Soft tissue lesion region active identification method and device based on separated weighted strain

Abstract

The application discloses a soft tissue lesion region active identification method and device based on separated weighted strain, and belongs to the technical field of data processing. The method comprises the following steps: estimating and predicting a stress balance equation based on an extended Kalman filter to obtain an updated parameter posterior covariance matrix and establishing an active perception target function of separated weighted strain; solving the target function based on a gradient descent method to obtain an optimal interactive action; controlling a robot to execute the optimal interactive action to obtain a stiffness distribution map, and judging whether the uncertainty variance of a high stiffness lesion region in the stiffness distribution map is less than a safety convergence threshold value; if the uncertainty variance is less than the safety convergence threshold value, outputting a high stiffness lesion region identification result; if the uncertainty variance is greater than or equal to the safety convergence threshold value, returning to the step of collecting the multi-modal observation data of the probability evolution model and the robot interaction to perform iteration until the uncertainty variance is less than the safety convergence threshold value, and the accuracy of lesion positioning and stiffness identification is improved.

Description

Technical Field

[0001] This application belongs to the field of data processing technology, and in particular relates to a method and device for active identification of soft tissue lesion areas based on separation weighted strain. Background Technology

[0002] In minimally invasive surgery, the application of intelligent robotic systems is becoming increasingly widespread. However, current autonomous systems are largely limited by a vision-centric perception paradigm. In real surgical environments, visual data alone is insufficient to perceive the underlying physical properties of tissues, such as spatial heterogeneity caused by pathological changes (e.g., tumors) or hidden high-stiffness abnormalities. Human surgeons typically bridge this cognitive gap through the synergistic integration of visual macroscopic deformation and kinematic resistance feedback.

[0003] However, standard laparoscopic surgery often lacks direct tactile sensor arrays, requiring the induction of substantial physical deformation to acquire internal mechanical properties. This necessitates "active perception" by the robot, determining the optimal point of contact and force application to obtain the most informative macroscopic tissue response. Traditional physics-driven methods typically rely on iterative finite element method optimization, which is extremely computationally expensive in real-time closed-loop active control. While data-driven methods and neural implicit representations have been able to capture complex heterogeneity in recent years, they often require significant single-scene training time and exhibit poor generalization ability when faced with topological changes (such as cutting), making them unsuitable for online surgical execution.

[0004] Current research lacks a framework that inherently couples perception, active exploration, and subsequent operations, preventing robots from effectively "questioning" the organization to eliminate ambiguity in physical properties. Summary of the Invention

[0005] This application aims to address at least one of the technical problems existing in the prior art. To this end, this application proposes a method and apparatus for active identification of soft tissue lesion regions based on separated weighted strain. The identification of high-stiffness lesions is treated as a dynamic sequential decision-making and state estimation process. In the variance space, extended Kalman filtering combined with differentiable projection dynamics is used to achieve real-time tracking and uncertainty quantification of the heterogeneous stiffness field of flexible tissue. Based on this, separated weighted strain is introduced as the active sensing objective function, enabling the autonomous planning of the next optimal interaction action. This more accurately generates local physical strain in the suspected lesion region with the highest classification uncertainty, thereby maximizing stiffness information gain and rapidly separating hidden high-stiffness abnormal regions.

[0006] To address the aforementioned problems, according to a first aspect of this application, a method for active identification of soft tissue lesion regions based on separation-weighted strain is provided, the method comprising: Obtain the target flexible structure, perform spatial discretization modeling on the target flexible structure, and obtain the probabilistic evolution model of the stiffness field state transition of the target flexible structure; Multimodal observation data of probabilistic evolution model and robot interaction were collected to establish the force balance equation of the target flexible tissue; The force equilibrium equation is estimated and predicted based on the extended Kalman filter to obtain the updated posterior covariance matrix of stiffness parameters. An active sensing objective function for separating weighted strain is established based on the updated posterior covariance matrix of stiffness parameters. The optimal interaction action between the robot and the probabilistic evolution model is obtained by solving the active perception objective function based on the gradient descent method. The optimal interactive action is sent to the robot's controller, which controls the robot to perform the optimal interactive action on the target flexible tissue. This yields a stiffness distribution map of the target flexible tissue. The uncertainty variance of the high-stiffness lesion region in the stiffness distribution map is then determined to be less than the safe convergence threshold. If the uncertainty variance is less than the safe convergence threshold, the identification result of the high-stiffness lesion region of the target flexible tissue is output. If the uncertainty variance is greater than or equal to the safe convergence threshold, the process returns to the step of collecting multimodal observation data of the probabilistic evolution model and the robot interaction for iteration until the uncertainty variance is less than the safe convergence threshold.

[0007] According to one embodiment of this application, the step of obtaining the target flexible structure, performing spatial discretization modeling on the target flexible structure, and obtaining a probabilistic evolution model of the stiffness field state transition of the target flexible structure includes: Obtain the target flexible structure and divide the target flexible structure into a finite element mesh model including multiple nodes and multiple triangular elements; The stiffness inside a single mesh element in the finite element mesh model is set to be uniformly distributed, and the heterogeneous stiffness field of the target flexible tissue is constructed as a high-dimensional random variable. A stiffness field state transition model is established based on high-dimensional random variables, resulting in a probabilistic evolution model for the stiffness field state transition of the target flexible tissue.

