Minimally invasive spine fusion system based on artificial intelligence

Abstract

The invention discloses a minimally invasive spine fusion system based on artificial intelligence, and the system comprises the steps: obtaining a deviation early warning signal, carrying out the smooth filtering of instrument track data, obtaining a filtered track sequence, and matching the filtered track sequence with an updated model grid, dynamic path adjustment parameters are determined and built in the surgical robot to adjust vibration frequency parameters to improve the control experience level; processing subsequent image data by adopting the iterated model version, determining temporary deviation caused by instrument shaking, correcting the temporary deviation through a path planning algorithm, generating a corrected navigation instruction sequence, and feeding back mechanism application in real time to transmit visual deviation information; and driving the surgical instrument to adjust motion through the navigation instruction sequence, acquiring adjusted trajectory data, circularly comparing the deviation from the optimized path, and if the deviation is judged to be continuously lower than a threshold value, maintaining a current model and outputting a final navigation model to ensure reliability and consistency.

Description

Technical Field

[0001] This invention belongs to the field of computer technology, specifically a minimally invasive spinal fusion system based on artificial intelligence. Background Technology

[0002] In spinal surgery, precise instrument navigation and real-time path optimization are core requirements for improving surgical safety and success rates. However, the dynamic changes in the complex intraoperative environment pose significant challenges to current technologies. Currently, the movement trajectory of surgical instruments is easily affected by factors such as surgeon's hand tremors, tissue deformation, and equipment vibration, making it difficult to accurately quantify and correct trajectory deviations in real time. Simultaneously, the dynamic matching of intraoperative imaging data and 3D navigation models exhibits lag, failing to accurately reflect real-time changes in anatomical structures and impacting the stability and reliability of path planning.

[0003] How to accurately monitor and predict instrument deviations through real-time fusion and analysis of multimodal data, and dynamically update the navigation model to adapt to changes in anatomical structures, has become an urgent technical challenge.

[0004] In the prior art, for ultra-wideband positioning used for wireless ultrasonic tracking and communication, CN201780023990.2 proposes a method for designing orthopedic implants, which selects an implant shape that meets predetermined conditions by iteratively evaluating anatomical and kinematic shape constraints.

[0005] This method focuses on static design optimization, lacks the ability to monitor and correct instrument trajectory deviations during surgery in real time, and does not involve the fusion of multimodal data and dynamic updates of navigation models, making it difficult to adapt to the dynamic changes in anatomical structures during spinal surgery.

[0006] To address the aforementioned issues, a pressing technical challenge is how to achieve real-time, high-precision monitoring and dynamic path optimization of surgical instrument trajectories in complex surgical environments.

[0007] Specifically, there are difficulties in real-time synchronization between intraoperative image data and navigation models, which leads to deviations in projection coordinates and affects the accuracy of deviation warnings.

[0008] Meanwhile, the dynamic adjustment of the instrument trajectory is limited by data processing latency, making it difficult to generate stable path correction parameters in real time. Summary of the Invention

[0009] The purpose of this invention is to provide a minimally invasive spinal fusion system based on artificial intelligence, in order to solve the problem mentioned in the background art that the real-time synchronization of intraoperative image data and navigation model is difficult, resulting in projection coordinate deviation and affecting the accuracy of deviation warning.

[0010] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows:

[0011] An AI-based minimally invasive spinal fusion system, comprising:

[0012] Surgical instrument trajectory data and intraoperative image data are collected by sensors. The current position coordinates are extracted from the trajectory data and compared with the preset path coordinates to obtain the initial deviation value. The deviation angle, distance and speed are recorded and monitored in combination with historical data to optimize the evaluation standard rules.

[0013] By analyzing deviation patterns in historical surgical data through machine learning, a predictive model is built to identify potential deviation trends of instruments and generate compensation parameters in advance. The vibration intensity and frequency are dynamically adjusted according to changes in the degree of deviation to convey intuitive deviation information and handle fluctuations in complex scenarios to ensure reliable consistency.

[0014] The three-dimensional navigation model is updated based on the position of the deformed anatomical structure, an updated model mesh is generated, and the projection coordinates of the current position of the instrument in the model are calculated. If the projection coordinates exceed the preset boundary, a deviation warning signal is triggered and the deviation speed value is monitored in real time to reflect the severity of the deviation.

[0015] After obtaining the deviation warning signal, the instrument trajectory data is smoothed and filtered to obtain the filtered trajectory sequence and match it with the updated model mesh. The dynamic path adjustment parameters are determined and the vibration frequency parameters are adjusted in the surgical robot to improve the level of operation experience.

[0016] By integrating high-resolution intraoperative imaging equipment and artificial intelligence algorithms, a three-dimensional anatomical model of the spine is generated in real time and the surgical path is dynamically updated. The stability of the optimized path curve is verified by fusing multi-modal data based on multi-dimensional evaluation dimensions. If the stability index is lower than the preset threshold, the multi-modal data is re-fused to iteratively update the three-dimensional navigation model and obtain the iterative model version.

[0017] The iterative model version is used to process subsequent image data, determine the temporary deviation caused by instrument shaking and correct it through path planning algorithm, generate a corrected navigation command sequence and apply a real-time feedback mechanism to convey intuitive deviation information;

[0018] The surgical instruments are driven to adjust their movement by the navigation command sequence, the adjusted trajectory data is obtained and the deviation from the optimized path is compared cyclically. If the deviation is consistently below the threshold, the current model is maintained and the final navigation model is output to ensure reliable consistency.

[0019] According to the above technical solution, the step of collecting surgical instrument trajectory data and intraoperative image data through sensors, extracting the current position coordinates from the trajectory data and comparing them with preset path coordinates to obtain an initial deviation value, and combining historical data to record and monitor the deviation angle, distance, and speed to optimize the evaluation criteria rules, includes:

[0020] By collecting motion trajectory data of surgical instruments and intraoperative image data through sensors, the current position coordinates are extracted to obtain real-time position information;

[0021] If the Euclidean distance between the current position coordinates and the preset path coordinates exceeds a preset threshold, the initial deviation value is calculated to obtain the deviation data.

[0022] By combining the initial deviation value with historical data records, the trajectory data is smoothed using the Kalman filter algorithm to obtain the optimized deviation trend;

[0023] Based on the optimized deviation trend, the deviation angle and deviation distance are calculated to obtain the dynamic change values ​​of the angle and distance;

[0024] By combining dynamically changing values ​​with historical data records, the severity of the bias is classified using a support vector machine algorithm to obtain the bias level.

[0025] If the deviation level exceeds the preset level threshold, the weight parameters in the evaluation criteria rules are adjusted to obtain the updated rules;

[0026] Based on the updated rules and combined with motion speed data, a linear regression algorithm is used to predict future deviation trends, and the prediction results are obtained.

[0027] According to the above technical solution, the step of analyzing deviation patterns in historical surgical data through machine learning, constructing a predictive model to identify potential deviation trends of instruments and generating compensation parameters in advance, dynamically adjusting vibration intensity and frequency according to changes in the degree of deviation to convey intuitive deviation information, and handling fluctuations in complex scenarios to ensure reliable consistency includes:

[0028] Historical surgical data is acquired, and deviation patterns are analyzed using machine learning algorithms to obtain a deviation feature set.

