An ultrasonic robot personalized scanning path intelligent planning method

Abstract

This invention discloses a personalized intelligent planning method for ultrasound robot scanning paths, relating to the field of ultrasound robots. The method includes: collecting anatomical structure information and clinical scanning needs information of the target patient's area to be scanned; establishing a patient-specific basic information set; wherein the anatomical structure information includes the spatial distribution characteristics of various tissues in the area; and the clinical needs information includes preset key areas of focus and scanning accuracy indicators. This invention establishes a patient-specific basic information set by collecting anatomical structure information and clinical scanning needs information of the patient's area to be scanned, analyzes spatial correlation features, and embeds real-time physiological and anatomical position dynamic correlation rules to construct a dynamic anatomical correlation model, which can accurately match the individual anatomical dynamic changes of the patient.

Description

Technical Field

[0001] This invention relates to the field of ultrasonic robot technology, specifically to an intelligent planning method for personalized scanning paths of ultrasonic robots. Background Technology

[0002] Ultrasonic robot scanning path technology plans the optimal path based on medical imaging and anatomical structure data, and combines force / visual sensors to adjust the probe position in real time, ensuring that the probe fits the body surface and the scanning area is fully covered, while ensuring the clarity of ultrasound images, providing stable technical support for accurate diagnosis of abdominal, cardiovascular and other parts of the body.

[0003] Patent application No. 202410357585.6 discloses an adaptive force control method for an autonomous breast ultrasound scanning robot, comprising the following steps: Step 1: Scanning the breast with an ultrasound probe and calculating the scanning speed based on the scanning trajectory; Step 2: Establishing a mathematical model for deformation estimation, identifying and updating deformation parameters online; Step 3: Determining the initial expected force based on the deformation estimation model; Step 4: Establishing a quality evaluation model for the contact between the ultrasound probe and the breast, and correcting the expected force. This application aims to address the problem that "traditional breast ultrasound scanning faces challenges in adaptability and accuracy due to the varying sizes, shapes, and hardness of breasts in different patients."

[0004] However, ultrasound robots are widely used in physical examinations, and how to adaptively plan scanning paths to be compatible with various testing scenarios is a problem that urgently needs to be solved.

[0005] To address this, we propose an intelligent planning method for personalized scanning paths of ultrasonic robots. Summary of the Invention

[0006] In view of the above-mentioned shortcomings of the existing technology, the present invention provides an intelligent planning method for personalized scanning paths of ultrasonic robots, which can effectively solve the problems of the existing technology.

[0007] To achieve the above objectives, the present invention is implemented through the following technical solutions;

[0008] This invention discloses an intelligent planning method for personalized scanning paths of an ultrasonic robot, comprising:

[0009] The system collects anatomical information and clinical scanning needs of the target patient's area to be scanned, establishing a patient-specific basic information set. This set includes anatomical information such as the spatial distribution characteristics of various tissues in the area, and clinical needs information such as preset key areas of focus and scanning accuracy indicators. The system analyzes the spatial correlation characteristics of the anatomical information, constructs a dynamic anatomical correlation model of the area to be scanned, and embeds dynamic correlation rules between the patient's real-time physiological state and the location of anatomical structures into the model. Based on the dynamic anatomical correlation model and clinical scanning needs, personalized scanning constraint parameters for the ultrasound robot are set, including the spatial boundary range of probe movement, actively avoided sensitive anatomical areas, and the image quality benchmark maintained during the scanning process.

[0010] Using constraint parameters as boundary conditions, the optimization direction is to completely cover the target scanning area, minimize the total path length, and minimize probe posture changes. An initial personalized scanning path is generated by analyzing the spatial distribution characteristics of the target area. The initial scanning path is imported into a virtual scanning scene to simulate the robot's scanning actions, verifying the completeness of the path's coverage of the target area and the matching degree of the constraint parameters. The path parameters are iteratively optimized based on the verification results to output the optimal scanning path. When the ultrasound robot executes the optimal scanning path, probe position information and ultrasound image quality information are collected in real time. If anatomical structure position deviation or image quality does not meet the benchmark is detected, path correction parameters are generated in real time and dynamically adapted to the current path.

[0011] Furthermore, when collecting anatomical information of the target patient's area to be scanned, the preoperative CT or MRI three-dimensional image data of the target patient's area to be scanned are obtained, and the three-dimensional contours and spatial coordinates of each tissue in the area are extracted by voxel segmentation algorithm.

[0012] The ultrasound robot is activated in pre-scan mode to acquire real-time two-dimensional ultrasound image sequences of the area to be scanned. Based on image registration, the real-time two-dimensional contour is fused and calibrated with the preoperative three-dimensional contour to eliminate the deviation between the image data and the actual anatomical position, and to obtain anatomical structure information that includes tissue spatial distribution characteristics and real-time positional deviation.

[0013] Furthermore, the stage of analyzing the spatial correlation features of anatomical structural information and constructing a dynamic anatomical correlation model of the area to be scanned follows the following:

[0014] Based on the three-dimensional contours and spatial coordinates of each organization, a topological relationship matrix is ​​used to quantify the spatial relationship characteristics between organizations. The matrix element is defined as the combination of the shortest spatial distance and the relative azimuth angle between any two organizations.

[0015] Based on the mapping relationship between patients' historical physiological signal data and anatomical structure displacement, a dynamic association rule base is established. Each rule in the rule base takes the physiological signal value range as input and the displacement vector of the corresponding anatomical structure as output. The topological relationship matrix is ​​coupled with the dynamic association rule base to finally generate a dynamic anatomical association model.

