Respiratory tract image-based segmentation method and system

Abstract

The invention relates to the technical field of image segmentation, in particular to a segmentation method and system based on a respiratory tract image. The method comprises the following steps: acquiring an initial image of a respiratory tract; performing lung airway image feature extraction on the initial respiratory tract image to obtain lung airway feature data; performing lung airway structure area image segmentation on the respiratory tract image according to the lung airway feature data to generate a lung airway structure segmentation image; constructing an airway exclusion mask based on the lung airway structure segmentation image to obtain an airway exclusion mask; performing binary division on the respiratory tract image through the airway exclusion mask to generate an initial blood vessel region image; and performing blood vessel tree analysis on the initial blood vessel region image to generate respiratory tract blood vessel tree data. The precision and simplicity of the segmentation method of the respiratory tract image are improved by performing feature extraction, structure segmentation optimization, motion influence evaluation, lesion area detection and three-dimensional reconstruction technologies on the respiratory tract CT image.

Description

Technical Field

[0001] The present invention relates to the technical field of image segmentation, and in particular to a respiratory tract image-based segmentation method and system. Background Art

[0002] In the early days, traditional methods such as threshold segmentation and edge detection were widely used. However, due to the complexity of respiratory tract structures, these methods were limited in accuracy and robustness. With the advancement of computing power and the refinement of mathematical models, statistical methods such as active contour models and region-based segmentation methods have gradually emerged. These methods improve segmentation accuracy and stability by incorporating prior knowledge and shape constraints. The development of machine learning has brought new opportunities to respiratory tract image segmentation. In particular, supervised learning methods such as support vector machines (SVMs) and random forests have significantly improved segmentation performance by learning from large amounts of annotated data. However, these methods rely on feature engineering and have limited effectiveness when dealing with complex morphological structures. In recent years, the application of deep learning technologies, particularly convolutional neural networks (CNNs), has revolutionized respiratory tract image segmentation. CNN-based U-Net and its variants, through end-to-end training, can automatically extract multi-level features, significantly improving segmentation accuracy and generalization. Furthermore, the combination of transfer learning and data augmentation techniques has effectively alleviated the problem of scarce medical data. Combining a three-dimensional convolutional network (3D CNN) with an attention mechanism further improves the segmentation of complex respiratory structures. However, segmenting the airways and blood vessels in respiratory images is currently a challenge because the airways and blood vessels are often mixed in the image. Furthermore, breathing and body movement can cause image artifacts that affect segmentation, making respiratory image segmentation more difficult and resulting in lower segmentation accuracy. Summary of the Invention

[0003] Based on this, it is necessary to provide a respiratory image-based segmentation method and system to solve at least one of the above technical problems.

[0004] To achieve the above object, a respiratory tract image-based segmentation method is provided, the method comprising the following steps: Step S1: Acquire an initial respiratory tract image; perform lung airway image feature extraction on the initial respiratory tract image to obtain lung airway feature data; perform lung airway structure region image segmentation on the respiratory tract image based on the lung airway feature data to generate a lung airway structure segmentation image; Step S2: constructing an airway exclusion mask based on the lung airway structure segmentation image to obtain an airway exclusion mask; performing binary segmentation on the respiratory image using the airway exclusion mask to generate an initial vascular region image; performing vascular tree analysis on the initial vascular region image to generate respiratory vascular tree data; performing vascular structure region image segmentation on the initial vascular region image using the respiratory vascular tree data to generate a lung vascular structure segmentation image; Step S3: Acquire user posture data and user respiratory rate data; perform user motion impact assessment based on the user posture data and user respiratory rate data to generate user motion impact assessment data; perform image artifact optimization on the lung airway structure segmentation image and the lung vascular structure segmentation image using the user motion impact assessment data to generate a respiratory segmentation artifact optimized image; Step S4: performing abnormality detection on the respiratory tract segmentation artifact optimized image to generate lesion area positioning data; performing lesion area segmentation on the respiratory tract segmentation artifact optimized image according to the lesion area positioning data to generate a respiratory tract lesion area segmentation image; performing three-dimensional image reconstruction on the lung airway structure segmentation image, the lung vascular structure segmentation image and the respiratory tract lesion area segmentation image to generate a respiratory tract structure segmentation three-dimensional image.

[0005] The present invention acquires an initial image of the respiratory tract and then extracts features from the lung airway image. These features include lung morphology, duct diameter, and branching structure. Feature extraction quantifies and describes the structure of the respiratory tract, providing foundational data for subsequent segmentation and analysis. This helps physicians accurately identify and analyze the health of a patient's airway, such as detecting stenosis, obstruction, or other abnormalities. By constructing an airway exclusion mask, the airway portion of the respiratory tract image is effectively isolated. This allows subsequent image processing to focus on analyzing the vascular region, eliminating the influence of the airways. This is crucial for accurately analyzing the branching structure, density, and location of blood vessels, and helps provide precise vascular tree data. By analyzing the user's posture and respiratory rate data, the impact of user motion during image acquisition can be assessed and understood. This data provides key insights into the impact of image quality. Optimizing the segmented images of the lung airway and vascular structure reduces or eliminates artifacts and blur caused by user motion, thereby improving image clarity and resolution. Abnormality detection performed on the optimized segmented images of the respiratory tract allows for rapid and accurate localization of lesions, specifically masses, inflammation, or other abnormal structures. By accurately locating and segmenting lesion areas, doctors can identify patients' health problems in advance. Therefore, the present invention improves the accuracy and simplicity of respiratory image segmentation methods by performing feature extraction, structural segmentation optimization, motion impact assessment, lesion area detection, and 3D reconstruction on respiratory tract CT images.