[0029] The random forest algorithm is used to extract key deviation patterns from the deviation feature set and determine the deviation trend model.

[0030] If the deviation trend model predicts that the device deviation exceeds the preset threshold, then the corresponding compensation parameters are generated.

[0031] Based on the compensation parameters, the vibration intensity and frequency are dynamically adjusted to transmit deviation information;

[0032] Real-time fluctuation data in complex scenarios is acquired, and the fluctuations are processed using the Kalman filter algorithm to obtain a stable deviation signal;

[0033] If the stable deviation signal matches the deviation trend model, then consistency is maintained by adjusting the vibration intensity and frequency.

[0034] Based on the consistency maintenance results, update the deviation feature set and optimize the prediction model.

[0035] According to the above technical solution, the step of updating the three-dimensional navigation model based on the position of the deformed anatomical structure, generating an updated model mesh, calculating the projection coordinates of the current position of the instrument in the model, and determining whether the projection coordinates exceed a preset boundary, triggering a deviation warning signal and monitoring the deviation speed value in real time to reflect the severity of the deviation includes:

[0036] The deformed anatomical structure position data is obtained, and an updated 3D navigation model is generated through a 3D reconstruction algorithm to obtain the model mesh;

[0037] Based on the updated model mesh, the stereoscopic projection method is used to calculate the projected coordinates of the current position of the instrument in the model, and the projected coordinate values ​​are obtained.

[0038] If the projected coordinate value exceeds the preset boundary, a deviation warning signal is triggered by comparing the difference between the coordinate value and the boundary threshold, and the warning signal status is obtained.

[0039] Based on the warning signal status, obtain the changes in projected coordinates at continuous time points, calculate the deviation velocity value, and obtain the deviation velocity value.

[0040] By analyzing the deviation from the speed value, a preset classification algorithm is used to determine the severity of the deviation and obtain the deviation level.

[0041] Based on the deviation level, adjust the display parameters of the 3D navigation model, generate real-time visual feedback, and obtain the updated display interface;

[0042] If the deviation level continues to increase, the preset boundary is dynamically updated by analyzing the trend of the deviation speed value, and the optimized boundary threshold is obtained.

[0043] According to the above technical solution, after obtaining the deviation warning signal, the instrument trajectory data is smoothed and filtered to obtain a filtered trajectory sequence, which is then matched with the updated model mesh. Dynamic path adjustment parameters are determined, and vibration frequency parameters are adjusted within the surgical robot to improve the user experience. This includes:

[0044] Obtain deviation warning signals, extract real-time trajectory data from instrument sensors, and generate an initial trajectory dataset;

[0045] For the initial trajectory dataset, the Kalman filter algorithm is used for smoothing to obtain the filtered trajectory sequence;

[0046] The updated model grid is extracted from the pre-established model grid database, and the filtered trajectory sequence is spatially mapped and matched with the model grid to determine the trajectory deviation value.

[0047] If the trajectory deviation value exceeds the preset threshold, the dynamic path adjustment parameters are calculated based on the deviation value, and a path adjustment vector is generated.

[0048] Based on the path adjustment vector, update the motion control parameters in the surgical robot control module and adjust the robot joint motion trajectory;

[0049] By calculating the changing trend of vibration frequency parameters through the adjusted motion trajectory, an optimized set of vibration frequency parameters is generated.

[0050] By using the optimized set of vibration frequency parameters, the vibration frequency configuration of the surgical robot's execution unit is updated, resulting in an improved level of control experience.

[0051] According to the above technical solution, the process involves integrating high-resolution intraoperative imaging equipment and artificial intelligence algorithms to generate a three-dimensional anatomical model of the spine in real time and dynamically update the surgical path. The stability of the optimized path curve is verified by fusing multi-modal data across multiple dimensions. If the stability index is lower than a preset threshold, the multi-modal data is re-fused to iteratively update the three-dimensional navigation model, resulting in an iterative model version. This includes:

[0052] Intraoperative multimodal data were acquired in real time using high-resolution imaging equipment to generate the first dataset;

[0053] Based on the first dataset, a three-dimensional anatomical model of the spine was constructed using a convolutional neural network algorithm to obtain the first three-dimensional model.

[0054] The surgical path curve is extracted from the first 3D model, and the path is dynamically adjusted using a geometric optimization algorithm to generate the first path curve.

[0055] For the first path curve, calculate the stability index. If the stability index is lower than the preset threshold, acquire new multimodal data and generate the second dataset.

[0056] Using the second dataset, the first 3D model is updated by incorporating the random forest algorithm to generate the second 3D model;

[0057] The surgical path curve is re-extracted from the second 3D model, and the path is adjusted using a geometric optimization algorithm to generate the second path curve;

[0058] For the second path curve, the stability index is recalculated to obtain an optimized version of the 3D navigation model.

[0059] According to the above technical solution, the step of processing subsequent image data using the iterative model version, determining the temporary deviation caused by instrument shaking and correcting it through a path planning algorithm, generating a corrected navigation command sequence, and applying a real-time feedback mechanism to convey intuitive deviation information includes:

[0060] Real-time multimodal data was acquired from intraoperative imaging data, and the data was processed using a convolutional neural network algorithm to obtain a pre-processed image dataset.

[0061] Using the pre-processed image dataset, temporary deviations caused by instrument vibration are detected, deviation vectors are calculated, and deviation distribution data are obtained.

[0062] The deviation distribution data is analyzed using a path planning algorithm to generate a corrected path curve, thus obtaining optimized path data.

[0063] If the stability index of the optimized path data is lower than the preset threshold, the new multimodal data is fused through the random forest algorithm to update the 3D navigation model and obtain the updated model version.

[0064] Extract the corrected path curves from the updated model version, generate a navigation instruction sequence, and obtain a real-time navigation instruction set;

[0065] By using a real-time navigation command set and a feedback mechanism to transmit deviation correction information, a dynamically adjusted navigation output is obtained.

[0066] For the dynamically adjusted navigation output, the stability of the path curve is verified, and the final navigation instruction sequence is obtained.

[0067] According to the above technical solution, it also includes: obtaining image data processing results based on dynamic image analysis, determining temporary deviations caused by instrument vibration through jitter signal extraction, generating a corrected navigation command sequence using a path optimization algorithm, transmitting intuitive deviation information through a real-time feedback mechanism, and obtaining dynamic adjustment results of the navigation command sequence, specifically including:

[0068] The image data processing results are obtained through dynamic image analysis, and the instrument region is separated by image segmentation technology to obtain the instrument location information;

[0069] The jitter signal is extracted from the instrument position information, and the signal frequency is analyzed using Fourier transform to determine the temporary deviation value;

[0070] If the temporary deviation value exceeds the preset threshold, the deviation trajectory is fitted by the least squares method to obtain the deviation change trend;

[0071] Based on the trend of deviation change, the A algorithm is used to generate a corrected navigation instruction sequence to obtain the optimized instruction set.

[0072] The optimized instruction set is transmitted to the control system through a real-time feedback mechanism to obtain the navigation data after deviation correction; dynamic adjustment parameters are extracted from the navigation data after deviation correction, and the navigation instruction sequence is smoothed by Kalman filtering algorithm to obtain the final navigation instruction sequence;

[0073] Based on the final navigation command sequence, update the control system parameters to obtain real-time adjusted navigation output.