[0008] According to one embodiment of this application, the process of collecting multimodal observation data from the interaction between the probabilistic evolution model and the robot to establish the force balance equation of the target flexible tissue includes: Multimodal observation data of the interaction between the probabilistic evolution model and the robot are collected. The multimodal observation data includes surface node deformation data of the target flexible tissue and contact force data of the robot end effector. Stiffness field parameters of the target flexible tissue are determined based on a probabilistic evolution model; Based on the surface node deformation data of the target flexible tissue and the contact force data of the robot end, an observation matrix is ​​established by linearizing the differentiable projection dynamics, and the composite effective observation noise data corresponding to the observation matrix is ​​calculated. Based on stiffness field parameters, observation matrix, and composite effective observation noise data, the force balance equation of the target flexible tissue is established.

[0009] According to one embodiment of this application, the step of estimating and predicting the force equilibrium equation based on extended Kalman filtering to obtain the updated stiffness parameter posterior covariance matrix includes: The internal force prediction values ​​are obtained by estimating and predicting the force equilibrium equations based on the extended Kalman filter. Calculate the force residual between the external contact force and the predicted internal force, and solve the Kalman gain of the force equilibrium equation based on the force residual; The heterogeneous stiffness field distribution of the target flexible tissue is iteratively updated based on the force residual and Kalman gain to obtain the updated posterior covariance matrix of stiffness parameters.

[0010] According to one embodiment of this application, the step of establishing an active sensing objective function for separating weighted strain based on the updated stiffness parameter posterior covariance matrix includes: Based on the updated posterior covariance matrix of stiffness parameters, the stiffness estimation uncertainty value of each mesh element in the finite element mesh model is extracted; Based on the uncertainty value of stiffness estimation for each mesh element, mesh elements with uncertainty values ​​greater than a preset threshold are selected as a set of suspected high-stiffness elements. The physical strain generated by the preset interactive action in each grid cell of the suspected high stiffness element set is calculated. The stiffness estimation uncertainty of each grid cell in the suspected high stiffness element set is coupled with the physical strain of each grid cell in the suspected high stiffness element set to establish an active sensing objective function of separated weighted strain.

[0011] According to one embodiment of this application, the calculation formula for the objective function is as follows: in, Denotes the Frobenius norm of a matrix. Let be the posterior covariance matrix of the parameters. To determine based on candidate actions The attention-weighted strain matrix generated in the prediction is... Here, k represents the spatial attention weights, and k represents the time step. The objective function is denoted as .

[0012] According to one embodiment of this application, the step of solving the active perception objective function based on the gradient descent method to obtain the optimal interaction action between the robot and the probabilistic evolution model includes: The robot's kinematic and dynamic constraints are used as the solution boundary for the active perception objective function; Within the solution boundary, the active perception objective function is iteratively calculated using the gradient descent method to select interactive action combinations that meet the solution boundary. Calculate the separate weighted strain information gain for each interaction action in the combination of interaction actions, and take the interaction action with the largest separate weighted strain information gain as the optimal interaction action between the robot and the probabilistic evolution model.

[0013] According to a second aspect of this application, an active identification device for soft tissue lesion regions based on separation-weighted strain is provided, the device comprising: The acquisition module is used to acquire the target flexible structure, perform spatial discretization modeling on the target flexible structure, and obtain the probabilistic evolution model of the stiffness field state transition of the target flexible structure. The first processing module is used to collect multimodal observation data of the interaction between the probabilistic evolution model and the robot, and to establish the force balance equation of the target flexible tissue. The second processing module is used to estimate and predict the force equilibrium equation based on the extended Kalman filter to obtain the updated stiffness parameter posterior covariance matrix. The third processing module is used to establish an active sensing objective function for separating weighted strain based on the updated posterior covariance matrix of stiffness parameters. The fourth processing module is used to solve the active perception objective function based on the gradient descent method to obtain the optimal interaction action between the robot and the probabilistic evolution model. The fifth processing module sends the optimal interactive action to the robot's controller, controls the robot to perform the optimal interactive action on the target flexible tissue, obtains the stiffness distribution map of the target flexible tissue, and determines whether the uncertainty variance of the high-stiffness lesion region in the stiffness distribution map is less than the safe convergence threshold. If the uncertainty variance is less than the safe convergence threshold, the identification result of the high-stiffness lesion region of the target flexible tissue is output. If the uncertainty variance is greater than or equal to the safe convergence threshold, the process returns to the step of collecting multimodal observation data of the probabilistic evolution model and the robot interaction for iteration until the uncertainty variance is less than the safe convergence threshold.

[0014] According to a third aspect of this application, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the active identification method for soft tissue lesion regions based on separation-weighted strain as described in the first aspect above.

[0015] According to a fourth aspect of this application, a non-transitory computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, implements the active identification method for soft tissue lesion regions based on separation-weighted strain as described in the first aspect above.

[0016] According to a fifth aspect of this application, a chip is provided, the chip including a processor and a communication interface coupled to the processor, the processor being used to run a program or instructions to implement the active identification method for soft tissue lesion regions based on separation-weighted strain as described in the first aspect.

[0017] According to a sixth aspect of this application, a computer program product is provided, comprising a computer program that, when executed by a processor, implements the active identification method for soft tissue lesion regions based on separation-weighted strain as described in the first aspect above.