[0006] Preferably, step S1 includes the following steps: Step S11: Scan the user's respiratory tract using a CT machine to obtain an initial respiratory tract image; Step S12: performing image standardization on the initial respiratory tract image to obtain a standard respiratory tract image, wherein the image preprocessing includes image denoising, image brightness enhancement, and image smoothing; Step S13: extracting lung airway image features from the standard respiratory tract image to obtain lung airway feature data; Step S14: performing lung airway structure region image segmentation on the standard respiratory image according to the lung airway feature data to generate a lung airway structure segmentation image.

[0007] The present invention uses a CT machine to scan the user's respiratory tract, acquiring high-resolution initial respiratory tract images and providing accurate baseline data for subsequent processing. Image normalization (including denoising, brightness enhancement, and smoothing) effectively improves image quality, eliminating noise and inhomogeneities, resulting in clearer images and facilitating subsequent feature extraction and segmentation. Pulmonary airway image feature extraction on the standardized respiratory tract images accurately captures the morphological characteristics of the pulmonary airways and produces high-quality pulmonary airway feature data. Using this extracted pulmonary airway feature data to segment the pulmonary airway structure region in the image, accurate pulmonary airway structure segmentation images are generated, clearly displaying the anatomical structure of the pulmonary airways. These high-quality segmented images and feature data can significantly improve the accuracy and efficiency of medical imaging, helping doctors better identify and assess respiratory-related diseases and abnormalities. Accurate pulmonary airway structure segmentation images and feature data also provide a solid foundation for subsequent medical image analysis and processing, facilitating more complex image analysis tasks.

[0008] Preferably, step S14 includes the following steps: Step S131: performing airway pixel category probability calculation on the standard airway image to obtain lung airway pixel category data; performing initial airway lung region segmentation on the standard airway image based on the lung airway pixel category data to obtain an initial airway lung region image; Step S132: removing small-area impurities from the initial respiratory tract and lung region image using cell tissue morphology theory to obtain an impurity-removed image of the respiratory tract and lung region; Step S133: performing 3D voxelization on the impurity-removed image of the respiratory tract lung region to generate respiratory tract lung region voxels; confirming the center points of the respiratory tract lung region voxels to obtain the main airway center points of the respiratory tract lung region; performing lung airway routing search on the respiratory tract lung region voxels using the main airway center points of the respiratory tract lung region to generate lung airway routing search data; Step S134: extracting airway branch points from the respiratory tract lung region impurity removal image using the lung airway route search data to obtain respiratory tract lung bronchial path data; performing lung bronchial topology analysis on the respiratory tract lung bronchial path data to generate respiratory tract lung bronchial topology data; Step S135: performing lung airway structure region image segmentation on the respiratory tract lung region impurity removal image according to the respiratory tract lung bronchial topological structure data and the main airway center point of the respiratory tract lung region to generate a lung airway structure segmentation image.

[0009] The present invention accurately identifies the pixel categories of the lung airway region through the pixel category probability calculation in step S131, ensuring the accuracy of the initial respiratory lung region segmentation. Using cell tissue morphology theory, small-area impurity removal (step S132) is performed on the initial respiratory lung region image to remove non-airway components from the image, improving the purity and reliability of the segmentation results. The 3D voxelization and center point confirmation performed in step S133 convert the 2D image into a 3D voxel model, precisely locating the main airway center point in the respiratory lung region and laying the foundation for subsequent airway routing search. Airway routing search of lung region voxels using the center point (step S133) accurately generates lung airway routing data, providing a basis for branch point extraction and topological structure analysis. Airway branch point extraction and bronchial topological structure analysis based on the airway routing search data provide a comprehensive understanding of the path and structural relationships of the lung bronchi, generating detailed topological structure data. By using the bronchial topological structure data and the main airway center point (step S135 ), the lung region can be finely segmented for the airway structure, thereby generating a high-quality lung airway structure segmentation image.

[0010] Preferably, step S2 includes the following steps: Step S21: constructing an airway exclusion mask based on the lung airway structure segmentation image to obtain an airway exclusion mask; Step S22: eliminating the influence of the airway position on the standard airway image using the airway exclusion mask, thereby generating a first residual image of the airway; Step S23: performing blood vessel pixel category probability calculation on the first residual image of the respiratory tract to obtain blood vessel pixel category data; performing binary division on the first residual image of the respiratory tract based on the blood vessel category data to generate an initial blood vessel region image; Step S24: Extract the vascular centerline of the initial vascular region image to obtain vascular centerline data; perform vascular tree analysis on the initial vascular region image based on the vascular centerline data to generate respiratory vascular tree data; and segment the vascular structure region image of the initial vascular region image using the respiratory vascular tree data to generate a pulmonary vascular structure segmentation image.

[0011] The present invention constructs an airway exclusion mask in step S21, effectively removing the impact of the airway on the image, allowing subsequent processing to focus more on the vascular region. The standard respiratory image is processed using the airway exclusion mask (step S22) to generate a first residual image of the respiratory tract, thereby reducing interference from the airway structure on vascular identification and improving the accuracy of vascular segmentation. In step S23, vascular pixel category probabilities are calculated on the first residual image of the respiratory tract, enabling accurate identification of vascular pixels and generating a high-quality initial vascular region image. In step S24, vascular centerline extraction is performed on the initial vascular region image to accurately determine the path and direction of the vessels, laying the foundation for subsequent vascular tree analysis. Vascular tree analysis (step S24) is performed based on the vascular centerline data to generate detailed respiratory vascular tree data, which facilitates a comprehensive understanding of the structure and distribution of pulmonary vessels. The initial vascular region image is segmented into vascular structures using the respiratory vascular tree data to generate a high-quality segmented image of the pulmonary vascular structure, providing accurate vascular distribution information.