[0074] According to the above technical solution, the step of driving the surgical instrument to adjust its movement through the navigation command sequence, acquiring the adjusted trajectory data and cyclically comparing the deviation with the optimized path, and determining whether the deviation is consistently below a threshold, then maintaining the current model and outputting the final navigation model to ensure reliable consistency, includes:

[0075] Initial motion control signals are generated by a sequence of navigation instructions to drive the surgical instruments to perform adjustment actions, thereby obtaining the adjusted trajectory data.

[0076] The position coordinate sequence is extracted from the adjusted trajectory data, and the deviation between the trajectory data and the optimized path is calculated using Euclidean distance.

[0077] If the deviation value is lower than the preset threshold, the current deviation value is recorded and the next loop comparison is performed to obtain the deviation change trend;

[0078] Based on the trend of deviation change, the gradient descent algorithm is used to adjust the parameters of the navigation command sequence and generate new motion control signals;

[0079] The surgical instruments are driven to perform adjustment actions again by new motion control signals to obtain updated trajectory data;

[0080] If the deviation between the updated trajectory data and the optimized path remains below a preset threshold, then the current navigation instruction sequence is determined as the final navigation model.

[0081] The final navigation model generates a standardized motion control sequence, which outputs navigation commands to drive surgical instruments.

[0082] According to the above technical solution, by acquiring the trajectory data of the surgical instruments after adjustment, iteratively comparing the deviation with the optimized path, and if the deviation remains below a threshold, maintaining the current model and outputting the final navigation model, a reliable and consistent navigation result is obtained, specifically including:

[0083] The trajectory data of the surgical instruments after movement is acquired, and the position coordinates are collected in real time by sensors to obtain the initial trajectory dataset;

[0084] Using the initial trajectory dataset, the least squares method is used to fit the expected path, the deviation between the actual path and the expected path is calculated, and the path deviation value is determined.

[0085] If the path deviation value is lower than the preset threshold, the current path model is retained, the optimized path parameters are output, and the current model parameters are obtained.

[0086] Based on the current model parameters, the Kalman filter algorithm is used to smooth the trajectory data, reduce noise interference, and obtain a smoothed trajectory dataset.

[0087] By using a smooth trajectory dataset, the deviation between the current path and the target navigation path is iteratively compared, the path model is updated, and the updated navigation model is obtained.

[0088] If the deviation of the updated navigation model is lower than the preset threshold, the model parameters are fixed, the final navigation model is generated, and a stable navigation result is obtained.

[0089] By stabilizing navigation results, the movement trajectory of surgical instruments is adjusted in real time, and consistent navigation commands are output to obtain accurate navigation output.

[0090] Compared with the prior art, the present invention has the following beneficial effects:

[0091] This invention discloses a surgical navigation optimization method based on multimodal data fusion and machine learning. Addressing the challenges of surgical instrument trajectory deviation and dynamic changes in complex anatomical structures, the method collects real-time surgical instrument movement trajectories and intraoperative image data, extracts current position coordinates, compares them with a preset path, and constructs a predictive model using historical data. This model identifies deviation trends, generates compensation parameters, and dynamically adjusts vibration intensity to provide intuitive feedback. Simultaneously, it updates the 3D navigation model based on high-resolution images, calculates instrument projection coordinates, triggers deviation warnings, optimizes the trajectory through smoothing filtering, and generates navigation commands by matching the updated model. Iterative model fusion of multimodal data corrects instrument vibration deviations, ensuring path stability. Through iterative comparison and optimization, this invention ultimately outputs a reliable navigation model, achieving precise surgical navigation and dynamic path adjustment, significantly improving the surgical experience and consistency. Attached Figure Description

[0092] Figure 1 This is a flowchart of the artificial intelligence-based minimally invasive spinal fusion system of the present invention. Detailed Implementation

[0093] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0094] Example 1

[0095] like Figure 1 As shown, the minimally invasive spinal fusion system based on artificial intelligence may specifically include the following steps:

[0096] Surgical instrument trajectory data and intraoperative image data are collected by sensors. The current position coordinates are extracted from the trajectory data and compared with the preset path coordinates to obtain the initial deviation value. The deviation angle, distance and speed are recorded and monitored in combination with historical data to optimize the evaluation criteria rules.

[0097] This embodiment provides a specific implementation method, which involves: acquiring motion trajectory data of surgical instruments and intraoperative image data through sensors, extracting the current position coordinates to obtain real-time position information. If the Euclidean distance between the current position coordinates and the preset path coordinates exceeds a preset threshold, an initial deviation value is calculated to obtain deviation data. The initial deviation value is combined with historical data records, and the trajectory data is smoothed using a Kalman filter algorithm to obtain an optimized deviation trend. Based on the optimized deviation trend, the deviation angle and deviation distance are calculated to obtain dynamic change values ​​for the angle and distance. Using the dynamic change values ​​and historical data records, a support vector machine algorithm is used to classify the severity of the deviation to obtain a deviation level. If the deviation level exceeds a preset level threshold, the weight parameters in the evaluation criteria rules are adjusted to obtain updated rules. Based on the updated rules and motion speed data, a linear regression algorithm is used to predict the future deviation trend to obtain the prediction result.

[0098] Specifically, the surgical instrument motion trajectory data and intraoperative image data are collected by sensors. First, a three-axis accelerometer and gyroscope mounted on the surgical instrument are used to acquire real-time motion trajectory data at a sampling rate of 100Hz. Simultaneously, an intraoperative optical imaging system acquires image data at 30 frames per second, generating a data stream containing spatial coordinates (x, y, z) and a timestamp, for example, coordinates at a certain moment are (10.5, 20.3, 15.2) cm. Next, the current position coordinates are extracted from the trajectory data, and the sensor data is smoothed using a Kalman filter algorithm to eliminate noise and obtain accurate coordinates, such as (10.52, 20.28, 15.18) cm. Then, it is compared with the preset path coordinates (10.0, 20.0, 15.0) cm to calculate the initial deviation value. The deviation formula is the Euclidean distance: √[(10.52-10.0)²+(20.28-20.0)²+(15.18-15.0)²]=0.38 cm. Combining historical data records, the coordinate sequence of the most recent 10 seconds is stored. The deviation angle is calculated using the vector angle formula cosθ=(a·b) / (|a||b|), where a is the current vector and b is the preset path vector, yielding an angle of 5.2 degrees. The deviation distance is calculated as the average of consecutive deviation values, resulting in 0.35 cm. The speed is calculated by dividing the difference between adjacent timestamp coordinates by the time interval, yielding an average speed of 0.1 cm / s. When optimizing the evaluation criteria, a weighted average algorithm is used, with weights allocated as follows: angle 0.4, distance 0.4, and speed 0.2. The comprehensive deviation score = 0.4×5.2 + 0.4×0.35 + 0.2×0.1 = 2.24. A score below the threshold of 3.0 is considered to meet the standard; otherwise, a warning is triggered and recorded in the database for postoperative analysis. These steps are automatically completed by a real-time data processing system to ensure surgical accuracy. Logically, a closed loop is formed from data acquisition to deviation calculation and then to comprehensive evaluation, optimizing surgical path planning.

