Puncture navigation method and system guided by light vision

Abstract

The invention discloses a visual guidance puncture navigation method and system. The method comprises the following steps: defining a double-curvature optical projection plane equation in a non-uniform optical field model; performing multi-mode optical mark tracking to obtain an optimized double-curvature optical projection plane equation; according to a preset puncture target and obstacle information collected in real time, constructing a path functional under a multi-constraint condition in the non-uniform optical field model; solving a path functional by using an improved gradient descent algorithm, and generating a puncture path with optimal characteristics; according to the error between the actual needle inserting track and the planned puncture path, a feedback correction value is constructed, and the puncture projection is corrected. The system comprises an optical field model construction module, a multi-mode optical mark tracking module, a path functional construction module, a puncture path calculation module and a puncture projection correction module. The precise optical field model can be constructed, and high-precision and stable puncture navigation is realized in combination with an optimized path planning algorithm and self-adaptive feedback.

Description

Technical Field

[0001] The present invention relates to the technical field of puncture navigation, and particularly to a light vision-guided puncture navigation method and system. Background Art

[0002] Puncture navigation technology has extensive application value in precision operation fields (such as industrial manufacturing, minimally invasive operation, robot guidance, etc.). Its core goal is to optimize the puncture path in a complex environment, improve the puncture accuracy, and reduce the risk of misoperation caused by path deviation. Traditional puncture navigation methods mainly rely on methods such as image guidance, mechanical control, and sensing feedback. However, these methods still face many challenges in path optimization, real-time feedback, and environmental adaptability, which limit their wide application.

[0003] The current puncture navigation methods are mainly divided into the following categories:

[0004] (1) Image-guided puncture navigation, which uses CT, MRI, or ultrasound images to image the target area and combines preoperative planning to calculate the optimal puncture path. The main advantage of this method is to provide high-resolution anatomical information, but its real-time performance is poor, and there are data registration errors in the image reconstruction process. In addition, the volume of the imaging device is large, and it is not suitable for portable or micro-space operation scenarios.

[0005] (2) Machine force feedback control-based puncture navigation, which combines a force sensor and an intelligent control algorithm to optimize the puncture trajectory by adjusting the puncture force and the needle insertion depth in real time. This type of method improves the path accuracy, but has a strong dependence on the hardware of the device, high cost, and there is an error accumulation problem in the force feedback in non-rigid media.

[0006] (3) Optically tracked puncture navigation, which obtains the relative position between the device and the target through markers and optical sensors to achieve path planning and real-time adjustment. This method has the characteristics of non-contact and strong real-time performance, but still faces technical bottlenecks in aspects such as environmental light interference, marker occlusion, and compensation for light field distortion.

[0007] The existing image-guided methods mentioned above have coordinate registration errors during data acquisition and processing. Especially during multi-modal fusion, there are deficiencies in image distortion and spatial transformation accuracy, resulting in puncture path deviation. Traditional path planning methods usually adopt geometric models or heuristic algorithms, lacking adaptability to complex environments and making it difficult to achieve optimal path calculation under multiple constraints. In addition, the lack of a real-time feedback mechanism for dynamic environmental changes leads to lag in path adjustment and affects operation stability. Due to the non-uniform distribution of the light field and the interference of ambient light, the tracking accuracy of optical markers is limited. Especially for optical projection calculations on curved or irregular surfaces, there are large errors, which affect the stability and reliability of puncture navigation. Most existing puncture navigation systems perform punctures based on a preset path, but lack an intraoperative error feedback and adaptive correction mechanism, resulting in the puncture path possibly deviating from the planned trajectory during actual operation and affecting the final operation accuracy. Summary of the Invention

[0008] In view of this, the purpose of the embodiments of the present invention is to provide a puncture navigation method and system based on optical vision guidance, which can overcome the deficiencies of the prior art in puncture accuracy, environmental adaptability, and real-time feedback by optimizing the path planning and real-time correction mechanism of optical vision guidance, and provide a more efficient and accurate navigation method for the field of precision operations.

[0009] The embodiments of the present invention are implemented as follows:

[0010] A puncture navigation method based on optical vision guidance, which includes:

[0011] Using the position data of optical markers in the device coordinate system, construct a non-uniform optical field model, and define a bi-curvature optical projection surface equation describing the spatial light field distribution in the non-uniform optical field model.

[0012] Introduce a filtering response constraint into the bi-curvature optical projection surface equation to perform multi-modal optical marker tracking and obtain an optimized bi-curvature optical projection surface equation.

[0013] According to the preset puncture target and the obstacle information collected in real time, construct a path functional under multiple constraints in the non-uniform optical field model.

[0014] Use an improved gradient descent algorithm to solve the path functional and generate a puncture path with optimal characteristics.

[0015] According to the error between the actual needle insertion trajectory and the planned puncture path, construct a feedback correction amount to correct the puncture projection.

[0016] In a preferred embodiment of the present invention, in the above-mentioned optical vision-guided puncture navigation method, the non-uniform optical field model is constructed by using the position data of the optical markers in the device coordinate system. The hyperbolic curvature optical projection plane equation for describing the spatial light field distribution defined in the non-uniform optical field model includes:

[0017] Set the device coordinate system O-XYZ, where O is the center of the device, the Z-axis direction is perpendicular to the device surface and consistent with the puncture direction, and the XY axes are parallel to the device surface, defining the horizontal distribution of the optical field.