[0018] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application.

[0019] This application provides a method for active identification of soft tissue lesion regions based on separation-weighted strain, which has the following advantages over existing technologies: (1) This application improves the real-time performance and engineering feasibility of soft tissue lesion identification by focusing on suspected high-rigidity areas to perform directional optimization calculations and planning robot interaction actions based on maximizing information gain. Combined with closed-loop iterative convergence judgment to complete lesion identification more accurately, it can stably adapt to flexible tissue scenarios with different deformation characteristics, effectively improving the scenario adaptability and clinical application value of the technical solution.

[0020] (2) This application constructs the force balance equation by using differentiable projection dynamics, iteratively updates the stiffness field and quantifies the posterior covariance by extended Kalman filtering. Compared with traditional soft tissue identification methods, it is more adaptable in lesion localization and stiffness estimation scenarios, effectively improving the problems of low localization accuracy and inability to quantify uncertainty in traditional identification methods, and further improving the accuracy of soft tissue lesion area identification and the reliability of state estimation. Based on the separation weighted strain to construct an active sensing objective function for action optimization, it can more effectively plan the optimal interactive action in the stiffness uncertainty space, better suppress the interference of tissue deformation and the influence of sensing noise, reduce the impact of invalid interaction on identification efficiency, further improve the convergence speed and anti-interference ability of lesion identification, realize efficient real-time tracking of unknown stiffness field, and improve real-time performance and robustness.

[0021] (3) This application organically combines grid modeling, multimodal observation, stiffness field filtering estimation and separation weighted strain active optimization to form a complete closed-loop process for active identification of soft tissue lesion areas. While ensuring higher computational efficiency, it achieves more accurate lesion localization and stiffness identification, providing more efficient and robust technical support for soft tissue exploration and accurate identification of lesion areas by medical robots. Attached Figure Description

[0022] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which: Figure 1 This is one of the flowcharts illustrating the active identification method for soft tissue lesion regions based on separation-weighted strain provided in this application embodiment; Figure 2 This is the second flowchart of the active identification method for soft tissue lesion regions based on separation-weighted strain provided in the embodiments of this application; Figure 3 This is a schematic diagram of the structure of the active identification device for soft tissue lesion areas based on separation-weighted strain provided in the embodiments of this application; Figure 4 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0023] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0024] The terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such use of data can be interchanged where appropriate to allow embodiments of this application to be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class and the number of objects is not limited; for example, a first object can be one or more. Furthermore, in the specification and claims, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0025] The following description, in conjunction with the accompanying drawings, details the active identification method, device, electronic device, and readable storage medium for soft tissue lesion regions based on separation-weighted strain, provided in this application, through specific embodiments and application scenarios.

[0026] Among them, the active identification method for soft tissue lesion areas based on separation weighted strain can be applied to the terminal, specifically executed by the hardware or software in the terminal.

[0027] The terminal includes, but is not limited to, portable communication devices such as mobile phones or tablets with touch-sensitive surfaces (e.g., touchscreen displays and / or touchpads). It should also be understood that, in some embodiments, the terminal may not be a portable communication device, but rather a desktop computer with touch-sensitive surfaces (e.g., touchscreen displays and / or touchpads).

[0028] The following embodiments describe a terminal including a display and a touch-sensitive surface. However, it should be understood that the terminal may include one or more other physical user interface devices such as a physical keyboard, mouse, and joystick.

[0029] The active identification method for soft tissue lesion regions based on separation-weighted strain provided in this application embodiment can be executed by an electronic device or a functional module or entity within an electronic device that can implement the active identification method for soft tissue lesion regions based on separation-weighted strain. The electronic devices mentioned in this application embodiment include, but are not limited to, mobile phones, tablets, computers, cameras, and wearable devices. The active identification method for soft tissue lesion regions based on separation-weighted strain provided in this application embodiment will be described below using an electronic device as the execution subject.

[0030] Figure 1 This is one of the flowcharts illustrating the active identification method for soft tissue lesion regions based on separation-weighted strain provided in this application embodiment, such as... Figure 1 As shown, the active identification method for soft tissue lesion regions based on separation-weighted strain includes steps 110, 120, 130, 140, 150, and 160.

[0031] Step 110: Obtain the target flexible structure, perform spatial discretization modeling on the target flexible structure, and obtain the probabilistic evolution model of the stiffness field state transition of the target flexible structure. In some embodiments, the step of acquiring the target flexible tissue, spatially discretizing and modeling the target flexible tissue to obtain a probabilistic evolution model of the stiffness field state transition of the target flexible tissue includes: Obtain the target flexible structure and divide the target flexible structure into a finite element mesh model including multiple nodes and multiple triangular elements; The stiffness inside a single mesh element in the finite element mesh model is set to be uniformly distributed, and the heterogeneous stiffness field of the target flexible tissue is constructed as a high-dimensional random variable. A stiffness field state transition model is established based on high-dimensional random variables, resulting in a probabilistic evolution model for the stiffness field state transition of the target flexible tissue.