[0016] Furthermore, in setting the personalized scanning constraint parameters for the ultrasound robot, the spatial boundary range of the probe's movement is determined by the following formula:

[0017] ;

[0018] In the formula: This refers to the spatial boundary range within which the probe can move. The real-time position coordinates of the probe in a three-dimensional rectangular coordinate system, determined with the geometric center of the area to be scanned as the origin; The spatial coordinates of the geometric center of the area to be scanned; This refers to the minimum safe distance between the probe and the surface of the sensitive anatomical area; The equivalent radius of the ultrasonic probe tip; This represents the limit range of probe movement along the vertical direction; The set of spaces occupied by sensitive anatomical areas to be actively avoided;

[0019] Among them, the spatial boundary range It also meets the requirements of probe posture for maintaining image quality benchmarks during the scanning process.

[0020] Furthermore, the process for generating the initial personalized scanning path is as follows:

[0021] Based on the spatial topology of the area to be scanned output by the dynamic anatomical association model, the area to be scanned is discretized into three-dimensional grid cells using the grid method. A multi-objective optimization function is constructed with the optimization objectives of completely covering the target scanning area, minimizing the total path length, and minimizing probe pose changes.

[0022] ;

[0023] In the formula: To optimize the weighting coefficients; This is the coverage integrity coefficient of the raster cell; This is the normalized value of the total path length; This is the normalized value of the number of probe attitude changes;

[0024] Then, based on the particle swarm optimization algorithm, the above function is solved to obtain the result that makes The grid traversal order with the minimum value is selected, and this order is converted into a spatial coordinate sequence of the probe to generate an initial personalized scanning path.

[0025] Furthermore, when verifying the coverage integrity of the target area by the verification path, in a virtual scanning scenario, all ultrasonic scanning sections corresponding to the initial scanning path are extracted to quantify the coverage integrity verification index. ;

[0026] In the formula: For effective coverage; This represents the total number of scan sections included in the initial scan path; Let be the overlap area between the i-th scanned section and its adjacent sections; Let be the area of ​​the i-th scanned section; Weighting coefficients for key areas of focus; The number of valid sampling points within the preset key focus area in the initial scanning path; The minimum number of sampling points required to achieve the scanning accuracy target in the key areas of focus;

[0027] like If the coverage value is less than the preset coverage threshold, the coverage integrity is deemed unsatisfactory, triggering path iteration optimization.

[0028] Furthermore, the dynamic association rule between the patient's real-time physiological state and the location of anatomical structures embedded in the dynamic anatomical association model is as follows:

[0029] Data on anatomical displacement of the patient's anatomical site under different physiological conditions were collected, and a mapping relationship between physiological signal parameters and anatomical displacement vectors was established. , Represents the displacement vector of the anatomical structure. Indicates the respiratory phase angle. Indicates the percentage of heart rate cycles;

[0030] Based on this mapping relationship, a real-time parameter update interface is set in the dynamic anatomical association model. When the ultrasound robot performs a scan, the physiological signal acquisition module acquires the data in real time. and The numerical values ​​are substituted into the mapping relationship to calculate the displacement vector of the current anatomical structure, so as to dynamically correct the spatial coordinates of each tissue in the model.

[0031] Furthermore, the image quality benchmark maintained during the scanning process, when set, conforms to:

[0032] Based on the scanning accuracy indicators in the clinical scanning needs information, image quality benchmark parameters are set for different tissue types in the area to be scanned, including:

[0033] Echo intensity threshold range, signal-to-noise ratio threshold, and boundary sharpness index;

[0034] The echo intensity threshold range, signal-to-noise ratio threshold, and boundary sharpness index are integrated into an image quality assessment matrix, which is used as a benchmark for image quality monitoring during the scanning process.

[0035] Furthermore, the steps for generating path correction parameters in real time and dynamically adapting to the current path when anatomical structure positional deviations or image quality failing to meet benchmarks are as follows:

[0036] When an anatomical structure positional shift is detected, a positional correction amount is calculated. , These represent the anatomical offset correction factor and the image quality correction factor, respectively. Represents the actual offset vector of the anatomical structure. Represents the image quality deviation vector;

[0037] Will The corrected real-time path is generated by superimposing it onto the coordinate sequence of the current probe path. At the same time, if the image quality does not meet the benchmark, the probe gain parameters and focusing depth are adjusted synchronously to ensure that the corrected path meets the image quality benchmark requirements.

[0038] Furthermore, when iteratively optimizing path parameters based on the verification results, if the coverage integrity in the virtual verification does not meet the standard, then... The grid traversal density of the initial path is adjusted based on the difference from the preset coverage threshold; that is, the larger the difference, the higher the grid traversal density is increased.

[0039] If the number of probe attitude changes exceeds the preset range, the probe attitude angles of adjacent grid cells in the path are smoothed, and the transition trajectory of adjacent attitude angles is fitted by Bezier curve.

[0040] If the total path length exceeds the preset optimization target, delete the raster cells that are repeatedly covered in the path and retain the redundant coverage cells in the key areas of interest.

[0041] After each parameter adjustment, the scanning path is regenerated and a virtual verification is performed until the path meets all constraint parameter requirements, at which point the iteration terminates.

[0042] Compared with the known prior art, the technical solution provided by this invention has the following beneficial effects:

[0043] This invention provides an intelligent planning method for personalized ultrasound robot scanning paths. During execution, this method establishes a unique basic information set by collecting anatomical structures of the patient's scanned area and clinical scanning needs. It analyzes spatial correlation features and embeds real-time physiological and anatomical position dynamic correlation rules to construct a dynamic anatomical correlation model. This model can accurately match the dynamic changes in the patient's individual anatomy. The set scanning constraint parameters clearly define the probe's activity boundaries and actively avoid sensitive anatomical areas while ensuring image quality standards. An initial path is generated with the goals of complete coverage, shortest path, and minimal posture changes, and then iteratively optimized through virtual verification. This improves path rationality and scanning efficiency. During execution, real-time acquisition of position and image quality information and dynamic path correction enable timely responses to anatomical deviations and substandard quality issues. Ultimately, this achieves safe, accurate, and efficient personalized ultrasound scanning, meeting diverse clinical scanning needs, reducing operational risks, and improving scanning effectiveness and reliability. Attached Figure Description

[0044] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0045] Figure 1 This is a flowchart illustrating a method for intelligent planning of personalized scanning paths for ultrasonic robots. Detailed Implementation

[0046] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0047] The present invention will be further described below with reference to embodiments.