[0018] Collect the position data of the optical markers in the device coordinate system where (x k , y k , z k ) are three-dimensional space coordinates, α k is the reflection intensity weight, α k ∈[0, 1], n is the number of optical markers distributed on the device surface, and k is the marker.

[0019] Adopt a Gaussian weighting function to convert the discrete optical marker data into a continuous light field, and construct a non-uniform optical field model for describing the light field intensity at the spatial point (x, y, z) where σ z is the depth-adaptive Gaussian kernel width, σ z =β·(z max -z), β is the adjustment coefficient, and z max is the maximum detection depth.

[0020] Adopt the hyperbolic tangent function to adjust the light field distribution in the depth direction where tanh(·) is the hyperbolic tangent function, γ is the curvature adjustment parameter, and the hyperbolic curvature optical projection plane equation is obtained

[0021] Its technical effects are as follows: The discrete optical marker data is converted into a continuous light field by using a Gaussian weighting function, avoiding the interpolation error caused by sparse marker points in the traditional method, making the light field intensity distribution smoother and more continuous, and improving the accuracy of optical projection surface modeling; By using a depth-adaptive Gaussian kernel width, the light field model can automatically adjust the resolution according to the target depth, that is, providing more refined optical guidance in the shallower area, while improving stability in the deeper area and reducing the influence of signal attenuation; The hyperbolic tangent function is used to adjust the light field distribution in the depth direction. Through the nonlinear characteristics of the tanh function, the optical projection has a high resolution in the area close to the center of the device, while smoothly transitioning in the area far from the center, reducing distortion and improving the adaptability to complex human anatomical structures; By constructing a bi-curvature optical projection surface equation, the distribution characteristics of the optical projection in a non-uniform optical field can be more accurately described, providing accurate optical gradient information for subsequent puncture path optimization.

[0022] In a preferred embodiment of the present invention, in the above-mentioned light vision-guided puncture navigation method, introducing a filtering response constraint into the bi-curvature optical projection surface equation and performing multi-modal optical marker tracking to obtain an optimized bi-curvature optical projection surface equation includes:

[0023] Introducing a filtering response constraint into the bi-curvature optical projection surface equation and performing multi-modal optical marker tracking

[0024] Wherein, is the correlation filtering response map, represents the matching degree between the input image and the filtering template, and is calculated by the inverse Fourier transform F -1 Calculated.

[0025] is the filtering template The conjugate complex number of, is the filtering template, is the Fourier transform of the input image, is the conjugate Fourier transform of the target template, and λ is the regularization coefficient.

[0026] is the maximum matching value of the correlation filtering.

[0027] is the L2 norm of the filtering template, used to normalize the response intensity.

[0028] Obtain the optimized bi-curvature optical projection surface equation

[0029] Its technical effects are as follows: By calculating the matching degree between the input image and the filtering template through correlation filtering, a filtering response map is obtained, which can accurately identify the position of the optical marker in a complex environment and improve the recognition accuracy of the optical marker; Using Fourier transform and inverse transform to calculate the matching degree enables marker tracking to not only rely on spatial domain information but also combine frequency domain features, reducing noise interference and enhancing the stability and robustness of the optical marker; It is applicable not only to single-modal optical markers but also supports multi-modal fusion, and can stably track optical markers under different lighting, tissue depth, and environmental change conditions; Through the optimized double-curvature optical projection plane equation, the information of multiple optical markers is fused in a unified projection coordinate system, improving the adaptability to complex surgical environments and ensuring the stability of optical guidance; Using the method of correlation filtering to calculate the marker matching degree has a lower computational complexity compared to traditional template matching or deep learning-based methods and is applicable to real-time puncture navigation; Due to the use of Fourier transform to accelerate the calculation, large-scale image matching can be completed in a short time, enhancing the response speed of optical marker tracking, enabling the intraoperative navigation system to quickly adjust the puncture path, and improving surgical efficiency.

[0030] In a preferred embodiment of the present invention, in the above-mentioned puncture navigation method guided by optical vision, the construction of the path functional under multiple constraints in the non-uniform optical field model according to the preset puncture target and the obstacle information collected in real time includes:

[0031] Based on the optimized double-curvature optical projection plane equation S″(x, y, z), calculate the gradient field of the spatial light field to obtain the gradient vector field of the optical field Wherein, Reflects the optical gradient change in the X direction, Reflects the optical gradient change in the Y direction, Reflects the optical gradient change in the Z direction.

[0032] According to the preset puncture target and the obstacle information collected in real time, set the optical gradient guidance term Path curvature constraint term w2κ 2 、Obstacle avoidance term w3d -2 And time dynamic speed constraint term Wherein, w1 is the weight factor of the optical gradient constraint, κ is the path curvature, w2 is the curvature control weight factor, d is the distance from the current path to the nearest obstacle, w3 is the obstacle avoidance weight factor, λ(t) is the time dynamic weight function, Is the instantaneous speed of the puncture path Γ at time t, V max Is the preset maximum allowable puncture speed.

