Injection path guidance method, system, device, and medium based on ultrasound and laser combined navigation

Abstract

This invention provides an injection path guidance method, system, device, and medium based on ultrasound and laser combined navigation, belonging to the field of medical injection technology. The method includes: denoising an initial sound wave to obtain a denoised wave; identifying the boundary points of the denoised wave based on the mean square relative fitting error to obtain an anatomical structure image of the target muscle; performing image segmentation on the anatomical structure image to obtain a target structure image of the target muscle; extracting feature information at different scales from the target structure image and then performing adaptive clustering through variable convolution to obtain a three-dimensional result of the target; registering and comparing the target three-dimensional result with an anatomical atlas database, marking key avoidance structures to determine a safe puncture boundary; determining the puncture target point and alternative paths in the target three-dimensional result based on the safe puncture boundary; and projecting light spot markers and the target path onto the target muscle using a near-infrared laser navigation module based on spatial transformation relationships between the puncture target point and the alternative paths.

Description

Technical Field

[0001] This invention relates to the field of medical injection technology, and in particular to an injection path guidance method, system, device and medium based on ultrasound and laser combined navigation. Background Technology

[0002] With the rapid development of cosmetic and therapeutic injection techniques, the use of botulinum toxin and other methods for cosmetic shaping or treating muscle spasms is becoming increasingly widespread. However, existing injection techniques still face many challenges in practice, mainly including insufficient positioning accuracy leading to uneven or inaccurate therapeutic effects; significant pain during the procedure affecting patient experience and cooperation; and potential neurovascular damage risks that may cause unnecessary complications. Therefore, there is an urgent need for an injection path guidance method that can achieve precise positioning, simplify the procedure, and significantly reduce pain and risks. Summary of the Invention

[0003] The main objective of this invention is to provide an injection path guidance method, system, device, and medium based on ultrasound and laser combined navigation, aiming to solve the problems of insufficient injection positioning accuracy, strong puncture pain, and easy damage to surrounding nerves and blood vessels in existing injection techniques.

[0004] In a first aspect, embodiments of the present invention provide an injection path guidance method based on combined ultrasound and laser navigation, comprising: The initial sound wave obtained by the ultrasound device detecting the target muscle is obtained, and the initial sound wave is denoised to obtain the corresponding denoised sound wave. Based on the mean square relative fitting error, the boundary point of the denoised wave is identified to obtain the anatomical structure image corresponding to the target muscle. The anatomical structure image is segmented to obtain the target structure image corresponding to the target muscle; Feature information at different scales is extracted from the target structure image, and then adaptive clustering is performed through variable convolution to obtain the target three-dimensional result corresponding to the target muscle. The target 3D result is registered and compared with a preset anatomical atlas database to mark key avoidance structures and determine safe puncture boundaries; Based on the safe puncture boundary, the puncture target point and alternative paths are determined in the target three-dimensional result of the target muscle; The near-infrared laser navigation module projects corresponding light spot marks and target paths onto the target muscle based on the spatial transformation relationship between the puncture target point and the alternative paths.

[0005] Secondly, embodiments of the present invention provide an injection path guidance system based on combined ultrasound and laser navigation, comprising: The data acquisition module is used to acquire the initial sound wave obtained by the ultrasound device detecting the target muscle, and to perform noise reduction processing on the initial sound wave to obtain the corresponding noise-reduced sound wave. The data recognition module is used to identify the boundary points of the denoised wave based on the mean square relative fitting error to obtain the anatomical structure image corresponding to the target muscle. The segmentation processing module is used to perform image segmentation processing on the anatomical structure image to obtain the target structure image corresponding to the target muscle; The data analysis module is used to extract feature information at different scales from the target structure image and then perform adaptive clustering through variable convolution to obtain the target three-dimensional result corresponding to the target muscle. The comparison and analysis module is used to register and compare the target three-dimensional results with a preset anatomical atlas database, mark key avoidance structures, and determine safe puncture boundaries; The path selection module is used to determine the puncture target point and alternative paths in the target three-dimensional result of the target muscle based on the safe puncture boundary. The information conversion module is used to project corresponding light spot marks and target paths onto the target muscle based on the spatial conversion relationship between the puncture target point and the alternative paths through the near-infrared laser navigation module. Thirdly, embodiments of the present invention also provide a terminal device, the terminal device including a processor, a memory, a computer program stored in the memory and executable by the processor, and a data bus for realizing communication between the processor and the memory, wherein when the computer program is executed by the processor, it implements the steps of any of the injection path guidance methods based on ultrasound and laser combined navigation provided in this specification.

[0006] Fourthly, embodiments of the present invention also provide a storage medium for computer-readable storage, characterized in that the storage medium stores one or more programs, which can be executed by one or more processors to implement the steps of any of the injection path guidance methods based on ultrasound and laser combined navigation provided in this specification.