[0048] Example:

[0049] This embodiment presents an intelligent planning method for personalized scanning paths of an ultrasonic robot, such as... Figure 1 As shown, it includes:

[0050] Collect anatomical structure information and clinical scanning needs information of the target patient's scanned area, and establish a patient-specific basic information set. The anatomical structure information includes the spatial distribution characteristics of various tissues in the area, and the clinical needs information includes preset key areas of focus and scanning accuracy indicators.

[0051] When collecting anatomical information of the site to be scanned in the target patient, obtain the preoperative CT or MRI three-dimensional image data of the site to be scanned in the target patient, and extract the three-dimensional contour and spatial coordinates of each tissue in the site through the voxel segmentation algorithm.

[0052] The ultrasound robot is activated in pre-scanning mode to acquire real-time two-dimensional ultrasound image sequences of the area to be scanned. Based on image registration, the real-time two-dimensional contour is fused and calibrated with the preoperative three-dimensional contour to eliminate the deviation between the image data and the actual anatomical position, and to obtain anatomical structure information including tissue spatial distribution characteristics and real-time positional deviation.

[0053] The spatial correlation features of anatomical structure information are analyzed, a dynamic anatomical correlation model of the area to be scanned is constructed, and dynamic correlation rules between the patient's real-time physiological state and the location of anatomical structures are embedded in the model.

[0054] The stage of analyzing the spatial correlation features of anatomical structural information and constructing a dynamic anatomical correlation model of the area to be scanned follows the principle of:

[0055] Based on the three-dimensional contours and spatial coordinates of each organization, a topological relationship matrix is ​​used to quantify the spatial relationship characteristics between organizations. The matrix element is defined as the combination of the shortest spatial distance and the relative azimuth angle between any two organizations.

[0056] Based on the mapping relationship between patients' historical physiological signal data and anatomical structure displacement, a dynamic association rule base is established. Each rule in the rule base takes the range of physiological signal values ​​as input and the displacement vector of the corresponding anatomical structure as output. The topological relationship matrix is ​​coupled with the dynamic association rule base to finally generate a dynamic anatomical association model.

[0057] The dynamic association rules between the patient's real-time physiological state and the location of anatomical structures embedded in the dynamic anatomical association model are as follows:

[0058] Data on anatomical displacement of the patient's anatomical site under different physiological conditions were collected, and a mapping relationship between physiological signal parameters and anatomical displacement vectors was established. , Represents the displacement vector of the anatomical structure. Indicates the respiratory phase angle. Indicates the percentage of heart rate cycles;

[0059] Based on this mapping relationship, a real-time parameter update interface is set in the dynamic anatomical association model. When the ultrasound robot performs a scan, the physiological signal acquisition module acquires the data in real time. and The numerical values ​​are substituted into the mapping relationship to calculate the displacement vector of the current anatomical structure, so as to dynamically correct the spatial coordinates of each tissue in the model;

[0060] Among them, the patient's scanned area is in different physiological states, including different respiratory phases and heart rate cycle stages, and the physiological signal parameters include respiratory phase angle and heart rate cycle percentage.

[0061] Based on the dynamic anatomical correlation model and clinical scanning needs, personalized scanning constraint parameters for the ultrasound robot are set, including the spatial boundary range of the probe's movement, the sensitive anatomical areas to be actively avoided, and the image quality benchmark maintained during the scanning process.

[0062] The image quality baseline maintained during the scanning process, when set, follows the following:

[0063] Based on the scanning accuracy indicators in the clinical scanning needs information, image quality benchmark parameters are set for different tissue types in the area to be scanned, including:

[0064] Echo intensity threshold range, signal-to-noise ratio threshold, and boundary sharpness index;

[0065] The echo intensity threshold range, signal-to-noise ratio threshold, and boundary sharpness index are integrated into an image quality assessment matrix, which is used as a benchmark for image quality monitoring during the scanning process.

[0066] Among them, the echo intensity threshold range is set based on the normal ultrasound echo characteristics of the tissue. If it exceeds the range, the image quality is judged to be substandard. The signal-to-noise ratio threshold is set as the minimum ratio of the ultrasound image signal intensity to the noise intensity of the tissue. If it is lower than the threshold, the image quality is judged to be substandard. The boundary clarity index is calculated by the edge detection algorithm to calculate the gradient value of the tissue contour and set a minimum threshold for the gradient value. If it is lower than the threshold, the image quality is judged to be substandard.

