Grinding control method and device, robot and storage medium

Abstract

The invention relates to the technical field of spine vertebral plate grinding, and discloses a grinding control method and device, a robot and a storage medium. The method comprises the steps that a contact force signal in the feeding direction in the spine grinding process is collected in real time; preprocessing the contact force signal in the feeding direction to obtain a standardized time sequence data fragment, and processing the standardized time sequence data fragment through a pre-trained bone feature extraction model to output a corresponding grinding physical state category; consistency verification is conducted on output results of the grinding physical state categories, and the grinding physical state categories continuously reaching the preset frame number serve as effective grinding physical state categories; according to the effective grinding physical state category, determining a logic grinding state with surgical semantics at the current moment by utilizing an anatomical structure sequence and a state transition rule of a spinal tissue set in a finite-state machine; and generating a grinding control instruction according to the logic grinding state to control the grinding robot to execute corresponding grinding operation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of spinal lamina grinding, and in particular to a grinding control method and device, a robot, and a storage medium. BACKGROUND

[0002] In spinal surgery, lamina grinding is often used to achieve spinal cord decompression, but due to the important tissues such as spinal cord and nerve root adjacent to the spinal cord, the operation precision requirement is extremely high. Existing robot-assisted grinding technology mostly relies on force / torque sensors for feedback control, and the bone tissue type is judged by setting a threshold value. However, the spinal lamina has a sandwich structure of cortical bone-cancellous bone-cortical bone, and the mechanical properties of the inner and outer cortical bones are similar, so it is difficult to distinguish them only by instantaneous force signals, which may lead to misjudgment. In addition, the traditional method lacks dynamic tracking ability of the grinding process, and may fail to identify when the cancellous bone enters the inner cortical bone again, which may cause penetration risk. SUMMARY

[0003] Therefore, the embodiments of the present application provide a grinding control method and device, a spinal grinding robot and a computer readable storage medium, which can realize accurate and safe grinding of the spinal lamina.

[0004] In a first aspect, the embodiments of the present application provide a grinding control method, comprising: real-time acquisition of a feed direction contact force signal in a spinal grinding process; preprocessing of the feed direction contact force signal to obtain a standardized time series data segment, and processing of the standardized time series data segment by a pre-trained bone quality feature extraction model to output a corresponding grinding physical state category; consistency verification of the output result of the grinding physical state category, and taking a grinding physical state category that continuously reaches a preset frame number as an effective grinding physical state category; determination of a logical grinding state with surgical semantics at the current time according to the anatomical structure sequence and state transition rule of the spinal tissue set in a finite state machine; generation of a grinding control instruction according to the logical grinding state to control the spinal grinding robot to perform a corresponding grinding operation.

[0005] In an optional implementation, the consistency verification of the output result of the grinding physical state category, and taking a grinding physical state category that continuously reaches a preset frame number as an effective grinding physical state category, comprises: setting a continuous counter, and setting the initial value of the counter to zero; if the grinding physical state category output at the current time is the same as that of the previous frame, the counter is incremented by one; if the current output of the grinding physical state category is not the same as the previous frame, the count of the counter is reset to one; When the duration count of the counter reaches a preset threshold, the current grinding physical state category is taken as the effective grinding physical state category.

[0006] In an optional implementation, the preprocessing of the feed direction contact force signal to obtain the standardized time series data segment includes: The feed direction contact force signal is smoothed, and the smoothed feed direction contact force signal is sliced according to a preset time window to generate a time series data segment; The time series data segment is dynamically standardized to obtain the standardized time series data segment.

[0007] In an optional implementation, the grinding physical state category includes an uncontacted state, a high resistance state, and a low resistance state. The uncontacted state represents a state of no force; the high resistance state represents that the grinding object is dense bone tissue; and the low resistance state represents that the grinding object is loose bone tissue or a penetration region.

[0008] In an optional implementation, the anatomical structure sequence and state transition rule of the finite state machine is modeled based on the cortical bone-cancellous bone-cortical bone anatomical structure of the spinal lamina tissue, so that the same grinding physical state category is mapped to different logical grinding states under different context time sequence conditions.

[0009] In an optional implementation, the logical grinding state includes an outer cortical bone grinding state, a cancellous bone grinding state, an inner cortical bone grinding state, and a penetration state. The anatomical structure sequence and state transition rule of the spinal tissue set in the finite state machine is used to determine the logical grinding state with surgical semantics at the current time according to the effective grinding physical state category, including: Starting from the grinding initial state, if the high resistance state is detected for the first time, the outer cortical bone grinding state is determined to be entered; When in the outer cortical bone grinding state and the high resistance state is detected to change to the low resistance state, the cancellous bone grinding state is determined to be entered; When in the cancellous bone grinding state and the low resistance state is detected to change to the high resistance state again, the inner cortical bone grinding state is determined to be entered; When in the inner cortical bone grinding state and the high resistance state is detected to mutate to the low resistance state, the penetration state is determined to be entered.