[0007] This invention provides an injection path guidance method, system, device, and medium based on ultrasound and laser combined navigation. The method includes: obtaining an initial sound wave from an ultrasound device detecting a target muscle, and denoising the initial sound wave to obtain a corresponding denoised wave; identifying the boundary point of the denoised wave based on the mean square relative fitting error to obtain an anatomical structure image corresponding to the target muscle; performing image segmentation on the anatomical structure image to obtain a target structure image corresponding to the target muscle; extracting feature information of different scales from the target structure image and then performing adaptive clustering through variable convolution to obtain a target three-dimensional result corresponding to the target muscle; registering and comparing the target three-dimensional result with a preset anatomical atlas database, marking key avoidance structures, and determining a safe puncture boundary; determining the puncture target point and alternative paths in the target three-dimensional result of the target muscle based on the safe puncture boundary; and projecting corresponding light spot markers and target paths onto the target muscle using a near-infrared laser navigation module based on spatial transformation relationships between the puncture target point and alternative paths. This invention denoises and intelligently analyzes the initial sound waves to reconstruct a high-precision, high-definition three-dimensional result corresponding to the target muscle, thus significantly improving the positioning accuracy of the injection target. Secondly, by automatically identifying anatomical structures, marking key avoidance areas, and determining safe puncture boundaries, damage to surrounding nerves and blood vessels can be effectively avoided, significantly reducing operational risks and the incidence of complications. Finally, based on the target's three-dimensional result, a path is planned, and real-time, intuitive spatial projection navigation is achieved using near-infrared laser, making the puncture process more precise and convenient. This not only reduces the physician's workload but also effectively reduces tissue damage and patient pain caused by repeated adjustments to the puncture site, thus improving overall treatment safety and patient experience. It solves the problems of insufficient injection positioning accuracy, strong puncture pain, and easy damage to surrounding nerves and blood vessels in existing injection techniques. Attached Figure Description

[0008] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0009] Figure 1 A schematic flowchart of an injection path guidance method based on combined ultrasound and laser navigation provided in an embodiment of the present invention; Figure 2 A schematic diagram of the module structure of an injection path guidance system based on combined ultrasound and laser navigation provided in an embodiment of the present invention; Figure 3 This is a schematic block diagram of a terminal device provided in an embodiment of the present invention. Detailed Implementation

[0010] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0011] The flowchart shown in the attached diagram is for illustrative purposes only and does not necessarily include all content and operations / steps, nor does it necessarily have to be performed in the order described. For example, some operations / steps can be broken down, combined, or partially merged, so the actual execution order may change depending on the actual situation.

[0012] It should be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0013] This invention provides an injection path guidance method, system, device, and medium based on combined ultrasound and laser navigation. The injection path guidance method based on combined ultrasound and laser navigation can be applied to a terminal device, which can be an electronic device such as a tablet computer, laptop computer, desktop computer, personal digital assistant, or wearable device. The terminal device can be a server or a server cluster.

[0014] The following detailed description of some embodiments of the present invention is provided in conjunction with the accompanying drawings. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0015] Please refer to Figure 1 , Figure 1 This is a flowchart illustrating an injection path guidance method based on combined ultrasound and laser navigation, provided as an embodiment of the present invention.

[0016] like Figure 1 As shown, the injection path guidance method based on ultrasound and laser combined navigation includes steps S101 to S107.

[0017] Step S101: Obtain the initial sound wave obtained by the ultrasound device detecting the target muscle, and perform noise reduction processing on the initial sound wave to obtain the corresponding noise-reduced sound wave.

[0018] For example, an initial sound wave is obtained by emitting sound waves into the target muscle region using a high-frequency ultrasound probe and receiving the reflected echoes. This initial sound wave contains noise from the instrument itself, tissue scattering, and the external environment. Preprocessing of the initial sound wave, such as gain adjustment and bandwidth filtering, is then performed to initially improve the signal-to-noise ratio. Subsequently, adaptive filters or wavelet transforms are used to separate and filter out noise based on its statistical characteristics and signal features, resulting in a denoised wave, laying the foundation for subsequent high-precision imaging and structural analysis.

[0019] In some embodiments, the step of denoising the initial sound wave to obtain the corresponding denoised wave includes: performing a standardization preprocessing on the initial sound wave to calculate the dynamic mean and standard deviation of the initial sound wave to obtain standardized information corresponding to the initial sound wave; performing empirical mode decomposition on the standardized information, detecting local extrema of the standardized information, and constructing upper and lower envelopes to obtain multiple intrinsic mode function components; generating a corresponding noise reference signal for each intrinsic mode function component, and determining a reference benchmark for adaptive noise cancellation based on the standardized information and the frequency domain characteristics of the noise reference signal; performing noise filtering on the intrinsic mode function components based on the reference benchmark to obtain filtering information corresponding to the intrinsic mode function components; and linearly superimposing the filtering information with the residual components obtained from empirical mode decomposition to reconstruct the denoised wave.

[0020] For example, the initial sound wave is preprocessed by standardization. The mean and standard deviation of the signal are calculated dynamically and recursively, and the initial sound wave is transformed into standardized information with zero mean and unit variance, so as to eliminate the difference in amplitude dimensions and improve the numerical stability of subsequent processing.

[0021] For example, empirical mode decomposition is performed on the standardized information to detect all local extrema in the standardized information. The upper and lower envelopes are fitted respectively, and the mean value corresponding to the standardized information is calculated. Through an iterative screening process, a series of intrinsic mode function components that meet the intrinsic mode conditions are gradually separated until the remaining signal becomes a monotonic residual component.

[0022] For example, for each intrinsic mode function component, the original noisy signal is processed by a Butterworth filter to generate a corresponding noise reference signal; the correlation between the two in the frequency domain is analyzed by combining the spectral distribution characteristics of the normalized signal and the noise reference signal, thereby determining the reference reference required in the adaptive noise canceller.