[0033] Construct the path functional

[0034] Its technical effects are as follows: The optical gradient guidance term makes the puncture path follow the direction of the rising optical gradient, guiding the puncture needle towards the optimal navigation path and improving the navigation accuracy. Since the optical field gradient information is derived from the optimized double-curvature optical projection plane equation, the path planning can be adaptively adjusted based on the dynamic optical environment to ensure that the puncture path always points towards the target area and reduce the deviation caused by tissue deformation or optical signal interference. The path curvature constraint term controls the smoothness of the path, preventing unreasonable sharp turns in the path, making the puncture trajectory more in line with physiological requirements, improving the feasibility and safety of surgical operations. By adjusting the curvature control weight factor, the compliance of the path can be flexibly adjusted according to surgical needs, suitable for different types of puncture tasks. The obstacle avoidance term keeps the puncture path away from obstacles, using the dynamic distance from the path to the nearest obstacle as the constraint index to make the obstacle avoidance control more intelligent and capable of adapting to changes in the intraoperative environment in real time, such as minor patient movements or tissue deformations. The time-dynamic speed constraint term controls the puncture speed to ensure the dynamic smoothness of the path, avoiding tissue damage or path tracking errors caused by too fast speed. By adjusting the puncture speed through the time-dynamic weight function, the puncture rate can be adaptively adjusted when the puncture path enters different tissue regions, such as accelerating when entering soft tissues and decelerating when approaching the target area, improving the stability and safety of the puncture.

[0035] In a preferred embodiment of the present invention, in the above-mentioned puncture navigation method guided by optical vision, the step of using the improved gradient descent algorithm to solve the path functional and generate a puncture path with optimal characteristics includes:

[0036] Using the improved gradient descent algorithm to establish a path update formula Γ k is the current puncture path, η is the learning rate, is the path gradient term, D cos is the anatomical structure matching term, and μ is the weight controlling the strength of the anatomical matching constraint.

[0037] Output the optimized path direction to obtain a puncture path Γ with optimal characteristics opt .

[0038] Its technical effects are as follows: The gradient descent strategy is adopted to optimize the path, and the path update is guided by the path gradient term, so that the puncture path is gradually optimized towards the global optimal solution, improving the computational efficiency of path planning. The optimized step size is dynamically adjusted through the improved learning rate. In the initial stage of path optimization, a larger step size speeds up the convergence speed. When the path is close to the optimal solution, a smaller step size avoids oscillation and improves the refined control ability of the path.

[0039] In a preferred embodiment of the present invention, in the above-mentioned puncture navigation method guided by optical vision, the path gradient term The calculation formula of is:

[0040] Among them, is the second-order gradient of the light field, which is used for path adjustment.

[0041] In a preferred embodiment of the present invention, in the above-mentioned light vision-guided puncture navigation method, the anatomical structure matching item D cos The calculation formula is:

[0042]

[0043] Among them, Φ CT (Γ k ) is the CT image feature vector corresponding to the current puncture path.

[0044] Φ MR is the preoperative MRI image reference feature vector.

[0045] ||Φ CT (Γ k )|| is the L2 norm of the CT image feature vector.

[0046] ||Φ MR || is the L2 norm of the preoperative MRI image reference feature vector.

[0047] In a preferred embodiment of the present invention, in the above-mentioned light vision-guided puncture navigation method, the construction of the feedback correction amount according to the error between the actual needle insertion trajectory and the planned puncture path, and the correction of the puncture projection include:

[0048] Obtain the needle insertion trajectory Γ real in real time, and calculate the overall trajectory error between it and the planned puncture path Γ opt Among them, L is the total path length. Among them, L is the total path length.

[0049] Based on the gradient of the overall trajectory error, combined with Gaussian smoothing filtering and light field gradient information, calculate the error feedback correction amount Among them, p i is the actual position of the i-th sampling point, is the gradient of the error point, G σ (p i ) is the Gaussian smoothing filter kernel, η′ is the feedback adjustment step size, v d is the preset puncture direction, represents the projection error direction.

[0050] Use the error feedback correction amount ΔP to adjust the intraoperative projection, and use the spatial transformation network to calculate the dynamic projection correction distribution Λ′(x,y) = STN(Λ(x,y)|θ)·diag(1 + α stn), where Λ(x, y) is the original optical projection distribution, STN(Λ(x, y)|θ) is the spatial transformation network, θ is the STN transformation parameter, and α stn is the forehead deformation compensation term.

[0051] Use the dynamic projection correction distribution Λ′(x, y) to update the optically guided puncture projection, and recalculate the gradient vector field ▽S″ of the optical field according to the corrected puncture projection to obtain an optimized puncture path.

[0052] Its technical effects are as follows: During the operation, due to tissue elasticity, operation errors, or external interference, the needle insertion trajectory may deviate from the original planned path. By constructing the overall trajectory error, the deviation between the needle insertion trajectory and the planned path can be quantified, providing accurate data for subsequent correction. Through the error gradient information, the local error trend is calculated, making the correction process more accurate, rather than a global linear adjustment, thereby reducing the accumulation of local errors; by introducing a Gaussian smoothing filter kernel to process the error gradient data, high-frequency noise is removed, making the error feedback correction amount smoother and avoiding path instability caused by drastic adjustments. By adjusting the feedback step size to control the error correction intensity, the path adjustment is made more stable, avoiding new errors caused by overcorrection; using a spatial transformation network, the dynamic projection correction distribution is calculated using the error feedback correction amount, and the optical projection is dynamically adjusted, which can adjust the optical projection in real time, making the projection information always align with the target path and improving the accuracy of intraoperative guidance.

[0053] In a preferred embodiment of the present invention, in the above optically vision-guided puncture navigation method, the update formula for the STN transformation parameter θ is θ = θ0 + k θ ·ΔP, where θ0 is the initial affine transformation parameter and k θ is the learning rate.

