Carotid artery position estimation system and method based on three-dimensional point cloud
Through three-dimensional point cloud technology and binocular infrared structured light camera, the carotid artery position is automatically estimated and the ultrasound equipment is guided, solving the problem of operation dependence and low efficiency of manual ultrasound scan, achieving high accuracy, standardization and patient comfort carotid artery examination.
Patent Information
- Application Number
- CN202410577568.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-05-10
- Publication Date
- 2024-11-01
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The existing manual ultrasound scan of carotid artery has problems such as operator dependence, high time consumption, poor patient comfort, strong subjectivity of image interpretation, and equipment limitation, resulting in low diagnostic efficiency and accuracy.
Using a carotid artery position estimation system based on three-dimensional point clouds, the head and neck three-dimensional imaging and positioning are performed through binocular infrared structured light cameras, and the neck central axis is reconstructed using parallax and geometric algorithms, and the robot ultrasound is automatically guided by combining prior structural knowledge. Move to the carotid artery position for examination.
Improves the accuracy and consistency of diagnosis, reduces operation time, enhances patient comfort, provides quantitative analysis and standardized diagnosis, reduces workload for medical staff, and supports remote diagnosis.
Smart Images

Figure CN118887285A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of medical imaging technology, and in particular to a system and method for estimating the position of a carotid artery based on three-dimensional point clouds. Background Art
[0002] Point cloud image processing is a technology that processes and analyzes large amounts of point data obtained from 3D scanners or other similar devices. These points represent the surface of the scanned object, and each point contains location information (usually X, Y, Z coordinates) and may include other data such as color, intensity, etc. Point cloud image processing is very important in many applications, such as manufacturing, geographic information systems (GIS), architecture, archeology, and robotics. The following are several key background technologies for point cloud image processing:
[0003] 1. 3D Scanning Technology The first step of point cloud image processing usually involves using 3D scanning technology to obtain physical surface data of physical objects. These scanning technologies include laser scanning, structured light scanning, stereo vision and other methods. Different technologies have their specific application scope and advantages. For example, laser scanning is suitable for large-scale and long-distance measurement, while structured light scanning is suitable for obtaining local data with higher accuracy.
[0004] 2. Data Preprocessing The raw point cloud data obtained by scanning often contains noise, outliers, and incomplete parts. Data preprocessing steps include filtering and denoising, point cloud simplification (downsampling), and outlier removal, which are crucial for subsequent data analysis.
[0005] 3. Point cloud registration In the case of multiple views or multiple scans, different point cloud data sets need to be aligned into a unified coordinate system. Point cloud registration techniques, such as the Iterative Closest Point (ICP) algorithm, are used to optimize the relative position and orientation between point clouds to achieve precise alignment.
[0006] 4. Point cloud segmentation and classification Point cloud segmentation is the process of dividing the points in a point cloud into multiple subsets representing different entities or surfaces. Classification further analyzes these subsets and labels them into different categories (such as buildings, vehicles, pedestrians, etc.). This step is particularly important for automated processing and interpretation of point cloud data.
[0007] 5. 3D Reconstruction and Modeling Based on the processed point cloud data, a 3D model of an object or environment can be constructed. This involves surface reconstruction techniques such as polygon mesh generation, surface fitting, etc. 3D models are very useful for visualization, simulation, and further analysis.
[0008] 6. Machine Learning and Artificial Intelligence With the development of technology, machine learning and deep learning technologies are increasingly being used in point cloud image processing to automatically identify and classify complex patterns and improve processing efficiency and accuracy. These technologies are particularly important in the fields of self-driving cars, robot navigation, etc.
[0009] 7. Visualization and Analysis Finally, visualization and analysis of point cloud data are crucial for users to understand and apply data. Advanced visualization tools can help users extract valuable information from complex data and provide decision support.
[0010] These background technologies of point cloud image processing provide powerful tools for various industries, making it possible to extract useful information from complex three-dimensional data.