[0099] In another embodiment, deviation patterns in historical surgical data are analyzed using machine learning to construct a predictive model that identifies potential instrument deviation trends and generates compensation parameters in advance. The vibration intensity and frequency are dynamically adjusted based on changes in the degree of deviation to convey intuitive deviation information, handling fluctuations in complex scenarios to ensure reliable consistency. Specifically:

[0100] This embodiment provides a specific implementation method, which involves: acquiring historical surgical data, analyzing deviation patterns using machine learning algorithms to obtain a deviation feature set; employing a random forest algorithm to extract key deviation patterns from the deviation feature set to determine a deviation trend model; if the deviation trend model predicts that the instrument deviation exceeds a preset threshold, generating corresponding compensation parameters; dynamically adjusting vibration intensity and frequency based on the compensation parameters to transmit deviation information; acquiring real-time fluctuation data in complex scenarios, processing the fluctuations using a Kalman filter algorithm to obtain a stable deviation signal; if the stable deviation signal matches the deviation trend model, maintaining consistency through adjusted vibration intensity and frequency; updating the deviation feature set and optimizing the prediction model based on the consistency maintenance result.

[0101] Specifically, machine learning was used to analyze deviation patterns in historical surgical data. First, a dataset containing 10,000 surgical records was collected. Each record included instrument coordinates (x, y, z), timestamps, deviation distance (in millimeters, ranging from 0-5), and surgical type. The data was cleaned using Python's Pandas library, removing missing and outlier values ​​(e.g., deviation distance > 5 mm), and the coordinate data was standardized. In feature engineering, time-series features (e.g., deviation velocity, acceleration) and contextual features (e.g., surgical duration, instrument type) were extracted. A random forest algorithm (n_estimators=100, max_depth=10) was used to analyze deviation patterns. The training and test sets were split in an 8:2 ratio, achieving a model accuracy of 85%. For predictive model construction, an LSTM neural network (2 layers, 64 units per layer, learning rate 0.001) was used to predict deviation trends within the next 10 seconds. The input was the coordinate and velocity sequence of the past 30 seconds, and the output was the predicted deviation distance (mean squared error 0.02 mm). To generate compensation parameters, a control algorithm is designed to calculate the compensation vector based on the predicted deviation distance. For example, when the deviation is 2mm, a reverse displacement (-2,0,0) is generated. Vibration intensity and frequency are dynamically adjusted. Vibration intensity is defined as proportional to the deviation distance (intensity = 0.5 * deviation distance, unit mN), and frequency varies with the deviation speed (0.1mm / s corresponds to 50Hz, 0.5mm / s corresponds to 100Hz). A PID controller (Kp=0.5, Ki=0.1, Kd=0.05) is used for real-time adjustment to ensure intuitive transmission of deviation information. In handling complex scene fluctuations, a Kalman filter is introduced to smooth sensor noise (process noise covariance Q=0.01, measurement noise covariance R=0.1), combined with an anomaly detection algorithm (Isolation Forest, contamination=0.05) to eliminate abnormal fluctuations and maintain consistency. All steps are integrated through automated scripts, with seamless data flow from acquisition to output, ensuring real-time performance and reliability.

[0102] In another embodiment, the three-dimensional navigation model is updated based on the position of the deformed anatomical structure, an updated model mesh is generated, and the projected coordinates of the instrument's current position in the model are calculated. If the projected coordinates exceed a preset boundary, a deviation warning signal is triggered, and the deviation speed value is monitored in real time to reflect the severity of the deviation. Specifically:

[0103] This embodiment provides a specific implementation method, which involves: acquiring the position data of the deformed anatomical structure; generating an updated 3D navigation model using a 3D reconstruction algorithm to obtain a model mesh; calculating the projection coordinates of the instrument's current position in the model using a stereoscopic projection method based on the updated model mesh; obtaining projection coordinate values ​​if the projection coordinate values ​​exceed a preset boundary, triggering a deviation warning signal by comparing the difference between the coordinate values ​​and the boundary threshold; obtaining a warning signal status based on the warning signal status; acquiring the changes in projection coordinates at continuous time points and calculating the deviation velocity values ​​based on the deviation velocity values; determining the severity of the deviation using a preset classification algorithm based on the deviation velocity values; adjusting the display parameters of the 3D navigation model based on the deviation level to generate real-time visual feedback and obtain an updated display interface; and dynamically updating the preset boundary by analyzing the trend of the deviation velocity values ​​to obtain an optimized boundary threshold based on the deviation velocity values.

[0104] Specifically, based on the input deformed anatomical structure position data, the system first updates the 3D navigation model using a point cloud registration algorithm. Assuming the original model contains 10,000 vertices and the deformed point cloud data contains 12,000 points, the Iterative Closest Point (ICP) algorithm is used for registration. The Euclidean distance between the point clouds is calculated, and convergence occurs after 10 iterations with an average registration error of 0.05 mm. An updated model mesh containing 15,000 triangular faces is generated, and the vertex coordinates are adjusted using linear interpolation, stored in OBJ format. Next, the system calculates the projected coordinates of the instrument's current position in the model. Assuming the instrument's real-time position is (x,y,z) = (10.2,15.3,20.1) mm, a ray casting algorithm is used to emit a ray from the instrument's position and intersect it with the model surface, obtaining the projected point coordinates as (10.0,15.0,20.0) mm. The calculation time is 0.02 seconds. Subsequently, it is determined whether the projected coordinates exceed the preset boundary, which is defined as a cuboid region: x∈[0,20], y∈[0,30], z∈[0,40] mm. If the current projected point is within the boundary, no warning is triggered. If the projected point exceeds the boundary, such as (25.0,15.0,20.0) mm, the system triggers a deviation warning signal and sends a JSON-formatted alarm to the monitoring system via API. The alarm includes the deviation distance, such as 5.0 mm. The deviation speed is monitored in real time. Assuming two consecutive frames have projected points of (25.0,15.0,20.0) and (26.0,15.0,20.0) mm respectively, with a time interval of 0.1 seconds, the deviation speed is calculated to be 10.0 mm / s. When the speed exceeds the threshold of 5.0 mm / s, a severity level of "high" is generated and displayed on the front-end interface using color coding (red), prompting the system to adjust the instrument trajectory. This process is implemented through automated algorithms, and the data stream is processed in the cloud to ensure real-time performance and accuracy.

[0105] In another embodiment, after acquiring the deviation warning signal, the instrument trajectory data is smoothed and filtered to obtain a filtered trajectory sequence, which is then matched with the updated model mesh. Dynamic path adjustment parameters are determined, and vibration frequency parameters are integrated into the surgical robot to improve the user experience. Specifically:

[0106] This embodiment provides a specific implementation method, which involves: acquiring a deviation warning signal, extracting real-time trajectory data from instrument sensors, and generating an initial trajectory dataset. For the initial trajectory dataset, a Kalman filter algorithm is used for smoothing to obtain a filtered trajectory sequence. An updated model mesh is extracted from a pre-established model mesh database, and the filtered trajectory sequence is spatially mapped and matched with the model mesh to determine the trajectory deviation value. If the trajectory deviation value exceeds a preset threshold, dynamic path adjustment parameters are calculated based on the deviation value to generate a path adjustment vector. Based on the path adjustment vector, the motion control parameters in the surgical robot control module are updated, and the robot joint motion trajectory is adjusted. Through the adjusted motion trajectory, the changing trend of vibration frequency parameters is calculated to generate an optimized vibration frequency parameter set. Using the optimized vibration frequency parameter set, the vibration frequency configuration of the surgical robot execution unit is updated to generate an improved level of control experience.