[0067] In setting the personalized scanning constraint parameters for the ultrasound robot, the spatial boundary range of the probe's movement is determined by the following formula:

[0068] ;

[0069] In the formula: The spatial boundary range in which the probe can move is represented by a three-dimensional rectangular coordinate system, with the origin of the coordinate system being the geometric center of the part to be scanned. The real-time position coordinates of the probe in a three-dimensional rectangular coordinate system, determined with the geometric center of the area to be scanned as the origin; The spatial coordinates of the geometric center of the area to be scanned; The minimum safe distance between the probe and the surface of sensitive anatomical areas is set based on clinical safety standards to achieve spatial isolation of sensitive anatomical areas that are actively avoided (such as areas with dense blood vessels, nerve bundles, etc.). The equivalent radius of the ultrasonic probe tip; This refers to the limit range of probe movement along the vertical direction (human body axis); The set of spaces occupied by sensitive anatomical regions to be actively avoided, the boundaries of which are defined by the three-dimensional contour coordinates of the sensitive regions in the anatomical structure information;

[0070] The above formula is based on a three-dimensional rectangular coordinate system, with the geometric center of the area to be scanned set as the origin. It incorporates key parameters such as the real-time position coordinates of the probe, the spatial coordinates of the geometric center, the minimum safe distance between the probe and the sensitive anatomical area, the equivalent radius of the probe tip, the vertical movement limit range, and the spatial set of the sensitive anatomical area. At the same time, it takes into account the requirements of the image quality benchmark for the probe posture, ensuring that the probe can actively avoid sensitive areas such as densely vascularized areas and nerve bundles within the defined range, while maintaining the preset imaging angle with the tissue to be scanned, so that the echo signal intensity meets the quality threshold, achieving dual protection of safety and imaging quality.

[0071] Among them, the spatial boundary range At the same time, it meets the requirements of the probe posture for maintaining the image quality benchmark during the scanning process, that is, the probe at any position within this range can maintain the preset imaging angle with the tissue to be scanned, so as to ensure that the echo signal intensity meets the threshold range in the image quality benchmark.

[0072] Using constraint parameters as boundary conditions, with the optimization direction being complete coverage of the target scanning area, shortest total path length, and fewest probe attitude changes, an initial personalized scanning path is generated by analyzing the spatial distribution characteristics of the target area.

[0073] The initial personalized scanning path generation process is as follows:

[0074] Based on the spatial topology of the area to be scanned output by the dynamic anatomical association model, the area to be scanned is discretized into three-dimensional grid cells using the grid method. A multi-objective optimization function is constructed with the optimization objectives of completely covering the target scanning area, minimizing the total path length, and minimizing probe pose changes.

[0075] ;

[0076] In the formula: To optimize the weighting coefficients; This is the coverage integrity coefficient of the raster cell; This is the normalized value of the total path length; This is the normalized value of the number of probe attitude changes;

[0077] The above formula is based on the area to be scanned after being discretized by the grid method. It introduces three core indicators: coverage integrity coefficient, normalized total path length, and normalized probe attitude change number. By setting positive optimization weight coefficients with a sum of 1, the requirements for complete coverage of the target area, shortening the total path length, and reducing probe attitude changes are balanced. The coverage integrity coefficient is the ratio of the number of covered grids to the total number of grids. The normalized path length is the ratio of the actual path length to the maximum diagonal length of the area to be scanned. The normalized attitude change number is the ratio of the actual number of changes to the theoretical maximum number of changes. This ensures that the quantification standards of each indicator are consistent. Subsequently, the optimal grid traversal order can be efficiently solved by combining the particle swarm optimization algorithm.

[0078] in, All are positive numbers, and their sum is 1. , Take the ratio of the number of covered grid cells to the total number of grid cells. Take the ratio of the actual path length to the maximum diagonal length of the area to be scanned. Take the ratio of the actual number of transformations to the theoretical maximum number of transformations;

[0079] Then, based on the particle swarm optimization algorithm, the above function is solved to obtain the result that makes The grid traversal order with the minimum value is selected, and this order is converted into a spatial coordinate sequence of the probe to generate an initial personalized scanning path;

[0080] The initial scanning path is imported into the virtual scanning scene to simulate the robot performing scanning actions. The completeness of the path's coverage of the target area and the matching degree of the constraint parameters are verified. Based on the verification results, the path parameters are iteratively optimized to output the optimal scanning path.

[0081] When verifying the coverage integrity of the target area by the scanning path, all ultrasonic scanning sections corresponding to the initial scanning path are extracted in the virtual scanning scenario to quantify the coverage integrity verification index. ;

[0082] In the formula: For effective coverage; This represents the total number of scan sections included in the initial scan path; Let be the overlap area between the i-th scanned section and its adjacent sections; Let be the area of ​​the i-th scanned section; Weighting coefficients for key areas of focus; The number of valid sampling points within the preset key focus area in the initial scanning path; The minimum number of sampling points required to achieve the scanning accuracy target in the key areas of focus;

[0083] The above formula quantifies the path coverage integrity from two aspects: the overlap of scanning sections and the sampling of key areas of concern. It calculates the ratio of the effective overlap area to the total area of ​​all scanning sections, combines the ratio of the number of effective sampling points in key areas of concern to the minimum number of sampling points required, and sets a weight coefficient of not less than 1 for key areas of concern to highlight the scanning priority of key areas in clinical needs. At the same time, it clarifies the rules that the overlap area is counted as 0 when the overlap rate does not reach the preset threshold and that sampling points are not included in the effective sampling points when the sampling point spacing exceeds the accuracy requirements. This ensures that the quantitative results accurately reflect the actual coverage effect, provides a clear and clinically relevant judgment standard for path verification, and avoids the shortcomings of traditional coverage assessment that ignore key areas and the effectiveness of overlap.

[0084] like If the coverage value is less than the preset coverage threshold, the coverage integrity is deemed unsatisfactory, triggering path iteration optimization.

[0085] Among them, for The overlap rate must not be lower than a preset threshold, otherwise The value is 0. As a preset value, ≥1, in determining At the same time, the spacing between sampling points must not be greater than the maximum spacing required by the scanning accuracy index; otherwise, it will not be counted.

[0086] When the ultrasound robot executes the optimal scanning path, it collects probe position information and ultrasound image quality information in real time. If it detects a shift in the position of the anatomical structure or an image quality that does not meet the benchmark, it generates path correction parameters in real time and dynamically adapts to the current path.