[0011] It is clear that current manual ultrasound scanning of the carotid artery has the following problems:
[0012] 1. Operator Dependence The quality of carotid ultrasound scans is highly dependent on the operator's skills and experience. Different operators may obtain different scan results, and this difference may be caused by factors such as the operator's technique, angle selection, pressure control, etc. Therefore, the repeatability and consistency of manual ultrasound scans may be low.
[0013] 2. Time consumption Manual ultrasound scans often take a long time to complete, especially when different parts of the carotid artery need to be examined in detail. This can mean a high time cost for doctors and patients, especially in busy medical environments.
[0014] 3. Patient comfort During manual ultrasound scanning, in order to obtain a clear image, the operator may need to apply some pressure to the patient's neck, which may cause discomfort to the patient. In addition, maintaining a fixed posture for a long time may also cause discomfort to the patient.
[0015] 4. Subjectivity of Image Interpretation Even experienced ultrasound technicians can have errors in image interpretation. The quality and interpretation of ultrasound images are affected by many factors, including equipment settings, differences in patient anatomy, and the complexity of the image itself.
[0016] 5. Device Restrictions The performance of the ultrasound equipment may also limit the quality of the scan. For example, a low-resolution device may not provide images clear enough to diagnose small vascular lesions. In addition, the portability and ease of use of the device may also affect the scanning process.
[0017] 6. Difficulty in quantitative analysis Manual ultrasound scans rely primarily on qualitative analysis, while quantitative analysis is often more complex and time-consuming. Although modern ultrasound equipment may include some automated quantitative analysis tools, the accuracy and reliability of these tools in certain complex cases remains a challenge.
[0018] In response to these problems, the development of automated ultrasound scanning systems has become an important research direction. This system aims to reduce operator dependence, improve scanning consistency and efficiency, and improve patient comfort and overall diagnosis and treatment experience. By integrating advanced imaging technology and artificial intelligence algorithms, automated systems can simplify the operation process and provide more accurate diagnostic support while ensuring high-quality images. Summary of the invention
[0019] In order to solve the above problems, the object of the present invention is to provide a system and method for estimating the position of the carotid artery based on three-dimensional point cloud.
[0020] Step 1: Design and adapt the baseline length and spot size of binocular structured light for human body imaging. Because the human body imaging distance is short, a shorter baseline (30mm-90mm) and a smaller structured light spot are selected. Therefore, the three-dimensional depth image can be effectively reconstructed within the imaging distance range of 100mm-1000mm.
[0021] Step 2: We will use an infrared binocular structured light camera to perform 3D imaging and positioning of the head and neck. Turn on the binocular infrared structured light and perform depth imaging of the binocular camera. The structured light is designed to be a dense light spot with alternating light and dark to increase the texture information of the imaged human body surface and increase the binocular camera's capture and registration of the same point.
[0022] Step 3: The infrared binocular structured light can perform three-dimensional reconstruction of each matching point on the surface of human tissue through the parallax and geometric algorithm of the binocular camera at the same point, and use the three-dimensional algorithm based on the parallax principle to reconstruct each point on the skin surface in three dimensions, thereby obtaining a three-dimensional point cloud image of the main area of the head and neck.
[0023] Step 4: Get the connected domain of the head and neck of the whole image through the connected domain of the depth value of the 3D point cloud image within a certain range. Based on this connected domain, a straight line of the head and neck can be fitted. This straight line is approximately the central axis of the head and neck. Step 5: Calculate the perpendicular line of the head and neck central axis and all tangents of the head and neck connected domain. The head and neck are divided into several tangents. The lengths of the tangents of the neck have strong length similarity and position clustering. All tangents in the neck area are classified by clustering algorithm. Thus, the neck area is obtained.
[0024] Step 6: Calculate the central axis of the neck through the neck area. Through the edge clustering characteristics of the point cloud depth value of the central axis, the accuracy of the central axis of the neck can be corrected and the precise central axis can be finally determined.