[0012] Preferably, performing vascular tree analysis on the initial vascular region image based on the vascular centerline data includes: Perform data preprocessing on the vascular centerline data to generate a standard vascular centerline coordinate sequence, wherein the data preprocessing includes data cleaning, data missing value filling and data standardization; Performing topological analysis on the initial vascular region image according to the standard vascular centerline coordinate sequence to generate vascular region topology data; using a tree structure method to connect the vascular region topology data according to the standard vascular centerline coordinate sequence to generate a vascular site network; Performing a blood vessel diameter size analysis on the blood vessel network to generate blood vessel diameter size data; comparing the blood vessel diameter size data with a preset blood vessel diameter size threshold; and when the blood vessel diameter size data is greater than or equal to the preset blood vessel diameter size threshold, marking the corresponding blood vessel in the blood vessel network as main blood vessel data; When the blood vessel diameter size data is smaller than the preset blood vessel diameter size threshold, the corresponding blood vessel in the blood vessel area network is marked as branch vessel data; the initial blood vessel area image is analyzed for blood vessel tree nodes using the trunk vessel data and branch vessel data to generate respiratory tract blood vessel tree data.

[0013] The present invention ensures the integrity and consistency of the data by performing data cleaning, missing value filling and data standardization on the vascular centerline data, generating a standard vascular centerline coordinate sequence, and providing a high-quality data foundation for subsequent analysis. Based on the standard vascular centerline coordinate sequence, the initial vascular region image is topologically analyzed to generate vascular region topological data, and the connection relationship and structural characteristics of the blood vessels are fully understood. The vascular region topological data is analyzed for vascular connectivity using a tree structure method to construct a vascular site network to ensure the integrity and connectivity of the blood vessels, laying the foundation for subsequent diameter analysis and classification. The vascular site network is analyzed for vascular diameter size to generate detailed vascular diameter size data and provide accurate information on vascular size. The vascular diameter size data is compared with a preset diameter size threshold, and the main vessels and branch vessels are classified and marked to generate main vessel data and branch vessel data, providing a basis for subsequent vascular tree node analysis. The initial vascular region image is analyzed for vascular tree nodes using the main vessel data and branch vessel data to generate high-quality respiratory vascular tree data, which comprehensively displays the structure and classification of the pulmonary vessels.

[0014] Preferably, step S3 includes the following steps: Step S31: Acquire user posture data and user breathing rate data; Step S32: performing posture change analysis on the user posture data to generate user posture change data; performing breathing intensity analysis on the user breathing frequency data to generate user breathing intensity data; Step S33: performing a user motion impact assessment based on the user posture change data and the user breathing intensity data to generate user motion impact assessment data; performing a CT image part disturbance analysis on the user motion impact assessment data to obtain user motion part disturbance data; Step S34: performing motion displacement region segmentation on the lung airway structure segmentation image and the lung vascular structure segmentation image based on the user motion part disturbance data to generate a respiratory tract associated artifact part image; performing image artifact optimization on the respiratory tract segmentation artifact part image to generate a respiratory tract part segmentation artifact optimized image.

[0015] By acquiring user posture data and respiratory rate data, the present invention ensures a comprehensive understanding of the user's movement and respiratory status, providing the necessary data foundation for subsequent analysis. Posture change analysis is performed on the user posture data to generate user posture change data, and respiratory rate analysis is performed on the respiratory intensity data to generate user respiratory intensity data. These data accurately reflect changes in the user's movement and breathing during the scan. A user motion impact assessment is performed based on the user posture change data and respiratory intensity data (step S33), generating user motion impact assessment data. This data allows a comprehensive assessment of the impact of user motion on the CT image. CT image region disturbance analysis is performed on the user motion impact assessment data to generate user motion region disturbance data, clearly identifying image disturbance areas caused by user motion. Based on the user motion region disturbance data, motion displacement region segmentation is performed on the pulmonary airway structure segmentation image and the pulmonary vascular structure segmentation image to generate a respiratory tract-associated artifact region image, accurately marking artifact areas caused by motion. The image artifacts of the respiratory tract segmentation artifact area are optimized (step S34) to generate an artifact-optimized respiratory tract segmentation image. This optimization process reduces or eliminates artifacts, improving image clarity and accuracy. These processing steps significantly improve CT image quality, reduce artifact interference caused by user movement and breathing, and make the final segmented image clearer and more accurate.

[0016] Preferably, step S34 includes the following steps: Step S341: extracting a time series image sequence from the pulmonary airway structure segmentation image and the pulmonary vascular structure segmentation image to obtain CT image time series data; performing respiratory phase division on the CT image time series data to generate respiratory phase division data; Step S342: performing lung motion vector field analysis on the lung airway structure segmentation image and the lung vascular structure segmentation image using the respiratory phase division data to generate a lung motion vector field; performing motion pixel matching on the lung airway structure segmentation image and the lung vascular structure segmentation image using the lung motion vector field to generate motion matching pixels; Step S343: performing motion displacement region segmentation on the lung airway structure segmentation image and the lung blood vessel structure segmentation image according to the motion matching pixels to generate a respiratory tract associated artifact portion image; Step S344: performing respiratory motion region masking correction on the respiratory tract segmentation artifact part image based on the lung motion vector field to generate a respiratory tract segmentation artifact optimized image.