[0087] The steps for generating path correction parameters in real time and dynamically adapting to the current path when anatomical structure positional deviations or image quality failing to meet benchmarks are as follows:

[0088] When an anatomical structure positional shift is detected, a positional correction amount is calculated. , These represent the anatomical offset correction factor and the image quality correction factor, respectively. Represents the actual offset vector of the anatomical structure. Represents the image quality deviation vector;

[0089] Will The corrected real-time path is generated by superimposing it onto the coordinate sequence of the current probe path. At the same time, if the image quality does not meet the benchmark, the gain parameters and focus depth of the probe are adjusted synchronously to make the corrected path meet the image quality benchmark requirements.

[0090] The above formula uses anatomical offset correction coefficient and image quality correction coefficient to associate the actual offset vector of the anatomical structure with the image quality deviation vector, respectively. The anatomical offset vector is calculated by the dynamic anatomical association model as the difference between the current and initial anatomical coordinates. The image quality deviation vector is obtained by weighted combination of the difference between the current image quality parameters and the benchmark. At the same time, the anatomical offset correction coefficient is adjusted according to the minimum safe distance between the tissue and the probe. The smaller the safe distance, the larger the coefficient. The image quality correction coefficient is adjusted according to the scanning accuracy index. The higher the accuracy requirement, the larger the coefficient. The correction amount is superimposed on the current probe path coordinate sequence, and the probe gain parameters and focusing depth are adjusted simultaneously to ensure dynamic and accurate adaptation when the anatomical position is offset and the image quality is substandard.

[0091] in, It is calculated in real time by the dynamic anatomical correlation model, that is, the difference between the current anatomical structure coordinates and the initial model coordinates. It is obtained by quantizing the difference between the current image quality parameters and the quality benchmark, such as a weighted combination of echo intensity deviation and signal-to-noise ratio deviation. The value range is (0,1], and the smaller the minimum safe distance S between the tissue and the probe, the better. The larger the value, the lower the value. The smaller the value, the higher the required scanning accuracy. The larger the value, the lower the value. The smaller the value;

[0092] When iteratively optimizing path parameters based on the verification results, if the coverage integrity in the virtual verification does not meet the standard, then... The grid traversal density of the initial path is adjusted based on the difference from the preset coverage threshold; that is, the larger the difference, the higher the grid traversal density is increased.

[0093] If the number of probe attitude changes exceeds the preset range, the probe attitude angles of adjacent grid cells in the path are smoothed, and the transition trajectory of adjacent attitude angles is fitted by Bezier curve.

[0094] If the total path length exceeds the preset optimization target, delete the raster cells that are repeatedly covered in the path and retain the redundant coverage cells in the key areas of interest.

[0095] After each parameter adjustment, the scanning path is regenerated and a virtual verification is performed until the path meets all constraint parameter requirements, at which point the iteration terminates.

[0096] In this embodiment, in the scenario of ultrasound robot scanning path planning, a personalized ultrasound scanning path can be planned by combining the anatomical features of the patient's area to be scanned with clinical needs. It can actively avoid sensitive areas to ensure safety, while ensuring complete coverage of the target area. During the scan, it can adapt to anatomical position deviations and image quality deviations in real time, and can also reduce probe posture changes and path lengths by optimization, taking into account both scanning accuracy and efficiency, and providing reliable support for the accurate acquisition of ultrasound information in clinical practice.

[0097] Referring to the methods in the above embodiments, the following are application examples of the methods in the above embodiments:

[0098] Example 1:

[0099] A hospital used this method to perform a liver ultrasound examination on a patient who needed to undergo the procedure, as follows:

[0100] First, information on the patient's liver anatomy and clinical scanning needs were collected: Preoperative CT 3D images of the patient's liver were obtained, and the 3D contours and spatial coordinates of the liver parenchyma, portal vein (a densely vascularized area), and right lobe lesion (a pre-defined area of ​​focus) were extracted using a voxel segmentation algorithm. The clinical requirement was clearly defined as focusing on scanning lesions approximately 2cm in diameter in the right lobe of the liver, with scanning accuracy ensuring that the spacing between sampling points in the lesion area did not exceed 1mm. Based on this information, a patient-specific basic information set was established.

[0101] Subsequently, the ultrasound robot pre-scan mode was activated to acquire real-time two-dimensional ultrasound image sequences of the liver. Through image registration, the real-time two-dimensional contour was fused and calibrated with the preoperative three-dimensional contour to eliminate the deviation between the image data and the actual anatomical position, and finally obtain anatomical structural information containing the spatial distribution characteristics of various liver tissues and real-time positional deviations.

[0102] Next, a dynamic anatomical correlation model of the liver was constructed: based on the three-dimensional contours and spatial coordinates of the liver parenchyma, portal vein, and lesion, the spatial correlation characteristics of the three were quantified through a topological relationship matrix (e.g., the shortest spatial distance between the liver parenchyma and the portal vein is 8mm and the relative azimuth angle is 35°, and the shortest spatial distance between the liver parenchyma and the lesion is 5mm and the relative azimuth angle is 120°). Combined with the patient's historical respiratory and heart rate data, a mapping relationship between physiological signals and anatomical structure displacement was established (e.g., when the respiratory phase angle is 0° and the heart rate cycle percentage is 50%, the liver displacement vector is 0.3mm vertically downward). A real-time parameter update interface was set in the model. During scanning, the real-time respiratory phase angle and heart rate cycle percentage can be obtained through the physiological signal acquisition module, and the current liver displacement vector can be calculated to correct the spatial coordinates of each tissue.