[0025] Step 7: Based on the prior knowledge of the structure of the carotid artery in the neck, a mapping relationship between the position of the normal carotid artery and the midline of the neck is established. The position of the midline of the neck obtained in step 6 can be mapped to the position of the carotid artery. This can automatically guide the robot ultrasound to move to the position of the carotid artery for imaging, carotid artery examination, and carotid plaque screening.
[0026] Compared with the prior art, the present invention has the following beneficial effects: The introduction of automated carotid scanning robots is a revolutionary step in the field of ultrasound scanning, providing multiple benefits, including improved diagnostic accuracy, consistency, efficiency, and an enhanced patient and provider experience. These benefits are discussed in detail below:
[0027] 1. Improve accuracy and repeatability The automatic scanning robot can accurately locate and scan the carotid artery through high-precision control and advanced imaging technology, reducing errors caused by human operation. The use of the robot reduces the variability between scanning results and improves the repeatability and reliability of the examination.
[0028] 2. Increase operational consistency The automatic scanning robot uses standardized scanning procedures to ensure that each scan is performed according to the same standards, which not only facilitates comparison during follow-up of the same patient, but also makes the scan results between different patients more comparable. This is particularly important for clinical research and disease monitoring.
[0029] 3. Improve efficiency The automated scanning robot is able to scan quickly, reducing the time required for manual operation. This automated process can significantly improve the workflow efficiency of medical facilities, allowing doctors and technicians to spend more time on diagnosing and treating patients instead of repetitive scanning operations.
[0030] 4. Reduce the workload of doctors and technicians Prolonged manual ultrasound scanning may cause muscle fatigue and strain on the operator. The use of automatic scanning robots can reduce the physical burden and reduce the risk of occupational-related diseases.
[0031] 5. Enhanced patient experience Robotic manipulation reduces physical intervention on patients, such as pressure and discomfort caused by manual adjustments, thereby improving patient comfort. The automated process also reduces the waiting time for patients in the hospital.
[0032] 6. Provide quantitative analysis Some advanced automatic scanning robotic systems are equipped with software that can provide quantitative analysis, such as plaque measurement and accurate calculation of blood flow velocity. These data are very useful for early diagnosis of diseases and formulation of treatment plans.
[0033] 7. Improve diagnostic standardization Automated scanning robots can follow fixed protocols and help standardize the diagnostic process. This is particularly important for ensuring quality standards for medical services globally or across regions.
[0034] 8. Remote diagnostic capabilities Data from automated scanning robots can be easily integrated into remote diagnostic systems, allowing specialists to evaluate and diagnose scans at different locations, which is particularly important in areas with limited resources.
[0035] The application of automated carotid scanning robots is expected to further optimize clinical pathways, improve the quality of medical services, and reduce the overall cost of the medical system. As technology advances and becomes more popular, it is expected that these robots will be increasingly adopted around the world. BRIEF DESCRIPTION OF THE DRAWINGS
[0036] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0037] Figure 1 The present invention is a schematic flow chart of the steps for implementing the method.
[0038] Figure 2 Schematic diagram of the binocular infrared structured light imaging system of the method of the present invention.
[0039] Figure 3 Schematic diagram of the calculation process of calculating the three-dimensional point cloud image through the parallax of the binocular structured light camera. Implementation
[0040] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0041] See also Figure 1 As shown, the present invention provides a system and method for estimating the position of a carotid artery based on a three-dimensional point cloud, comprising:
[0042] Step 1: Design and adapt the baseline length and spot size of binocular structured light for human body imaging. Because the human body imaging distance is short, a shorter baseline (30mm-90mm) and a smaller structured light spot are selected. Therefore, the three-dimensional depth image can be effectively reconstructed within the imaging distance range of 100mm-1000mm.
