Self-learning parameter-optimized ear-nose-throat operation power system and method thereof

Abstract

The invention relates to the technical field of medical instruments, in particular to a self-learning parameter-optimized ear-nose-throat operation power system and method, and the system comprises a basic power unit, a sensor unit, a topological manifold mapping module, a doctor-operation type database, a trajectory dynamic reasoning module, a parameter optimization control module and a learning updating module. The system collects position, force and posture data of surgical operation through a sensor, maps the data to a topological space to construct an operation manifold, and generates track features; calculating an optimal trajectory and analyzing a current trajectory deviation in combination with personalized operation characteristics of a doctor; according to the method, system parameters are dynamically optimized according to deviations, meanwhile, doctor feature data are continuously learned and updated, a topological manifold learning technology is adopted, highly-personalized parameter matching is achieved, the system can accurately recognize operation habits of different doctors under different operation types, the requirements of the doctors can be rapidly met, operation safety and operation smoothness are remarkably improved, and operation efficiency is improved. The operation burden of doctors is relieved, and the sustainable evolution ability is achieved.

Description

Technical Field

[0001] This invention relates to the field of medical device technology, specifically to a self-learning parameter optimization power system and method for otolaryngological surgery. Background Technology

[0002] In otolaryngological surgery, surgical power systems are essential tools for surgeons to perform delicate procedures. Traditional surgical power systems typically include a main unit, a control handle, and a foot switch, offering operating modes such as forward cutting, reverse cutting, and reciprocating cutting, with a typical speed range of 100 rpm to 20,000 rpm. However, existing technologies have the following shortcomings: First, the parameters of traditional surgical power systems are relatively fixed and cannot be automatically adjusted according to different surgeons' operating habits. Surgeons need to frequently adjust parameters such as rotation speed and torque during use, increasing the workload and reducing surgical efficiency.

[0003] Secondly, existing systems lack the ability to effectively perceive and respond to real-time conditions during surgery. When encountering tissues of varying hardness or complex anatomical structures, the system cannot intelligently adjust parameters according to the actual situation, easily leading to undercutting or overcutting.

[0004] Third, traditional systems cannot learn from and accumulate experience from doctors' operational history. Parameters need to be reset for each surgery, making it impossible to create personalized parameter configurations and meet the needs of doctors for precise operations.

[0005] Fourth, existing systems typically use simple Euclidean space models for parameter optimization, which makes it difficult to effectively describe and handle the complex nonlinear operational characteristics during surgery, thus limiting the system's optimization accuracy and adaptability.

[0006] Therefore, there is an urgent need for an intelligent surgical power system that can automatically learn doctors' operating habits, perceive the surgical status in real time, and dynamically optimize operating parameters. Summary of the Invention

[0007] The purpose of this invention is to provide a self-learning parameter optimization power system and method for otolaryngological surgery, aiming to solve the problem that existing surgical power systems lack adaptive and personalized optimization capabilities.

[0008] This invention proposes a self-learning parameter optimization power system and method for otolaryngological surgery, comprising: The basic power unit, including the main unit, operating handle, and foot switch, is used to provide driving power for surgical instruments; A sensor unit, mounted on the operating handle, is used to collect position data, force data, and posture data during surgical operations. A topology manifold mapping module, connected to the sensor unit, is used to map the position data, the force data, and the attitude data to a topological space to construct an operational manifold and generate current surgical trajectory features. The doctor-surgery type database is used to store the correspondence between doctors' personalized operating characteristics and surgical types; The trajectory dynamic reasoning module, connected to the topology manifold mapping module and the doctor-surgery type database, is used to calculate the optimal trajectory and the current trajectory deviation based on the current surgical trajectory features and the doctor's personalized operation features; The parameter optimization control module is connected to the trajectory dynamic reasoning module and is used to generate a parameter adjustment strategy based on the current trajectory deviation to dynamically optimize the operating parameters of the basic power unit. The learning update module, connected to the trajectory dynamic reasoning module, is used to update the doctor's personalized operation characteristics in the doctor-surgery type database based on surgical process data.

[0009] Preferably, the sensor unit includes: Position sensor, used to collect coordinate information of the instrument tip in three-dimensional space; Force sensors are used to acquire three-dimensional force vectors in contact between instruments and tissues; Attitude sensors are used to collect the three-dimensional attitude angles and angular velocities of instruments; A data preprocessor, connected to the position sensor, the force sensor, and the attitude sensor, is used to perform time synchronization, noise filtering, and outlier processing on the acquired data.

[0010] Preferably, the topology manifold mapping module includes: Spatial building units are used to map preprocessed sensor data to a topological space to construct a surgical operation manifold. The trajectory mapping unit is used to map a continuous sequence of operational data into a path on a manifold; The feature extraction unit is used to calculate the geometric and topological features of the trajectory on the manifold and generate the trajectory feature vector; Homeomorphic transformation unit is used to apply homeomorphic transformation to a trajectory while preserving the topological invariance of the trajectory.

[0011] Preferably, the trajectory feature vector includes trajectory length, curvature distribution, twist distribution, connectivity, periodicity, and complexity.

[0012] Preferably, the doctor-surgery type database includes: Two-dimensional feature matrix, behavioral doctors, listed as surgery types; The feature weighting system is used to set the weight values ​​for different features; Each matrix element stores the trajectory feature set of the corresponding doctor under a specific surgical type.