[0107] In another embodiment, by integrating high-resolution intraoperative imaging equipment with artificial intelligence algorithms, a three-dimensional anatomical model of the spine is generated in real time and the surgical path is dynamically updated. The stability of the optimized path curve is verified based on multi-dimensional evaluation and multimodal data fusion. If the stability index is lower than a preset threshold, multimodal data is re-fused to iteratively update the three-dimensional navigation model, resulting in an iterative model version. Specifically:

[0108] This embodiment provides a specific implementation method, which involves: acquiring intraoperative multimodal data in real time using high-resolution imaging equipment to generate a first dataset. Based on the first dataset, a three-dimensional anatomical model of the spine is constructed using a convolutional neural network algorithm to obtain a first three-dimensional model. Surgical path curves are extracted from the first three-dimensional model, and the path is dynamically adjusted using a geometric optimization algorithm to generate a first path curve. For the first path curve, a stability index is calculated. If the stability index is lower than a preset threshold, new multimodal data is acquired to generate a second dataset. Using the second dataset, the first three-dimensional model is updated by fusing a random forest algorithm to generate a second three-dimensional model. Surgical path curves are re-extracted from the second three-dimensional model, and the path is adjusted using a geometric optimization algorithm to generate a second path curve. For the second path curve, the stability index is recalculated to obtain an optimized version of the three-dimensional navigation model.

[0109] Specifically, high-resolution intraoperative imaging equipment such as intraoperative CT and MRI are used to acquire raw image data of the spinal region with a resolution of 0.5 mm, generating 256x256x256 pixel three-dimensional stereomicroscopic images. Artificial intelligence algorithms employ the U-Net architecture within convolutional neural networks (CNNs) to perform semantic segmentation of the images, identifying spinal bones, soft tissues, and neural structures with a segmentation accuracy exceeding 95%, generating an initial three-dimensional anatomical model. Based on real-time intraoperative data, the surgical path is dynamically updated, and the A* algorithm is used to search for the optimal path within the three-dimensional model, controlling the path length error within 0.2 mm. Combined with intraoperative force feedback sensor data (pressure range 0-10 N), the path curvature radius is optimized to a minimum of 2 mm. Multimodal data fusion, including image data, force feedback, and pose information from intraoperative navigation instruments, is performed using a Kalman filter algorithm with weights allocated as follows: image 0.6, force feedback 0.3, pose 0.1, generating a fused dataset. When evaluating path stability, the second derivative of the path curve is calculated, and the stability index is the rate of change of curvature, with a threshold set to 0.01. If the data falls below a threshold, an iterative update is triggered. Multimodal data is re-input into the deep learning model (ResNet-50), and the 3D navigation model is updated through gradient descent optimization (learning rate 0.001). After iteration, the path deviation of the model version is reduced to 0.1 mm. Each iteration takes less than 500 milliseconds to ensure real-time performance. The logic chain is as follows: high-resolution images generate the initial model; dynamic path optimization relies on real-time data; multimodal data is fused to improve accuracy; stability evaluation drives model iteration; and ultimately, navigation accuracy and surgical safety are ensured.

[0110] In another embodiment, the iterative model version is used to process subsequent image data, determine temporary deviations caused by instrument jitter, correct them using a path planning algorithm, generate a corrected navigation command sequence, and implement a real-time feedback mechanism to convey intuitive deviation information. Specifically:

[0111] This embodiment provides a specific implementation method, which involves: acquiring real-time multimodal data from intraoperative imaging data; processing the data using a convolutional neural network algorithm to obtain a pre-processed image dataset; detecting temporary deviations caused by instrument jitter using the pre-processed image dataset; calculating the deviation vector to obtain deviation distribution data; analyzing the deviation distribution data using a path planning algorithm to generate a corrected path curve, resulting in optimized path data; if the stability index of the optimized path data is lower than a preset threshold, fusing new multimodal data using a random forest algorithm to update the 3D navigation model, resulting in an updated model version; extracting the corrected path curve from the updated model version to generate a navigation command sequence, resulting in a real-time navigation command set; using the real-time navigation command set, employing a feedback mechanism to transmit deviation correction information, resulting in dynamically adjusted navigation output; and verifying the stability of the path curve for the dynamically adjusted navigation output to obtain the final navigation command sequence.

[0112] Specifically, when processing subsequent image data based on the iterative 3D navigation model, the instrument position is first captured using an intraoperative high-frame-rate optical tracking system at a frame rate of 120 frames per second and a resolution of 0.3 mm, generating a real-time position sequence. For temporary deviations caused by instrument jitter, a Fast Fourier Transform (FFT) is used to analyze the frequency spectrum of the position sequence, identifying the jitter frequency range (2-10 Hz), calculating the deviation amplitude, and controlling the average deviation within 0.15 mm. Next, based on the deviation data, an improved RRT* (Rapidly-exploring Random Tree Star) algorithm is used to replan the path, setting the sampling point density to 100 points per cubic millimeter, optimizing the path length error to 0.1 mm, and maintaining the radius of curvature above 2.5 mm to ensure smooth instrument movement. A corrected navigation command sequence is generated, and the path points are converted into an instrument joint angle sequence using an inverse kinematics algorithm (with an maximum of 50 iterations), achieving an angle accuracy of 0.01 radians, with command generation time controlled within 300 milliseconds. To achieve real-time feedback, a deviation information encoding mechanism was designed, mapping deviation values ​​to an RGB color space (0-0.2 mm deviation corresponds to a green-to-red gradient), and transmitting the data to the intraoperative display system via a visual signal interface at a refresh rate of 60Hz. The logic chain is as follows: optical tracking captures real-time position, FFT analysis identifies jitter deviations, RRT* algorithm optimizes the path, inverse motion student generates command sequences, and RGB encoding transmits deviation information, ensuring accurate navigation commands and intuitive feedback.

[0113] Based on the image data processing results obtained from dynamic image analysis, the temporary deviation caused by instrument vibration is determined by extracting the jitter signal. The path optimization algorithm is used to generate a corrected navigation command sequence. The intuitive deviation information is transmitted through a real-time feedback mechanism to obtain the dynamic adjustment result of the navigation command sequence.

[0114] Image data processing results are obtained through dynamic image analysis. Image segmentation technology is used to separate the instrument region and obtain instrument position information. Jitter signals are extracted from the instrument position information, and Fourier transform is used to analyze the signal frequency to determine temporary deviation values. If the temporary deviation value exceeds a preset threshold, the deviation trajectory is fitted using the least squares method to obtain the deviation change trend. Based on the deviation change trend, the A algorithm is used to generate a corrected navigation command sequence, resulting in an optimized command set. The optimized command set is transmitted to the control system through a real-time feedback mechanism to obtain navigation data after deviation correction. Dynamic adjustment parameters are extracted from the deviation-corrected navigation data, and the Kalman filter algorithm is used to smooth the navigation command sequence, obtaining the final navigation command sequence. Based on the final navigation command sequence, the control system parameters are updated to obtain real-time adjusted navigation output.