[0043] A1. Selection of baseline length The baseline length, i.e. the distance between the two cameras, is very important in determining the parallax sensitivity and maximum detection distance of the system. For human imaging: Short baseline (30mm-90mm): Advantages: Improve the accuracy of close-range imaging and reduce parallax errors. Suitable for use in scenarios that require detailed observation of human parts (such as faces, hands, etc.) Disadvantages: The imaging effect will decrease at longer distances because the small baseline length limits the parallax range, which may lead to inaccurate depth information of distant objects.
[0044] A2.Size of structured light spot Structured light systems work by projecting a known pattern of light (spots or stripes) onto an object and using a camera to capture the deformation of this pattern to calculate depth information. The choice of spot size affects the details of the image: Small spot: Advantages: Suitable for fine 3D reconstruction, especially in close-range imaging, it can capture more surface details, which is very important for human body imaging Disadvantages: Small light spots may be affected by high ambient light. In strong light environments, it may be necessary to increase the contrast of the light spot or use a light source with a specific wavelength.
[0045] A3. Influence of imaging distance (100mm-1000mm) In the imaging distance range of 100mm to 1000mm, the short baseline and small spot size are selected to optimize the accuracy and resolution of close-range imaging. This distance range is very suitable for applications such as medical imaging, body art measurement, and biometric recognition.
[0046] A4. System design considerations Optical design: Ensure that the optical system of the light source and camera can provide clear images and sufficient illumination over the entire imaging distance range Camera and light synchronization: It is important to ensure that the camera and the projected light source are highly synchronized to prevent blur or inconsistency in dynamic imaging Software and Algorithms: Develop efficient image processing and 3D reconstruction algorithms to process data captured from structured light systems, improving the speed and accuracy of imaging Through the above design and adaptation, the binocular structured light system can effectively perform 3D imaging of the human body at a shorter imaging distance, which is very valuable for many medical and biometric applications. The design of such a system must take into account the actual application environment of imaging and the characteristics of the target object to ensure that the expected imaging quality and accuracy can be achieved.
[0047] Step 2: We will use an infrared binocular structured light camera for 3D imaging and positioning of the head and neck. Turn on the binocular infrared structured light and perform depth imaging with the binocular camera. The structured light is designed to be a dense light spot with alternating light and dark to increase the texture information of the imaged human body surface and increase the binocular camera's capture and registration of the same point. The structured light intensity distribution formula is: For applications of 3D imaging and positioning of the head and neck using an infrared binocular structured light camera, the use of a dense spot design with alternating light and dark is an effective strategy. This approach can significantly improve the binocular camera's ability to capture and align on surfaces without obvious textures, especially for areas of the human body such as the head and neck, which usually have less texture. The following are several key aspects of the system design and implementation:
[0048] B1. Selection of infrared light source Selecting the right infrared wavelength is crucial to ensure the clarity of the light spot and its harmlessness to the human body. Compared with visible light, infrared light is less susceptible to interference from ambient light and is suitable for use under a variety of lighting conditions. Usually, the wavelength of the infrared light source is selected between 850 nm and 940 nm, which can reduce the potential impact on the human eye and improve the concealment of the imaging.
[0049] B2. Generation of dense light spots A specific spot generator, such as a digital micromirror device (DMD) or a customized optical slide, is used to generate a dense and regular pattern of alternating light and dark spots. This pattern helps to improve the imaging system's sensitivity to tiny bumps on the human body surface, thereby improving the resolution and accuracy of the depth map.
[0050] B3. Image Capture and Processing Synchronous capture: Ensure that the binocular camera is highly synchronized with the infrared structured light emitter to accurately capture the tiny deformations caused by the spot pattern Image registration: The captured left and right views are registered through advanced image processing algorithms to accurately calculate the disparity of each pixel and then convert it into depth information.
[0051] B4. Depth Calculation and 3D Reconstruction Depth is calculated using disparity information, and 3D reconstruction technology is used to convert depth data into an accurate 3D model. This step usually requires complex algorithms to process large amounts of data and solve possible errors caused by occlusion, reflection, and other factors.