[0023] For example, each intrinsic mode function component is used as the desired signal, and the aforementioned reference benchmark is used as input. An improved recursive least squares adaptive filtering algorithm is employed for recursive noise filtering. A regularization term is introduced during the weight update process to suppress overfitting, thereby obtaining the denoised filtered components. Finally, all filtered intrinsic mode function components are linearly superimposed with the residual components obtained from empirical mode decomposition in chronological order to reconstruct the complete denoised waveform.

[0024] Specifically, by using standardized preprocessing and empirical mode decomposition to adaptively separate signal components, combined with improved adaptive filtering techniques, the detailed features of the disturbed signal are significantly preserved while effectively suppressing noise, thus improving denoising accuracy and signal fidelity. The adaptive processing mechanism enhances the adaptability to non-stationary signals and complex disturbances, thereby improving robustness.

[0025] In some implementations, the step of performing noise filtering on the intrinsic mode function (IMF) components based on the reference benchmark to obtain the filtering information corresponding to the IMF components includes: determining the current time corresponding to the IMF components and determining the covariance matrix corresponding to the previous time of the current time; determining the gain vector corresponding to the current time based on the covariance matrix and the reference benchmark, wherein the gain vector is used to determine the degree of influence of the IMF components at the current time on the update of the initial weight vector; predicting the prediction output corresponding to the current time based on the initial weight vector, and determining the prior error based on the prediction output and the IMF components; using the prior error and the gain vector, synchronously and recursively updating the initial weight vector and the covariance matrix to obtain the updated target weight vector and target matrix; and performing instantaneous filtering on the IMF components based on the target weight vector and the target matrix to obtain the filtering information corresponding to the IMF components.

[0026] For example, the values ​​of the intrinsic mode function components at the current time are determined, and the covariance matrix updated at the previous time is called. Then, based on the covariance matrix and the pre-established reference signal, the gain vector at the current time is calculated. The gain vector quantifies the contribution of the current input to the weight adjustment.

[0027] For example, calculating the gain vector is crucial for efficient recursive filtering, aiming to accurately quantify the influence of the current noise reference on the filter weight adjustment. First, upon entering the calculation loop at the current time (time n), the covariance matrix P(n-1) updated from the previous time (time n-1) is already available. This matrix is ​​a symmetric positive definite matrix. Then, the current input vector u(n) is obtained, which is a vector composed of the reference signal. This vector represents the characteristics of the noise reference signal at the current sampling point. The covariance matrix P(n-1) is then multiplied with the current input vector u(n) to obtain the first vector. Next, a second scalar is calculated based on P(n-1) and u(n) to represent the energy or significance of the current input vector under the metric defined by the historical covariance matrix. A forgetting factor λ is then added to this second scalar to track changes in the non-stationary system. Finally, the first vector is divided by the second vector to obtain the gain vector at the current time.

[0028] For example, the output is predicted using the initial weight vector. The prior error is obtained by comparing the predicted value with the actual value of the intrinsic mode function component. The prior error is then combined with the gain vector to synchronously and recursively update the initial weight vector and covariance matrix. The initial weight vector update introduces a regularization term to enhance stability, thereby obtaining the updated target weight vector and target covariance matrix, which is the target matrix.

[0029] For example, the updated target weight vector is used to perform instantaneous filtering calculations on the intrinsic mode function components at the current time, and the denoised filtering information is directly output, providing a basis for subsequent signal reconstruction.

[0030] Step S102: Based on the mean square relative fitting error, identify the boundary point of the denoised wave to obtain the anatomical structure image corresponding to the target muscle.

[0031] For example, a sliding window is used to extract local data segments along the denoised waveform. For the data within each window, a least-squares fit is performed using a polynomial or spline curve, and the mean square relative fitting error of the data in that window (i.e., the ratio of the sum of squared fitting residuals to the energy of the original data) is calculated. This error value sensitively reflects the degree of deviation between the local signal morphology and the smooth fitting curve. When the sliding window crosses the anatomical boundaries of different muscle tissues (such as muscle bundles and fascia), the local signal morphology will undergo nonlinear changes due to abrupt changes in the acoustic properties of the tissue, resulting in a significant increase in the fitting error of that window, forming an error peak. By detecting the local extreme points of the fitting error on the entire denoised waveform, and combining the threshold criterion of the error amplitude with the spatial continuity analysis of adjacent extreme points, the anatomical boundary points marking different tissue characteristics can be accurately identified. Finally, by mapping and connecting these boundary points in two-dimensional or three-dimensional spatial coordinates, an anatomical structure image that clearly reflects the muscle fascia layer, muscle bundle direction, and texture differences can be reconstructed.

[0032] In some implementations, the step of identifying the boundary points of the denoised wave based on the mean square relative fitting error to obtain the anatomical structure image corresponding to the target muscle includes: establishing a mathematical relationship equation between signal points in the denoised wave, and performing data analysis through the mathematical relationship equation to obtain the signal feature parameters corresponding to the denoised wave; segmenting the denoised wave according to the mean square relative fitting error, starting from the signal starting point in the denoised wave, adding only one data point at a time, calculating the fitting error of the current segment, and if the fitting error is less than a threshold, continuing to add points; if the fitting error exceeds the threshold, recording the data point as a potential segmentation point; increasing the search step size to a preset maximum value based on the potential segmentation point, continuing to monitor the fitting error until the fitting error exceeds the threshold, reducing the search step size until a target step size and the precise segmentation point corresponding to the target step size are determined; and generating an image based on the precise segmentation point and the denoised wave to obtain the anatomical structure image corresponding to the target muscle.