[0054] The forehead deformation compensation term α stn has the following calculation formula where is the optical projection intensity of the current tissue pixel point, is the position of the corresponding point in the preoperative image, which is used to calculate the influence of intraoperative tissue deformation on the optical projection and compensate for non-linear distortion.

[0055] An optically vision-guided puncture navigation system, which includes:

[0056] An optical field model construction module, which is used to construct a non-uniform optical field model using the position data of optical markers in the device coordinate system, and define a bi-curvature optical projection surface equation describing the spatial light field distribution in the non-uniform optical field model.

[0057] A multimodal optical marker tracking module is used to introduce a filtering response constraint into the bi-curvature optical projection plane equation, perform multimodal optical marker tracking, and obtain an optimized bi-curvature optical projection plane equation.

[0058] A path functional construction module is used to construct a path functional under multiple constraints in the non-uniform optical field model according to a preset puncture target and real-time acquired obstacle information.

[0059] A puncture path calculation module is used to solve the path functional by using an improved gradient descent algorithm and generate a puncture path with optimal characteristics.

[0060] A puncture projection correction module is used to construct a feedback correction amount according to the error between the actual needle insertion trajectory and the planned puncture path and correct the puncture projection.

[0061] The beneficial effects of the embodiments of the present invention are as follows:

[0062] The present invention constructs a non-uniform optical field model by using optical marker position data and adopts a bi-curvature optical projection plane equation, so that the optical projection can more accurately describe the spatial light field distribution, improving the adaptability to complex tissue structures; the Gaussian weighting function is combined with the hyperbolic tangent function to adjust the light field distribution in the depth direction, enabling the optical field to dynamically adapt to different tissue depths and improving the applicability of the model; through non-uniformity modeling of the light field, the resolution ability of optical signals to the target area is improved, and the influence of light scattering on the navigation accuracy is reduced.

[0063] The present invention introduces a filtering response constraint and uses the method of correlation filtering for multimodal optical marker tracking, effectively enhancing the detection stability of the marker points and reducing the false recognition rate; the Fourier transform is used to calculate the filtering response map, making the marker tracking more adaptable to light changes, partial occlusion, and tissue deformation, improving the stability of the navigation system; the target matching degree is calculated through the inverse Fourier transform and combined with the L2 norm to normalize the response intensity, improving the calculation efficiency of the algorithm and realizing real-time navigation.

[0064] The present invention constructs a path functional including optical gradient guidance, path curvature constraint, obstacle avoidance constraint, and time dynamic speed constraint to ensure that the puncture path meets clinical requirements; the path functional includes an obstacle avoidance term, which can dynamically adjust the path to avoid important tissue structures such as blood vessels and bones, improving puncture safety; a time dynamic speed constraint term is introduced to automatically adjust the puncture speed according to the puncture depth and tissue characteristics, reducing tissue damage and improving the surgical success rate.

[0065] The present invention uses an improved gradient descent algorithm to solve the path functional, dynamically adjusts the puncture trajectory through the path update formula, and realizes optimal path planning; introduces an anatomical structure matching term, matches the CT image feature vector with the MRI image reference feature vector, makes the puncture path accurately coincide with the patient's anatomical structure, and improves the surgical accuracy; combines deep learning methods during the path optimization process to adaptively adjust the anatomical features and improve the robustness of path optimization.

[0066] The present invention calculates the feedback correction amount based on the error between the actual needle insertion trajectory and the planned path, combines Gaussian smoothing filtering and light field gradient information, dynamically corrects the puncture projection, and reduces error accumulation; uses STN to calculate the dynamic projection correction distribution, adaptively adjusts the optical guidance projection, and improves the accuracy of intraoperative navigation; corrects the influence of intraoperative tissue deformation on the optical projection through the frontal shape deformation compensation term, reduces the deviation caused by the patient's slight movement or tissue deformation, and improves the puncture accuracy. Brief Description of the Drawings

[0067] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the embodiments. It should be understood that the following drawings only show some embodiments of the present invention, and therefore should not be regarded as limiting the scope. For those of ordinary skill in the art, without creative efforts, other related drawings can also be obtained based on these drawings.

[0068] Figure 1 It is a flowchart of the puncture navigation method with optical vision guidance of the present invention. Detailed Embodiments

[0069] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Usually, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0070] Please refer to Figure 1, the first embodiment of the present invention provides a light vision-guided puncture navigation method, which includes: using the position data of optical markers in the device coordinate system to construct a non-uniform optical field model, and defining a bi-curvature optical projection plane equation describing the spatial light field distribution in the non-uniform optical field model; introducing a filtering response constraint into the bi-curvature optical projection plane equation to perform multi-modal optical marker tracking, and obtaining an optimized bi-curvature optical projection plane equation; constructing a path functional under multiple constraints in the non-uniform optical field model according to a preset puncture target and real-time collected obstacle information; using an improved gradient descent algorithm to solve the path functional to generate a puncture path with optimal characteristics; constructing a feedback correction amount according to the error between the actual needle insertion trajectory and the planned puncture path, and correcting the puncture projection.