[0103] Next, personalized scanning constraint parameters were set: Within the probe's movement space boundary, with the liver's geometric center as the origin, the minimum safe distance between the probe and the portal vein was preset to 5mm, the equivalent radius of the probe tip to be 3mm, and the vertical movement limit to be "±20mm from the geometric center z-coordinate". Ultimately, the probe's movable x-coordinate range was determined to be ±2mm from the liver's geometric center x-coordinate, the y-coordinate range similarly, and the z-coordinate range to be ±2mm from the liver's geometric center z-coordinate, within the ±20mm vertical limit, while excluding the space occupied by densely populated portal vein areas. Regarding image quality benchmarks, for liver parenchyma, the echo intensity threshold range was set to 40-60dB, the signal-to-noise ratio threshold to be no less than 30:1, and the minimum threshold for boundary sharpness gradient value to be 25; for lesions in the right lobe of the liver, the echo intensity threshold range was set to 55-75dB, the signal-to-noise ratio threshold to be no less than 35:1, and the minimum threshold for boundary sharpness gradient value to be 30. These parameters were then integrated into an image quality assessment matrix.

[0104] The next step is to generate an initial personalized scanning path: The liver scanning area is discretized into three-dimensional grid cells using the grid method. Optimization weight coefficients are set (coverage integrity coefficient weight 0.5, path length normalization weight 0.3, and pose transformation number normalization weight 0.2). A multi-objective optimization function is constructed, and the grid traversal order corresponding to the minimum value of the function is obtained by solving the particle swarm optimization algorithm. This order is then converted into a probe spatial coordinate sequence to generate the initial scanning path.

[0105] Then, virtual scanning verification and iterative optimization were performed: 30 scanning cross-sections of the initial path were extracted in the virtual scene, and the effective coverage Ceff was calculated to be 92% (preset coverage threshold 85%), which met the coverage integrity standard; at the same time, the number of probe posture changes was detected to be 12 times (preset range ≤ 15 times) and the total path length was 80mm (preset optimization target ≤ 90mm), both of which met the requirements and no further adjustment was needed. The path was determined to be the optimal scanning path.

[0106] Finally, the optimal scanning path is executed: while the ultrasound robot scans along the path, it collects probe position and image quality information in real time. When the patient's breathing causes the liver to shift upward by 0.2 mm, the position correction amount ΔP is calculated to be 0.18 mm and superimposed on the current probe coordinate sequence correction path. At the same time, it is found that the signal-to-noise ratio of the ultrasound image in the lesion area drops to 32:1 (lower than the benchmark of 35:1). The probe gain parameters and focusing depth are immediately adjusted to restore the signal-to-noise ratio to 36:1, ensuring that the scanning process always meets the accuracy and safety requirements, and finally completing a high-quality personalized ultrasound scan of the liver.

[0107] Example 2:

[0108] Taking the thyroid nodule scan of adult patients as an example, the specific procedure for applying this method is as follows:

[0109] Establish a patient-specific basic information set: First, acquire the patient's preoperative CT 3D image data of the thyroid gland. Voxel segmentation is used to extract the 3D contours and spatial coordinates of the left and right thyroid lobes, isthmus, and common carotid artery (sensitive anatomical areas). Simultaneously, clinical scanning needs are identified, and the right thyroid lobe nodule is designated as a preset key area of ​​focus. The scanning accuracy is set to a sampling point spacing of no more than 0.8 mm. Then, the ultrasound robot pre-scan mode is activated to acquire real-time 2D ultrasound image sequences of the thyroid region. Image registration is used to fuse and calibrate the real-time 2D contours with the preoperative 3D contours, eliminating deviations between the image data and the actual anatomical location, and obtaining anatomical structural information including the spatial distribution characteristics of various thyroid tissues and real-time positional deviations.

[0110] A dynamic anatomical correlation model was constructed: Based on the three-dimensional contours and spatial coordinates of various thyroid tissues, a topological relation matrix was used to quantify the spatial correlation characteristics between tissues. For example, the shortest spatial distance between the right lobe of the thyroid gland and the common carotid artery is 5 mm, and the relative azimuth angle is 45°. Combining the mapping relationship between the patient's historical physiological signal data and the displacement of anatomical structures, a dynamic correlation rule base was established (taking the respiratory phase angle and heart rate cycle percentage as inputs, and outputting the thyroid tissue displacement vector accordingly). A real-time parameter update interface was set in the model. When the ultrasound robot performs a scan, the current respiratory phase angle of 28° and heart rate cycle percentage of 35% are obtained in real time through the physiological signal acquisition module. Substituting these values ​​into the mapping relationship, the current displacement vector of the right lobe of the thyroid gland is calculated (x direction +0.9 mm, y direction -0.6 mm, z direction +0.4 mm), and the spatial coordinates of each tissue in the model are dynamically corrected.

[0111] Personalized scanning constraint parameters were set: Regarding image quality benchmarks, based on clinical scanning accuracy indicators, an echo intensity threshold of 35-75 dB, a signal-to-noise ratio threshold of 28:1, and a boundary sharpness gradient threshold of 18 were set for the thyroid parenchyma. These three parameters were integrated into an image quality assessment matrix, used as a benchmark for quality monitoring during the scanning process. When setting the probe's movable spatial boundary range, the geometric center of the thyroid gland was used as the origin of a three-dimensional rectangular coordinate system. According to clinical safety standards, the minimum safe distance between the probe and the common carotid artery was set at 4 mm, and the equivalent radius of the ultrasound probe tip was 2.5 mm. The probe's vertical (human axis) movement limit range was from 1 cm below the mandible to 2 cm above the suprasternal notch. Simultaneously, the space occupied by the common carotid artery was excluded. Ultimately, the probe's movable x-range was determined to be -2.55 cm to 2.55 cm, y-range to 2.55 cm, and z-range to -1 cm to 2 cm. Within this range, the probe could maintain a preset imaging angle with the thyroid tissue, ensuring that the echo signal intensity met the quality thresholds.