[0017] The present invention extracts a time-series image sequence from the pulmonary airway structure segmentation image and the pulmonary vascular structure segmentation image to generate CT image time-series data, and then performs respiratory phase segmentation to generate respiratory phase segmentation data. This allows image processing to account for changes in the respiratory cycle, improving processing accuracy. Lung motion vector field analysis is performed on the pulmonary airway structure segmentation image and the pulmonary vascular structure segmentation image using the respiratory phase segmentation data (step S342) to generate a lung motion vector field. This step captures lung motion information in different respiratory phases, providing a basis for subsequent pixel matching. Motion pixel matching is performed on the pulmonary airway structure segmentation image and the pulmonary vascular structure segmentation image using the lung motion vector field to generate motion-matched pixels. This step ensures accurate matching of corresponding pixels in the images across different respiratory phases, improving segmentation accuracy. Motion displacement region segmentation is performed on the pulmonary airway structure segmentation image and the pulmonary vascular structure segmentation image based on the motion-matched pixels (step S343) to generate an image of respiratory-related artifact areas. This step identifies artifact areas caused by respiration and motion, providing target areas for subsequent optimization. Based on the lung motion vector field, respiratory motion masking is performed on the respiratory segmentation artifact image (step S344), generating an artifact-optimized respiratory segmentation image. This masking correction reduces the interference of motion artifacts and optimizes the final image quality. These processing steps significantly improve the clarity and accuracy of the segmented images of the lung airway and vascular structures, reduce artifacts caused by breathing and motion, and make the final segmented image more reliable.

[0018] Preferably, step S4 includes the following steps: Step S41: performing abnormality detection on the respiratory tract segmentation artifact optimized image to generate lesion area positioning data; performing lesion area segmentation on the respiratory tract segmentation artifact optimized image based on the lesion area positioning data to generate a respiratory tract lesion area segmentation image; Step S42: performing three-dimensional coordinate mapping on the pulmonary airway structure segmentation image, the pulmonary vascular structure segmentation image, and the respiratory tract lesion area segmentation image to generate three-dimensional point cloud data of the segmentation structure; Step S43: annotating and quantifying the segmented structure three-dimensional point cloud data to generate respiratory tract segmented structure quantitative data; reconstructing the segmented structure three-dimensional point cloud data using the respiratory tract segmented structure quantitative data to generate a respiratory tract structure segmented three-dimensional image.

[0019] The present invention generates lesion area positioning data by performing anomaly detection on the respiratory tract segmentation artifact optimized image; based on the lesion area positioning data, the lesion area is segmented on the respiratory tract segmentation artifact optimized image to generate a respiratory tract lesion area segmentation image. This step ensures the precise positioning and segmentation of the lesion area and improves the accuracy of lesion identification. The pulmonary airway structure segmentation image, the pulmonary vascular structure segmentation image, and the respiratory tract lesion area segmentation image are mapped in three-dimensional coordinates to generate segmentation structure three-dimensional point cloud data. Through three-dimensional coordinate mapping, the two-dimensional image data is converted into three-dimensional point cloud data, providing a basis for subsequent three-dimensional reconstruction. The segmentation structure three-dimensional point cloud data is annotated and quantified to generate respiratory tract segmentation structure quantitative data; the segmentation structure three-dimensional point cloud data is reconstructed into a three-dimensional image using the respiratory tract segmentation structure quantitative data to generate a respiratory tract structure segmentation three-dimensional image. This step enables the segmented structure to be presented in the form of a three-dimensional image, improving the three-dimensional sense and intuitiveness of the image. Through anomaly detection and lesion segmentation, the lesion area of ​​the respiratory tract can be accurately identified and segmented. 3D reconstructed respiratory structure images can intuitively display the spatial relationship between the lung airways, blood vessels, and lesion areas, enhancing the visualization of medical images and helping doctors better understand the lesion structure. Automated data processing and 3D reconstruction reduce human intervention and errors, improving processing efficiency and image reliability.

[0020] Preferably, step S41 includes the following steps: Step S411: performing pixel brightness detection on the respiratory tract segmentation artifact optimized image to obtain respiratory tract pixel brightness data; performing pixel clustering on the respiratory tract pixel brightness data to generate respiratory tract pixel brightness cluster data; Step S412: confirming the range of suspicious lesion points in the respiratory tract segmentation artifact optimized image based on the respiratory tract pixel brightness cluster data, and generating respiratory tract pixel suspicious lesion point range data; Step S413: Performing a connected domain analysis on the range data of suspicious lesion points in the respiratory tract pixels to generate respiratory tract abnormal point range data; performing a sliding window scan on the respiratory tract segmentation artifact optimized image based on the respiratory tract abnormal point range data to generate lesion area positioning data.

[0021] The present invention performs pixel brightness detection on the respiratory tract segmentation artifact-optimized image to obtain respiratory tract pixel brightness data. This data is then clustered to generate respiratory tract pixel brightness cluster data. This step effectively distinguishes pixel brightness between normal and abnormal areas, providing basic data for subsequent lesion detection. Based on the respiratory tract pixel brightness cluster data, the respiratory tract segmentation artifact-optimized image is used to confirm the range of suspected lesions (step S412), generating respiratory tract pixel range data. This step effectively screens possible lesions based on brightness differences, narrowing the detection range. Connected domain analysis is performed on the respiratory tract pixel range data to generate respiratory tract abnormal point range data. Connected domain analysis identifies and connects adjacent abnormal points to form a complete abnormal region, improving the accuracy of lesion detection. Based on the respiratory tract abnormal point range data, a sliding window scan is performed on the respiratory tract segmentation artifact-optimized image to generate lesion region location data. This sliding window scan progressively scans each region in the image, ensuring that all possible lesions are detected and their locations are accurately located. Through precise brightness detection and cluster analysis, as well as connected domain analysis and sliding window scanning, the accuracy and sensitivity of lesion detection can be significantly improved, and the probability of missed detection and false detection can be reduced.