[0115] Specifically, image data processing results are obtained through dynamic image analysis. A high-resolution camera can capture images of the surgical area at 30 frames per second, generating a video stream with a resolution of 1920×1080. A convolutional neural network (CNN) is used to segment the image in real time, extracting instrument edge features. A pixel threshold of 0.8 is set to distinguish instruments from the background, obtaining the instrument center coordinates (x, y), such as (500, 600). Temporary deviations caused by instrument jitter are determined by jitter signal extraction. Fast Fourier Transform (FFT) is used to analyze the change of instrument center coordinates over time. A sampling frequency of 100Hz is set, and a jitter signal with a frequency of 5Hz is identified, calculating the deviation amplitude to be ±2 mm. A path optimization algorithm is used to generate a corrected navigation command sequence. The A* algorithm (i.e., A-Star algorithm) is used to calculate the optimal path based on the current instrument position and the target position (e.g., (510, 610)). The grid resolution is set to 1 mm, and the weighting factor is the weighted sum of distance and deviation (weight ratio 0.7:0.3), generating a smooth command sequence, such as [(501, 601), (502, 602), ..., (510, 610)]. A real-time feedback mechanism is used to transmit intuitive deviation information. Using the WebSocket protocol, deviation data (±2 mm) and the corrected command sequence are transmitted to the control system at 10-millisecond intervals. This is displayed on the interface with green indicating a deviation <1 mm and yellow indicating 1-3 mm. The dynamic adjustment results of the navigation command sequence are obtained. Based on the real-time deviation feedback, a Kalman filter is used to predict the instrument position in the next frame. The predicted coordinates are fused with the commands generated by the A* algorithm, and the sequence is adjusted to adapt to deviation changes, such as correcting (502, 602) to (502.5, 602.3), ensuring the deviation is controlled within ±0.5 mm. The above steps form a closed-loop logic through image processing, signal analysis, path planning, and real-time communication to ensure navigation accuracy and stability.

[0116] In another embodiment, the surgical instruments are driven to adjust their movement via the navigation command sequence. The adjusted trajectory data is then acquired and its deviation from the optimized path is repeatedly compared. If the deviation remains below a threshold, the current model is maintained, and the final navigation model is output to ensure reliable consistency. Specifically:

[0117] This embodiment provides a specific implementation method, which involves: generating an initial motion control signal through a navigation command sequence to drive the surgical instrument to perform adjustment actions, thereby obtaining adjusted trajectory data. A position coordinate sequence is extracted from the adjusted trajectory data, and the deviation between the trajectory data and the optimized path is calculated using Euclidean distance. If the deviation value is lower than a preset threshold, the current deviation value is recorded, and the next iteration of the comparison is performed to obtain the deviation change trend. Based on the deviation change trend, the parameters of the navigation command sequence are adjusted using a gradient descent algorithm to generate a new motion control signal. The new motion control signal drives the surgical instrument to perform adjustment actions again, obtaining updated trajectory data. If the deviation between the updated trajectory data and the optimized path remains lower than the preset threshold, the current navigation command sequence is determined to be the final navigation model. A standardized motion control sequence is generated using the final navigation model, and navigation commands for driving the surgical instrument are output.

[0118] Specifically, by driving the movement of surgical instruments through a sequence of navigation commands, a control algorithm based on inverse kinematics can be used to generate precise joint angle commands.

[0119] For example, assuming the surgical instrument is a six-DOF robotic arm, the input navigation command is the target point coordinates (x=10.0cm, y=5.0cm, z=2.0cm). The inverse kinematics algorithm calculates the angles of each joint, outputting a command sequence (θ1=30°, θ2=45°, θ3=60°), which is then sent to the servo motor to drive the robotic arm. The adjusted trajectory data is obtained by using a laser tracking sensor installed at the end effector of the robotic arm to record the end effector position in real time at a sampling frequency of 100Hz, generating a trajectory point set {(x1,y1,z1), (x2,y2,z2), ...}, for example, {(10.1,5.2,2.1), (10.05,5.1,2.05)}. The deviation between the trajectory and the optimized path is compared iteratively, and the Euclidean distance is used to calculate the deviation between the actual trajectory points and the target path points.

[0120] For example, the deviation at a certain point is d = √((10.1-10.0)^2 + (5.2-5.0)^2 + (2.1-2.0)^2) = 0.245cm. If the deviation threshold is set to 0.3cm, the average deviation of each cycle is compared. If the average deviation of 10 consecutive cycles is lower than 0.3cm, for example, {0.245, 0.230, 0.210, ...}, the model is considered stable. While maintaining the current model, the inverse kinematics parameters and trajectory data are saved to the database in JSON format, including joint angles and trajectory points. The final output navigation model optimizes path planning through deep reinforcement learning, using the Q-learning algorithm with a learning rate of 0.01 and a discount factor of 0.9. After training for 1000 iterations, the optimal path is generated and saved as a navigation model file. This ensures the reliable consistency of surgical instruments in complex environments (such as obstacle avoidance). The output model accuracy is evaluated as a mean deviation of 0.15cm, which meets the surgical accuracy requirements.

[0121] By acquiring the trajectory data of the surgical instruments after adjustment, the deviation from the optimized path is compared and iterated. If the deviation is consistently below the threshold, the current model is maintained, and the final navigation model is output, resulting in a reliable and consistent navigation result.

[0122] The system acquires trajectory data of surgical instruments after movement, using sensors to collect position coordinates in real time, resulting in an initial trajectory dataset. Using this dataset, a least-squares method is employed to fit the expected path, calculating the deviation between the actual and expected paths to determine the path deviation value. If the deviation value is below a preset threshold, the current path model is retained, and optimized path parameters are output, yielding the current model parameters. Based on these parameters, a Kalman filter algorithm is used to smooth the trajectory data, reducing noise interference, resulting in a smoothed trajectory dataset. Using this smoothed dataset, the deviation between the current path and the target navigation path is iteratively compared, updating the path model to obtain an updated navigation model. If the updated navigation model deviation is below a preset threshold, the model parameters are fixed, generating the final navigation model and achieving a stable navigation result. Based on the stable navigation result, the surgical instrument trajectory is adjusted in real time, outputting consistent navigation commands for precise navigation output.

[0123] Specifically, the motion trajectory data of surgical instruments in three-dimensional space is acquired through sensors. Assuming the use of a high-precision inertial measurement unit (IMU) and optical tracking system, 1000 sets of data are collected per second. Each set includes x, y, and z coordinates and attitude angles, with a coordinate accuracy of 0.1 mm and an attitude angle accuracy of 0.01 degrees. The initial trajectory data is [(x1,y1,z1,θ1,φ1,ψ1), (x2,y2,z2,θ2,φ2,ψ2), ...], for example, [(10.0,5.0,3.0,0.1,0.2,0.3), (10.1,5.1,3.1,0.11,0.21,0.31), ...]. An iterative loop is used to compare the deviation between the actual trajectory and the preset path. The preset path is the ideal trajectory in the surgical plan, such as a straight path {(10.0,5.0,3.0), (10.5,5.5,3.5), ...}. Deviation is calculated using Euclidean distance, and the formula is:

[0124]

[0125] Where (x) i ,y i ,z i (x) represents the actual point. p ,y p ,z p The preset point is defined as follows: If the deviation of a point is d = 0.15 mm, which is below the threshold of 0.2 mm, it is recorded as a qualified point. The path deviation is optimized using a gradient descent algorithm, adjusting the instrument motion parameters (e.g., velocity v = 1 mm / s, angular velocity ω = 0.05 degrees / s) to minimize the deviation. After 100 iterations, if the average deviation decreases from 0.3 mm to 0.12 mm and remains below 0.2 mm, the current model is considered stable. The final navigation model fuses IMU and optical data using Kalman filtering to output a smooth trajectory, such as [(10.02, 5.01, 3.01, 0.101, 0.201, 0.301), ...], and verifies that its average deviation from the preset path is 0.11 mm, meeting the surgical accuracy requirements. Reliability and consistency are verified through repeated experiments; in 100 tests, the deviation was below 0.2 mm in 95 of them, demonstrating the model's robustness.