[0032] It's easy to understand that in practical applications such as surgery, flexible tissues exhibit continuous and complex spatial heterogeneity. To enable robots to perform computations and processing, it's first necessary to spatially discretize and model the target flexible tissue. This involves dividing the flexible tissue into... Each node and A finite element mesh model of triangular elements. Assuming the stiffness within each tiny element is uniformly distributed, the heterogeneous stiffness field of the entire structure can be abstracted as a high-dimensional random variable, as shown in the following expression: in, For a high-dimensional random variable matrix, For the m-th variable, Represents the current time step. The number of triangular units.

[0033] Considering that surgical procedures (such as cutting and suturing) continuously alter the topology and mechanical properties of tissues, the following probabilistic evolution model is established to describe the state transitions of the stiffness field: in, For the heterogeneous stiffness field at the next time step, This term represents the direct increment of local stiffness caused by surgical procedures (such as partial tissue removal). In the simple exploration phase where no topological change occurs, this term is a zero vector. Gaussian process noise with zero mean The covariance matrix of zero-mean Gaussian process noise is used to absorb small errors or environmental disturbances not modeled in the system dynamics model. It follows a Gaussian distribution.

[0034] Step 120: Collect multimodal observation data of the interaction between the probabilistic evolution model and the robot, and establish the force balance equation of the target flexible tissue; In some embodiments, the acquisition of multimodal observation data from the interaction between the probabilistic evolution model and the robot, and the establishment of the force balance equations for the target flexible tissue, includes: Multimodal observation data of the interaction between the probabilistic evolution model and the robot are collected. The multimodal observation data includes surface node deformation data of the target flexible tissue and contact force data of the robot end effector. Stiffness field parameters of the target flexible tissue are determined based on a probabilistic evolution model; Based on the surface node deformation data of the target flexible tissue and the contact force data of the robot end, an observation matrix is ​​established by linearizing the differentiable projection dynamics, and the composite effective observation noise data corresponding to the observation matrix is ​​calculated. Based on stiffness field parameters, observation matrix, and composite effective observation noise data, the force balance equation of the target flexible tissue is established.

[0035] After completing the basic modeling, external information needs to be acquired through the robot's physical interaction. However, the surface visual data and end-effector force data acquired by the sensors are noisy and nonlinear, making it impossible to directly invert the internal stiffness. Therefore, a differentiable projection dynamics technique is introduced to construct a linearized observation model capable of rapid differentiation. The robot's end-effector is controlled to contact the tissue, and a binocular vision sensor is used to extract the nodal deformation of the soft tissue surface. Simultaneously, force sensors are used to read the external contact force applied by the robot's end effector. Based on the principle of quasi-static force equilibrium, the internal forces resisting deformation within an organization. It must be in equilibrium with the external contact force. To meet the calculation requirements of the subsequent Kalman filter, the deformation location must be measured... By performing a first-order Taylor expansion on the nonlinear internal force function, the following linear observation equation is derived as the force equilibrium equation, and the calculation formula is shown below: in, External forces borne by soft tissue For the observation matrix, ( (This represents the number of observable surface nodes), reflecting the sensitivity mapping of changes in internal stiffness parameters to the forces acting on the surface nodes. For a high-dimensional random variable matrix, To integrate visual and force perception into a combined effective observation noise, The location of soft tissue surface nodes is obtained through a visual sensor.

[0036] The expression for the covariance matrix of the combined effective observation noise is shown below: in, The tangent stiffness matrix cleverly acts as a means to absorb visual position sampling noise. The mapping bridge projected onto the mechanical noise domain, and then combined with the pure force noise. Superimposed, Sampling noise for visual position, For pure force noise, This is the covariance matrix of the composite effective observation noise.

[0037] Step 130: Estimate and predict the force equilibrium equation based on extended Kalman filtering to obtain the updated posterior covariance matrix of stiffness parameters; In some embodiments, the step of estimating and predicting the force equilibrium equations based on extended Kalman filtering to obtain the updated posterior covariance matrix of stiffness parameters includes: The internal force prediction values ​​are obtained by estimating and predicting the force equilibrium equations based on the extended Kalman filter. Calculate the force residual between the external contact force and the predicted internal force, and solve the Kalman gain of the force equilibrium equation based on the force residual; The heterogeneous stiffness field distribution of the target flexible tissue is iteratively updated based on the force residual and Kalman gain to obtain the updated posterior covariance matrix of stiffness parameters.

[0038] It is easy to understand that, in order to solve the complex noise problem in the force equilibrium equation and the multiple solutions caused by the ill-conditioned inverse problem, the force equilibrium equation is not solved directly. Instead, an extended Kalman filter is used to continuously approximate the true stiffness distribution in the probability space. The system first calculates the actual measured external forces. The deviation between the internal model forces predicted based on the current prior stiffness estimate and the force residuals: in, For the sake of residual strength, External forces borne by soft tissue The internal forces within soft tissue that resist deformation. To extend the predicted soft tissue heterogeneous stiffness field in the Kalman filter, For node position deformation, Let be the posterior covariance matrix of the parameters. The calculated Kalman gain is responsible for dynamically balancing the confidence levels of model predictions and sensor observations. This is to expand the parameter covariance matrix predicted in the Kalman filter.

[0039] It is worth noting that the updated parameter posterior covariance matrix Its diagonal elements explicitly quantify the uncertainty (i.e., variance) of the stiffness estimate for each small unit in the network. This step extends the original single value estimate to the variance space, providing a mathematical basis for the robot to know where uncertainty still exists. Regions with high variance are considered to be potentially high-stiffness regions.