[0010] In an optional implementation, in the training process of the bone quality feature extraction model, all samples derived from the cortical bone region are uniformly labeled as the high resistance state.

[0011] In a second aspect, the embodiments of the present application provide a grinding control device, comprising: a signal acquisition module, configured to acquire a feed direction contact force signal in a spine grinding process in real time; a preprocessing module, configured to preprocess the feed direction contact force signal, obtain a standardized time series data segment, and process the standardized time series data segment through a pre-trained bone quality feature extraction model to output a corresponding grinding physical state category; a state confirmation module, configured to perform consistency verification on an output result of the grinding physical state category, and take a grinding physical state category that continuously reaches a preset frame number as an effective grinding physical state category; a logic control module, configured to determine a logical grinding state with surgical semantics at a current time according to a state transition rule and an anatomical structure sequence of a spine tissue set in a finite state machine, according to the effective grinding physical state category; a motion control module, configured to generate a grinding control instruction according to the logical grinding state, so as to control a spine grinding robot to perform a corresponding grinding operation.

[0012] In a third aspect, the embodiments of the present application provide a spine grinding robot, comprising a processor and a memory, the memory stores a computer program, and the processor is configured to execute the computer program to implement the grinding control method described above.

[0013] In a fourth aspect, the embodiments of the present application provide a computer readable storage medium, which stores a computer program, and the computer program is executed on a processor to implement the grinding control method described above.

[0014] The embodiments of the present application have the following beneficial effects: the present application acquires standardized time sequence data segments by collecting and preprocessing the feeding direction contact force signals in real time, and then analyzes the time sequence data by using a pre-trained bone feature extraction model, so as to accurately identify the physical state category in the current grinding process and enhance the perception ability of different bone tissue impedance characteristics. Further, the embodiments of the present application perform consistency verification on the state output by the model, and only when the same state lasts for a preset number of frames is it confirmed as a valid state, thereby reducing the risk of misjudgment caused by transient interference or prediction fluctuations and improving the reliability of system decision-making. Further, the confirmed valid physical state is input into a finite state machine, combined with the inherent anatomical structure sequence and state transition rules of the spinal tissue, to realize dynamic tracking and semantic analysis of the grinding process, so that the system can accurately understand the current surgical stage. Finally, control instructions are generated according to the logical grinding state with surgical semantics, and then the safety control of the spinal grinding robot is realized, and the precision, stability and clinical safety of the grinding operation are overall improved. BRIEF DESCRIPTION OF DRAWINGS

[0015] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments. It should be understood that the following drawings only show some of the embodiments of the present application, and therefore should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can also be obtained without creative labor on the basis of these drawings.

[0016] Figure 1 A first flowchart of the grinding control method of the embodiments of the present application is shown; Figure 2 A second flowchart of the grinding control method of the embodiments of the present application is shown; Figure 3 A third flowchart of the grinding control method of the embodiments of the present application is shown; Figure 4 A fourth flowchart of the grinding control method of the embodiments of the present application is shown; Figure 5 A structure diagram of the grinding control device of the embodiments of the present application is shown. DETAILED DESCRIPTION

[0017] The technical solutions of the embodiments of the present application will be described clearly and completely in the embodiments of the present application combined with the drawings. Obviously, the described embodiments are only some of the embodiments of the present application, not all the embodiments.

[0018] The components of the application embodiments described and illustrated herein can be arranged and designed in a wide variety of different configurations. Therefore, the following detailed description of the application, as provided in the context of a specific embodiments of the application, is not intended to limit the scope of the application, but is merely representative of selected embodiments of the application. The detailed description set forth below in connection with the appended drawings is intended as a description of the process and procedures in accordance with the embodiments of the application and is not intended to represent the only embodiments in which the application can be practiced. Each embodiment of the application can be implemented alone, or in combination with another embodiments of the application. The term "comprising" (and its variants), as used herein, means "including" or "containing", but not limited to, so that it is intended to cover also embodiments consisting of, consisting essentially of, or consisting of the listed steps, components, elements, members, or combinations thereof.

[0019] Hereinafter, the terms "include", "have", and their conjugates, as used in various embodiments of the present application, merely indicate the presence of the features, numbers, steps, operations, elements, components, or combinations thereof, and do not exclude the possibility of the presence or addition of one or more other features, numbers, steps, operations, elements, components, or combinations thereof. In addition, the terms "first", "second", "third", and the like are used only to distinguish the description, and cannot be understood as indicating or implying relative importance.

[0020] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which various embodiments of the present application belong. The terms (such as terms defined in a generally used dictionary) will be interpreted to have the same meaning as the contextual meaning in the relevant technical field and will not be interpreted to have an idealized or overly formal meaning unless clearly defined in various embodiments of the present application.

[0021] Some embodiments of the present application will be described in detail below with reference to the accompanying drawings. The following embodiments and features in the embodiments can be combined with each other without conflict.