[0022] In this specification, a respiratory image-based segmentation system is provided for executing the above-mentioned respiratory image-based segmentation method. The respiratory image-based segmentation system includes: The airway structure segmentation module is used to obtain an initial respiratory image; extract lung airway image features from the initial respiratory image to obtain lung airway feature data; and segment the respiratory image into lung airway structure regions based on the lung airway feature data to generate a lung airway structure segmentation image. A vascular structure segmentation module is used to construct an airway exclusion mask based on the lung airway structure segmentation image to obtain the airway exclusion mask; perform binary division of the respiratory image using the airway exclusion mask to generate an initial vascular region image; perform vascular tree analysis on the initial vascular region image to generate respiratory vascular tree data; perform vascular structure region image segmentation on the initial vascular region image using the respiratory vascular tree data to generate a lung vascular structure segmentation image; An artifact optimization module is used to obtain user posture data and user respiratory rate data; perform user motion impact assessment based on the user posture data and user respiratory rate data to generate user motion impact assessment data; perform image artifact optimization on the lung airway structure segmentation image and the lung vascular structure segmentation image using the user motion impact assessment data to generate a respiratory segmentation artifact optimized image; The three-dimensional reconstruction module is used to detect abnormalities in the respiratory tract segmentation artifact optimized image and generate lesion area positioning data; perform lesion area segmentation on the respiratory tract segmentation artifact optimized image based on the lesion area positioning data to generate a respiratory tract lesion area segmentation image; perform three-dimensional image reconstruction on the lung airway structure segmentation image, the lung vascular structure segmentation image and the respiratory tract lesion area segmentation image to generate a respiratory tract structure segmentation three-dimensional image.

[0023] The beneficial effect of the present invention is that by acquiring an initial image of the respiratory tract and performing lung airway image feature extraction, characteristic data related to the respiratory tract can be extracted for subsequent analysis and processing. Based on the lung airway feature data, the respiratory tract image is segmented into a lung airway structure region image to generate a lung airway structure segmentation image. This helps to accurately locate and segment the airway structure of the respiratory tract. Based on the lung airway structure segmentation image, an airway exclusion mask is constructed to exclude the airway from the respiratory tract image so as to more accurately analyze and process other structures. By performing a vascular tree analysis on the initial vascular region image, vascular tree data of the respiratory tract can be extracted to provide vascular structure information for subsequent analysis and processing. Based on the respiratory vascular tree data, the initial vascular region image is segmented into a vascular structure region image to generate a lung vascular structure segmentation image. This helps to accurately describe the vascular structure of the respiratory tract. By acquiring user posture data and respiratory rate data, the user's movement is evaluated to understand the degree of influence of the user's movement state on the respiratory tract image. Based on the user motion impact assessment data, the image artifact optimization is performed on the lung airway structure segmentation image and the lung blood vessel structure segmentation image to reduce the artifacts caused by user motion and improve the image quality. The respiratory part segmentation artifact optimized image is detected for abnormalities, the lesion area is located, and a respiratory lesion area segmentation image is generated. This helps to detect and locate the lesion area of ​​the respiratory tract. By performing three-dimensional image reconstruction on the lung airway structure segmentation image, the lung blood vessel structure segmentation image and the respiratory lesion area segmentation image, a three-dimensional image of the respiratory tract structure segmentation can be generated to provide more comprehensive respiratory tract structure information. Therefore, the present invention improves the accuracy and simplicity of the respiratory tract image segmentation method by performing feature extraction, structure segmentation optimization, motion impact assessment, lesion area detection and three-dimensional reconstruction technology on the respiratory tract CT image. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 Schematic diagram of the steps of a segmentation method based on respiratory images; Figure 2 for Figure 1 Detailed implementation steps of step S2 in FIG. Figure 3 for Figure 1 Detailed implementation steps of step S3 in FIG. Figure 4 for Figure 1 Detailed implementation steps of step S4 in FIG. The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0025] The following is a clear and complete description of the technical method of the present invention in conjunction with the accompanying drawings. It is obvious that the embodiments described are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making any creative efforts are within the scope of protection of the present invention.

[0026] In addition, the accompanying drawings are merely schematic illustrations of the present invention and are not necessarily drawn to scale. Identical reference numerals in the figures denote identical or similar parts, and thus repetitive descriptions thereof will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities that do not necessarily correspond to physically or logically separate entities. These functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor and / or microcontroller approaches.

[0027] It should be understood that although the terms "first," "second," and the like may be used herein to describe various elements, these elements should not be limited by these terms. These terms are used solely to distinguish one element from another. For example, a first element may be referred to as a second element, and similarly, a second element may be referred to as a first element, without departing from the scope of the exemplary embodiments. The term "and / or" as used herein includes any and all combinations of one or more of the listed associated items.

[0028] To achieve this, please refer to Figures 1 to 4 , a respiratory tract image-based segmentation method, the method comprising the following steps: Step S1: Acquire an initial respiratory tract image; perform lung airway image feature extraction on the initial respiratory tract image to obtain lung airway feature data; perform lung airway structure region image segmentation on the respiratory tract image based on the lung airway feature data to generate a lung airway structure segmentation image; Step S2: constructing an airway exclusion mask based on the lung airway structure segmentation image to obtain an airway exclusion mask; performing binary segmentation on the respiratory image using the airway exclusion mask to generate an initial vascular region image; performing vascular tree analysis on the initial vascular region image to generate respiratory vascular tree data; performing vascular structure region image segmentation on the initial vascular region image using the respiratory vascular tree data to generate a lung vascular structure segmentation image; Step S3: Acquire user posture data and user respiratory rate data; perform user motion impact assessment based on the user posture data and user respiratory rate data to generate user motion impact assessment data; perform image artifact optimization on the lung airway structure segmentation image and the lung vascular structure segmentation image using the user motion impact assessment data to generate a respiratory segmentation artifact optimized image; Step S4: performing abnormality detection on the respiratory tract segmentation artifact optimized image to generate lesion area positioning data; performing lesion area segmentation on the respiratory tract segmentation artifact optimized image according to the lesion area positioning data to generate a respiratory tract lesion area segmentation image; performing three-dimensional image reconstruction on the lung airway structure segmentation image, the lung vascular structure segmentation image and the respiratory tract lesion area segmentation image to generate a respiratory tract structure segmentation three-dimensional image.