[0052] B5. Software and User Interface Develop intuitive user interfaces and powerful backend software to support real-time processing and visualization of data. This should include data filtering, smoothing, and various tools for medical analysis.
[0053] B6. Safety and ergonomic considerations Ensure that the system design complies with the relevant safety standards for medical devices, especially for parts that come into contact with the skin for a long time. At the same time, the design should take into account the comfort and ergonomics of operation so that both patients and operators can remain comfortable during use.
[0054] Imaging the head and neck using an infrared binocular structured light camera is a technically challenging but promising area of application. Proper implementation of the above solution will significantly improve the accuracy and efficiency of head and neck medical imaging, helping to optimize diagnosis and treatment. This technology also has important application potential in plastic surgery, neurosurgery and even in some specialized oral medical procedures.
[0055] Step 3: The infrared binocular structured light can perform three-dimensional reconstruction of each matching point on the surface of human tissue through the parallax and geometric algorithm of the binocular camera at the same point, and use the three-dimensional algorithm based on the parallax principle to reconstruct each point on the skin surface in three dimensions, thereby obtaining a three-dimensional point cloud image of the main area of the head and neck. .
[0056] C1. Projection and capture of structured light Spot projection: The system uses an infrared light source to project a specific structured light pattern (such as a grid or coded stripes) onto a target area, such as the head and neck of a person. Infrared light is harmless to the human body and has little interference with ambient light. Image capture: The binocular camera captures the patterns formed by structured light illumination from two slightly different perspectives at the same time. These patterns will be slightly distorted due to the geometric characteristics of the surface (such as bumps, curvature, etc.)
[0057] C2. Disparity Calculation Image processing: The images received from the stereo camera are first pre-processed, including denoising and enhancement, to improve the accuracy of subsequent processing Parallax matching: By comparing the images from the left and right cameras, the system uses geometry and image processing algorithms to find the corresponding positions of the same feature points (i.e. image points generated by the same light spot) in the two views Disparity calculation: The system calculates the position difference (disparity) of these matching points in the two images. The size of the disparity is inversely proportional to the distance from the point to the camera.
[0058] C3. 3D Reconstruction Depth Mapping: Using parallax data, the exact depth of each point is calculated through triangulation. Point cloud generation: convert depth data into three-dimensional space coordinate points to generate detailed three-dimensional point cloud images Reconstruction optimization: Further processing of the generated point cloud, including filtering, outlier removal, and smoothing, to improve the quality of the 3D model
[0059] C4. Application Medical Applications: This detailed 3D point cloud image can be used for a variety of medical applications, including but not limited to surgical planning, lesion monitoring, cosmetic reconstruction, and the design of custom therapeutic devices Research and Education: Provide realistic 3D models of the human body for use in medical research and education, helping students and professionals better understand complex anatomy The infrared binocular structured light system has shown extremely high application value in the field of medical imaging, especially in the imaging of complex head and neck areas, due to its advantages of high precision, easy operation and harmlessness to the human body. The development and application of this technology has greatly promoted the progress of medical imaging technology.
[0060] Step 4: Get the connected domain of the head and neck of the whole image through the connected domain of the depth value of the 3D point cloud image within a certain range. Based on this connected domain, a straight line of the head and neck can be fitted. This straight line is approximately the central axis of the head and neck. .
[0061] D1. Acquisition of 3D point cloud images First, it is necessary to use a suitable imaging device (such as an infrared binocular structured light camera) to obtain the 3D point cloud data of the head and neck. This data needs to include the 3D spatial coordinates and corresponding depth value of each point.
[0062] D2. Connected domain analysis Depth threshold setting: According to the actual imaging environment and the typical depth range of the target area, set one or more depth thresholds to identify and separate the head and neck from other body parts or background. Connected domain extraction: Apply image processing algorithms (such as region growing, clustering algorithms, etc.) to identify connected domains with depth values within a specific range. These connected domains should roughly correspond to the actual location of the head and neck.