[0022] In spinal surgery, laminectomy is a common and critical operation, which is widely used in spinal cord decompression, tumor resection, and vertebral canal shaping. Due to the complex anatomical structure of the spine, important tissues such as the spinal cord, nerve roots, and large blood vessels are adjacent to the surrounding, the traditional manual grinding relies on the experience and feeling of the doctor, and there are problems such as low operation precision, risk of fatigue accumulation leading to mistakes, etc. In recent years, with the development of medical robot technology, robot-assisted spinal grinding systems have been gradually applied in clinical practice, aiming to improve the precision, stability, and safety of the operation.

[0023] Existing robot grinding control systems usually monitor the contact force information in the grinding process in real time through force / torque sensors installed on the end of the mechanical arm or the main shaft, and use it as a feedback signal for parameter adjustment. Some systems use a fixed force threshold to determine the type of tissue. When the force value detected exceeds a certain range, it is considered to be in the dense bone tissue area, and below a certain range it is considered to be in the loose bone or penetration state. In addition, some studies have tried to introduce vibration signals, spindle current or acoustic emission signals as auxiliary criteria to enhance the perception of bone changes.

[0024] However, the above methods still have obvious limitations. The spinal lamina has a typical cortical bone-cancellous bone-cortical bone sandwich structure, where the outer and inner cortical bones both exhibit high stiffness and high impedance mechanical properties, showing similar high amplitude responses on the real-time force signal. It is difficult to distinguish between the two by relying solely on the instantaneous force size, resulting in the system being unable to determine whether the current contact is the outer plate that can be safely removed or the inner plate that needs to be strictly protected adjacent to the spinal cord cavity wall. In addition, existing control strategies are mostly based on single observation values at the current time to make decisions, lacking the ability to remember the time sequence of the grinding process. When the tool enters the cancellous bone from the outer cortical bone, the resistance drops sharply, but continues to advance and encounters high resistance again. The system may mistakenly identify the high resistance signal after this as the initial cutting stage, failing to adjust the cutting parameters in time, and there is a risk of excessive grinding or even penetrating the inner cortical bone. At the same time, motor noise, tool chatter and local bone islands in the surgical environment can easily cause fluctuations in the force signal, resulting in model output oscillation. If directly used for control decisions, it may cause robot motion oscillation, affecting operation stability. Based on this, the present embodiment proposes a grinding control method that builds a dual recognition architecture of physical state classification and logical grinding state reasoning, integrates deep learning models, signal de-bouncing mechanisms and finite state machines based on anatomical constraints, and thus realizes safe and precise grinding control of the spinal lamina cortical bone-cancellous bone-cortical bone sandwich structure.

[0025] The grinding control method will be described below in conjunction with some specific embodiments.

[0026] Figure 1 A flowchart of the grinding control method of the present embodiment is shown. Exemplarily, the grinding control method includes steps S100-S500: Step S100, real-time acquisition of the contact force signal in the feed direction during the spinal grinding process.

[0027] Exemplarily, in the spinal column grinding process, the contact force signal in the feed direction along the Z-axis direction of the tool coordinate system is collected in real time through a six-dimensional force / torque sensor installed between the end effector of the spinal column grinding operation robot and the high-speed grinding drill. The direction is defined as the main force transmission direction of the grinding tool, and the dynamic change thereof directly reflects the tissue impedance characteristics of the current grinding area.

[0028] In step S200, the feed direction contact force signal is preprocessed to obtain a standardized time series data segment, and the standardized time series data segment is processed by a pre-trained bone quality feature extraction model to output a corresponding grinding physical state category.

[0029] Since the above-mentioned original force signal often mixes high-frequency noise such as motor vibration, mechanical resonance and environmental interference, it needs to be preprocessed to improve the signal-to-noise ratio and adapt to the subsequent model input requirements. Based on this, in this step, the original signal needs to be preprocessed first to generate a standardized time series data segment with a unified format and scale.

[0030] In some embodiments, as Figure 2 shown, preprocessing the feed direction contact force signal to obtain a standardized time series data segment includes steps S210-S220: In step S210, the feed direction contact force signal is smoothed, and the smoothed feed direction contact force signal is sliced according to a pre-set time window to generate a time series data segment.

[0031] Exemplarily, for the collected feed direction contact force signal, a Butterworth low-pass filter is first used for smoothing to filter out the high-frequency noise components introduced by motor vibration, tool vibration and environmental interference. In this embodiment, a second-order Butterworth filter with a cutoff frequency of 20 Hz can be selected, which has a maximum flat amplitude-frequency response characteristic and can minimize phase distortion while effectively suppressing high-frequency interference, thereby retaining the key dynamic characteristics of the force signal representing the bone quality change trend.