[0071] In a preferred embodiment of the present invention, in the above light vision-guided puncture navigation method, the step of using the position data of optical markers in the device coordinate system to construct a non-uniform optical field model and defining a bi-curvature optical projection plane equation describing the spatial light field distribution in the non-uniform optical field model includes: setting a device coordinate system O-XYZ, where O is the center of the device, the Z-axis direction is perpendicular to the device surface and is consistent with the puncture direction, and the XY axes are parallel to the device surface to define the horizontal distribution of the optical field; collecting the position data of optical markers in the device coordinate system where, (x k , y k , z k ) are three-dimensional space coordinates, α k is the reflection intensity weight, α k ∈ [0, 1], n is the number of optical markers distributed on the device surface, and k is the marker; using a Gaussian weighting function to convert discrete optical marker data into a continuous light field, and constructing a non-uniform optical field model describing the light field intensity at the spatial point (x, y, z) where, σ z is the depth-adaptive Gaussian kernel width, σ z = β·(z max - z), β is an adjustment coefficient, and z max is the maximum detection depth; using a hyperbolic tangent function to adjust the light field distribution in the depth direction where, tanh(·) is the hyperbolic tangent function, and γ is the curvature adjustment parameter, to obtain the bi-curvature optical projection plane equation

[0072] Its technical effects are as follows: The Gaussian weighting function is used to convert discrete optical marker data into a continuous light field, avoiding the interpolation error caused by sparse marker points in traditional methods, making the light field intensity distribution smoother and more continuous, and improving the accuracy of optical projection surface modeling; The depth-adaptive Gaussian kernel width is adopted, enabling the light field model to automatically adjust the resolution according to the target depth, that is, providing more refined optical guidance in shallower areas and improving stability in deeper areas to reduce the influence of signal attenuation; The hyperbolic tangent function is used to adjust the light field distribution in the depth direction. Through the nonlinear characteristics of the tanh function, the optical projection has a high resolution in the area near the device center and smoothly transitions in the area far from the center, reducing distortion and improving the adaptability to complex human anatomical structures; By constructing the double-curvature optical projection surface equation, the distribution characteristics of the optical projection in a non-uniform optical field can be more accurately described, providing accurate optical gradient information for subsequent puncture path optimization.

[0073] In a preferred embodiment of the present invention, in the above-mentioned light vision-guided puncture navigation method, introducing a filtering response constraint into the double-curvature optical projection surface equation and performing multi-modal optical marker tracking to obtain an optimized double-curvature optical projection surface equation includes: introducing a filtering response constraint into the double-curvature optical projection surface equation and performing multi-modal optical marker tracking Wherein, is the correlation filtering response map, indicating the matching degree between the input image and the filtering template, which is calculated by the inverse Fourier transform F -1 ; is the filtering template is the conjugate complex number of is the filtering template, is the Fourier transform of the input image, is the conjugate Fourier transform of the target template, and λ is the regularization coefficient; is the maximum matching value of the correlation filtering; is the L2 norm of the filtering template, which is used to normalize the response intensity; obtaining the optimized double-curvature optical projection surface equation

[0074] Its technical effects are as follows: By calculating the matching degree between the input image and the filtering template through correlation filtering, a filtering response map is obtained, which can accurately identify the position of the optical marker in a complex environment and improve the recognition accuracy of the optical marker; Using Fourier transform and inverse transform to calculate the matching degree enables marker tracking to not only rely on spatial domain information but also combine frequency domain features, reducing noise interference and enhancing the stability and robustness of the optical marker; It is applicable not only to single-modal optical markers but also supports multi-modal fusion, and can stably track optical markers under different lighting, tissue depth, and environmental change conditions; Through the optimized double-curvature optical projection plane equation, the information of multiple optical markers is fused in a unified projection coordinate system, improving the adaptability to complex surgical environments and ensuring the stability of optical guidance; Using the method of correlation filtering to calculate the marker matching degree has a lower computational complexity compared to traditional template matching or deep learning-based methods and is applicable to real-time puncture navigation; Due to the use of Fourier transform for accelerated calculation, large-scale image matching can be completed in a short time, enhancing the response speed of optical marker tracking, enabling the intraoperative navigation system to quickly adjust the puncture path, and improving surgical efficiency.

[0075] In a preferred embodiment of the present invention, in the above-mentioned puncture navigation method guided by optical vision, the construction of the path functional under multiple constraints in the non-uniform optical field model according to the preset puncture target and the obstacle information collected in real time includes: Based on the optimized double-curvature optical projection plane equation S″(x, y, z), calculating the gradient field of the spatial light field to obtain the gradient vector field of the optical field where reflects the optical gradient change in the X direction, reflects the optical gradient change in the Y direction, reflects the optical gradient change in the Z direction; According to the preset puncture target and the obstacle information collected in real time, setting the optical gradient guidance term the path curvature constraint term w2κ 2 and the obstacle avoidance term w3d -2 and the time dynamic speed constraint term where w1 is the weight factor of the optical gradient constraint, κ is the path curvature, w2 is the curvature control weight factor, d is the distance from the current path to the nearest obstacle, w3 is the obstacle avoidance weight factor, λ(t) is the time dynamic weight function, is the instantaneous speed of the puncture path Γ at time t, V max is the preset maximum allowable puncture speed; where the optical gradient guidance term is used to ensure that the path follows the ascending direction of the optical gradient and improve the navigation accuracy; the path curvature constraint term is used to control the path smoothness and prevent excessive bending; the obstacle avoidance term is used to ensure that the path is far from the obstacle and improve safety; the time dynamic speed constraint term is used to control the puncture speed and ensure the path is stable; The path functional is constructed.