[0029] The present invention acquires an initial image of the respiratory tract and then extracts features from the lung airway image. These features include lung morphology, duct diameter, and branching structure. Feature extraction quantifies and describes the structure of the respiratory tract, providing foundational data for subsequent segmentation and analysis. This helps physicians accurately identify and analyze the health of a patient's airway, such as detecting stenosis, obstruction, or other abnormalities. By constructing an airway exclusion mask, the airway portion of the respiratory tract image is effectively isolated. This allows subsequent image processing to focus on analyzing the vascular region, eliminating the influence of the airways. This is crucial for accurately analyzing the branching structure, density, and location of blood vessels, and helps provide precise vascular tree data. By analyzing the user's posture and respiratory rate data, the impact of user motion during image acquisition can be assessed and understood. This data provides key insights into the impact of image quality. Optimizing the segmented images of the lung airway and vascular structure reduces or eliminates artifacts and blur caused by user motion, thereby improving image clarity and resolution. Abnormality detection performed on the optimized segmented images of the respiratory tract allows for rapid and accurate localization of lesions, specifically masses, inflammation, or other abnormal structures. By accurately locating and segmenting lesion areas, doctors can identify patients' health problems in advance. Therefore, the present invention improves the accuracy and simplicity of respiratory image segmentation methods by performing feature extraction, structural segmentation optimization, motion impact assessment, lesion area detection, and 3D reconstruction on respiratory tract CT images.

[0030] In the embodiment of the present invention, reference Figure 1 The above is a schematic flow chart of the steps of a respiratory tract image segmentation method according to the present invention. In this example, the respiratory tract image segmentation method includes the following steps: Step S1: Acquire an initial respiratory tract image; perform lung airway image feature extraction on the initial respiratory tract image to obtain lung airway feature data; perform lung airway structure region image segmentation on the respiratory tract image based on the lung airway feature data to generate a lung airway structure segmentation image; In an embodiment of the present invention, high-resolution respiratory tract images are acquired using medical imaging equipment such as CT (computed tomography) and MRI (magnetic resonance imaging). The acquired respiratory tract images are preprocessed, including denoising, contrast enhancement, and image normalization, in preparation for feature extraction. Computer vision and image processing techniques, such as edge detection, feature point extraction, and texture analysis, are used to extract characteristic data of the lung airway structure. Based on the acquired lung airway characteristic data, a specific segmentation algorithm is employed to perform regional image segmentation. Commonly used segmentation methods include threshold-based segmentation, region growing, and active contour models (such as the Chan-Vese model) to ensure accurate segmentation of the lung airway structure. Combined with the output of the segmentation algorithm, clear segmented images of the lung airway structure are generated. These images display detailed structural information of the lung airways, providing accurate anatomical information for subsequent steps such as vascular analysis and lesion detection.

[0031] Step S2: constructing an airway exclusion mask based on the lung airway structure segmentation image to obtain an airway exclusion mask; performing binary segmentation on the respiratory image using the airway exclusion mask to generate an initial vascular region image; performing vascular tree analysis on the initial vascular region image to generate respiratory vascular tree data; performing vascular structure region image segmentation on the initial vascular region image using the respiratory vascular tree data to generate a lung vascular structure segmentation image; In an embodiment of the present invention, an airway exclusion mask is created by using a lung airway structure segmentation image through image processing techniques such as morphological operations or region labeling. This mask removes the airway region from the image and focuses on the analysis of blood vessels and other tissue structures. The respiratory image and the airway exclusion mask are binarized to obtain an initial vascular region image. In this image, the airway region is excluded, leaving a binary image containing only vascular structures. A vascular tree analysis is performed on the initial vascular region image. This involves vascular branch point detection, vascular diameter measurement, and even analysis of the integrity of the vascular network to obtain comprehensive vascular tree data. Based on the results of the vascular tree analysis, the initial vascular region image is further segmented using a specific segmentation algorithm such as a threshold-based, region growing, or a specific vascular segmentation algorithm. This will generate a clear segmented image of the lung vascular structure.

[0032] Step S3: Acquire user posture data and user respiratory rate data; perform user motion impact assessment based on the user posture data and user respiratory rate data to generate user motion impact assessment data; perform image artifact optimization on the lung airway structure segmentation image and the lung vascular structure segmentation image using the user motion impact assessment data to generate a respiratory segmentation artifact optimized image; In an embodiment of the present invention, appropriate sensors or monitoring devices are used to obtain the user's posture data during the image scan, including body position, posture, and motion trajectory. Simultaneously, the user's respiratory rate data is obtained, which can be obtained through a respiratory monitoring device or a respiratory signal monitoring system built into the image scanning device. The acquired posture data is used to analyze user motion introduced during the scan, such as breathing and body movement. The user's breathing pattern and its potential impact on image quality are evaluated in combination with the respiratory rate data. The posture data and respiratory rate data are integrated and analyzed to generate a detailed user motion impact assessment report. This report describes the impact of user motion on respiratory image segmentation, such as image blur and artifacts. Based on the motion impact assessment data, corresponding image processing algorithms are developed or applied to specifically optimize the lung airway structure segmentation image and the lung vascular structure segmentation image. Specific optimization measures include motion correction, image reconstruction algorithms, motion artifact removal techniques, etc., to reduce motion-induced artifacts and image blur. The optimized lung airway structure segmentation image and lung vascular structure segmentation image are combined to generate an artifact-optimized respiratory segmentation image. These optimized images retain high-quality anatomical information, reduce imaging artifacts caused by user motion, and improve image reliability and accuracy.