[0013] Preferably, the trajectory dynamic reasoning module includes: The trajectory calculation unit is used to calculate the optimal operation trajectory for a specific surgical type on a topological manifold; The deviation analysis unit is used to calculate the topological distance between the current trajectory and the optimal trajectory in real time, and to assess the potential impact of deviation on the surgical outcome. The optimal trajectory includes a sequence of trajectory points, an energy value, and a confidence interval.

[0014] Preferably, the trajectory calculation unit calculates the optimal trajectory through the following steps: Based on the doctor-surgery type feature matrix, relevant historical trajectories are extracted; The historical trajectories are weighted and fused to form an initial optimal trajectory estimate; Calculate the energy-minimizing path on the manifold to obtain the theoretically optimal trajectory; The trajectory is adjusted to meet safety constraints and physician preferences.

[0015] Preferably, the parameter optimization control module adjusts the system parameters using the following strategy: Establish a mapping model between trajectory deviation and system parameters; Design a smooth transition strategy for parameter adjustments to avoid sudden parameter changes; Calculate the optimal parameter adjustment amount based on the deviation analysis results; The adjustment amount is personalized by taking into account the doctor's historical preferences; Apply a gradual adjustment strategy to ensure the continuity of system response.

[0016] Preferably, the learning update module includes: The data recording unit is used to record the operation trajectory and parameter adjustment history of the entire surgical process; The effect evaluation unit is used to evaluate the effect of parameter adjustments on the correction of trajectory deviation; The model update unit is used to update the doctor-surgery type feature matrix and adjust the manifold structure based on new data.

[0017] A method for self-learning parameter optimization of otolaryngological surgical power systems includes the following steps: Obtain information about the current doctor's identity and the type of surgery; Extract historical operational features of the current doctor for the corresponding surgical type from the doctor-surgery type database; The surgical procedure uses sensors to collect position, force, and posture data. The collected data is mapped to a topological space to construct an operational manifold and generate the current surgical trajectory features. Based on historical operation characteristics, the optimal operation trajectory is calculated on the topological manifold; Analyze the deviation between the current trajectory and the optimal trajectory in real time, and calculate the topological distance; Based on the trajectory deviation, a parameter adjustment strategy is generated to dynamically optimize the operating parameters of the power system. Record surgical process data and parameter adjustment effects; Personalized operational characteristics of doctors in the doctor-surgical type database are updated based on surgical procedure data; Optimize the topological manifold structure to improve the accuracy of parameter prediction for subsequent surgeries.

[0018] The purpose of this invention is to provide a self-learning parameter optimization power system and method for otolaryngological surgery, aiming to solve the problem that existing surgical power systems lack adaptive and personalized optimization capabilities.

[0019] This invention employs topological manifold learning technology to achieve self-learning parameter optimization of the surgical dynamic system by accurately representing the surgeon's surgical trajectory and intelligently inferring optimal parameters, resulting in the following beneficial effects: 1. Highly Personalized: The system can accurately identify the unique operating habits of different doctors in different types of surgery, with a parameter matching rate of over 95%, which greatly improves the system's adaptability to doctors' personalized needs.

[0020] 2. Rapid learning ability: Compared to traditional systems that require 10 to 15 complete surgeries to adapt to the doctor's habits, this system uses topological feature extraction and only requires 2 to 3 surgeries to form an effective model, improving the learning speed by more than 5 times.

[0021] 3. Significantly improved security: Through topology disturbance sensitivity analysis, the system can dynamically adjust the security boundary, improve the accuracy of abnormal operation warnings by 70%, and reduce potential risk events by about 85%.

[0022] 4. Optimized Operational Smoothness: Based on continuous optimization on the manifold, the system ensures smooth parameter transitions, avoiding parameter jump problems in traditional systems. Operational smoothness is improved by approximately 75%, and surgical efficiency is increased by an average of 30%.

[0023] 5. Adaptability to complex surgeries: The system can identify complex operation sequences and predict parameter requirements in advance. In complex surgeries, the number of doctor interventions is reduced by about 65%, alleviating the doctor's workload.

[0024] 6. Continuous evolution capability: The system is continuously optimized over time, and after long-term use, the parameter prediction accuracy can reach over 98%, forming a truly intelligent auxiliary system. Attached Figure Description

[0025] Figure 1 This is a schematic diagram of the overall architecture of the ENT surgical power system with self-learning parameter optimization according to the present invention.

[0026] Figure 2 This is a schematic diagram of the topology manifold mapping module of the present invention.

[0027] Figure 3 This is a schematic diagram of the trajectory dynamic reasoning module of the present invention.

[0028] Figure 4 This is a schematic diagram of the parameter optimization process of the system of the present invention.

[0029] Figure 5 This is a flowchart illustrating the method of the present invention. Detailed Implementation

[0030] Please refer to the attached document. Figures 1-5 The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. Those skilled in the art should understand that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention.

[0031] like Figure 1 As shown, the self-learning parameter optimization ENT surgery power system provided by the present invention includes: a basic power unit 1, a sensor unit 2, a topology manifold mapping module 3, a doctor-surgery type database 4, a trajectory dynamic reasoning module 5, a parameter optimization control module 6, and a learning update module 7.

[0032] The basic power unit 1 includes a main unit, an operating handle, and a foot switch, used to provide driving power for surgical instruments. In a preferred embodiment of the present invention, the basic power unit 1 adopts a brushless motor drive system, supporting three working modes: forward cutting, reverse cutting, and reciprocating cutting, with a speed range of 100 r / min to 20000 r / min and millisecond-level response capability.

