for real-time detection of micro-vibration signals at the end of the surgical arm. It generates reverse micro-motion commands through an active compensation algorithm to counteract end-effector jitter, thereby achieving sub-millimeter-level positioning stability of the surgical instrument tip during intracranial operations.

8. The humanoid intelligent neurosurgical robot system according to claim 1, characterized in that, The system also includes an embodied intelligent decision-making module, which is connected to the intelligent assessment module, surgical planning module, humanoid neurosurgery execution module, and auxiliary rehabilitation module. It is used to autonomously perform cognitive understanding and decision-making reasoning of intracranial multimodal sensory information according to the neurosurgical task objectives, generate positioning and operation strategies for surgical instruments and send them to the humanoid neurosurgery execution module, and autonomously iterate and correct the strategies based on execution feedback until the surgical task objectives are achieved.

9. A control method for a humanoid intelligent neurosurgical robot system, applied to the humanoid intelligent neurosurgical robot system according to any one of claims 1 to 8, characterized in that, include: Based on the patient's medical data and physical signs, targeted disease assessments and neurosurgical indications are determined, generating and displaying personalized auxiliary assessment results. Based on the patient's multimodal neuroimaging data, a three-dimensional fusion model of the cranium and brain, including lesions, brain functional areas, and white matter fiber tracts, is constructed. Based on the three-dimensional fusion model of the cranium and brain, targeted neurosurgical planning is carried out for the patient, a personalized surgical plan is generated, and the plan is displayed. According to the surgical plan, the surgical instruments are precisely guided to the target intracranial location through stereotactic positioning, and the neurosurgical operation is performed based on intraoperative multimodal perception. Based on the patient's medical data and postoperative assessment results, a personalized postoperative rehabilitation plan is generated and displayed.

10. An electronic device, characterized in that, The system includes a processor and a memory, wherein the memory stores computer-executable instructions, and the processor executes the computer-executable instructions to implement the steps of the control method for the humanoid intelligent neurosurgical robot system of claim 9.

11. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the control method for the humanoid intelligent neurosurgical robot system of claim 9.

Images