[0032] In this embodiment, the filter is implemented in a digital IIR (infinite impulse response) structure, and the output sequence is obtained by the difference equation , where is a normalization coefficient, is a filter coefficient, where , , the filter order, sampling frequency and cutoff frequency are designed and fixed in advance; is the current time index, is a loop variable in the summation operation.

[0033] The filter can ensure that the transient glitches in the original signal are effectively smoothed, while retaining the slow-changing trends reflecting the changes in tissue hierarchy. Subsequently, the smoothed continuous signal stream is segmented using a sliding window mechanism. The window length L is set to correspond to a fixed time of historical data, for example, when the system sampling frequency is 100 Hz, L = 100 is set, that is, each window contains 100 consecutive sampling points in the last 1 second, and an independent time series data segment is output at each time.

[0034] In step S220, the time series data segment is dynamically standardized to obtain a standardized time series data segment.

[0035] Exemplarily, in order to eliminate the differences in bone density between different individuals and the reference offset problem caused by sensor zero drift, each time series data segment is normalized in this step. Specifically, based on the training data set constructed by a large number of ex vivo grinding experiments, the global mean and standard deviation of all sample force signals are obtained in advance.

[0036] For the current input time series data segment , the normalization transformation can be performed according to the formula ; wherein is the normalized output sequence. After normalization, each time series data segment is mapped to a unified order of magnitude range, so that the final output standardized time series data segment retains the time dynamic characteristics of the original signal and has good consistency and comparability.

[0037] After obtaining the standardized time series data segment through the above process, the standardized time series data segment is further processed by a pre-trained bone quality feature extraction model to output the corresponding grinding physical state category. Specifically, after obtaining the standardized time series data segment, the obtained standardized time series data segment is input into a pre-trained bone quality feature extraction model, and the dynamic mechanical mode contained in the input standardized time series data segment is analyzed by the model to output the corresponding grinding physical state category at the current time.

[0038] The grinding physical state category includes non-contact state, high resistance state and low resistance state; the non-contact state represents the force-free state (i.e., the grinding tool is in a suspended or idling state, and does not substantially contact the bone tissue, at which time the force signal amplitude is close to zero); the high resistance state represents that the grinding object is dense bone tissue, such as the outer or inner cortical bone, which exhibits a sustained high contact force level during grinding; the low resistance state represents that the grinding object is loose bone tissue or a penetration region (i.e., the case of penetrating the bone wall into the unsupported space), which typically exhibits a sudden drop or even disappearance of resistance.

[0039] Further, in the training process of the pre-trained bone quality feature extraction model, all samples from the cortical bone region are uniformly labeled as high resistance state.

[0040] Exemplarily, the bone quality feature extraction model is constructed based on a deep learning architecture, and in this embodiment, a TimesNet network can be used as the core model. TimesNet extracts the periodic components in the input signal through fast Fourier transform, reshapes the one-dimensional time series data into a two-dimensional tensor form, and captures multi-scale time period patterns using the Inception structure, thereby effectively identifying the characteristic force change law generated by different bone tissues.

[0041] In the model training stage, a large amount of real grinding data from the spine ex vivo experiment can be used for supervised learning. Each labeled sample contains standardized force signals and their corresponding artificial labels. In particular, when making the training set, all samples from the cortical bone region (including the outer layer and the inner layer) are uniformly labeled as high resistance state, regardless of their anatomical location. This labeling strategy aims to make the model focus on identifying the common physical characteristics of hard tissues, rather than performing semantic tasks to determine which layer of hard tissue it belongs to. Furthermore, the perception-reasoning separation mechanism formed enables subsequent state discrimination of logical grinding state based on temporal context. After training, the model can be solidified and deployed in the control system of the robot. In actual operation, the model receives the standardized time series data segments output from step S200 in real time, generates a probability distribution of three categories through forward reasoning, and outputs the highest probability category as the grinding physical state category at the current time.

[0042] In step S300, the output result of the grinding physical state category is verified for consistency, and the grinding physical state category that has reached a preset number of frames continuously is taken as the effective grinding physical state category.

[0043] Exemplarily, since the bone quality feature extraction model used in step S200 is implemented based on deep learning, it may have output jump phenomenon at the junction of different bone layers, for example, frequent oscillation between high resistance state and low resistance state. Such transient jitter, if directly passed to the subsequent control link, may cause the grinding robot to misjudge the tissue level, frequently adjust parameters, and even cause mechanical arm micro-vibration, affecting the safety and stability of the operation. Therefore, this embodiment introduces a consistency verification method based on a counter, which is used to smooth and confirm the effectiveness of the grinding physical state category output by the model. This method sets a time duration condition to ensure that only when a state has been stable for a long enough time is it recognized as a real tissue change, thereby filtering out false triggers caused by short-term noise or model prediction fluctuations.

[0044] In some embodiments, asFigure 3 As shown, step S300 includes steps S310-S340: Step S310, a persistence counter is set, and the initial value of the counter is set to zero.