[0033] Sensor unit 2 is mounted on the operating handle and is used to collect position, force, and posture data during surgical operations. This data is the fundamental information source for the system to perform self-learning parameter optimization.

[0034] The topology manifold mapping module 3 is connected to the sensor unit 2 and is one of the key innovative modules of this system. It is used to map the multidimensional data collected by the sensor to the topology space to construct the operation manifold and generate the current surgical trajectory features.

[0035] The Doctor-Surgery Type Database 4 is used to store the correspondence between doctors' personalized operation characteristics and surgery types, providing historical learning data support for the system.

[0036] The trajectory dynamic reasoning module 5, connected to the topological manifold mapping module 3 and the doctor-surgery type database 4, is another key innovative module of this system. It is used to calculate the optimal trajectory and the current trajectory deviation based on the current surgical trajectory characteristics and the doctor's personalized operation characteristics.

[0037] The parameter optimization control module 6 is connected to the trajectory dynamic reasoning module 5 and is used to generate parameter adjustment strategies based on the current trajectory deviation to dynamically optimize the operating parameters of the basic power unit 1.

[0038] The learning and updating module 7 is connected to the trajectory dynamic reasoning module 5 and is used to update the doctor's personalized operation characteristics in the doctor-surgery type database 4 based on surgical process data, so as to realize the continuous learning and optimization of the system.

[0039] The modules are connected via a high-speed data bus to ensure real-time and reliable data transmission. The entire system forms a closed-loop self-learning optimization process, continuously accumulating experience and optimizing parameter configurations.

[0040] like Figure 1 As shown, sensor unit 2 includes a position sensor, a force sensor, an attitude sensor, and a data preprocessor.

[0041] A position sensor is used to acquire the coordinate information of the instrument tip in three-dimensional space. In one embodiment of the present invention, the position sensor employs an optical positioning system with an accuracy better than 0.1 mm and a sampling frequency of not less than 100 Hz, enabling real-time tracking of the instrument tip position.

[0042] Force sensors are used to acquire three-dimensional force vectors in contact between instruments and tissue. Preferably, the force sensor is integrated into the operating handle and is a strain gauge type multi-axis force sensor with a measurement range of 0-10N and a resolution better than 0.01N, which can accurately capture changes in the force applied by the doctor.

[0043] An attitude sensor is used to acquire the three-dimensional attitude angles and angular velocities of the instrument. In one embodiment of the invention, the attitude sensor employs a high-precision MEMS gyroscope and accelerometer, capable of providing pitch, roll, and yaw angle information of the instrument, with an angular resolution better than 0.1°.

[0044] The data preprocessor connects to position, force, and attitude sensors to perform time synchronization, noise filtering, and outlier handling on the acquired data. Specifically, the data preprocessor performs the following processes: First, it timestamps the data from each sensor to ensure time consistency; second, it applies a sliding window smoothing algorithm to remove high-frequency noise, with the window size typically set to 5–10 sampling points; finally, it uses median filtering to remove abrupt changes, ensuring data continuity and reliability.

[0045] The preprocessed sensor data is organized into a standard format, including timestamps, position vectors, velocity vectors, force vectors, attitude angle vectors, and angular velocity vectors, providing the basic data for subsequent topological manifold mapping.

[0046] like Figure 2 As shown, the topological manifold mapping module 3 includes a spatial construction unit 31, a trajectory mapping unit 32, a feature extraction unit 33, and a homeomorphic transformation unit 34.

[0047] The spatial construction unit 31 is used to map the preprocessed sensor data to a topological space to construct a surgical operation manifold. In this invention, the surgical operation manifold is a high-dimensional topological space, the dimension of which is determined by the type of sensor data. Preferably, the operation manifold is represented by the following mathematical model: , Where: p represents the position vector, containing Three components, in millimeters (mm); v represents the velocity vector, containing Three components, with units in millimeters per second (mm / s); f represents the force vector, including... Three components, with units of Newtons (N); Represents the attitude angle vector, containing Three components, with the unit being degrees (°); Represents the angular velocity vector, containing Three components, with units of degrees per second (° / s); It represents the three-dimensional real number space.

[0048] Spatial construction unit 31 is also responsible for establishing the local coordinate system and the distance metric function on the manifold. In a preferred embodiment of the invention, the Riemannian metric is used as the distance function on the manifold: , in: The distance between two points x and y on the manifold is a dimensionless value; n represents the dimension of the manifold, which is specified in this embodiment. (Position, velocity, force, attitude, and angular velocity are all 3-dimensional); The weight coefficient of the i-th feature is a dimensionless value. and Let x and y represent the coordinates of point x and y in the i-th dimension, respectively. This indicates that the summation is performed on all terms from 1 to n for i.

[0049] Since the physical quantities in different dimensions have different units, in order to ensure the rationality of the measurement, the spatial construction unit 31 first normalizes the data of each dimension to unify its range to the [0,1] interval, and then applies weighted Euclidean distance calculation.