[0112] Initial personalized scanning path generation: The thyroid scan area is discretized into three-dimensional grid cells using a grid method. The optimization objectives are complete coverage of the target scan area, shortest total path length, and fewest probe posture changes. Optimization weight coefficients are set (coverage integrity coefficient weight 0.5, normalized total path length weight 0.3, normalized probe posture change count weight 0.2), and a multi-objective optimization function is constructed. This function is solved using a particle swarm optimization algorithm, yielding a grid traversal order with a coverage integrity coefficient of 0.94, a normalized total path length of 0.55, and a normalized probe posture change count of 0.38. This order is then converted into a sequence of probe spatial coordinates to generate the initial personalized scanning path.

[0113] Virtual verification and iterative optimization: The initial scanning path is imported into a virtual scanning scenario to simulate a robot performing scanning actions. 25 ultrasound scanning sections corresponding to the initial path are extracted, and the effective coverage Ceff is calculated: the average overlap area ratio of all scanning sections is 0.28 (all reaching the preset overlap threshold). There are 96 effective sampling points within the right thyroid lobe nodule (a key area of ​​focus) (the minimum number of sampling points required to meet scanning accuracy is 95). The weight coefficient for the key area of ​​focus is set to 1.2. The final calculated Ceff is 0.97, higher than the preset coverage threshold of 0.95. Furthermore, the number of probe posture changes and the total path length both meet the preset optimization targets, eliminating the need for iterative optimization and directly outputting the optimal scanning path.

[0114] Real-time path dynamic adaptation: When the ultrasound robot executes the optimal scanning path, it acquires probe position and image quality information in real time. During this process, it detects a positional shift of the thyroid gland due to swallowing (x-direction +1.2mm, y-direction -0.8mm), calculates the position correction (x-direction -1.2mm, y-direction +0.8mm, z-direction 0mm) and superimposes it onto the current probe path coordinate sequence; subsequently, it detects that the ultrasound image signal-to-noise ratio drops to 26:1 (below the quality benchmark threshold), and simultaneously adjusts the probe gain parameters and focusing depth to restore the image signal-to-noise ratio to 30:1, ensuring that the scanning process always meets the requirements for anatomical position accuracy and image quality.

[0115] In summary, the method described in the above embodiments establishes a dedicated basic information set by collecting information on the anatomical structure of the patient's area to be scanned and clinical scanning needs. It analyzes spatial correlation features and embeds real-time physiological and anatomical position dynamic correlation rules to construct a dynamic anatomical correlation model. This model can accurately match the dynamic changes in the individual anatomy of the patient. The set scanning constraint parameters can clearly define the boundaries of probe movement and actively avoid sensitive anatomical areas, while ensuring image quality benchmarks. An initial path is generated with the goals of complete coverage, shortest path, and minimal posture changes, and iteratively optimized through virtual verification, which can improve path rationality and scanning efficiency. During execution, position and image quality information are collected in real time and the path is dynamically corrected, which can promptly address anatomical deviation and substandard quality issues. Ultimately, this achieves safe, accurate, and efficient personalized ultrasound scanning, meets different clinical scanning needs, reduces operational risks, and improves scanning effects and reliability.

[0116] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An ultrasound robotic personalized scan path intelligent planning method, characterized in that, The application relates to a method for setting personalized scanning parameters of an ultrasonic robot. The method comprises the following steps: Collecting anatomical structure information and clinical scanning requirement information of a target patient's to-be-scanned part, wherein the anatomical structure information comprises spatial distribution characteristics of each tissue of the part, and the clinical scanning requirement information comprises a preset key attention area and a scanning accuracy index; Analyzing spatial correlation characteristics of the anatomical structure information, constructing a dynamic anatomical correlation model of the to-be-scanned part, and embedding dynamic correlation rules of real-time physiological states of the patient and anatomical structure positions in the model; The step of analyzing spatial correlation characteristics of the anatomical structure information and constructing a dynamic anatomical correlation model of the to-be-scanned part is subject to the following conditions: Based on three-dimensional contours and spatial coordinates of each tissue, a topological relationship matrix is used to quantize spatial correlation characteristics among the tissues, wherein a matrix element is defined as a combination of a shortest spatial distance and a relative orientation angle of any two tissues; Based on a mapping relationship between historical physiological signal data of the patient and anatomical structure displacement, a dynamic correlation rule database is established, each rule in the rule database takes a physiological signal value interval as input and takes a displacement vector of a corresponding anatomical structure as output, the topological relationship matrix is coupled with the dynamic correlation rule database, and finally a dynamic anatomical correlation model is generated; Combined with the dynamic anatomical correlation model and the clinical scanning requirement information, personalized scanning constraint parameters of the ultrasonic robot are set, including a spatial boundary range in which a probe can move, a sensitive anatomical region that is actively avoided, and an image quality benchmark that is maintained during a scanning process; The constraint parameters are used as boundary conditions, and the optimization direction is to completely cover the target scanning area, to minimize the total length of a path and to minimize the transformation of a probe posture, an initial scanning path is generated by analyzing spatial distribution characteristics of the target scanning area; The initial scanning path is imported into a virtual scanning scene, a scanning action of the robot is simulated, the coverage integrity of the path to the target scanning area, the total length of the path, the number of times of transformation of the probe posture and the matching degree of the constraint parameters are checked, and path parameters are iteratively optimized according to the checking result, so as to output an optimal scanning path; 2. The method of claim 1, wherein, When the ultrasonic robot executes the optimal scanning path, real-time probe position information and ultrasonic image quality information are collected, if it is detected that an anatomical structure position deviates or an image quality does not reach a benchmark, path correction parameters are generated in real time and are dynamically adapted to a current path. When the anatomical structure information of the target patient's to-be-scanned part is collected, preoperative CT or MRI three-dimensional image data of the target patient's to-be-scanned part is acquired, three-dimensional contours and spatial coordinates of each tissue of the part are extracted through a voxel segmentation algorithm; A pre-scanning mode of the ultrasonic robot is started synchronously, real-time two-dimensional ultrasonic image sequences of the to-be-scanned part are collected, real-time two-dimensional contours are fused and calibrated with the preoperative three-dimensional contours based on image registration, deviations of image data and actual anatomical positions are eliminated, and anatomical structure information containing spatial distribution characteristics of tissues and real-time position deviations is obtained; 3. The method of claim 1, wherein, The real-time two-dimensional contours are obtained by contour extraction on the real-time two-dimensional ultrasonic image sequences collected in the pre-scanning. ； In the formula: is the spatial boundary range in which the probe is movable; is the real-time position coordinate of the probe in the three-dimensional rectangular coordinate system determined with the geometric center of the part to be scanned as the origin; is the spatial coordinate of the geometric center of the part to be scanned; is the minimum safe distance between the probe and the surface of the sensitive anatomical region; is the equivalent radius of the front end of the ultrasound probe; is the motion limit range of the probe along the vertical direction; is the spatial set occupied by the sensitive anatomical region to be actively avoided; wherein the spatial boundary range while meeting the requirements of the image quality criteria maintained during the scan on the probe attitude.