[0126] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, 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 process, method, article, or apparatus.

[0127] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A minimally invasive spinal fusion system based on artificial intelligence, characterized in that: The system includes: Surgical instrument trajectory data and intraoperative image data are collected by sensors. The current position coordinates are extracted from the trajectory data and compared with the preset path coordinates to obtain the initial deviation value. The deviation angle, distance and speed are recorded and monitored in combination with historical data to optimize the evaluation standard rules. By analyzing deviation patterns in historical surgical data through machine learning, a predictive model is built to identify potential deviation trends of instruments and generate compensation parameters in advance. The vibration intensity and frequency are dynamically adjusted according to changes in the degree of deviation to convey intuitive deviation information and handle fluctuations in complex scenarios to ensure reliable consistency. The three-dimensional navigation model is updated based on the position of the deformed anatomical structure, an updated model mesh is generated, and the projection coordinates of the current position of the instrument in the model are calculated. If the projection coordinates exceed the preset boundary, a deviation warning signal is triggered and the deviation speed value is monitored in real time to reflect the severity of the deviation. After obtaining the deviation warning signal, the instrument trajectory data is smoothed and filtered to obtain the filtered trajectory sequence and match it with the updated model mesh. The dynamic path adjustment parameters are determined and the vibration frequency parameters are adjusted in the surgical robot to improve the level of operation experience. By integrating high-resolution intraoperative imaging equipment and artificial intelligence algorithms, a three-dimensional anatomical model of the spine is generated in real time and the surgical path is dynamically updated. The stability of the optimized path curve is verified by fusing multi-modal data based on multi-dimensional evaluation dimensions. If the stability index is lower than the preset threshold, the multi-modal data is re-fused to iteratively update the three-dimensional navigation model and obtain the iterative model version. The iterative model version is used to process subsequent image data, determine the temporary deviation caused by instrument shaking and correct it through path planning algorithm, generate a corrected navigation command sequence and apply a real-time feedback mechanism to convey intuitive deviation information; The surgical instruments are driven to adjust their movement by the navigation command sequence, the adjusted trajectory data is obtained and the deviation from the optimized path is compared cyclically. If the deviation is consistently below the threshold, the current model is maintained and the final navigation model is output to ensure reliable consistency.

2. The minimally invasive spinal fusion system based on artificial intelligence according to claim 1, characterized in that: The process involves collecting surgical instrument trajectory data and intraoperative image data via sensors, extracting the current position coordinates from the trajectory data and comparing them with preset path coordinates to obtain an initial deviation value, and combining this with historical data to record and monitor deviation angles, distances, and speeds to optimize evaluation criteria rules, including: By collecting motion trajectory data of surgical instruments and intraoperative image data through sensors, the current position coordinates are extracted to obtain real-time position information; If the Euclidean distance between the current position coordinates and the preset path coordinates exceeds a preset threshold, the initial deviation value is calculated to obtain the deviation data. By combining the initial deviation value with historical data records, the trajectory data is smoothed using the Kalman filter algorithm to obtain the optimized deviation trend; Based on the optimized deviation trend, the deviation angle and deviation distance are calculated to obtain the dynamic change values ​​of the angle and distance; By combining dynamically changing values ​​with historical data records, the severity of the bias is classified using a support vector machine algorithm to obtain the bias level. If the deviation level exceeds the preset level threshold, the weight parameters in the evaluation criteria rules are adjusted to obtain the updated rules; Based on the updated rules and combined with motion speed data, a linear regression algorithm is used to predict future deviation trends, and the prediction results are obtained.

3. The minimally invasive spinal fusion system based on artificial intelligence according to claim 1, characterized in that: The method involves analyzing deviation patterns in historical surgical data using machine learning to construct a predictive model that identifies potential instrument deviation trends and generates compensation parameters in advance. It dynamically adjusts vibration intensity and frequency based on changes in the degree of deviation to convey intuitive deviation information and handles fluctuations in complex scenarios to ensure reliable consistency. This includes: Historical surgical data is acquired, and deviation patterns are analyzed using machine learning algorithms to obtain a deviation feature set. The random forest algorithm is used to extract key deviation patterns from the deviation feature set and determine the deviation trend model. If the deviation trend model predicts that the device deviation exceeds the preset threshold, then the corresponding compensation parameters are generated. Based on the compensation parameters, the vibration intensity and frequency are dynamically adjusted to transmit deviation information; Real-time fluctuation data in complex scenarios is acquired, and the fluctuations are processed using the Kalman filter algorithm to obtain a stable deviation signal; If the stable deviation signal matches the deviation trend model, then consistency is maintained by adjusting the vibration intensity and frequency. Based on the consistency maintenance results, update the deviation feature set and optimize the prediction model.

4. The minimally invasive spinal fusion system based on artificial intelligence according to claim 1, characterized in that: The process of updating the 3D navigation model based on the deformed anatomical structure position, generating an updated model mesh, calculating the projected coordinates of the instrument's current position in the model, and determining if the projected coordinates exceed a preset boundary, triggering a deviation warning signal and monitoring the deviation speed value in real time to reflect the severity of the deviation, includes: The deformed anatomical structure position data is obtained, and an updated 3D navigation model is generated through a 3D reconstruction algorithm to obtain the model mesh; Based on the updated model mesh, the stereoscopic projection method is used to calculate the projected coordinates of the current position of the instrument in the model, and the projected coordinate values ​​are obtained. If the projected coordinate value exceeds the preset boundary, a deviation warning signal is triggered by comparing the difference between the coordinate value and the boundary threshold, and the warning signal status is obtained. Based on the warning signal status, obtain the changes in projected coordinates at continuous time points, calculate the deviation velocity value, and obtain the deviation velocity value. By analyzing the deviation from the speed value, a preset classification algorithm is used to determine the severity of the deviation and obtain the deviation level. Based on the deviation level, adjust the display parameters of the 3D navigation model, generate real-time visual feedback, and obtain the updated display interface; If the deviation level continues to increase, the preset boundary is dynamically updated by analyzing the trend of the deviation speed value, and the optimized boundary threshold is obtained.