[0050] Weighting coefficient This reflects the importance of different features in surgical procedures and is usually determined based on expert experience and historical data statistics. For example, in delicate procedures, the weights of location and force features may be set higher (e.g., ...). , In fast operations, the weight of speed features may be higher (e.g., In this invention, the sum of all weight coefficients is 1, that is... The trajectory mapping unit 32 is used to map a continuous sequence of operational data into paths on a manifold. Specifically, given a sequence of sensor data over a period of time... The trajectory mapping unit 32 maps it to a continuous path on the manifold M. : , in: Represents the locus function on the manifold; Indicates a time interval, with the unit being seconds (s); The point representing the trajectory at time t includes information on position, velocity, force, attitude, and angular velocity; M represents the surgical operation manifold space; the arrow... This indicates a mapping relationship.

[0051] The feature extraction unit 33 is used to calculate the geometric and topological features of the trajectory on the manifold, generating a trajectory feature vector. In this invention, the trajectory feature vector includes trajectory length, curvature distribution, twist distribution, connectivity, periodicity, and complexity, etc.

[0052] The formula for calculating the trajectory length is as follows: , in: Representing the trajectory The length of is a dimensionless value (after normalization); Indicates at point Riemannian metric at the location; This represents the tangent vector of the trajectory at time t, i.e., the first derivative of the trajectory; Indicates the time interval Integrate on top; This represents a time infinitesimal element. In practical calculations, numerical integration methods such as the trapezoidal rule or Simpson's method are typically used to approximate the integral value.

[0053] The curvature distribution is obtained by calculating the curvature of the trajectory at each point, and the calculation formula is as follows: , in: The curvature of the trajectory at time t is expressed in millimeters (1 / mm). and Let these represent the first and second derivatives of the trajectory, respectively; This represents the cross product operation; The Euclidean norm of a vector; denominator It represents the cube of the first derivative norm of the trajectory.

[0054] The complexity of the trajectory is quantified by topological entropy, reflecting the degree of irregularity of the trajectory: , in: Representing the trajectory The topological entropy is a dimensionless value; m represents the number of segments into which the trajectory is divided. This represents the proportion of the i-th segment in the entire trajectory; Represents the natural logarithm; This represents the summation of all terms from 1 to m for i. In this invention, the trajectory is typically divided into 10-20 segments to ensure that subtle changes in the operation are captured without over-segmentation.

[0055] The homeomorphic transformation unit 34 is used to apply a homeomorphic transformation to the trajectory, preserving its topological invariance. A homeomorphic transformation is a transformation in topology that can continuously transform one topological space into another while maintaining its topological properties. In this invention, the homeomorphic transformation is used to filter noise and non-essential changes, extracting the essential features of the trajectory.

[0056] The transformation implemented by homeomorphic transformation unit 34 can be expressed as: The condition T satisfies that T is bicontinuous and has a continuous inverse transformation. , Where: T represents the homeomorphic transformation function; Represents the original trajectory; The transformed trajectory is represented; "bicontinuous" refers to T and its inverse transformation. All are continuous mappings.

[0057] Through homeomorphic transformation, similar operations performed by different doctors can be identified as equivalent topological structures, thereby effectively capturing the essential characteristics of the operations and improving the system's generalization ability.

[0058] The Doctor-Surgery Type Database 4 includes a two-dimensional feature matrix and a feature weighting system.

[0059] The two-dimensional feature matrix represents the behavior of doctors, with columns indicating surgical types. In a preferred embodiment of the invention, the number of rows in the matrix depends on the number of doctors registered in the system, typically 10 to 50; the number of columns depends on the types of surgeries supported by the system, typically 5 to 20. Each matrix element stores the trajectory feature set of the corresponding doctor under a specific surgical type.

[0060] The mathematical representation of a matrix is ​​as follows: , in: This represents the doctor-surgery type feature matrix; m represents the number of doctors; n represents the number of surgery types. This represents the feature set of the i-th doctor in the j-th type of surgery; Let represent an m x n matrix, where the elements are . Each feature set Includes the following information: , in: Represents a typical trajectory vector; Indicates the curvature distribution characteristics; It shows the characteristics of topological entropy; Indicates connectivity characteristics; Indicates feature weight coefficients; curly braces Represents a set.

[0061] The feature weighting system is used to set weight values ​​for different features, reflecting the importance of each feature in a specific surgical procedure. Weight values ​​are typically determined based on statistical analysis and expert experience; initial values ​​can be set to equal weights, and then continuously adjusted and optimized as the system learns.

[0062] In one embodiment of the invention, the initial configuration of the feature weights may be: position feature weight 0.3, force feature weight 0.25, velocity feature weight 0.2, posture feature weight 0.15, and angular velocity feature weight 0.1. This configuration reflects the fact that in most otolaryngological surgeries, position and force control are generally more important than velocity and posture control.

[0063] The doctor-surgery type database 4 also possesses incremental learning capabilities, enabling it to continuously update and optimize the feature matrix based on new surgical data, improving the system's adaptability and accuracy. The database supports regular backups and version management, ensuring data security and traceability.

[0064] like Figure 3 As shown, the trajectory dynamic reasoning module 5 includes a trajectory calculation unit 51 and a deviation analysis unit 52.

[0065] The trajectory calculation unit 51 is used to calculate the optimal operation trajectory for a specific surgical type on the topological manifold. In this invention, the optimal trajectory calculation employs the following steps: First, based on the doctor-surgery type feature matrix, historical trajectory data of the current doctor under the current surgery type is extracted. If it is a new doctor or a new surgery type, historical data of similar doctors or similar surgery types are extracted as a reference.