[0033] For example, a mathematical relationship equation between signal points is constructed based on the denoised wave, and signal feature parameters such as amplitude, frequency and phase are extracted by analyzing the equation to describe the core properties of the waveform.

[0034] For example, the mean square relative fitting error is used as the segmentation criterion. Starting from the starting point of the denoised wave, the data segment is gradually expanded. The fitting error of the current segment is calculated for each additional data point. If the error is lower than a preset threshold, the expansion continues. Once the error exceeds the threshold, the point is recorded as a potential segmentation point. Then, based on the potential segmentation points, the search step size is increased to a preset maximum value to quickly scan subsequent signals, and the fitting error is continuously monitored. When the error exceeds the threshold again, the step size is gradually reduced for fine adjustment until a stable target step size and the corresponding accurate segmentation point are determined, ensuring that the segmentation accurately captures the characteristic changes of the signal.

[0035] For example, precise segmentation points are mapped to the time-amplitude sequence of the denoised wave. Each segmentation point defines a relatively stable time interval for signal features. Combined with signal feature parameters such as amplitude, frequency, and phase of the denoised wave within this interval, a series of signal units with clear start and end times and feature descriptions are formed. Then, image reconstruction algorithms such as back projection, iterative reconstruction, or deep learning generative models are used to map the features of these signal units into a preset muscle anatomy spatial model. The signal amplitude can be converted into the intensity or color gradient of image pixels, the frequency features reflect the fineness of tissue texture, and the phase information is used to adjust the continuity of the spatial structure. Then, through spatial interpolation and smoothing techniques, the discrete signal feature data are fused into a continuous, high-resolution anatomical structure image.

[0036] Step S103: Perform image segmentation processing on the anatomical structure image to obtain the target structure image corresponding to the target muscle.

[0037] For example, the anatomical structure image is preprocessed with contrast enhancement and noise smoothing, and then a deep learning-based semantic segmentation model such as U-Net is used to classify the pixels in the image into different categories such as target muscle, surrounding tissue and background through trained weights or preset parameters, generating an initial segmentation mask. Then, morphological operations are used to post-process the segmentation mask to smooth muscle boundaries, fill internal holes and remove discrete noise regions to ensure the continuity and integrity of the segmentation results. Finally, the mask is applied to the anatomical structure image to obtain the target structure image.

[0038] In some embodiments, the step of performing image segmentation processing on the anatomical structure image to obtain the target structure image corresponding to the target muscle includes: calculating the average gray value of the anatomical structure image and determining an initial threshold based on the average gray value; segmenting the anatomical structure image according to the initial threshold to obtain a first target region and a first background region; sequentially attempting to reclassify each gray level in the gray histogram of the first target region according to the initial threshold, calculating the weighted variance after each adjustment, and obtaining an adjustment threshold; obtaining the average rate of change and intra-class variance of the image gray pixels of the first target region under the adjustment threshold; further segmenting the first target region according to the average rate of change and the intra-class variance to obtain a second target region and a second background region; and determining the target structure image corresponding to the target muscle based on the second target region, the first background region, and the second background region.

[0039] For example, the average gray value of the anatomical structure image is calculated, and an initial threshold slightly higher than the average gray value is set based on this to initially distinguish the target muscle from the background tissue.

[0040] For example, the anatomical structure image is segmented according to an initial threshold to obtain a first target region and a first background region corresponding to candidate muscle regions. Based on the gray-level histogram of the first target region, the classification method of gray levels is adjusted iteratively: each gray level is attempted to be reassigned to the target or background class in turn, and the weighted variance sum after each adjustment is calculated to determine an adjustment threshold that minimizes intra-class differences. Subsequently, the average change rate of gray pixels in the first target region image (i.e., the relative change of the average gray level of the region before and after segmentation) and the change of intra-class variance are evaluated under the adjustment threshold. Based on these two indicators, it is determined whether to continue segmentation: if the average change rate is significant and the intra-class variance changes little, it indicates that the current segmentation is sufficient, and the iteration stops; otherwise, the first target region is segmented again to further separate a more accurate second target region and a second background region.

[0041] For example, the second target region, the first background region, and the second background region are integrated to generate a clear target structure image corresponding to the target muscle, thus completing the image segmentation process.

[0042] Step S104: Extract feature information at different scales from the target structure image and then perform adaptive clustering through variable convolution to obtain the target three-dimensional result corresponding to the target muscle.

[0043] For example, by constructing an image pyramid or using a multi-scale convolutional neural network layer to extract multi-scale feature information from the target structure image, local details and global contour features of muscle tissue can be captured at different resolutions.

[0044] For example, a variable convolutional network is applied to process multi-scale feature information. Its convolutional kernel can adaptively adjust the sampling position and flexibly adapt to the geometric deformation of muscle shape, thereby enhancing the spatial adaptability of feature representation. Based on the extracted features, adaptive clustering is performed, pixels or regions are automatically grouped according to feature similarity, muscle substructures are distinguished and noise is removed.

[0045] For example, by utilizing clustering results and multi-scale features, two-dimensional segmentation information is transformed into an accurate three-dimensional muscle model through three-dimensional reconstruction technology, thereby obtaining the target three-dimensional result corresponding to the target muscle.