[0040] To further improve the real-time performance of the system's online operation, at critical nodes such as significant topological changes in the organization, the standard recursive form will be transformed into one based on an information matrix. The batch update strategy avoids frequent matrix inversion operations at each time step. The expression for the batch update strategy is as follows: in, This is the information matrix at time k+t after the topological change. This is the information matrix at time k. For the observation matrix, This is the inverse matrix of the observation noise covariance.

[0041] In this embodiment, by performing target identification in the extended Kalman filter variance space, the intra-class variance is effectively reduced, and the interference of low-stiffness artifacts (false positives) caused by observation noise is suppressed to the greatest extent, resulting in superior statistical separability of high-stiffness targets.

[0042] Step 140: Establish an active sensing objective function for separated weighted strain based on the updated posterior covariance matrix of stiffness parameters; In some embodiments, establishing the active sensing objective function for separating weighted strain based on the updated stiffness parameter posterior covariance matrix includes: Based on the updated posterior covariance matrix of stiffness parameters, the stiffness estimation uncertainty value of each mesh element in the finite element mesh model is extracted; Based on the uncertainty value of stiffness estimation for each mesh element, mesh elements with uncertainty values ​​greater than a preset threshold are selected as a set of suspected high-stiffness elements. The physical strain generated by the preset interactive action in each grid cell of the suspected high stiffness element set is calculated. The stiffness estimation uncertainty of each grid cell in the suspected high stiffness element set is coupled with the physical strain of each grid cell in the suspected high stiffness element set to establish an active sensing objective function of separated weighted strain.

[0043] After obtaining the posterior distribution and uncertainty (variance matrix) of the stiffness field After that, the robot enters the active perception planning stage. Simply relying on open-loop contact often fails to elicit the mechanical characteristics of deep, high-stiffness regions due to insufficient deformation. To guide the robot to actively expose these hidden features, the next interaction action that generates the maximum information gain must be optimized. Since traditional objective functions based on minimizing global information entropy suffer from severe computational ill-conditioning and gradient vanishing problems, this application innovatively proposes a physically inspired heuristic objective function—SWS ​​(Separation-Weighted Strain). The resulting parameter uncertainty (variance) is coupled with the physical strain generated by the proposed action, and a spatial attention bias is applied.

[0044] In some embodiments, the objective function is calculated using the following formula: in, Denotes the Frobenius norm of a matrix. Let be the posterior covariance matrix of the parameters. To determine based on candidate actions The attention-weighted strain matrix generated in the prediction is... To predict what will happen in the first The principal strain vector induced by each element. Spatial attention weights designed specifically for this purpose For the first 1D unit vector It is an artificially set exploration bias factor. It is a constant. This is an indicator function that returns 1 when the condition is met. It is a collection of suspected high-stiffness elements.

[0045] This mechanism causes the analytical gradient of the objective function to be strongly biased towards actions that produce drastic deformation in lesion areas that are "highly suspicious and have high variance." This not only reduces ineffective repeated traction on unrelated healthy tissues, but also quickly and statistically separates genuine sclerotic lesions from false positive areas caused by noise.

[0046] In this embodiment, by designing a separate weighted strain objective function, the robot's movements can be guided to more precisely focus on highly suspicious lesion areas and apply effective deformation excitation. This allows areas that are misjudged as lesions due to noise to rapidly reduce variance and be eliminated, while the true high-stiffness areas are significantly enhanced. This mechanism maximizes information acquisition with minimal physical contact while reducing blind exploration and extensive traction of healthy tissue, thus improving the safety of minimally invasive procedures.

[0047] Step 150: Solve the active perception objective function based on the gradient descent method to obtain the optimal interaction action between the robot and the probabilistic evolution model; In some embodiments, solving the active perception objective function based on gradient descent to obtain the optimal interaction action between the robot and the probabilistic evolution model includes: The robot's kinematic and dynamic constraints are used as the solution boundary for the active perception objective function; Within the solution boundary, the active perception objective function is iteratively calculated using the gradient descent method to select interactive action combinations that meet the solution boundary. Calculate the separate weighted strain information gain for each interaction action in the combination of interaction actions, and take the interaction action with the largest separate weighted strain information gain as the optimal interaction action between the robot and the probabilistic evolution model.

[0048] With a clear objective function established, the robot's execution instructions are determined by optimizing the solver, forming a control closed loop. Based on the constructed SWS objective function, the optimal interactive actions under robot kinematic and dynamic constraints are solved using the gradient descent method. The calculation formula is as follows: in, This is the optimal interactive action.

[0049] The solution It includes the optimal spatial coordinates of the robot's contact with the tissue at the next moment, as well as the direction and magnitude of the contact force.

[0050] Step 160: Send the optimal interactive action to the robot's controller, control the robot to perform the optimal interactive action on the target flexible tissue, obtain the stiffness distribution map of the target flexible tissue, determine whether the uncertainty variance of the high stiffness lesion region in the stiffness distribution map is less than the safe convergence threshold. If the uncertainty variance is less than the safe convergence threshold, output the identification result of the high stiffness lesion region of the target flexible tissue. If the uncertainty variance is greater than or equal to the safe convergence threshold, return to the step of collecting multimodal observation data of the probability evolution model and robot interaction for iteration until the uncertainty variance is less than the safe convergence threshold.