[0076] Its technical effects are as follows: The optical gradient guiding term makes the puncture path along the upward direction of the optical gradient, guiding the puncture needle towards the optimal navigation path, improving the navigation accuracy. Since the optical field gradient information is derived from the optimized double-curvature optical projection surface equation, the path planning can be adaptively adjusted based on the dynamic optical environment to ensure that the puncture path always points towards the target area and reduce the deviation caused by tissue deformation or optical signal interference; the path curvature constraint term controls the smoothness of the path, prevents unreasonable sharp turns in the path, makes the puncture trajectory more in line with physiological requirements, improves the feasibility and safety of surgical operations, and by adjusting the curvature control weight factor, the compliance of the path can be flexibly adjusted according to surgical needs, suitable for different types of puncture tasks; the obstacle avoidance term makes the puncture path away from obstacles, using the dynamic distance from the path to the nearest obstacle as the constraint index, making the obstacle avoidance control more intelligent and capable of adapting to changes in the intraoperative environment in real time, such as minor patient movements or tissue deformations; the time-dynamic speed constraint term controls the puncture speed, ensures the dynamic stability of the path, avoids tissue damage or path tracking errors caused by too fast speed, and adjusts the puncture speed through the time-dynamic weight function, enabling the puncture path to adaptively adjust the puncture rate when entering different tissue regions, such as accelerating when entering soft tissues and decelerating when approaching the target area, improving the stability and safety of the puncture.

[0077] In a preferred embodiment of the present invention, in the above-mentioned optical vision-guided puncture navigation method, the use of the improved gradient descent algorithm to solve the path functional and generate a puncture path with optimal characteristics includes: using the improved gradient descent algorithm to establish a path update formula Γ k is the current puncture path, η is the learning rate, used to control the optimization step size, is the path gradient term, representing the optimization direction of the current path, D cos is the anatomical structure matching term, used to calculate the matching degree between the current path and the preoperative image to avoid accidentally injuring important tissues, and μ is the weight controlling the strength of the anatomical matching constraint; output the optimized path direction to obtain a puncture path Γ with optimal characteristics opt .

[0078] Its technical effects are as follows: Adopting the gradient descent strategy to optimize the path, guiding the path update through the path gradient term, gradually optimizing the puncture path towards the global optimal solution, and improving the computational efficiency of path planning; dynamically adjusting the optimization step size through the improved learning rate, at the initial stage of path optimization, a larger step size speeds up the convergence speed, and when the path is close to the optimal solution, a smaller step size avoids oscillation and improves the fine control ability of the path.

[0079] In a preferred embodiment of the present invention, in the above-mentioned optical vision-guided puncture navigation method, the path gradient term The calculation formula is: Wherein, is the second-order gradient of the optical field, which is used for path adjustment.

[0080] In a preferred embodiment of the present invention, in the above-mentioned puncture navigation method guided by optical vision, the anatomical structure matching item D cos The calculation formula is: Wherein, Φ CT (Γ k ) is the CT image feature vector corresponding to the current puncture path; Φ MR is the preoperative MRI image reference feature vector; ||Φ CT (Γ k )|| is the L2 norm of the CT image feature vector; ||Φ MR || is the L2 norm of the preoperative MRI image reference feature vector.

[0081] In a preferred embodiment of the present invention, in the above-mentioned puncture navigation method guided by optical vision, constructing a feedback correction amount according to the error between the actual needle insertion trajectory and the planned puncture path, and correcting the puncture projection includes: obtaining the needle insertion trajectory Γ in real time real , calculating the overall trajectory error between it and the planned puncture path Γ opt Wherein, L is the total path length; based on the gradient of the overall trajectory error, combined with Gaussian smoothing filtering and optical field gradient information, calculating the error feedback correction amount Wherein, p is the actual position of the i-th sampling point, i is the gradient of the error point, which is used to represent the local change trend of the error at this point, G (p σ ) is the Gaussian smoothing filter kernel, which is used to remove noise, η′ is the feedback adjustment step size, v i is the preset puncture direction, d is the projection error direction; using the error feedback correction amount ΔP to adjust the intraoperative projection, and calculating the dynamic projection correction distribution Λ′(x,y) = STN(Λ(x,y)|θ)·diag(1 + α ) by using a spatial transformation network, wherein, Λ(x,y) is the original optical projection distribution, STN(Λ(x,y)|θ) is the spatial transformation network, θ is the STN transformation parameter, α stn is the frontal shape deformation compensation item; using the dynamic projection correction distribution Λ′(x,y) to update the optical-guided puncture projection, and re-calculating the gradient vector field of the optical field according to the corrected puncture projection stn to obtain an optimized puncture path.

[0082] ​Its technical effects are as follows: During the operation, due to tissue elasticity, operation errors, or external interference, the needle insertion trajectory may deviate from the original planned path. By constructing the overall trajectory error, the deviation between the needle insertion trajectory and the planned path can be quantified, providing accurate data for subsequent correction. Through the error gradient information, the local error trend is calculated, making the correction process more precise, rather than a global linear adjustment, thereby reducing the accumulation of local errors. By introducing a Gaussian smoothing filter kernel to process the error gradient data, high-frequency noise is removed, making the error feedback correction amount smoother and avoiding path instability caused by drastic adjustments. By adjusting the feedback step size to control the error correction intensity, the path adjustment is made more stable, avoiding new errors caused by overcorrection. A spatial transformation network is adopted to calculate the dynamic projection correction distribution using the error feedback correction amount and dynamically adjust the optical projection, enabling real-time adjustment of the optical projection so that the projection information always aligns with the target path, improving the accuracy of intraoperative guidance.