[0045] Step S320, if the current output grinding physical state category is the same as the previous frame, the counter is incremented by one.

[0046] Step S330, if the current output grinding physical state category is not the same as the previous frame, the count of the counter is reset to one.

[0047] Step S340, when the persistence count of the counter reaches a preset threshold, the current grinding physical state category is taken as the valid grinding physical state category.

[0048] Specifically, first, a persistence counter is set to record the number of frames that the current corresponding grinding physical state is continuously output by the model, and the count of the counter is zero in the initial state. During system operation, the output results from the bone quality feature extraction model are received frame by frame. When it is detected that the state category output by the current frame is consistent with that of the previous frame, it indicates that the grinding physical state has continuity, and the counter value is incremented by one. For example, if three consecutive frames all output a high resistance state, the counter value is increased from 1 to 2 and then to 3. When the state output by the model switches, that is, the current frame is inconsistent with the previous frame, it indicates that a new tissue region may be entered or interference is received. Since the count value of the counter has not reached the preset threshold, the grinding physical state counted by the current counter is taken as an invalid grinding physical state, and the count of the counter is reset to one, starting the persistence observation of the new grinding physical state. This way can avoid state misconfirmation caused by temporary fluctuations, and thus improve the anti-interference ability of the system.

[0049] In this embodiment, a preset threshold is set, which can be determined according to the actual sampling frequency and the clinical response requirement, for example, 5-10 frames. Only when the counter of a state is continuously incremented and reaches the preset threshold, the system formally confirms it as the current valid physical state and takes it as the input of the next stage, otherwise, the system maintains the valid grinding physical state confirmed at the last time. For example, when entering the cancellous bone region after the outer plate grinding is completed, the force signal is converted from high resistance to low resistance. Although there may be individual frame misjudgment as high resistance at the initial stage, the preset threshold is not reached, and the low resistance is correctly identified as the valid state when the count of the low resistance reaches the preset threshold. Conversely, if it is only a transient disturbance, it cannot meet the persistence requirement and will not trigger the state transition.

[0050] Through the consistency checking manner, the embodiment can realize the de-bounce processing on the original model output, and further construct a stable and reliable intermediate state output layer, thereby guaranteeing the safety of the entire control system under a dynamic environment.

[0051] In step S400, the logical grinding state with the surgical semantics in the current moment is determined according to the anatomical structure sequence and the state transition rule of the spine tissue set in the finite state machine according to the effective grinding physical state category.

[0052] Demonstratively, after obtaining the stable and effective physical state, the system enters the logical reasoning stage. The finite state machine constructed based on the anatomical prior knowledge converts the physical state from the perception layer into the logical grinding state with clear clinical significance.

[0053] The finite state machine not only depends on the current input state information, but also combines the running history of the system to make context judgment, thereby realizing the dynamic tracking and semantic analysis of the grinding process. The core idea is that the same physical signal feature may correspond to completely different tissue level meanings in different grinding timing backgrounds. For example, the high resistance state indicates the beginning of contacting the outer cortical bone in the initial stage, and when it appears again after the cancellous bone, it should be identified as a key warning signal for approaching the inner cortical bone.

[0054] In the embodiment, the logical grinding state includes the outer cortical bone grinding state, the cancellous bone grinding state, the inner cortical bone grinding state, and the penetration state. These states are directly related to the operation intention and safety strategy in the surgery, and constitute the basis of the autonomous control decision of the robot.

[0055] Further, in order to ensure that the state recognition conforms to the real anatomical law, the anatomical structure sequence and the state transition rule of the finite state machine are modeled based on the typical cortical bone-cancellous bone-cortical bone sandwich structure of the spine lamina tissue. This modeling method fully utilizes the fixed and unique layer sequence relationship of the spine bone tissue in space, so that the same grinding physical state category can be mapped to different logical grinding states in different timing contexts, thereby improving the accuracy and reliability of the tissue level discrimination.

[0056] Demonstratively, as shown in Figure 4 , step S400 includes steps S410-S440: Step S410, from the grinding initial state, if the high resistance state is detected for the first time, it is determined to enter the outer cortical bone grinding state.

[0057] Step S420, when being in the outer cortical bone grinding state and detecting that the high resistance state changes to the low resistance state, it is determined to enter the cancellous bone grinding state.

[0058] Step S430, when in the cancellous bone grinding state and detecting again the transition from low resistance state to high resistance state, it is determined to enter the inner cortical bone grinding state.

[0059] Step S440, when in the inner cortical bone grinding state and detecting the sudden change from high resistance state to low resistance state, it is determined to enter the penetration state.

[0060] Specifically, in the initial stage of the operation, the grinding tool is located outside the bone tissue in the non-contact or low force state. When the system first detects the transition from the non-contact state to the high resistance state, it is determined that the hard tissue contacted at this time must be the outer cortical bone according to the anatomical initial sequence uniqueness principle. Therefore, the finite state machine performs state transition, enters the outer cortical bone grinding state, and starts the corresponding normal cutting parameters.