[0063] D3. Fitting of the central axis Point set processing: Select representative point sets from the extracted connected domain. Noise reduction and filtering may be required to improve the accuracy of subsequent fitting. Central axis fitting: A straight line is calculated based on the extracted point set using geometric fitting techniques (such as the least squares method), which can be approximated as the central axis of the head and neck. The direction and position of this line should be consistent with the anatomical structure of the head and neck.
[0064] D4. Analysis and Application Verification and adjustment: The fitted central axis needs to be verified to ensure that it is consistent with anatomy and clinical observations. Parameter adjustment and algorithm optimization may be required based on preliminary results. Clinical applications: The determined central axis can be used in a variety of medical applications, including surgical navigation, radiotherapy planning, disease monitoring, etc.
[0065] D5. Software and algorithm development Software tools: Develop dedicated software tools to automatically perform the above process, including image acquisition, data processing, connected domain extraction, central axis fitting and other functions User Interface: An intuitive user interface is designed to enable doctors and technicians to easily operate the software, adjust parameters, and view results
[0066] The successful implementation of this technology relies on accurate 3D imaging technology, efficient image processing algorithms, and a deep understanding of the anatomical structure of the head and neck. Further research and development can improve the accuracy and practicality of this method, providing important support for the diagnosis and treatment of head and neck diseases.
[0067] Step 5: Calculate the perpendicular line of the head and neck central axis and all tangents of the head and neck connected domain. The head and neck are divided into several tangents. The lengths of the tangents of the neck have strong length similarity and position clustering. All tangents of the neck area are classified by clustering algorithm. Thus, the area of the neck is obtained. .
[0068] E1. Extraction of the central axis First, the central axis of the head and neck is calculated from the 3D point cloud. This can be achieved by using adaptive algorithms such as principal component analysis (PCA) to determine the direction in which the point cloud data mainly extends and use this direction as the central axis.
[0069] E2. Calculate the vertical tangent At each point along the median axis, planes perpendicular to the median axis are calculated, and then tangents perpendicular to the median axis are extracted on these planes. These tangents pass through the point cloud of the head and neck to form line segments that can be used for further analysis.
[0070] E3. Tangent length and position measurement The length of each tangent line is a key parameter because it reflects the width of the head and neck at that particular section. The position of each tangent line is also recorded, which helps in subsequent cluster analysis.
[0071] E4. Cluster analysis Use a clustering algorithm, such as K-means or DBSCAN, to cluster the tangents based on their length and location. This step is to identify clusters of tangents that are close to each other and have similar lengths, with each cluster representing a specific area of the neck.
[0072] E5. Determination of the neck region The clustering results can help determine the specific location and extent of the neck region in the head and neck. These regions can be directly visualized based on the clustered tangent groups, or further analyzed to extract more biomedical information.
[0073] E6. Data Visualization and Validation Finally, the extracted neck region needs to be visualized and verified against clinical data. This step is critical to ensure that the extracted data accurately reflects the anatomical structure.
[0074] E7. Application and Optimization Based on feedback from actual applications, the algorithm is adjusted and optimized to improve accuracy and reliability. At the same time, this technology can be expanded to other anatomical regions to support a wider range of clinical applications.
[0075] Step 6: Calculate the central axis of the neck through the neck area. Through the edge clustering characteristics of the point cloud depth value of the central axis, the accuracy of the central axis of the neck can be corrected and the precise central axis can be finally determined.
[0076] F1. Preliminary estimation of the central axis First, it is necessary to generate a preliminary central axis of the neck based on the 3D point cloud data. This is usually done through geometric methods or shape analysis based on point cloud segmentation. The preliminary central axis provides the basic direction and path of the neck structure.