[0083] In a preferred embodiment of the present invention, in the above optical vision-guided puncture navigation method, the update formula for the STN transformation parameter θ is θ = θ0 + k θ ·ΔP, where θ0 is the initial affine transformation parameter and k θ is the learning rate; the external deformation compensation term α stn is calculated by the formula where is the optical projection intensity of the current tissue pixel point, is the position of the corresponding point in the preoperative image, which is used to calculate the influence of intraoperative tissue deformation on the optical projection and compensate for non-linear distortion.

[0084] The second embodiment of the present invention provides an optical vision-guided puncture navigation system, which includes: an optical field model construction module for constructing a non-uniform optical field model using the position data of optical markers in the device coordinate system and defining a bi-curvature optical projection surface equation describing the spatial light field distribution in the non-uniform optical field model; a multi-modal optical marker tracking module for introducing a filter response constraint into the bi-curvature optical projection surface equation to perform multi-modal optical marker tracking and obtain an optimized bi-curvature optical projection surface equation; a path functional construction module for constructing a path functional under multiple constraints in the non-uniform optical field model according to a preset puncture target and real-time collected obstacle information; a puncture path calculation module for using an improved gradient descent algorithm to solve the path functional and generate a puncture path with optimal characteristics; a puncture projection correction module for constructing a feedback correction amount based on the error between the actual needle insertion trajectory and the planned puncture path and correcting the puncture projection.

[0085] The computer program product of the optical vision-guided puncture navigation method and device provided by the embodiments of the present invention includes a computer-readable storage medium storing program codes. The instructions included in the program codes can be used to execute the methods in the foregoing method embodiments. For specific implementation, reference can be made to the method embodiments, which will not be elaborated here.

[0086] Specifically, the storage medium can be a general storage medium, such as a mobile disk, a hard disk, etc. When the computer program on the storage medium is run, it can execute the above-mentioned optical vision-guided puncture navigation method, so as to be able to construct a more accurate optical field model, and combine an optimized path planning algorithm with adaptive feedback to achieve a higher-precision and more stable puncture navigation.

[0087] If the above functions are implemented in the form of software function units and sold or used as independent products, they can be stored in a non-volatile computer-readable storage medium executable by a processor. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM), random access memories (RAM), magnetic disks, or optical discs that can store program codes.

[0088] Finally, it should be noted that the above-mentioned embodiments are only specific implementation manners of the present invention, used to illustrate the technical solutions of the present invention, rather than limiting them. The protection scope of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: any person skilled in the art within the technical scope disclosed by the present invention can still modify the technical solutions recorded in the foregoing embodiments, or can easily think of changes, or perform equivalent replacements for some of the technical features; and these modifications, changes, or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be covered by the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. A light vision-guided puncture navigation method, characterized in that, Including: Using the position data of the optical marker in the device coordinate system, constructing a non-uniform optical field model, and defining a bi-curvature optical projection surface equation describing the spatial light field distribution in the non-uniform optical field model; Introducing a filtering response constraint into the bi-curvature optical projection surface equation, performing multi-modal optical marker tracking, and obtaining an optimized bi-curvature optical projection surface equation; According to the preset puncture target and the obstacle information collected in real time, constructing a path functional under multiple constraints in the non-uniform optical field model; Using an improved gradient descent algorithm to solve the path functional and generating a puncture path with optimal characteristics; According to the error between the actual needle insertion trajectory and the planned puncture path, constructing a feedback correction amount to correct the puncture projection.