[0033] Step S4: performing abnormality detection on the respiratory tract segmentation artifact optimized image to generate lesion area positioning data; performing lesion area segmentation on the respiratory tract segmentation artifact optimized image according to the lesion area positioning data to generate a respiratory tract lesion area segmentation image; performing three-dimensional image reconstruction on the lung airway structure segmentation image, the lung vascular structure segmentation image and the respiratory tract lesion area segmentation image to generate a respiratory tract structure segmentation three-dimensional image.

[0034] In embodiments of the present invention, abnormality detection is performed on an optimized respiratory tract segmentation image using a computer-aided design (CAD) system or a deep learning model. This can include detecting abnormal areas such as lesions, nodules, and spots. Lesion location data is generated from the detected abnormal areas, recording information such as the location, size, and shape of the lesion. Utilizing this lesion location data, the optimized respiratory tract segmentation image is further segmented using region growing, morphological processing, or a deep learning segmentation network to accurately segment the lesion. The resulting segmented image of the respiratory tract lesion region displays the detailed structure of the lesion region. The optimized segmented images of the lung airway structure, lung vascular structure, and respiratory lesion region are integrated and reconstructed in three dimensions. Three-dimensional reconstruction algorithms, such as voxelization and marching cubes, are used to convert the two-dimensional image data into a three-dimensional respiratory tract structure segmentation image. The resulting three-dimensional image provides a more intuitive and comprehensive display of the lung's anatomical structure, vascular network, and lesion location.

[0035] Preferably, step S1 includes the following steps: Step S11: Scan the user's respiratory tract using a CT machine to obtain an initial respiratory tract image; Step S12: performing image standardization on the initial respiratory tract image to obtain a standard respiratory tract image, wherein the image preprocessing includes image denoising, image brightness enhancement, and image smoothing; Step S13: extracting lung airway image features from the standard respiratory tract image to obtain lung airway feature data; Step S14: performing lung airway structure region image segmentation on the standard respiratory image according to the lung airway feature data to generate a lung airway structure segmentation image.

[0036] In an embodiment of the present invention, a computed tomography (CT) device is used to scan the user's airway to obtain a high-resolution initial airway image. CT scans typically provide detailed anatomical information, including clear visualization of the airways and blood vessels. The acquired initial airway image is then subjected to image preprocessing, including using filters or denoising algorithms to remove noise from the image to improve image quality. The image brightness and contrast are adjusted to highlight the details of the airway structure. A smoothing filter, such as a Gaussian filter, is applied to reduce graininess and unnecessary detail in the image, resulting in a smoother and more continuous image. Computer vision and image processing techniques are used to extract features from the standardized airway image. Feature extraction can include methods such as edge detection, texture analysis, and morphological feature extraction to obtain quantitative feature data of the lung airway structure. Based on the obtained lung airway feature data, a specific image segmentation algorithm is used to segment the lung airway structure region in the standardized airway image. Common segmentation methods include threshold segmentation, region growing, and active contour models to accurately segment the lung airway structure. The generated lung airway structure segmentation image shows a clear airway structure, providing important anatomical information for subsequent medical image analysis and clinical practice.

[0037] Preferably, step S14 includes the following steps: Step S131: performing airway pixel category probability calculation on the standard airway image to obtain lung airway pixel category data; performing initial airway lung region segmentation on the standard airway image based on the lung airway pixel category data to obtain an initial airway lung region image; Step S132: removing small-area impurities from the initial respiratory tract and lung region image using cell tissue morphology theory to obtain an impurity-removed image of the respiratory tract and lung region; Step S133: performing 3D voxelization on the impurity-removed image of the respiratory tract lung region to generate respiratory tract lung region voxels; confirming the center points of the respiratory tract lung region voxels to obtain the main airway center points of the respiratory tract lung region; performing lung airway routing search on the respiratory tract lung region voxels using the main airway center points of the respiratory tract lung region to generate lung airway routing search data; Step S134: extracting airway branch points from the respiratory tract lung region impurity removal image using the lung airway route search data to obtain respiratory tract lung bronchial path data; performing lung bronchial topology analysis on the respiratory tract lung bronchial path data to generate respiratory tract lung bronchial topology data; Step S135: performing lung airway structure region image segmentation on the respiratory tract lung region impurity removal image according to the respiratory tract lung bronchial topological structure data and the main airway center point of the respiratory tract lung region to generate a lung airway structure segmentation image.

[0038] In an embodiment of the present invention, a machine learning or deep learning model is utilized to classify each pixel in a standard respiratory image, calculating its probability of belonging to an airway or non-airway, and generating an airway pixel category probability map. Based on the airway pixel category probability map, the standard respiratory image is initially segmented to produce an initial respiratory lung region image containing both airways and lung regions. Using cellular tissue morphology theory, small impurities in the initial respiratory lung region image are identified and eliminated, removing noise and unnecessary details from the image. The result is a clearer, impurity-removed image of the respiratory lung region that preserves the primary airway and lung structures. The impurity-removed image of the respiratory lung region is converted into three-dimensional voxel data representing the three-dimensional structure of the airways and lung regions. The generated voxel data is analyzed to identify the center points of the major airways, which are typically located at the geometric center of the airways. Based on the major airway center points, a lung airway routing search is performed to generate detailed lung airway network data containing airway path information. Using the lung airway routing search data, airway branch points are extracted from the impurity-removed image of the respiratory lung region. The extracted bronchial pathway data is subjected to topological analysis to generate topological data for the airway network, including bronchial connectivity and distribution patterns. Combining this bronchial topological data with information about the main airway center points, the final lung airway structural region image segmentation is performed on the impurity-removed lung region image. The resulting lung airway structural segmentation image accurately displays the distribution and connectivity of the airways, providing a reliable foundation for further medical imaging analysis and clinical practice.