[0046] In some implementations, the target structure image includes target images from different viewpoints. The step of extracting feature information at different scales from the target structure image and then adaptively clustering it using deformable convolution to obtain the target 3D result corresponding to the target muscle includes: using a feature extraction layer of a depth optimization model to adaptively aggregate features at different scales from the target images at different viewpoints through deformable convolution operations to obtain target fusion features corresponding to the target images; using an information determination layer of the depth optimization model to map features from other viewpoint images onto multiple hypothetical depth planes of a reference viewpoint through a differentiable homography transformation, constructing a matching cost body, which characterizes the matching degree of each pixel under different depth hypotheses; using a cost self-learning layer of the depth optimization model to further smooth and optimize the matching cost body based on the target fusion features using deformable convolution, obtaining a regression depth map corresponding to the optimized matching cost body; and fusing information based on the regression depth map and the target structure image to obtain the target 3D result corresponding to the target muscle.

[0047] For example, the input is a target image acquired from multiple perspectives and the acquisition angles. A feature extraction layer of a deep optimization model adaptively fuses multi-scale features from the image using deformable convolutions to generate a fused feature map containing rich contextual information.

[0048] For example, the information determination layer of the depth optimization model maps the feature maps of target images from different viewpoints onto multiple hypothetical depth planes of the reference viewpoint through a differentiable homography transformation, constructing a matching cost body. This matching cost body characterizes the matching confidence of each pixel under different depth hypotheses. In other words, based on geometric principles, the information determination layer of the depth optimization model maps the feature maps of all other viewpoints onto a series of hypothetical depth planes of the reference viewpoint through a differentiable homography transformation, and constructs a three-dimensional matching cost body by calculating the feature variance. This matching cost body quantifies the matching confidence of each pixel at different depths.

[0049] For example, the deep optimization model does not directly use fused features. Instead, it first regularizes the initial matching cost volume using a 3D convolutional network to smooth noise, and then enables its cost self-learning module. This module again utilizes the adaptive sampling capability of deformable convolutions to aggregate information from its reliable neighborhood for each position in the matching cost volume to correct outliers, thereby obtaining an optimized matching cost volume. By converting this optimized matching cost volume into a probability distribution and performing regression calculations, the regression depth map of the reference view can be obtained.

[0050] For example, by matching the regression depth map and the target structure image one-to-one and fusing the information, the target three-dimensional result corresponding to the target muscle is obtained.

[0051] Step S105: Register and compare the target three-dimensional result with the preset anatomical atlas database, mark the key avoidance structures, and determine the safe puncture boundary.

[0052] For example, the reconstructed 3D target result is spatially aligned with a standard 3D model in an anatomical atlas database. This process typically involves two stages: coarse registration and fine registration. Coarse registration performs initial alignment based on the overall shape or key anatomical landmarks, while fine registration uses optimization algorithms to minimize the distance between corresponding surfaces of the two models, thereby achieving millimeter-level or even sub-millimeter-level precise matching.

[0053] For example, the spatial information of key blood vessels, nerve bundles, important organs, and other avoidance structures pre-annotated in the registered anatomical atlas is automatically mapped and superimposed onto the target 3D result. This allows for the calculation of the 3D spatial distance between the reconstructed surface and these avoidance structures, automatically identifying high-risk areas. Based on preset safety rules, such as maintaining a distance of at least 3 millimeters from key nerves and the geometry of the avoidance structures, a safe puncture boundary is generated in the target 3D result.

[0054] Step S106: Based on the safe puncture boundary, determine the puncture target point and alternative paths in the target three-dimensional result of the target muscle.

[0055] For example, the puncture target point is located inside the safe puncture boundary. Then, using this puncture target point as the endpoint and the pre-defined safe needle insertion point on the skin as the starting point, one or more alternative paths are planned within the virtual channel formed by the three-dimensional safe boundary using a path search algorithm. The core principle of the planning is that the path must be completely within the safe boundary, avoiding all marked critical avoidance structures throughout the process, while striving for the shortest path, the straightest angle, and the fewest tissue layers traversed, in order to improve operational efficiency and safety.

[0056] Step S107: Using the near-infrared laser navigation module, the puncture target point and the alternative path are projected onto the target muscle according to the spatial transformation relationship.

[0057] For example, the output power of the near-infrared laser navigation module is controlled to ≤1mW (Class 1 safety laser), and the wavelength is selected in the near-infrared band of 780-850nm to ensure eye safety and form a clearly visible light spot on the skin surface. During the preoperative planning stage, based on the puncture target points and alternative paths determined in the target's three-dimensional structure in previous steps, they are mapped to physical space through spatial coordinate transformation: First, using an optical positioning system or pre-performed camera-laser joint calibration, a spatial transformation relationship is established between the virtual three-dimensional model coordinate system and the actual patient's body surface coordinate system; based on the spatial transformation relationship, the corresponding projection position of each puncture target point and alternative path on the patient's skin surface is calculated. The control system of the near-infrared laser navigation module receives these position coordinates, drives the laser to emit a low-power near-infrared beam, and scans or statically projects it through an optical lens group to accurately draw light spot markers representing the target points and the target path on the skin.

[0058] In some embodiments, the method further includes: during the puncture of the target muscle according to the light spot marker and the target path, continuously receiving real-time images and dynamically tracking the three-dimensional position of the echo-enhanced puncture needle tip, comparing the three-dimensional position with the target path to obtain a target comparison result; and providing real-time offset prompts and correction guidance in the fusion display interface according to the target comparison result to obtain a target guidance result.