[0077] F2. Point cloud depth value analysis Depth value extraction: Extract depth values from the point cloud corresponding to the preliminary central axis. These depth values reflect the distance of each point in the point cloud relative to the observation plane. Edge detection: Apply edge detection algorithms (such as gradient operators or other edge detection techniques) to identify significant changes in depth values, which often correspond to the edges of anatomical structures
[0078] F3. Cluster analysis Feature selection: Select appropriate features for cluster analysis. Common features include depth change rate, spatial distribution of points, etc. Application of clustering algorithms: Use clustering algorithms (such as K-means, hierarchical clustering, DBSCAN, etc.) to classify depth values. The goal is to cluster depth values into two or more categories to distinguish different anatomical structures on both sides of the midline of the neck.
[0079] F4. Centerline correction Central axis adjustment: Based on the clustering results, adjust the position and orientation of the central axis to more accurately reflect the center of the cervical anatomy. This may involve mathematical optimization methods such as least squares or dynamic programming to determine the best path Smoothing: In order to ensure the smoothness of the central axis and the authenticity of physiological anatomy, it may be necessary to smooth the adjusted central axis, such as using a B-spline curve or other curve fitting methods
[0080] F5. Verification and Application Result verification: Verify the accuracy of the corrected central axis through medical imaging data (such as MRI or CT) Clinical application: The corrected central axis can be used in clinical diagnosis, surgical planning, personalized medical device design and other fields to provide key anatomical references
[0081] F6. Visualization and user feedback Visualization: Provides an intuitive graphical interface for doctors and researchers to display the central axis and point cloud data, as well as the clustering and correction process User feedback: Further adjust the algorithm and processing flow based on user feedback to meet the needs of actual applications
[0082] This central axis correction method based on point cloud depth value and edge clustering characteristics provides strong technical support for detailed research and application of the neck and other anatomical parts by improving the accuracy of the central axis.
[0083] Step 7: Based on the prior knowledge of the structure of the carotid artery in the neck, establish a mapping relationship between the normal human carotid artery and the position of the neck midline. The position of the neck midline obtained in step 6 can be mapped to the position of the carotid artery. This can automatically guide the robot ultrasound to move to the carotid artery position for imaging, carotid artery examination, and carotid artery plaque screening. .
[0084] G1. Integration of prior knowledge of cervical anatomy First, it is necessary to collect and organize detailed anatomical knowledge about the structure of the neck, especially the position of the carotid artery relative to the midline of the neck. This can be achieved through anatomical literature, clinicians' experience, or existing medical imaging data.
[0085] G2. Establishing a mapping model Using statistical methods or machine learning techniques, a prediction model is established based on the neck and carotid artery data of the normal population, which can predict the accurate location of the carotid artery based on the location of the neck midline. This may involve multiple steps such as feature extraction, model training and verification.
[0086] G3. Integration of Ultrasound Robot The predictive model is integrated into the control system of the ultrasound robot. This requires close coordination of software and hardware to ensure that the robot can move accurately to the corresponding position based on the carotid artery position data provided by the model.
[0087] G4. Development of automated imaging protocols Develop a protocol for automatically guiding the ultrasound probe for contiguous imaging. This includes automated control of probe placement angle, pressure, and scanning speed to optimize image quality and minimize discomfort to the patient.
[0088] G5. System testing and optimization Test the system in an actual clinical environment to evaluate its performance, especially accuracy and reliability. Make necessary adjustments and optimizations based on the test results.
[0089] G6. Clinical implementation and monitoring Apply the system to actual carotid artery examination and plaque screening, monitor its performance in actual use, and collect feedback from users (doctors and patients) to further improve the system.
[0090] G7. Data Analysis and Feedback Learning Using data collected from actual applications, the prediction model is retrained and updated to improve its prediction accuracy. This process can adopt a continuous learning approach, and the performance of the model will continue to improve as the amount of data increases.
[0091] Through the above steps, a highly automated carotid artery ultrasound examination system can be realized, which not only improves the efficiency and accuracy of diagnosis, but also significantly improves the patient experience. In addition, the development and implementation of this system will also promote the development of medical automation technology and have a profound impact on the future medical service model.