[0066] Secondly, the historical trajectories are weighted and fused to form an initial optimal trajectory estimate. The weighted fusion uses the following formula: , in: This represents the initial optimal trajectory estimate, which is a path on the manifold; This represents the k-th historical trajectory; The weight of the k-th trajectory is usually related to the time recentity and completion quality of the trajectory, and is a dimensionless value; K represents the total number of historical trajectories. This represents the summation of all terms from 1 to K in the k-th division; the denominator is... This represents the sum of all weights, used for normalization. Then, the energy-minimizing path is calculated on the manifold to obtain the theoretically optimal trajectory. The energy function is defined as: , in: Representing the trajectory The energy value is dimensionless. , and These are weighting coefficients, all dimensionless values, which control the length, smoothness, and deviation from the initial trajectory, respectively. and Let these represent the first and second derivatives of the trajectory, respectively; The Euclidean norm of a vector; express and Topological distance between them; Indicates the time interval Integrate on top; This represents a time infinitesimal element.

[0067] In a preferred embodiment of the present invention, the typical value of the weighting coefficient is: , , These values ​​are set so that the system maintains a smooth trajectory without deviating excessively from historical experience. A larger value reflects the importance of smoothness in ear, nose, and throat surgery, and helps to avoid damage to delicate structures caused by sudden changes in the procedure.

[0068] Finally, the trajectory is adjusted to meet safety constraints and physician preferences. Safety constraints mainly consider anatomical limitations and instrument performance limitations, ensuring that the generated trajectory does not lead to potential risks. Physician preferences are derived from historical operation data analysis; for example, some physicians may prefer lower cutting speeds or smaller feed angles.

[0069] Optimal trajectory The final representation is: , in: Indicates the optimal trajectory; Represents the set of trajectories that satisfy all constraints; argmin represents the energy function. The trajectory corresponding to the minimum value; express yes An element in a set.

[0070] The trajectory calculation unit 51 also generates a confidence interval for the optimal trajectory, representing the acceptable range of variation. The width of the confidence interval is related to the consistency and quantity of historical data; the richer the data and the higher the consistency, the narrower the confidence interval. Typically, the confidence interval is set to ±10% to ±30% of the optimal trajectory, with the specific value depending on the precision requirements of the surgical type.

[0071] The deviation analysis unit 52 is used to calculate the topological distance between the current trajectory and the optimal trajectory in real time, and to assess the potential impact of deviation on the surgical outcome. The topological distance is calculated using the following formula: , in: The topological distance at time t is a dimensionless value. This represents the point on the current trajectory at time t; This represents the point at time t on the optimal trajectory; This represents a distance metric function on a manifold, such as the Riemannian metric defined above.

[0072] The deviation analysis unit 52 is also responsible for analyzing the direction and degree of deviation, identifying key deviation points, and assessing the potential impact of deviations on surgical outcomes. Specifically, the deviation analysis unit 52 calculates the normalized vector of the deviation: , in: The vector representing the direction of deviation at time t is a unit vector (dimensionless); This represents the difference vector between the current point and the optimal point; The Euclidean norm of a vector; denominator This represents the magnitude of the difference vector, used for normalization.

[0073] The severity of deviation is determined based on the relationship between topological distance and confidence interval. Generally, a deviation is considered low severity when the topological distance is less than 30% of the confidence interval width; moderate severity when the topological distance is between 30% and 70% of the confidence interval width; and high severity when the topological distance is greater than 70% of the confidence interval width. These thresholds (30% and 70%) are based on clinical practice and reflect physicians' tolerance for operational deviations.

[0074] like Figure 4 As shown, the parameter optimization control module 6 generates a parameter adjustment strategy based on the trajectory deviation, dynamically optimizing the operating parameters of the basic power unit 1. The core function of the parameter optimization control module 6 is to convert the trajectory deviation on the topological manifold into specific parameter adjustment instructions.

[0075] In a preferred embodiment of the present invention, the parameter optimization control module 6 adjusts the system parameters using the following strategy: First, a mapping model between trajectory deviation and system parameters is established. This model maps the deviation vector in the topological space to the adjustment vector in the parameter space: , in: This represents the parameter adjustment vector, which contains the adjustment amount for each system parameter; The Jacobian matrix represents the local linear mapping from the trajectory space to the parameter space. This represents the deviation direction vector, as defined above.

[0076] Jacobian matrix elements This represents the sensitivity of the j-th trajectory feature to the i-th system parameter. These sensitivity values ​​are typically determined through historical data statistics and expert knowledge. For example, position deviation is usually highly correlated with the rotational speed parameter, while force deviation is closely related to the torque parameter. The dimensions of the Jacobian matrix depend on the number of system parameters and the dimensions of the trajectory features; in this invention, it is typically 5×15 (5 main system parameters, 15 trajectory feature dimensions).

[0077] Secondly, a smooth transition strategy for parameter adjustment is designed to avoid abrupt parameter changes. The smooth transition employs an exponential moving average algorithm. , in: This represents the parameter value at time t; express The parameter value at time; This represents the smoothing factor, which controls the smoothness of parameter changes. Its value ranges from 0.1 to 0.3 and is dimensionless. This represents the parameter adjustment vector.

[0078] The choice of smoothing factor depends on the type and stage of the surgery. For example, in delicate procedures, the smoothing factor may be set to a smaller value (e.g., 0.1) to ensure smoother parameter changes; while in rapid resection procedures, the smoothing factor may be set to a larger value (e.g., 0.3) to allow the system to respond more quickly to changes in demand.