2. The light vision-guided puncture navigation method according to claim 1, wherein The step of using the position data of the optical marker in the device coordinate system to construct a non-uniform optical field model and defining a bi-curvature optical projection surface equation describing the spatial light field distribution in the non-uniform optical field model includes: Setting a device coordinate system O-XYZ, where O is the center of the device, the Z-axis direction is perpendicular to the device surface and is consistent with the puncture direction, and the XY axes are parallel to the device surface to define the horizontal distribution of the optical field; Collect the position data of the optical markers in the device coordinate system where (x k , y k , z k ) are three-dimensional space coordinates, α k is the reflection intensity weight, α k ∈ [0, 1], n is the number of optical markers distributed on the device surface, and k is the marker; Using a Gaussian weighting function, the discrete optical marker data is converted into a continuous optical field, and a non-uniform optical field model describing the optical field intensity at the spatial point (x, y, z) is constructed. Among them, σ z is the depth-adaptive Gaussian kernel width, σ z = β · (z max - z), β is the adjustment coefficient, z max is the maximum detection depth. Adjust the light field distribution in the depth direction using the hyperbolic tangent function where tanh(·) is the hyperbolic tangent function and γ is the curvature adjustment parameter, and the double-curvature optical projection surface equation is obtained 3. The light vision-guided puncture navigation method according to claim 2, wherein The step of introducing a filtering response constraint into the bi-curvature optical projection surface equation, performing multi-modal optical marker tracking, and obtaining an optimized bi-curvature optical projection surface equation includes: Introduce a filtering response constraint into the bi-curvature optical projection plane equation for multi-modal optical marker tracking Among them, is the correlation filtering response map, indicating the matching degree between the input image and the filtering template, which is calculated by the inverse Fourier transform F -1 and obtained; is the filtering template is the conjugate complex number of is the filtering template is the Fourier transform of the input image is the conjugate Fourier transform of the target template, and λ is the regularization coefficient; is the maximum matching value of the correlation filtering; is the L2 norm of the filtering template and is used to normalize the response intensity; Obtain the optimized bi-curvature optical projection surface equation 4. The light vision-guided puncture navigation method according to claim 3, wherein The step of constructing a path functional under multiple constraints in the non-uniform optical field model according to the preset puncture target and the obstacle information collected in real time includes: Based on the optimized double-curvature optical projection surface equation S″(x, y, z), calculate the gradient field of the spatial light field to obtain the gradient vector field of the optical field where reflects the optical gradient change in the X direction, reflects the optical gradient change in the Y direction, reflects the optical gradient change in the Z direction; Set the optical gradient guidance term \(w_1\|\nabla S''\|\) according to the preset puncture target and the obstacle information collected in real time 2 and the path curvature constraint term \(w_2\kappa\) 2 and the obstacle avoidance term \(w_3d\) -2 and the time dynamic speed constraint term where \(w_1\) is the weight factor of the optical gradient constraint, \(\kappa\) is the path curvature, \(w_2\) is the curvature control weight factor, \(d\) is the distance from the current path to the nearest obstacle, \(w_3\) is the obstacle avoidance weight factor, \(\lambda(t)\) is the time dynamic weight function is the instantaneous speed of the puncture path \(\Gamma\) at time \(t\), \(V\) max is the preset maximum allowable puncture speed The path functional is constructed 5. The light vision-guided puncture navigation method according to claim 4, wherein, The step of using an improved gradient descent algorithm to solve the path functional and generating a puncture path with optimal characteristics includes: Using an improved gradient descent algorithm, establish a path update formula Γ k is the current puncture path, η is the learning rate, is the path gradient term, D cos is the anatomical structure matching term, and μ is the weight controlling the strength of the anatomical matching constraint; Output the optimized path direction to obtain the puncture path Γ with optimal characteristics opt .

6. The optical vision-guided puncture navigation method according to claim 5, wherein The path gradient term has the following calculation formula: Among them, ▽ 2 S″ is the second-order gradient of the optical field and is used for path adjustment.

7. The light vision-guided puncture navigation method according to claim 5, characterized in that The anatomical structure matching item D cos The calculation formula is as follows: Among them, Φ CT (Γ k ) is the CT image feature vector corresponding to the current puncture path; Φ MR is the preoperative MRI image reference feature vector; ||Φ CT (Γ k )|| is the L2 norm of the CT image feature vector; ||Φ MR ||is the L2 norm of the preoperative MRI image reference feature vector.

8. The light vision-guided puncture navigation method according to claim 5, wherein The step of constructing a feedback correction amount according to the error between the actual needle insertion trajectory and the planned puncture path and correcting the puncture projection includes: Obtain the needle insertion trajectory Γ in real time real , and calculate the overall trajectory error between the obtained needle insertion trajectory Γ opt and the planned puncture path Γ where L is the total path length; Based on the gradient of the overall trajectory error, combined with Gaussian smoothing filtering and light field gradient information, calculate the error feedback correction amount where p i is the actual position of the i-th sampling point, is the gradient of the error pair point, G σ (p i ) is the Gaussian smoothing filter kernel, η′ is the feedback adjustment step size, v d is the preset puncture direction, sgn(▽S″×v d ) represents the projection error direction; Adjust the intraoperative projection using the error feedback correction amount ΔP, and calculate the dynamic projection correction distribution Λ′(x, y) = STN(Λ(x, y)|θ)·diag(1 + α stn ), where Λ(x, y) is the original optical projection distribution, STN(Λ(x, y)|θ) is the spatial transformation network, θ is the STN transformation parameter, and α stn is the forehead deformation compensation term; Using the dynamic projection correction distribution Λ′(x, y), updating the optical-guided puncture projection, and according to the corrected puncture projection, recalculating the gradient vector field ▽S″ of the optical field to obtain an optimized puncture path.

9. The optical vision-guided puncture navigation method according to claim 8, characterized in that The update formula for the STN transformation parameter θ is θ = θ0 + k θ ·ΔP, where θ0 is the initial affine transformation parameter and k θ is the learning rate; The forehead shape deformation compensation term α stn has the following calculation formula where is the optical projection intensity of the current tissue pixel point, is the position of the corresponding point in the preoperative image, which is used to calculate the influence of intraoperative tissue deformation on optical projection and compensate for non - linear distortion.

10. A light vision-guided puncture navigation system, characterized in that, Including: An optical field model construction module, configured to use the position data of the optical marker in the device coordinate system to construct a non-uniform optical field model, and define a bi-curvature optical projection surface equation describing the spatial light field distribution in the non-uniform optical field model; A multi-modal optical marker tracking module, configured to introduce a filtering response constraint into the bi-curvature optical projection surface equation, perform multi-modal optical marker tracking, and obtain an optimized bi-curvature optical projection surface equation; A path functional construction module, configured to construct a path functional under multiple constraints in the non-uniform optical field model according to the preset puncture target and the obstacle information collected in real time; A puncture path calculation module, configured to use an improved gradient descent algorithm to solve the path functional and generate a puncture path with optimal characteristics; A puncture projection correction module, configured to construct a feedback correction amount according to the error between the actual needle insertion trajectory and the planned puncture path, and correct the puncture projection.

Images