[0059] For example, during the puncture procedure, real-time navigation is achieved through the fusion of near-infrared optical positioning and ultrasound imaging. Specifically, a real-time two-dimensional ultrasound image stream of the target muscle is continuously acquired using an ultrasound probe, and an embedded computer vision algorithm is used to automatically identify and track the pixel position of the highly echogenic puncture needle tip in each frame. Combined with pre-calibrated ultrasound probe spatial pose data, a three-dimensional reconstruction algorithm is used to fuse and calculate the continuous two-dimensional needle tip positions in real time, thereby dynamically obtaining the precise three-dimensional coordinates of the needle tip in the patient's anatomical space. This real-time three-dimensional coordinate is continuously compared with a pre-planned digital target path in a unified spatial coordinate system, and the spatial offset and angular deviation of the needle tip relative to the ideal path are calculated to obtain the target comparison result.

[0060] For example, in the fusion display interface (which typically overlays real-time ultrasound images, 3D planned paths, and needle tip tracking positions), real-time offset prompts are provided intuitively in a visual manner, such as dynamic arrows, deviation rulers, and highlighted color warnings. At the same time, based on preset tolerance thresholds and kinematic models, specific target guidance results are generated, thereby assisting doctors to correct the puncture needle in real time and accurately and keep it within the safe path.

[0061] In some application scenarios, it also includes preoperative positioning and calibration: The patient's injection site is disinfected, an ultrasound coupling agent is applied, the ultrasound probe is placed against the skin, and the high-frequency ultrasound positioning module is activated to scan the target muscle and determine the muscle's central target point; the near-infrared laser navigation module is activated, and the laser spot is precisely projected onto the skin surface corresponding to the target point through the laser calibration component, completing laser-ultrasound target point coordinate matching; puncture path planning and fixation: The echo-enhanced ultra-fine puncture needle is installed onto the fixed support slide rail, and the angle positioning knob is adjusted to determine the needle insertion angle and locked according to the muscle depth displayed in the ultrasound image. Position the stent to ensure the puncture path aligns with the laser marking direction; Real-time navigation and injection during the procedure: Under the real-time guidance of the image fusion processing module, slowly advance the puncture needle, observing the movement of the hyperechoic spot at the needle tip towards the target muscle in the ultrasound image; When the needle tip reaches the target position, stop needle insertion and inject botulinum toxin solution at a rate of 0.1-0.3 ml / s, continuously monitoring the ultrasound image to ensure the injection solution diffuses within the target muscle; Post-operative care: After injection, slowly withdraw the puncture needle, apply pressure to the injection site to stop bleeding, turn off the equipment, and clean the ultrasound probe, puncture needle, and laser emitter.

[0062] Please see Figure 2 , Figure 2This application provides an injection path guidance system 200 based on ultrasound and laser combined navigation. The system includes a data acquisition module 201, a data recognition module 202, a segmentation processing module 203, a data analysis module 204, a comparison analysis module 205, a path filtering module 206, and an information conversion module 207. The data acquisition module 201 acquires the initial sound wave obtained by the ultrasound device detecting the target muscle and performs noise reduction processing on the initial sound wave to obtain a corresponding denoised sound wave. The data recognition module 202 identifies the boundary points of the denoised sound wave based on the mean square relative fitting error to obtain an anatomical structure image corresponding to the target muscle. The segmentation processing module 203 processes the anatomical structure image... The system performs image segmentation to obtain the target structure image corresponding to the target muscle; a data analysis module 204 is used to extract feature information of different scales from the target structure image and then perform adaptive clustering through variable convolution to obtain the target three-dimensional result corresponding to the target muscle; a comparison analysis module 205 is used to register and compare the target three-dimensional result with a preset anatomical atlas database, mark key avoidance structures, and determine the safe puncture boundary; a path selection module 206 is used to determine the puncture target point and alternative paths in the target three-dimensional result of the target muscle based on the safe puncture boundary; and an information conversion module 207 is used to project corresponding light spot marks and target paths onto the target muscle according to the spatial conversion relationship between the puncture target point and the alternative paths through a near-infrared laser navigation module. In some implementations, the injection path guidance system 200 based on combined ultrasound and laser navigation can be applied to terminal devices.

[0063] It should be noted that those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the injection path guidance system 200 based on ultrasound and laser combined navigation described above can be referred to the corresponding process in the aforementioned embodiments of the injection path guidance method based on ultrasound and laser combined navigation, and will not be repeated here.

[0064] Please see Figure 3 , Figure 3 This is a schematic block diagram of the structure of a terminal device provided in an embodiment of the present invention.

[0065] like Figure 3 As shown, the terminal device 300 includes a processor 301 and a memory 302, which are connected via a bus 303, such as an I2C (Inter-integrated Circuit) bus.

[0066] Specifically, processor 301 provides computing and control capabilities to support the operation of the entire terminal device. Processor 301 can be a Central Processing Unit (CPU), but it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor.

[0067] Specifically, the memory 302 can be a Flash chip, a read-only memory (ROM) disk, an optical disk, a USB flash drive, or a portable hard drive, etc.

[0068] Those skilled in the art will understand that Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the embodiments of the present invention, and does not constitute a limitation on the terminal device to which the embodiments of the present invention are applied. A specific server may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0069] The processor is used to run a computer program stored in a memory, and when executing the computer program, implements any of the injection path guidance methods based on ultrasound and laser combined navigation provided in the embodiments of the present invention.