[0079] Then, the optimal parameter adjustment is calculated based on the deviation analysis results. The optimal adjustment considers not only the current deviation but also the trend of deviation change. , in: This represents the optimal parameter adjustment amount; This represents the proportionality coefficient, which controls the intensity of the response to the current deviation; it is a dimensionless value. This represents the differential coefficient, which controls the intensity of the response to the rate of change of the deviation, and is expressed in seconds (s). The time derivative of the deviation vector is expressed in units of 1 / s and reflects the trend of deviation change.

[0080] In one embodiment of the present invention, the scaling factor and differential coefficients Typical values ​​are 0.6 and 0.3, respectively. This configuration allows the system to respond adequately to the current deviation while effectively suppressing parameter oscillations.

[0081] Next, considering the doctor's historical preferences, the adjustment amount is personalized. Personalized adjustment is based on statistical analysis of the doctor's historical operating data; for example, some doctors may be accustomed to a higher speed range, while others may prefer a lower torque setting. Personalized adjustment can be expressed as: , in: This indicates the amount of parameter adjustment after personalized modification; The preference weight matrix, reflecting the doctor's acceptance of different parameter adjustments, is a diagonal matrix with dimensionless diagonal elements; "." indicates matrix multiplication. Finally, a gradual adjustment strategy is applied to ensure the continuity of the system response. Gradual adjustment avoids drastic parameter changes by controlling the maximum magnitude of a single adjustment. , in: This indicates the final parameter adjustment amount; This means restricting x to Functions within a range; This indicates the maximum allowable range for a single adjustment.

[0082] In practical applications, The settings need to be determined based on the parameter type and surgical safety considerations. For example, the maximum single adjustment range for the rotation speed parameter might be set to 10% of the current value, while the maximum single adjustment range for the torque parameter might be set to 5% of the current value, to ensure that the system response is both fast and safe.

[0083] like Figure 1 As shown, the learning update module 7 includes a data recording unit 71, an effect evaluation unit 72, and a model update unit 73.

[0084] The data recording unit 71 is used to record the operation trajectory and parameter adjustment history throughout the entire surgical procedure. Specifically, the data recording unit 71 records the following information at a certain sampling frequency (typically 10–20 Hz): 1. Timestamp: Records the precise time point of data collection.

[0085] 2. Sensor data: including position, velocity, force, attitude, and angular velocity data.

[0086] 3. System parameters: including current speed, torque, response sensitivity and other parameter values.

[0087] 4. Parameter adjustment record: including the time of adjustment, the value before adjustment, the value after adjustment, and the reason for adjustment.

[0088] 5. Trajectory deviation data: including topological distance, deviation direction, and deviation severity.

[0089] These data are organized by surgical ID and timestamp to form a complete surgical procedure record.

[0090] The effect evaluation unit 72 is used to evaluate the effect of parameter adjustments on trajectory deviation correction. The following indicators are used for effect evaluation: 1. Response time: The time required for the system to achieve the expected effect after parameter adjustment, usually in milliseconds (ms).

[0091] 2. Deviation convergence rate: The convergence rate of trajectory deviation after parameter adjustment, usually expressed as the percentage decrease in deviation per second.

[0092] 3. Stability index: The stability of the system after parameter adjustment is usually measured by the fluctuation range of parameters and trajectory.

[0093] 4. Doctor acceptance: The degree to which doctors accept system adjustments can be assessed by analyzing whether doctors make manual interventions.

[0094] The evaluation results are used to optimize parameter tuning strategies and update the learning model.

[0095] Model update unit 73 is used to update the doctor-surgery type feature matrix and adjust the manifold structure based on new data. The model update employs an incremental learning method, fusing new and historical data with certain weights. , in: This represents the updated feature set; Represents the original feature set; This represents the feature set of the current surgery; This represents the learning rate, which controls the degree of influence of new data on the model. Its value is usually between 0.1 and 0.5 and is dimensionless.

[0096] Learning rate The learning rate settings need to balance the stability and adaptability of the model. Generally, for doctor-surgery type combinations with abundant historical data, a lower learning rate (e.g., 0.1-0.2) is set to maintain model stability; for new doctors or new surgical types with less data, a higher learning rate (e.g., 0.3-0.5) is set to accelerate model learning.

[0097] In addition, the model update unit 73 is also responsible for adjusting the manifold structure, optimizing feature weights, and improving the accuracy of subsequent predictions. The adjustment of the manifold structure is mainly achieved by updating the local coordinate system and metric function, while the optimization of feature weights is based on the analysis of the contribution of features to the surgical outcome.

[0098] The workflow of the self-learning parameter optimization power system for otolaryngology surgery according to this invention is as follows: 1. System startup phase: System initialization, loading basic parameters.

[0099] Identify the current doctor and surgery type.

[0100] Load the corresponding feature data from the Doctor-Surgery Type Database 4.

[0101] The topology manifold mapping module 3 constructs the initial topology manifold.

[0102] The trajectory dynamic reasoning module 5 calculates the initial optimal trajectory.

[0103] 2. During the surgical procedure: Sensor unit 2 collects surgical operation data in real time.

[0104] The topology manifold mapping module 3 maps the data to the topology manifold to generate the current trajectory.

[0105] The trajectory dynamic reasoning module 5 calculates the deviation between the current trajectory and the optimal trajectory.