Claims
1. A system and method for estimating the position of a carotid artery based on a three-dimensional point cloud, characterized in that: include: Step 1. Design and adapt the baseline length and spot size of the binocular structured light for human body imaging. Because the imaging distance of the human body is short, a shorter baseline (30mm-90mm) and a smaller structured light spot are selected to effectively reconstruct the three-dimensional depth image within the imaging distance range of 100mm-1000mm. Step 2. Use an infrared binocular structured light camera for three-dimensional imaging and positioning of the head and neck. Turn on the binocular infrared structured light for depth imaging of the binocular camera. The design of structured light is a dense spot of alternating light and dark to increase the texture information of the imaged human body surface to increase the capture and alignment of the binocular camera for the same point. Structured light intensity distribution formula: Step 3: Infrared binocular structured light can perform 3D reconstruction of each matching point on the surface of human tissue through the disparity and geometric algorithm of the binocular camera at the same point. The 3D algorithm based on the parallax principle is used to perform 3D reconstruction of each point on the skin surface, thereby obtaining a 3D point cloud image of the main area of the head and neck. Step 4: Get the connected domain of the head and neck of the whole image through the connected domain of the depth value of the 3D point cloud image within a certain range. Based on this connected domain, a straight line of the head and neck can be fitted. This straight line is approximately the central axis of the head and neck. Step 5: Calculate all tangents between the perpendicular line of the head and neck central axis and the head and neck connected domain. The head and neck are divided into several tangents. Since the lengths of the tangents of the neck have strong length similarity and position clustering, all tangents in the neck area are classified by clustering algorithm, thus obtaining the neck area. . Step six, calculate the central axis of the neck through the neck area. Through the edge clustering characteristics of the point cloud depth value of the central axis, correct the accuracy of the central axis of the neck and finally determine the precise central axis. Step seven, through the prior structural knowledge of the carotid artery in the neck, establish a mapping relationship between the position of the normal carotid artery and the central axis of the neck. The position of the carotid artery can be mapped to the position of the central axis of the neck obtained in step six. Thereby, the robot ultrasound can be automatically guided to move to the position of the carotid artery for fitting imaging, thereby realizing carotid artery examination and carotid plaque screening. .
2. A binocular infrared structured light system as claimed in claim 1, characterized in that: The system is specially designed for human vascular imaging, including multi-band (700nm-1200nm) structured light design to increase imaging quality and robustness, while enhancing the contrast of vascular imaging and the skin tissue penetration ability of the light source. The structured light design of the alternating light and dark sine-shaped spot increases the skin surface texture and greatly improves the accuracy of three-dimensional reconstruction.
3. A binocular infrared reconstruction algorithm as claimed in claim 1, characterized in that: The three-dimensional reconstruction algorithm uses the spot information to match the same point, and then reconstructs the depth of the point based on the parallax of the binocular camera. After reconstructing the depth of all points, the depth of the skin surface can be obtained, and filtering and pathological point processing must be performed at the same time.
4. After obtaining the point cloud image as described in claim 1, the connected domain of the head and neck is extracted and segmented to obtain the connected domain of the point cloud image of the head and neck, and the central axis of the head and neck is obtained by fitting.
5. According to the point cloud image based on the head and neck as described in claim 1, the head and neck point cloud is divided into several tangents according to the perpendicular tangent of the central axis of the head and neck, and the tangent of the neck is extracted based on the length position clustering algorithm of the tangent, so as to obtain the connected domain position of the neck.
6. The point cloud image based on the head and neck as claimed in claim 1 estimates the position and alignment information of the carotid artery based on the prior anatomical position mapping relationship between the neck midline and the carotid artery.
Citation Information
Cited By
Complex wiring preprocessing method and system based on 3D point cloud measurement and registration technology
CN119810321A