[0070] In one embodiment, the processor is configured to run a computer program stored in memory, and when executing the computer program, perform the following steps: The initial sound wave obtained by the ultrasound device detecting the target muscle is obtained, and the initial sound wave is denoised to obtain the corresponding denoised sound wave. Based on the mean square relative fitting error, the boundary point of the denoised wave is identified to obtain the anatomical structure image corresponding to the target muscle. The anatomical structure image is segmented to obtain the target structure image corresponding to the target muscle; Feature information at different scales is extracted from the target structure image, and then adaptive clustering is performed through variable convolution to obtain the target three-dimensional result corresponding to the target muscle. The target 3D result is registered and compared with a preset anatomical atlas database to mark key avoidance structures and determine safe puncture boundaries; Based on the safe puncture boundary, the puncture target point and alternative paths are determined in the target three-dimensional result of the target muscle; The near-infrared laser navigation module projects corresponding light spot marks and target paths onto the target muscle based on the spatial transformation relationship between the puncture target point and the alternative paths.

[0071] It should be noted that those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process of the terminal device described above can be referred to the corresponding process in the aforementioned embodiment of the injection path guidance method based on ultrasound and laser combined navigation, and will not be repeated here.

[0072] This invention also provides a storage medium for computer-readable storage, wherein the storage medium stores one or more programs that can be executed by one or more processors to implement the steps of any of the injection path guidance methods based on ultrasound and laser combined navigation as provided in the specification of this invention.

[0073] The storage medium can be an internal storage unit of the terminal device described in the foregoing embodiments, such as the hard drive or memory of the terminal device. Alternatively, the storage medium can be an external storage device of the terminal device, such as a plug-in hard drive, Smart Media Card (SMC), Secure Digital (SD) card, or Flash Card equipped on the terminal device.

[0074] Those skilled in the art will understand that all or some of the steps, systems, or apparatuses disclosed above, and their functional modules / units, can be implemented as software, firmware, hardware, or suitable combinations thereof. In hardware embodiments, the division between functional modules / units mentioned in the above description does not necessarily correspond to the division of physical components; for example, a physical component may have multiple functions, or a function or step may be performed collaboratively by several physical components. Some or all physical components may be implemented as software executed by a processor, such as a central processing unit, digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit. Such software may be distributed on a computer-readable medium, which may include computer storage media (or non-transitory media) and communication media (or transient media). As is known to those skilled in the art, the term computer storage media includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data). Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disc (DVD) or other optical disc storage, magnetic cartridges, magnetic tape, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and can be accessed by a computer. Furthermore, it is well known to those skilled in the art that communication media typically contain computer-readable instructions, data structures, program modules, or other data in modulated data signals such as carrier waves or other transmission mechanisms, and may include any information delivery medium.

[0075] It should be understood that the term "and / or" as used in this specification and the appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes such combinations. It should be noted that, herein, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system 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 system. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element.

[0076] The sequence numbers of the above embodiments of the present invention are merely for descriptive purposes and do not represent the superiority or inferiority of the embodiments. The above descriptions are only specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for guiding injection pathways based on combined ultrasound and laser navigation, characterized in that, include: The initial sound wave obtained by the ultrasound device detecting the target muscle is obtained, and the initial sound wave is denoised to obtain the corresponding denoised sound wave. Based on the mean square relative fitting error, the boundary point of the denoised wave is identified to obtain the anatomical structure image corresponding to the target muscle. The anatomical structure image is segmented to obtain the target structure image corresponding to the target muscle; Feature information at different scales is extracted from the target structure image, and then adaptive clustering is performed through variable convolution to obtain the target three-dimensional result corresponding to the target muscle. The target 3D result is registered and compared with a preset anatomical atlas database to mark key avoidance structures and determine safe puncture boundaries; Based on the safe puncture boundary, the puncture target point and alternative paths are determined in the target three-dimensional result of the target muscle; The near-infrared laser navigation module projects corresponding light spot marks and target paths onto the target muscle based on the spatial transformation relationship between the puncture target point and the alternative paths.

2. The method according to claim 1, characterized in that, The step of denoising the initial sound wave to obtain the corresponding denoised sound wave includes: The initial sound wave is preprocessed by standardization to calculate the dynamic mean and standard deviation of the initial sound wave and obtain the standardized information corresponding to the initial sound wave. The standardized information is subjected to empirical mode decomposition, local extreme points of the standardized information are detected, and upper and lower envelopes are constructed to obtain multiple intrinsic mode function components. A corresponding noise reference signal is generated for each of the intrinsic mode function components, and a reference benchmark for adaptive noise cancellation is determined based on the normalization information and the frequency domain characteristics of the noise reference signal. The intrinsic mode function components are subjected to noise filtering based on the reference benchmark to obtain the filtering information corresponding to the intrinsic mode function components; The filtered information is linearly superimposed with the residual components obtained from empirical mode decomposition to reconstruct the denoised wave.