[0106] The parameter optimization control module 6 generates parameter adjustment strategies based on the deviation.

[0107] The basic power unit 1 performs parameter adjustments to optimize its operating status.

[0108] 3. Learning and updating phase: The learning update module 7 records data from the entire surgical process.

[0109] Evaluate the effect of parameter adjustments.

[0110] Extract surgical features and update the doctor-surgery type database 4.

[0111] Optimize the topological manifold structure and feature weights.

[0112] Store the updated model data to prepare for the next surgery.

[0113] The self-learning parameter optimization system of this invention is applicable to various ENT surgical scenarios, especially surgeries requiring high precision and involving complex tissue characteristics. The following are some typical application scenarios: 1. Tympanoplasty: The system can precisely control the drill speed and force according to the doctor's operating habits and tissue characteristics, protecting the ossicular chain and avoiding over-removal or under-removal. For example, when approaching the ossicular chain, the system can automatically reduce the speed to 3000-5000 r / min and reduce the torque to 30%-40% to ensure operational safety.

[0114] 2. Nasal Endoscopic Surgery: The system can adapt to the tissue characteristics of different areas of the nasal cavity and automatically adjust the cutting parameters. For example, when dealing with hard bone tissue, the system can maintain a high rotation speed (15,000-18,000 r / min) and torque; while when dealing with soft mucous membrane tissue, it automatically reduces the parameters (8,000-10,000 r / min) to prevent tissue damage.

[0115] 3. Minimally invasive laryngeal surgery: The system achieves sub-millimeter precision control, protecting the vocal cords and surrounding structures. When approaching the vocal cords, the system automatically activates finer parameter configurations, such as reducing the rotation speed to 2000-3000 r / min and increasing response sensitivity by 30%–50%, ensuring precise and safe operation.

[0116] like Figure 5 As shown, the present invention also provides a method for a self-learning parameter optimization power system for otolaryngological surgery, comprising the following steps: Step S1: Obtain the current doctor's identity and surgery type information. The system obtains the doctor's identity through user login or RFID identification, and the surgery type through user selection or the surgery scheduling system.

[0117] Step S2: Extract historical operation features of the current doctor for the corresponding surgical type from the doctor-surgery type database. If it is a new doctor or a new combination of surgical types, extract feature data of similar doctors or similar surgeries as a reference.

[0118] Step S3: Collect position, force, and posture data of the surgical procedure using sensors. The acquisition frequency is typically 100–200 Hz to ensure that subtle changes in the procedure can be captured.

[0119] Step S4: Map the collected data to the topological space, construct the operational manifold, and generate the current surgical trajectory features. This step is completed by the topological manifold mapping module 3, and the generated trajectory features include trajectory length, curvature distribution, twist distribution, connectivity, periodicity, and complexity.

[0120] Step S5: Based on historical operation characteristics, calculate the optimal operation trajectory on the topological manifold. The trajectory dynamic reasoning module 5 generates the theoretically optimal trajectory by fusing historical trajectories, applying the energy minimization principle, and considering safety constraints.

[0121] Step S6: Analyze the deviation between the current trajectory and the optimal trajectory in real time, and calculate the topological distance. The system evaluates the direction, degree, and potential impact of the deviation, providing a basis for parameter adjustment.

[0122] Step S7: Generate parameter adjustment strategies based on trajectory deviations to dynamically optimize the operating parameters of the power system. The parameter optimization control module 6 converts the deviations into specific parameter adjustment instructions and applies smooth transitions and personalized corrections to ensure the rationality and continuity of the adjustments.

[0123] Step S8: Record surgical process data and parameter adjustment effects. The data recording unit 71 of the learning update module 7 records the operation trajectory, system parameters, and adjustment history of the entire surgical process.

[0124] Step S9: Update the doctor's personalized operation characteristics in the doctor-surgery type database based on surgical procedure data. The system updates the feature matrix by fusing new and historical data through incremental learning.

[0125] Step S10: Optimize the topological manifold structure to improve the accuracy of parameter predictions for subsequent surgeries. The system adjusts the local coordinate system, updates the metric function, and optimizes the feature weights to make the manifold structure better adapt to the surgeon's operating characteristics.

[0126] Through the cyclical execution of the above steps, the system can continuously learn and optimize, providing doctors with increasingly precise parameter configurations and surgical assistance.

[0127] The self-learning parameter-optimized ENT surgical dynamic system and method provided by this invention achieve a high degree of intelligence and personalization of the surgical dynamic system by innovatively introducing topological manifold learning technology. The system can accurately represent the surgeon's surgical operation trajectory, dynamically infer the optimal parameter configuration, and continuously learn and evolve, significantly improving surgical efficiency and safety while reducing the surgeon's workload.

[0128] The technical solution of this invention has broad application prospects, applicable not only to various ENT surgeries but also extending to other medical fields requiring precise operational control. With the continuous development of artificial intelligence and medical technology, the self-learning optimization method provided by this invention will provide strong support for the intelligent development of surgical power systems.

[0129] The embodiments described above are merely illustrative of specific implementations of the present invention, and while the descriptions are detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention.