4. The method of claim 1, wherein, In the step of setting the personalized scanning constraint parameters of the ultrasonic robot, the spatial boundary range in which the probe can move is determined by calculation according to the following formula: The generation process of the initial scanning path is as follows: Based on the spatial topology of the region to be scanned output by the dynamic anatomical correlation model, the region to be scanned is discretized into three-dimensional grid cells using a grid method; Taking complete coverage of the target scanning region, shortest total path length, and least probe attitude transformation as optimization objectives, a multi-objective optimization function is constructed: ； In the formula: is an optimized weight coefficient; is a coverage integrity coefficient of the grid unit; is a normalized value of the total length of the path; is a normalized value of the number of probe attitude transformations; Solving the above function based on particle swarm optimization algorithm, the grid traversal order that makes the minimum value is obtained, and the order is converted into the spatial coordinate sequence of the probe to generate the initial scan path. Solving the above function based on particle swarm optimization algorithm, the grid traversal order that makes the minimum value is obtained, and the order is converted into the spatial coordinate sequence of the probe to generate the initial scan path.

5. The method of claim 1, wherein, The coverage integrity of the target scanning area is checked in the virtual scanning scene, all ultrasound scanning sections corresponding to the initial scanning path are extracted to quantify the coverage integrity checking index ; In the formula: is the effective coverage degree; is the total number of scanning sections included in the initial scanning path; is the overlapping area of the i-th scanning section and the adjacent scanning section; is the area of the i-th scanning section; is the weight coefficient of the preset key attention area; is the number of effective sampling points of the initial scanning path in the preset key attention area; is the minimum number of sampling points required for the preset key attention area to reach the scanning accuracy index. If If the coverage integrity is not up to the standard, the path iteration optimization is triggered.

6. The method of claim 1, wherein, The dynamic correlation rules between the real-time physiological state of the patient and the position of the anatomical structure embedded in the dynamic anatomical correlation model are: Collecting anatomical structure displacement data of a patient to-be-scanned part under different physiological states, and establishing a mapping relationship between physiological signal parameters and anatomical structure displacement vectors , indicates an anatomical structure displacement vector, indicates a respiratory phase angle, indicates a heart rate period proportion; Based on the mapping relationship, a real-time parameter updating interface is set in the dynamic anatomical correlation model, when the ultrasound robot performs scanning, the physiological signal acquisition module is used to acquire the values of the physiological signals in real time and The displacement vector of the current anatomical structure is calculated by substituting the values into the mapping relationship, so as to dynamically correct the spatial coordinates of each tissue in the model.

7. The method of claim 1, wherein, When the image quality criteria maintained during the scanning process are set, they are subject to: Based on the scanning accuracy indicators in the clinical scanning requirement information, the image quality criteria parameters are set for different tissue types in the region to be scanned, including: echo intensity threshold range, signal-to-noise ratio threshold, and boundary definition index; The echo intensity threshold range, signal-to-noise ratio threshold, and boundary definition index are integrated into an image quality evaluation matrix, which is used as a reference for image quality monitoring during the scanning process.

8. The method of claim 1, wherein, The step of generating path correction parameters in real time and dynamically adapting the current path when the position of the anatomical structure deviates or the image quality does not meet the criteria is: When the anatomical structure position offset is detected, a position correction amount is calculated , respectively represent an anatomical offset correction coefficient, an image quality correction coefficient, represents an actual offset vector of the anatomical structure, represents an image quality deviation vector; Will The corrected real-time path is generated by superimposing it onto the coordinate sequence of the current probe path. At the same time, if the image quality does not meet the benchmark, the probe gain parameters and focusing depth are adjusted synchronously to ensure that the corrected path meets the image quality benchmark requirements.

9. The method of claim 1, wherein, When iteratively optimizing path parameters based on the verification results, if the coverage integrity fails to meet the standard during verification, then... The grid traversal density of the initial scanning path is adjusted based on the difference from the preset coverage threshold. That is, the larger the difference, the higher the grid traversal density is increased. If the number of probe attitude transformations exceeds the preset range, the probe attitude angles of adjacent grid cells in the initial scanning path are smoothed, and a Bezier curve is used to fit the transition trajectory of adjacent attitude angles; If the total path length exceeds the preset optimization objective, the grid cells that are repeatedly covered in the initial scanning path are deleted, and redundant coverage cells in the preset key focus area are retained; After each parameter adjustment, the scanning path is regenerated and verified until the path meets all the constraint parameter requirements, and the iteration is terminated.

Images