[0061] As the grinding goes deeper, the tool gradually penetrates the outer cortical bone and enters the cancellous bone area with loose internal structure. This process shows that the continuous high resistance suddenly drops to low resistance. Since the previous state is the outer cortical bone grinding state, this resistance drop is not abnormal, but an expected interlayer transition. Therefore, the state machine recognizes this change as a safety event and enters the cancellous bone grinding state, while reducing the feed speed to prepare for detecting the next layer.

[0062] When the system is working in the cancellous bone area, if the transition from low resistance state to high resistance state is detected again, it means that the grinding tool has encountered a new dense barrier. Since there are only two layers of cortical bone in the spinal lamina structure, and the outer layer has been ground off, it can be determined that it is the inner cortical bone adjacent to the wall of the spinal cord cavity under the condition of conforming to the typical lamina anatomical structure. According to the anatomical uniqueness, the state machine determines to enter the inner cortical bone grinding state, and immediately triggers the protection mechanism to further reduce the cutting amount or switch to the micro pulse grinding mode to prevent further intrusion.

[0063] When the system is in the inner cortical bone grinding state, any sudden disappearance of resistance means that the key barrier has been broken through, and it is very likely that the inner plate has been ground through and entered the unsupported lateral space of the spinal cord. This situation belongs to a high-risk penetration event. Once the finite state machine detects a rapid jump from high resistance to low resistance, it determines the penetration state and triggers the emergency response process.

[0064] In the above manner, the embodiment realizes the conversion from physical perception to surgical semantics. The finite state machine not only can distinguish between tissue levels with similar mechanical properties but different clinical meanings, but also can dynamically update the grinding state based on the time sequence, thereby supporting the subsequent precise control of the robot.

[0065] Step S500, generating a grinding control instruction according to the logical grinding state to control the spinal column grinding robot to perform corresponding grinding operation.

[0066] In this step, according to different logical grinding states, the corresponding grinding control instructions are dynamically generated, and the working parameters of the six-degree-of-freedom robot and the high-speed grinding drill are adjusted in real time through the robot motion controller, so that safe and adaptive bone tissue grinding operation is realized.

[0067] Specifically, when the system determines to be in the outer cortical bone grinding state, it indicates that normal cutting operation is being performed on the target area of the surface dense bone. At this time, the tissue structure is complete and far away from the nerve sensitive area, allowing higher processing efficiency. Therefore, the control system maintains the default feed speed (for example, 0.2 mm / s) and spindle speed (for example, 8000 rpm) to ensure the smoothness of the operation and the progress of the operation.

[0068] When the state machine enters the cancellous bone grinding state, it indicates that the grinding tool has penetrated the outer cortical bone and entered the internal loose bone area. The mechanical impedance of this area decreases significantly, and if high-intensity cutting is continued, it may cause tool sinking or excessive invasion. Therefore, the control system automatically reduces the feed speed and reduces the single cutting depth, while the grinding path can be adjusted to a scanning reciprocating motion to ensure smooth transition and prepare for subsequent detection of the inner layer barrier.

[0069] When the inner cortical bone grinding state is detected, it indicates that the grinding site has approached the spinal cord cavity wall and is in a high-risk area. At this time, even a small overcut may cause nerve damage. Therefore, the control system immediately starts the safety protection strategy: further reduces the feed speed, and even switches to a pulse-type micro-grinding mode, that is, only the grinding drill is started and a very small step is advanced under certain trigger conditions, and then paused and evaluated the feedback signal, realizing fine control and minimizing the risk of mispenetration.

[0070] Once the state machine determines to enter the penetration state, it indicates that the system detects that the inner cortical bone barrier suddenly disappears, and there is a serious risk of perforation. At this time, the control system immediately executes the emergency response process: first, cut off the power output of the grinding drill and stop all rotating actions; second, control the robot arm to quickly retreat a preset distance (for example, 0.5 mm) in the opposite direction of the feed to escape the dangerous area; at the same time, trigger the audible and visual alarm device to give visual and audible cues to the surgical personnel; finally, lock the automatic feed function, which must be confirmed by the doctor before continuing operation to prevent secondary injury.

[0071] The above control instructions can be transmitted to the robot motion controller through the communication interface, and the motion controller analyzes the instruction content and drives the servo motor, spindle driving unit and safety interlock module to cooperate and execute, thereby realizing precise control.