[0051] Furthermore, the solution will be obtained The data is sent to the robot's underlying servo controller for physical execution. Subsequently, it is evaluated whether the uncertainty variance of the high-stiffness lesion region in the current stiffness distribution map has decreased below a preset safe convergence threshold. If the threshold has not been reached, deformation and force data after executing the new action are re-acquired, and the process returns to step 120 to begin the next round of "perception-planning-execution" iteration. If convergence has been achieved, it means that the boundary and parameters of the implicit high-stiffness region have been fully activated and clearly identified, ending the active exploration phase and outputting the final high-precision 3D stiffness distribution map for use by doctors or subsequent cutting algorithms.

[0052] Figure 2 This is the second flowchart illustrating the active identification method for soft tissue lesion regions based on separation-weighted strain provided in this application embodiment, as follows: Figure 2 As shown, the method includes the following steps: S1. The target flexible tissue in the surgical scenario is discretized into a grid, and the construction and initialization configuration of the soft tissue finite element mesh model are completed. The basic discretization representation of the tissue stiffness field is established, providing a mesh carrier for subsequent mechanical calculations and state estimation.

[0053] S2. Perform interactive operation to acquire multimodal observation data and control the surgical robot end effector to perform physical interaction operation with the meshed flexible tissue; synchronously collect multimodal observation data, including visual deformation node data of the flexible tissue surface and contact force data when the robot end effector contacts the tissue.

[0054] S3. Construct force equilibrium equations and use differentiable projection dynamics for local linearization calculation of effective observation matrix and noise. Based on the quasi-static force equilibrium principle, combine the collected visual deformation node data and contact force data to construct the tissue force equilibrium equations; use differentiable projection dynamics to perform local linearization processing on the internal force functions in the force equilibrium equations, calculate the effective observation matrix for solving the adaptive stiffness field, and simultaneously calculate the composite observation noise parameters that fuse visual and force perception.

[0055] S4. Update the stiffness field distribution of soft tissue and explicitly quantify the posterior covariance. Use extended Kalman filtering to estimate the state of the linearized force equilibrium equation; calculate the force residual between the measured contact force and the model predicted internal force, and iteratively update the heterogeneous stiffness field distribution of the flexible tissue based on the force residual; at the same time, explicitly quantize and output the posterior covariance matrix corresponding to the stiffness field to characterize the degree of uncertainty of the stiffness estimation of each grid element.

[0056] S5. Based on the uncertainty values ​​of each unit in the posterior covariance matrix, regions with higher uncertainty are selected as suspected high-stiffness anomaly regions. Taking the suspected high-stiffness anomaly regions as the core, an active perception objective function based on separated weighted strain is constructed to guide the planning of the robot's subsequent optimal interactive actions.

[0057] S6. Active perception optimization planning: Maximize the information gain of the separated weighted strain, and solve for the optimal operation position and direction at the next moment. With maximizing the information gain of the separated weighted strain as the optimization objective, active perception optimization planning is performed. Within the robot motion constraint range, the optimal operation action at the next moment is obtained. The optimal operation action includes the optimal contact position, force direction and force parameters of the robot's interaction with the tissue.

[0058] S7. Convergence Judgment: Determine whether the confidence level of the current distribution of high stiffness regions in flexible tissue has converged and meets the preset identification conditions. If it has not converged and does not meet the preset conditions, return to step S2 and execute a new round of interactive observation, mechanical calculation, state estimation and action planning to perform closed-loop iterative identification. If it has converged and meets the preset conditions, output the final high stiffness region identification result and complete the active identification process of flexible tissue lesion regions.

[0059] The active identification method for soft tissue lesion regions based on separation-weighted strain provided in this application can be executed by a device for active identification of soft tissue lesion regions based on separation-weighted strain. This application uses the example of a device for active identification of soft tissue lesion regions based on separation-weighted strain executing the active identification method for soft tissue lesion regions based on separation-weighted strain to illustrate the device provided in this application.

[0060] This application also provides an active identification device for soft tissue lesion regions based on separation-weighted strain, such as... Figure 3 As shown, the active identification device for soft tissue lesion areas based on separation weighted strain includes: an acquisition module 310, a first processing module 320, a second processing module 330, a third processing module 340, a fourth processing module 350, and a fifth processing module 360.

[0061] The acquisition module 310 is used to acquire the target flexible structure, perform spatial discretization modeling on the target flexible structure, and obtain the probabilistic evolution model of the stiffness field state transition of the target flexible structure. The first processing module 320 is used to collect multimodal observation data of the interaction between the probabilistic evolution model and the robot, and to establish the force balance equation of the target flexible tissue; The second processing module 330 is used to estimate and predict the force equilibrium equation based on the extended Kalman filter to obtain the updated stiffness parameter posterior covariance matrix. The third processing module 340 is used to establish an active sensing objective function for separating weighted strain based on the updated stiffness parameter posterior covariance matrix. The fourth processing module 350 is used to solve the active perception objective function based on the gradient descent method to obtain the optimal interaction action between the robot and the probabilistic evolution model. The fifth processing module 360 ​​is used to send the optimal interactive action to the robot's controller, control the robot to perform the optimal interactive action on the target flexible tissue, obtain the stiffness distribution map of the target flexible tissue, determine whether the uncertainty variance of the high stiffness lesion region in the stiffness distribution map is less than the safe convergence threshold, if the uncertainty variance is less than the safe convergence threshold, output the identification result of the high stiffness lesion region of the target flexible tissue, if the uncertainty variance is greater than or equal to the safe convergence threshold, return to the step of collecting multimodal observation data of the probability evolution model and robot interaction for iteration, until the uncertainty variance is less than the safe convergence threshold.