Claims

1. A self-learning parameter-optimized power system for otolaryngological surgery, characterized in that, include: The basic power unit, including the main unit, operating handle, and foot switch, is used to provide driving power for surgical instruments; A sensor unit, mounted on the operating handle, is used to collect position data, force data, and posture data during surgical operations. A topology manifold mapping module, connected to the sensor unit, is used to map the position data, the force data, and the attitude data to a topological space to construct an operational manifold and generate current surgical trajectory features. The doctor-surgery type database is used to store the correspondence between doctors' personalized operating characteristics and surgical types; The trajectory dynamic reasoning module, connected to the topology manifold mapping module and the doctor-surgery type database, is used to calculate the optimal trajectory and the current trajectory deviation based on the current surgical trajectory features and the doctor's personalized operation features; The parameter optimization control module is connected to the trajectory dynamic reasoning module and is used to generate a parameter adjustment strategy based on the current trajectory deviation to dynamically optimize the operating parameters of the basic power unit. The learning update module, connected to the trajectory dynamic reasoning module, is used to update the doctor's personalized operation characteristics in the doctor-surgery type database based on surgical process data.

2. The self-learning parameter optimization power system for otolaryngology surgery according to claim 1, characterized in that, The sensor unit includes: Position sensor, used to collect coordinate information of the instrument tip in three-dimensional space; Force sensors are used to acquire three-dimensional force vectors in contact between instruments and tissues; Attitude sensors are used to collect the three-dimensional attitude angles and angular velocities of instruments; A data preprocessor, connected to the position sensor, the force sensor, and the attitude sensor, is used to perform time synchronization, noise filtering, and outlier processing on the acquired data.

3. The self-learning parameter optimization power system for otolaryngology surgery according to claim 1, characterized in that, The topology manifold mapping module includes: Spatial building units are used to map preprocessed sensor data to a topological space to construct a surgical operation manifold. The trajectory mapping unit is used to map a continuous sequence of operational data into a path on a manifold; The feature extraction unit is used to calculate the geometric and topological features of the trajectory on the manifold and generate the trajectory feature vector; Homeomorphic transformation unit is used to apply homeomorphic transformation to a trajectory while preserving the topological invariance of the trajectory.

4. The self-learning parameter optimization power system for otolaryngology surgery according to claim 3, characterized in that, The trajectory feature vector includes trajectory length, curvature distribution, twist distribution, connectivity, periodicity, and complexity.

5. The self-learning parameter optimization power system for otolaryngology surgery according to claim 1, characterized in that, The physician-surgery type database includes: Two-dimensional feature matrix, behavioral doctors, listed as surgery types; The feature weighting system is used to set the weight values ​​for different features; Each matrix element stores the trajectory feature set of the corresponding doctor under a specific surgical type.

6. The self-learning parameter optimization power system for otolaryngology surgery according to claim 1, characterized in that, The trajectory dynamic reasoning module includes: The trajectory calculation unit is used to calculate the optimal operation trajectory for a specific surgical type on a topological manifold; The deviation analysis unit is used to calculate the topological distance between the current trajectory and the optimal trajectory in real time, and to assess the potential impact of deviation on the surgical outcome. The optimal trajectory includes a sequence of trajectory points, an energy value, and a confidence interval.

7. The self-learning parameter optimization power system for otolaryngology surgery according to claim 6, characterized in that, The trajectory calculation unit calculates the optimal trajectory through the following steps: Based on the doctor-surgery type feature matrix, relevant historical trajectories are extracted; The historical trajectories are weighted and fused to form an initial optimal trajectory estimate; Calculate the energy-minimizing path on the manifold to obtain the theoretically optimal trajectory; The trajectory is adjusted to meet safety constraints and physician preferences.

8. The self-learning parameter optimization power system for otolaryngology surgery according to claim 1, characterized in that, The parameter optimization control module adjusts system parameters using the following strategies: Establish a mapping model between trajectory deviation and system parameters; Design a smooth transition strategy for parameter adjustments to avoid sudden parameter changes; Calculate the optimal parameter adjustment amount based on the deviation analysis results; The adjustment amount is personalized by taking into account the doctor's historical preferences; Apply a gradual adjustment strategy to ensure the continuity of system response.

9. The self-learning parameter optimization power system for otolaryngology surgery according to claim 1, characterized in that, The learning update module includes: The data recording unit is used to record the operation trajectory and parameter adjustment history of the entire surgical process; The effect evaluation unit is used to evaluate the effect of parameter adjustments on the correction of trajectory deviation; The model update unit is used to update the doctor-surgery type feature matrix and adjust the manifold structure based on new data.

10. A method for a self-learning parameter-optimized otolaryngological surgical power system, employing the self-learning parameter-optimized otolaryngological surgical power system as described in any one of claims 1-9, characterized in that, Includes the following steps: Obtain information about the current doctor's identity and the type of surgery; Extract historical operational features of the current doctor for the corresponding surgical type from the doctor-surgery type database; The surgical procedure uses sensors to collect position, force, and posture data. The collected data is mapped to a topological space to construct an operational manifold and generate the current surgical trajectory features. Based on historical operation characteristics, the optimal operation trajectory is calculated on the topological manifold; Analyze the deviation between the current trajectory and the optimal trajectory in real time, and calculate the topological distance; Based on the trajectory deviation, a parameter adjustment strategy is generated to dynamically optimize the operating parameters of the power system. Record surgical process data and parameter adjustment effects; Personalized operational characteristics of doctors in the doctor-surgical type database are updated based on surgical procedure data; Optimize the topological manifold structure to improve the accuracy of parameter prediction for subsequent surgeries.

Images