[0072] The embodiment realizes dynamic identification and safety control of tissue levels in the grinding process by real-time acquisition of the contact force signal in the feed direction during the spine grinding process, combined with signal preprocessing, deep learning model inference, state consistency verification and anatomical constraint finite state machine logical judgment. Compared with the control strategy based on single force value threshold or instantaneous feedback in the prior art, the embodiment not only focuses on the physical stress level at the current time, but also introduces the context information in the time dimension, thereby constructing the mapping relationship from physical perception to surgical semantics. In the specific implementation process, the high-frequency noise interference caused by motor vibration and mechanical resonance is effectively suppressed by Butterworth low-pass filtering and smoothing processing of the original force signal, and the signal-to-noise ratio is improved; through sliding window slicing and time series standardization processing, the input data has good consistency and comparability, which adapts to the influence of different individual bone quality differences and sensor reference drift. The pre-trained TimesNet model is used to analyze the standardized time series segment, which can capture the multi-scale periodic patterns implied in the force signal, thereby accurately identifying basic physical states such as non-contact, high resistance and low resistance, especially showing strong feature extraction capability in the cortical bone and cancellous bone junction area. Further, by setting a consistency verification mechanism based on a counter, the state output by the model is de-bounced, and only when a certain state is continuously maintained for a preset number of frames is it confirmed as valid input, avoiding misjudgment caused by local bone islands, instantaneous disturbances or model prediction fluctuations, thereby improving the stability of the system in a dynamic environment. On this basis, the valid physical state confirmed is input into the finite state machine based on the sequential modeling of the spinal lamina anatomical structure, and the inherent layer sequence uniqueness is used to realize accurate inference of the logical state. This way makes the same mechanical characteristics have different clinical meanings in different time series backgrounds, fundamentally solving the key technical problem that the prior art cannot distinguish between the inner and outer cortical bones. Finally, according to the determined logical grinding state, the corresponding control instruction is generated, and then precise regulation is realized, that is, in the outer cortical bone stage, efficient cutting is maintained, in the cancellous bone stage, the feed is reduced to smoothly transition, in the inner cortical bone stage, pulse-type micro-grinding is switched to prevent overcutting, and when the penetration risk is detected, the power is immediately cut off, the system is quickly withdrawn, and an audible and light alarm is triggered, thereby ensuring the safety of the operation.

[0073] Figure 5 A structural schematic diagram of a grinding control device of an embodiment of the application is shown. Exemplarily, the grinding control device comprises: The signal acquisition module 100 is configured to acquire the contact force signal in the feed direction during the spine grinding process in real time.

[0074] The preprocessing module 200 is configured to preprocess the contact force signal in the feed direction, obtain a standardized time series data segment, and process the standardized time series data segment through a pre-trained bone quality feature extraction model to output a corresponding grinding physical state category.

[0075] The state confirmation module 300 is configured to perform consistency verification on the output result of the grinding physical state category, and take the grinding physical state category that continuously reaches a preset frame number as a valid grinding physical state category.

[0076] The logic control module 400 is configured to input the valid grinding physical state category at the current moment into a finite state machine, so that the finite state machine determines a logical grinding state with a surgical semantic at the current moment according to the current valid grinding physical state category and a state transition rule of the anatomical structure of the spinal column.

[0077] The motion control module 500 is configured to generate a grinding control instruction according to the logical grinding state, so as to control the spinal column grinding robot to perform corresponding grinding work.

[0078] It can be understood that the device of the embodiment corresponds to the grinding control method of the above-mentioned embodiment, and the optional items in the above-mentioned embodiment are also applicable to the embodiment, and thus will not be described here again.

[0079] The application further provides a spinal column grinding robot, which exemplarily comprises a processor and a memory, wherein the memory stores a computer program, and the processor executes the computer program, so that the spinal column grinding robot performs the functions of the grinding control method or each module of the grinding control device.

[0080] The processor can be an integrated circuit chip with a signal processing capability. The processor can be a general-purpose processor, including a central processing unit (CPU), a graphics processing unit (GPU), and a network processor (NP), a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, or at least one of the above. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc., which can implement or execute the disclosed methods, steps, and logic block diagrams in the embodiments of the application.

[0081] The memory can be, but is not limited to, a random access memory (RAM), a read only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electric erasable programmable read-only memory (EEPROM), and the like. Among them, the memory is used to store a computer program, and the processor can execute the computer program correspondingly after receiving an execution instruction.

[0082] The application further provides a computer readable storage medium for storing the computer program used in the spinal grinding robot. For example, the computer readable storage medium can include, but is not limited to, a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various storage program codes.

[0083] In several embodiments provided in the present application, it should be understood that the disclosed apparatus and method can also be implemented by other ways. The apparatus embodiments described above are only schematic, for example, the flow chart and block diagram in the drawings show the possible implementation architecture, function and operation of the apparatus, method and computer program product according to the embodiments of the present application. In this regard, each block in the flow chart or block diagram can represent a module, a program segment or a part of code, which contains one or more executable instructions for implementing the specified logical function. It should also be noted that, in alternative implementation ways, the functions noted in the block can also occur in different order from that noted in the drawings. For example, two consecutive blocks can actually be executed substantially in parallel, and sometimes they can be executed in reverse order, depending on the functions involved. It should also be noted that each block in the structural diagram and / or flow chart, and the combination of blocks in the structural diagram and / or flow chart, can be implemented by a dedicated hardware-based system for executing the specified function or action, or can be implemented by a combination of special hardware and computer instructions.