3. The method according to claim 2, characterized in that, The step of performing noise filtering on the intrinsic mode function components based on the reference benchmark to obtain the filtering information corresponding to the intrinsic mode function components includes: Determine the current time corresponding to the intrinsic mode function component, and determine the covariance matrix corresponding to the previous time of the current time; The gain vector corresponding to the current time is determined based on the covariance matrix and the reference benchmark. The gain vector is used to determine the degree of influence of the intrinsic mode function components at the current time on the update of the initial weight vector. The prediction output corresponding to the current time is predicted based on the initialized weight vector, and the prior error is determined based on the prediction output and the intrinsic mode function components. Using the prior error and the gain vector, the initial weight vector and the covariance matrix are synchronously and recursively updated to obtain the updated target weight vector and target matrix. Based on the target weight vector and the target matrix, perform instantaneous filtering on the intrinsic mode function components to obtain the filtering information corresponding to the intrinsic mode function components.

4. The method according to claim 1, characterized in that, The step of identifying the boundary points of the denoised wave based on the mean square relative fitting error to obtain the anatomical structure image corresponding to the target muscle includes: Establish mathematical relationship equations between signal points in the denoised wave, and obtain signal characteristic parameters corresponding to the denoised wave through data analysis using the mathematical relationship equations. The denoised wave is segmented based on the mean square relative fitting error. Starting from the signal starting point in the denoised wave, only one data point is added at a time, and the fitting error of the current segment is calculated. If the fitting error is less than the threshold, the points are added again. If the fitting error exceeds the threshold, the data point is recorded as a potential segmentation point. The search step size is increased to a preset maximum value based on the potential segmentation points. The fitting error is monitored until it exceeds the threshold. Then, the search step size is reduced until the target step size and the corresponding precise segmentation points are determined. Based on the precise segmentation points and the denoised waveform, an image of the anatomical structure corresponding to the target muscle is generated.

5. The method according to claim 1, characterized in that, The step of performing image segmentation processing on the anatomical structure image to obtain the target structure image corresponding to the target muscle includes: Calculate the average gray value of the anatomical structure image, and determine an initial threshold based on the average gray value; The anatomical structure image is segmented according to the initial threshold to obtain a first target region and a first background region; Based on the initial threshold, the gray levels in the gray histogram of the first target region are reclassified sequentially, the weighted variance after each adjustment is calculated, and the adjustment threshold is obtained. Obtain the average rate of change and intra-class variance of the gray pixels in the first target region under the adjusted threshold; The first target region is further segmented based on the average rate of change and the intra-class variance to obtain a second target region and a second background region. The target structure image corresponding to the target muscle is determined based on the second target region, the first background region, and the second background region.

6. The method according to claim 1, characterized in that, The target structure image includes target images from different viewpoints. The step of extracting feature information at different scales from the target structure image and then performing adaptive clustering through variable convolution to obtain the target 3D result corresponding to the target muscle includes: Based on the feature extraction layer of the deep optimization model, the target images from different perspectives are adaptively aggregated with features of different scales through deformable convolution operations to obtain the target fusion features corresponding to the target images; Based on the information determination layer of the depth optimization model, the target images from different viewpoints are transformed by differentiable homography to map the features of other viewpoint images onto multiple hypothetical depth planes of the reference viewpoint, thereby constructing a matching cost body. The matching cost body is used to characterize the matching degree of each pixel under different depth assumptions. Based on the cost self-learning layer of the deep optimization model, the deformable convolution is used again to smooth and optimize the outlier value of the matching cost body according to the target fusion feature, so as to obtain the regression depth map corresponding to the optimized matching cost body. Information is fused from the regression depth map and the target structure image to obtain the target three-dimensional result corresponding to the target muscle.

7. The method according to any one of claims 1-6, characterized in that, The method further includes: During the puncture of the target muscle based on the light spot markers and the target path, real-time images are continuously received, and the three-dimensional position of the echo-enhanced puncture needle tip is dynamically tracked. The three-dimensional position is compared with the target path to obtain the target comparison result. Based on the target comparison results, real-time offset prompts and correction guidance are provided in the fusion display interface to obtain target guidance results.

8. An injection path guidance system based on combined ultrasound and laser navigation, characterized in that, The system includes: The data acquisition module is used to acquire the initial sound wave obtained by the ultrasound device detecting the target muscle, and to perform noise reduction processing on the initial sound wave to obtain the corresponding noise-reduced sound wave. The data recognition module is used to identify the boundary points of the denoised wave based on the mean square relative fitting error to obtain the anatomical structure image corresponding to the target muscle. The segmentation processing module is used to perform image segmentation processing on the anatomical structure image to obtain the target structure image corresponding to the target muscle; The data analysis module is used to extract feature information at different scales from the target structure image and then perform adaptive clustering through variable convolution to obtain the target three-dimensional result corresponding to the target muscle. The comparison and analysis module is used to register and compare the target three-dimensional results with a preset anatomical atlas database, mark key avoidance structures, and determine safe puncture boundaries; The path selection module is used to determine the puncture target point and alternative paths in the target three-dimensional result of the target muscle based on the safe puncture boundary. The information conversion module is used to project corresponding light spot marks and target paths onto the target muscle based on the spatial conversion relationship between the puncture target point and the alternative paths through the near-infrared laser navigation module.

9. A terminal device, characterized in that, The terminal device includes a processor and a memory; The memory is used to store computer programs; The processor is used to execute the computer program and, in executing the computer program, implement the injection path guidance method based on ultrasound and laser combined navigation as described in any one of claims 1 to 7.

10. A computer storage medium for computer storage, characterized in that, The computer storage medium stores one or more programs, which can be executed by one or more processors to implement the steps of the injection path guidance method based on ultrasound and laser combined navigation as described in any one of claims 1 to 7.

Images