[0062] The active identification method for soft tissue lesions based on separated weighted strain provided in this application treats the identification of high-stiffness lesions as a dynamic sequential decision-making and state estimation process. Within the variance space, extended Kalman filtering combined with differentiable projection dynamics is used to achieve real-time tracking and uncertainty quantification of the heterogeneous stiffness field of flexible tissue. Furthermore, separated weighted strain is introduced as the active sensing objective function, enabling the method to autonomously plan the next optimal interaction action. This more accurately generates local physical strain in suspected lesion areas with the highest classification uncertainty, thereby maximizing stiffness information gain and rapidly separating hidden high-stiffness abnormal regions.

[0063] The soft tissue lesion region active identification device based on separation-weighted strain provided in this application embodiment can achieve... Figures 1 to 2 The various processes implemented in the embodiment of the active identification method for soft tissue lesion regions based on separation weighted strain will not be described in detail here to avoid repetition.

[0064] In some embodiments, such as Figure 4 As shown, this application embodiment also provides an electronic device 400, including a processor 401, a memory 402, and a computer program stored in the memory 402 and executable on the processor 401. When the program is executed by the processor 401, it implements the various processes of the above-described active identification method for soft tissue lesion areas based on separation weighted strain, and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0065] It should be noted that the electronic devices in the embodiments of this application include the mobile electronic devices and non-mobile electronic devices described above.

[0066] This application also provides a non-transitory computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the various processes of the above-described active identification method for soft tissue lesion areas based on separation-weighted strain, and achieves the same technical effect. To avoid repetition, it will not be described again here.

[0067] The processor is the processor in the electronic device described in the above embodiments. The readable storage medium includes computer-readable storage media, such as computer read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk.

[0068] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method for active identification of soft tissue lesion regions based on separation-weighted strain.

[0069] The processor is the processor in the electronic device described in the above embodiments. The readable storage medium includes computer-readable storage media, such as computer read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk.

[0070] This application also provides a chip, which includes a processor and a communication interface. The communication interface and the processor are coupled. The processor is used to run programs or instructions to implement the various processes of the above-described active identification method for soft tissue lesion areas based on separation weighted strain, and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0071] It should be understood that the chip mentioned in the embodiments of this application may also be referred to as a device-level chip, device chip, chip device, or on-chip device chip, etc.

[0072] It should be noted that, in this document, 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 a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. Furthermore, it should be noted that the scope of the methods and apparatuses in the embodiments of this application is not limited to performing functions in the order shown or discussed, but may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.

[0073] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a computer software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, or network device, etc.) to execute the active identification method for soft tissue lesion regions based on separation weighted strain of the various embodiments of this application.

[0074] In the description of this application, "first feature" and "second feature" may include one or more of the features.

[0075] In the description of this application, "multiple" means two or more.

[0076] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims, and all of these forms are within the protection scope of this application.

[0077] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "illustrative embodiment," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0078] Although embodiments of this application have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of this application, the scope of which is defined by the claims and their equivalents.

Claims

1. A method for active identification of soft tissue lesion regions based on separation-weighted strain, characterized in that, The method includes: Obtain the target flexible structure, perform spatial discretization modeling on the target flexible structure, and obtain the probabilistic evolution model of the stiffness field state transition of the target flexible structure; Multimodal observation data of probabilistic evolution model and robot interaction were collected to establish the force balance equation of the target flexible tissue; The force equilibrium equation is estimated and predicted based on the extended Kalman filter to obtain the updated posterior covariance matrix of stiffness parameters. An active sensing objective function for separating weighted strain is established based on the updated posterior covariance matrix of stiffness parameters. The optimal interaction action between the robot and the probabilistic evolution model is obtained by solving the active perception objective function based on the gradient descent method. The optimal interactive action is sent to the robot's controller, which controls the robot to perform the optimal interactive action on the target flexible tissue. This yields a stiffness distribution map of the target flexible tissue. The uncertainty variance of the high-stiffness lesion region in the stiffness distribution map is then determined to be less than the safe convergence threshold. If the uncertainty variance is less than the safe convergence threshold, the identification result of the high-stiffness lesion region of the target flexible tissue is output. If the uncertainty variance is greater than or equal to the safe convergence threshold, the process returns to the step of collecting multimodal observation data of the probabilistic evolution model and the robot interaction for iteration until the uncertainty variance is less than the safe convergence threshold.

2. The method for active identification of soft tissue lesion regions based on separation-weighted strain according to claim 1, characterized in that, The process of acquiring the target flexible structure, spatially discretizing and modeling the target flexible structure to obtain a probabilistic evolution model of the stiffness field state transition of the target flexible structure includes: Obtain the target flexible structure and divide the target flexible structure into a finite element mesh model including multiple nodes and multiple triangular elements; The stiffness inside a single mesh element in the finite element mesh model is set to be uniformly distributed, and the heterogeneous stiffness field of the target flexible tissue is constructed as a high-dimensional random variable. A stiffness field state transition model is established based on high-dimensional random variables, resulting in a probabilistic evolution model for the stiffness field state transition of the target flexible tissue.