[0084] In addition, each functional module or unit in the embodiments of the present application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0085] If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application essentially or the parts that contribute to the prior art or parts of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a number of instructions for causing a computer device (which can be a smart phone, a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application.

[0086] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be covered within the protection scope of the present application.

Claims

1. A grinding control method, characterized in that, include: Real-time acquisition of contact force signals in the feed direction during the spinal grinding process; The feed direction contact force signal is preprocessed to obtain a standardized time-series data segment, and the standardized time-series data segment is processed by a pre-trained bone feature extraction model to output the corresponding grinding physical state category. The output results of the grinding physical state category are checked for consistency, and the grinding physical state category that continuously reaches the preset number of frames is taken as the valid grinding physical state category. Based on the effective grinding physical state category, the logical grinding state with surgical semantics at the current moment is determined by using the anatomical structure sequence of the spinal tissue and the state transition rules set in the finite state machine. Grinding control commands are generated based on the logical grinding state to control the spinal grinding robot to perform corresponding grinding operations.

2. The grinding control method according to claim 1, characterized in that, The process of performing consistency verification on the output results of the grinding physical state category, and taking grinding physical state categories that continuously reach a preset number of frames as valid grinding physical state categories, includes: Set a continuous counter and set the initial value of the counter to zero; If the current output of the grinding physical state category is the same as that of the previous frame, the counter is incremented by one; If the current output of the grinding physical state category is different from that of the previous frame, then the counter count is reset to one. When the counter's continuous count reaches a preset threshold, the current grinding physical state category is taken as the valid grinding physical state category.

3. The grinding control method according to claim 1, characterized in that, The preprocessing of the contact force signal in the feed direction to obtain standardized time-series data segments includes: The feed direction contact force signal is smoothed, and the smoothed feed direction contact force signal is sliced ​​according to a preset time window to generate time-series data segments. The time-series data segment is dynamically standardized to obtain the standardized time-series data segment.

4. The grinding control method according to claim 1, characterized in that, The grinding physical state categories include non-contact state, high resistance state, and low resistance state; Wherein, the non-contact state indicates a state without force; the high resistance state indicates that the grinding object is dense bone tissue; and the low resistance state indicates that the grinding object is loose bone tissue or a penetrating area.

5. The grinding control method according to claim 4, characterized in that, The anatomical sequence and state transition rules of the finite state machine are modeled based on the anatomical structure of the spinal vertebral lamina, which allows the same grinding physical state category to be mapped to different logical grinding states under different contextual temporal conditions.

6. The grinding control method according to claim 4, characterized in that, The logical grinding states include outer cortical bone grinding state, cancellous bone grinding state, inner cortical bone grinding state, and penetration state; The step of determining the logical grinding state with surgical semantics at the current moment based on the effective grinding physical state category and using the anatomical structure sequence and state transition rules of the spinal tissue set in the finite state machine includes: Starting from the initial grinding state, if the high resistance state is detected for the first time, it is determined to switch to the outer cortical bone grinding state; When the outer cortical bone is in the grinding state and a change from the high resistance state to the low resistance state is detected, it is determined that the process has entered the grinding state of the cancellous bone. When the cancellous bone is in the grinding state and a change from the low resistance state to the high resistance state is detected again, it is determined that the process will switch to the inner cortical bone grinding state. When the inner cortical bone is in the grinding state and a sudden change from the high resistance state to the low resistance state is detected, it is determined that the process has entered the penetration state.

7. The grinding control method according to claim 4, characterized in that, During the training process of the bone feature extraction model, all samples originating from the cortical bone region are uniformly labeled as the high-resistance state.

8. A grinding control device, characterized in that, include: The signal acquisition module is used to acquire the contact force signal in the feed direction during the spinal grinding process in real time. The preprocessing module is used to preprocess the contact force signal in the feed direction to obtain standardized time-series data segments, and to process the standardized time-series data segments through a pre-trained bone feature extraction model to output the corresponding grinding physical state category. The status confirmation module is used to perform consistency verification on the output results of the grinding physical status category, and to take the grinding physical status category that continuously reaches a preset number of frames as the valid grinding physical status category. The logic control module is used to determine the logical grinding state with surgical semantics at the current moment based on the effective grinding physical state category and the anatomical structure sequence and state transition rules of the spinal tissue set in the finite state machine. The motion control module is used to generate grinding control commands based on the logical grinding state to control the spinal grinding robot to perform corresponding grinding operations.

9. A spinal grinding robot, characterized in that, The spinal grinding robot includes a processor and a memory, the memory storing a computer program, and the processor executing the computer program to implement the grinding control method according to any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, It stores a computer program that, when executed on a processor, implements the grinding control method according to any one of claims 1-7.

Images