5. The minimally invasive spinal fusion system based on artificial intelligence according to claim 1, characterized in that: After acquiring the deviation warning signal, the instrument trajectory data is smoothed and filtered to obtain a filtered trajectory sequence, which is then matched with the updated model mesh. Dynamic path adjustment parameters are determined, and vibration frequency parameters are integrated into the surgical robot to improve the user experience. This includes: Obtain deviation warning signals, extract real-time trajectory data from instrument sensors, and generate an initial trajectory dataset; For the initial trajectory dataset, the Kalman filter algorithm is used for smoothing to obtain the filtered trajectory sequence; The updated model grid is extracted from the pre-established model grid database, and the filtered trajectory sequence is spatially mapped and matched with the model grid to determine the trajectory deviation value. If the trajectory deviation value exceeds the preset threshold, the dynamic path adjustment parameters are calculated based on the deviation value, and a path adjustment vector is generated. Based on the path adjustment vector, update the motion control parameters in the surgical robot control module and adjust the robot joint motion trajectory; By calculating the changing trend of vibration frequency parameters through the adjusted motion trajectory, an optimized set of vibration frequency parameters is generated. By using the optimized set of vibration frequency parameters, the vibration frequency configuration of the surgical robot's execution unit is updated, resulting in an improved level of control experience.

6. The minimally invasive spinal fusion system based on artificial intelligence according to claim 1, characterized in that: The process involves integrating high-resolution intraoperative imaging equipment with artificial intelligence algorithms to generate a real-time three-dimensional anatomical model of the spine and dynamically update the surgical path. The stability of the optimized path curve is verified by fusing multi-modal data across multiple dimensions. If the stability index falls below a preset threshold, the multi-modal data is re-fused to iteratively update the three-dimensional navigation model, resulting in an iterative model version. This includes: Intraoperative multimodal data were acquired in real time using high-resolution imaging equipment to generate the first dataset; Based on the first dataset, a three-dimensional anatomical model of the spine was constructed using a convolutional neural network algorithm to obtain the first three-dimensional model. The surgical path curve is extracted from the first 3D model, and the path is dynamically adjusted using a geometric optimization algorithm to generate the first path curve. For the first path curve, calculate the stability index. If the stability index is lower than the preset threshold, acquire new multimodal data and generate the second dataset. Using the second dataset, the first 3D model is updated by incorporating the random forest algorithm to generate the second 3D model; The surgical path curve is re-extracted from the second 3D model, and the path is adjusted using a geometric optimization algorithm to generate the second path curve; For the second path curve, the stability index is recalculated to obtain an optimized version of the 3D navigation model.

7. The minimally invasive spinal fusion system based on artificial intelligence according to claim 1, characterized in that: The process involves using the iterative model version to process subsequent image data, determining temporary deviations caused by instrument jitter and correcting them using a path planning algorithm, generating a corrected navigation command sequence, and applying a real-time feedback mechanism to convey intuitive deviation information, including: Real-time multimodal data was acquired from intraoperative imaging data, and the data was processed using a convolutional neural network algorithm to obtain a pre-processed image dataset. Using the pre-processed image dataset, temporary deviations caused by instrument vibration are detected, deviation vectors are calculated, and deviation distribution data are obtained. The deviation distribution data is analyzed using a path planning algorithm to generate a corrected path curve, thus obtaining optimized path data. If the stability index of the optimized path data is lower than the preset threshold, the new multimodal data is fused through the random forest algorithm to update the 3D navigation model and obtain the updated model version. Extract the corrected path curves from the updated model version, generate a navigation instruction sequence, and obtain a real-time navigation instruction set; By using a real-time navigation command set and a feedback mechanism to transmit deviation correction information, a dynamically adjusted navigation output is obtained. For the dynamically adjusted navigation output, the stability of the path curve is verified, and the final navigation instruction sequence is obtained.

8. The minimally invasive spinal fusion system based on artificial intelligence according to claim 7, characterized in that: Also includes: Based on the image data processing results obtained from dynamic image analysis, temporary deviations caused by instrument vibration are determined by jitter signal extraction. A path optimization algorithm is used to generate a corrected navigation command sequence, and intuitive deviation information is transmitted through a real-time feedback mechanism to obtain the dynamic adjustment results of the navigation command sequence, specifically including: The image data processing results are obtained through dynamic image analysis, and the instrument region is separated by image segmentation technology to obtain the instrument location information; The jitter signal is extracted from the instrument position information, and the signal frequency is analyzed using Fourier transform to determine the temporary deviation value; If the temporary deviation value exceeds the preset threshold, the deviation trajectory is fitted by the least squares method to obtain the deviation change trend; Based on the trend of deviation change, the A algorithm is used to generate a corrected navigation instruction sequence to obtain the optimized instruction set. The optimized instruction set is transmitted to the control system through a real-time feedback mechanism to obtain the navigation data after deviation correction; dynamic adjustment parameters are extracted from the navigation data after deviation correction, and the navigation instruction sequence is smoothed by Kalman filtering algorithm to obtain the final navigation instruction sequence; Based on the final navigation command sequence, update the control system parameters to obtain real-time adjusted navigation output.

9. The minimally invasive spinal fusion system based on artificial intelligence according to claim 1, characterized in that: The process of driving the surgical instrument to adjust its movement through the navigation command sequence, acquiring the adjusted trajectory data and cyclically comparing the deviation with the optimized path, and determining whether the deviation is consistently below a threshold, involves maintaining the current model and outputting the final navigation model to ensure reliable consistency. This includes: Initial motion control signals are generated by a sequence of navigation instructions to drive the surgical instruments to perform adjustment actions, thereby obtaining the adjusted trajectory data. The position coordinate sequence is extracted from the adjusted trajectory data, and the deviation between the trajectory data and the optimized path is calculated using Euclidean distance. If the deviation value is lower than the preset threshold, the current deviation value is recorded and the next loop comparison is performed to obtain the deviation change trend; Based on the trend of deviation change, the gradient descent algorithm is used to adjust the parameters of the navigation command sequence and generate new motion control signals; The surgical instruments are driven to perform adjustment actions again by new motion control signals to obtain updated trajectory data; If the deviation between the updated trajectory data and the optimized path remains below a preset threshold, then the current navigation instruction sequence is determined as the final navigation model. The final navigation model generates a standardized motion control sequence, which outputs navigation commands to drive surgical instruments.

10. The minimally invasive spinal fusion system based on artificial intelligence according to claim 1, characterized in that: By acquiring trajectory data of the surgical instruments after adjustment, an iterative comparison is made between the data and the optimized path. If the deviation remains below a threshold, the current model is maintained, and the final navigation model is output, resulting in a reliable and consistent navigation result. Specifically, this includes: The trajectory data of the surgical instruments after movement is acquired, and the position coordinates are collected in real time by sensors to obtain the initial trajectory dataset; Using the initial trajectory dataset, the least squares method is used to fit the expected path, the deviation between the actual path and the expected path is calculated, and the path deviation value is determined. If the path deviation value is lower than the preset threshold, the current path model is retained, the optimized path parameters are output, and the current model parameters are obtained. Based on the current model parameters, the Kalman filter algorithm is used to smooth the trajectory data, reduce noise interference, and obtain a smoothed trajectory dataset. By using a smooth trajectory dataset, the deviation between the current path and the target navigation path is iteratively compared, the path model is updated, and the updated navigation model is obtained. If the deviation of the updated navigation model is lower than the preset threshold, the model parameters are fixed, the final navigation model is generated, and a stable navigation result is obtained. By stabilizing navigation results, the movement trajectory of surgical instruments is adjusted in real time, and consistent navigation commands are output to obtain accurate navigation output.

Images