3. The method for active identification of soft tissue lesion regions based on separation-weighted strain according to claim 1, characterized in that, The multimodal observation data obtained from the acquisition of the probabilistic evolution model and robot interaction are used to establish the force balance equation of the target flexible tissue, including: Multimodal observation data of the interaction between the probabilistic evolution model and the robot are collected. The multimodal observation data includes surface node deformation data of the target flexible tissue and contact force data of the robot end effector. Stiffness field parameters of the target flexible tissue are determined based on a probabilistic evolution model; Based on the surface node deformation data of the target flexible tissue and the contact force data of the robot end, an observation matrix is ​​established by linearizing the differentiable projection dynamics, and the composite effective observation noise data corresponding to the observation matrix is ​​calculated. Based on stiffness field parameters, observation matrix, and composite effective observation noise data, the force balance equation of the target flexible tissue is established.

4. The active identification method for soft tissue lesion regions based on separation-weighted strain according to claim 3, characterized in that, The method of estimating and predicting the force equilibrium equation based on extended Kalman filtering to obtain the updated posterior covariance matrix of stiffness parameters includes: The internal force prediction values ​​are obtained by estimating and predicting the force equilibrium equations based on the extended Kalman filter. Calculate the force residual between the external contact force and the predicted internal force, and solve the Kalman gain of the force equilibrium equation based on the force residual; The heterogeneous stiffness field distribution of the target flexible tissue is iteratively updated based on the force residual and Kalman gain to obtain the updated posterior covariance matrix of stiffness parameters.

5. The active identification method for soft tissue lesion regions based on separation-weighted strain according to claim 4, characterized in that, The active sensing objective function for separating weighted strain is established based on the updated posterior covariance matrix of stiffness parameters, including: Based on the updated posterior covariance matrix of stiffness parameters, the stiffness estimation uncertainty value of each mesh element in the finite element mesh model is extracted; Based on the uncertainty value of stiffness estimation for each mesh element, mesh elements with uncertainty values ​​greater than a preset threshold are selected as a set of suspected high-stiffness elements. The physical strain generated by the preset interactive action in each grid cell of the suspected high stiffness element set is calculated. The stiffness estimation uncertainty of each grid cell in the suspected high stiffness element set is coupled with the physical strain of each grid cell in the suspected high stiffness element set to establish an active sensing objective function of separated weighted strain.

6. The active identification method for soft tissue lesion regions based on separation-weighted strain according to claim 5, characterized in that, The formula for calculating the objective function is as follows: in, Denotes the Frobenius norm of a matrix. Let be the posterior covariance matrix of the parameters. To determine based on candidate actions The attention-weighted strain matrix generated in the prediction is... Here, k represents the spatial attention weights, and k represents the time step. The objective function is denoted as .

7. The active identification method for soft tissue lesion regions based on separation-weighted strain according to claim 6, characterized in that, The method of solving the active perception objective function based on gradient descent to obtain the optimal interaction action between the robot and the probabilistic evolution model includes: The robot's kinematic and dynamic constraints are used as the solution boundary for the active perception objective function; Within the solution boundary, the active perception objective function is iteratively calculated using the gradient descent method to select interactive action combinations that meet the solution boundary. Calculate the separate weighted strain information gain for each interaction action in the combination of interaction actions, and take the interaction action with the largest separate weighted strain information gain as the optimal interaction action between the robot and the probabilistic evolution model.

8. A device for active identification of soft tissue lesion regions based on separation-weighted strain, implemented using the active identification method for soft tissue lesion regions based on separation-weighted strain as described in any one of claims 1 to 7, characterized in that, The device includes: The acquisition module is used to acquire the target flexible structure, perform spatial discretization modeling on the target flexible structure, and obtain the probabilistic evolution model of the stiffness field state transition of the target flexible structure. The first processing module is used to collect multimodal observation data of the interaction between the probabilistic evolution model and the robot, and to establish the force balance equation of the target flexible tissue. The second processing module is used to estimate and predict the force equilibrium equation based on the extended Kalman filter to obtain the updated stiffness parameter posterior covariance matrix. The third processing module is used to establish an active sensing objective function for separating weighted strain based on the updated posterior covariance matrix of stiffness parameters. The fourth processing module is used to solve the active perception objective function based on the gradient descent method to obtain the optimal interaction action between the robot and the probabilistic evolution model. The fifth processing module sends the optimal interactive action to the robot's controller, controls the robot to perform the optimal interactive action on the target flexible tissue, obtains the stiffness distribution map of the target flexible tissue, and determines whether the uncertainty variance of the high-stiffness lesion region in the stiffness distribution map is less than the safe convergence threshold. If the uncertainty variance is less than the safe convergence threshold, the identification result of the high-stiffness lesion region of the target flexible tissue is output. If the uncertainty variance is greater than or equal to the safe convergence threshold, the process returns to the step of collecting multimodal observation data of the probabilistic evolution model and the robot interaction for iteration until the uncertainty variance is less than the safe convergence threshold.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the active identification method for soft tissue lesion regions based on separation-weighted strain as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the active identification method for soft tissue lesion regions based on separation-weighted strain as described in any one of claims 1 to 7.

Images