[0039] As an example of the present invention, refer to Figure 2 As shown, in this example, step S2 includes: Step S21: constructing an airway exclusion mask based on the lung airway structure segmentation image to obtain an airway exclusion mask; Step S22: eliminating the influence of the airway position on the standard airway image using the airway exclusion mask, thereby generating a first residual image of the airway; Step S23: performing blood vessel pixel category probability calculation on the first residual image of the respiratory tract to obtain blood vessel pixel category data; performing binary division on the first residual image of the respiratory tract based on the blood vessel category data to generate an initial blood vessel region image; Step S24: Extract the vascular centerline of the initial vascular region image to obtain vascular centerline data; perform vascular tree analysis on the initial vascular region image based on the vascular centerline data to generate respiratory vascular tree data; and segment the vascular structure region image of the initial vascular region image using the respiratory vascular tree data to generate a pulmonary vascular structure segmentation image.

[0040] In this embodiment of the present invention, pixels in the airway region are identified from the lung airway structure segmentation image generated in step S14. A binary mask is created, in which pixels in the airway region are marked as 1 and pixels in the non-airway region are marked as 0. This generated airway exclusion mask is used in subsequent steps to eliminate the influence of the airway on other structures. A pixel-level operation is performed on the airway exclusion mask and the standard airway image. The pixel values ​​of the standard airway image corresponding to areas where the mask is 1 are set to 0, generating a first residual image of the airway. This first residual image of the airway primarily retains blood vessels and other non-airway structures, providing basic data for subsequent vascular analysis. Using a machine learning model such as a convolutional neural network (CNN) or random forest, each pixel in the first residual image of the airway is classified, calculating its probability of belonging to a blood vessel or non-blood vessel, and generating a vascular pixel class probability map. The vascular pixel class probability map is binarized using the Otsu method or adaptive threshold segmentation method to generate an initial vascular region image and identify the vascular regions. A skeletonization algorithm or a centerline extraction algorithm is used to extract vascular centerlines from the initial vascular region image. Generate vascular centerline data, representing the main vascular path. Based on the extracted vascular centerline data, perform vascular tree structure analysis to identify vascular branches and connections. Generate respiratory vascular tree data, including vascular topology information. Using this vascular tree data, further segment the initial vascular region image to determine the specific vascular boundaries. Generate a segmented image of the pulmonary vascular structure, displaying the detailed vascular network structure.

[0041] Preferably, performing vascular tree analysis on the initial vascular region image based on the vascular centerline data includes: Perform data preprocessing on the vascular centerline data to generate a standard vascular centerline coordinate sequence, wherein the data preprocessing includes data cleaning, data missing value filling and data standardization; Performing topological analysis on the initial vascular region image according to the standard vascular centerline coordinate sequence to generate vascular region topology data; using a tree structure method to connect the vascular region topology data according to the standard vascular centerline coordinate sequence to generate a vascular site network; Performing a blood vessel diameter size analysis on the blood vessel network to generate blood vessel diameter size data; comparing the blood vessel diameter size data with a preset blood vessel diameter size threshold; and when the blood vessel diameter size data is greater than or equal to the preset blood vessel diameter size threshold, marking the corresponding blood vessel in the blood vessel network as main blood vessel data; When the blood vessel diameter size data is smaller than the preset blood vessel diameter size threshold, the corresponding blood vessel in the blood vessel area network is marked as branch vessel data; the initial blood vessel area image is analyzed for blood vessel tree nodes using the trunk vessel data and branch vessel data to generate respiratory tract blood vessel tree data.

[0042] In embodiments of the present invention, a skeletonization algorithm or a centerline extraction algorithm is used to extract vascular centerlines from an initial vascular region image to generate vascular centerline data. The vascular centerline data is cleaned to remove noise and outliers to ensure data accuracy and consistency. Missing values ​​in the data are filled using interpolation or other appropriate methods to generate complete vascular centerline data. The vascular centerline data is standardized to have the same scale and units, generating a standard vascular centerline coordinate sequence. Based on the standard vascular centerline coordinate sequence, a topological analysis is performed on the initial vascular region image to identify vascular branching points and connections, generating vascular region topology data. Using a tree structure method, the vascular region topology data is interconnected based on the standard vascular centerline coordinate sequence to generate a complete vascular location network. Vessel diameter analysis is performed on the vascular location network, measuring the diameter of each vessel to generate vascular diameter data. The vascular diameter data is then compared with a preset vascular diameter threshold. When the vessel diameter data is greater than or equal to the preset vascular diameter threshold, the corresponding vessel is marked as a trunk vessel. When the vessel diameter data is less than the preset vascular diameter threshold, the corresponding vessel is marked as a branch vessel. Using the trunk and branch vessel data, the initial vascular region image is analyzed for vascular tree nodes, identifying and labeling the main nodes and branches of the vessels. This generates respiratory vascular tree data containing detailed vascular topology and classification information.

[0043] As an example of the present invention, refer to Figure 3 As shown, in this example, step S3 includes: Step S31: Acquire user posture data and user breathing rate data; Step S32: performing posture change analysis on the user posture data to generate user posture change data; performing breathing intensity analysis on the user breathing frequency data to generate user breathing intensity data; Step S33: performing a user motion impact assessment based on the user posture change data and the user breathing intensity data to generate user motion impact assessment data; performing a CT image part disturbance analysis on the user motion impact assessment data to obtain user motion part disturbance data; Step S34: performing motion displacement region segmentation on the lung airway structure segmentation image and the lung vascular structure segmentation image based on the user motion part disturbance data to generate a respiratory tract associated artifact part image; performing image artifact optimization on the respiratory tract segmentation artifact part image to generate a respiratory tract part segmentation artifact optimized image.

[0044] In an embodiment of the present invention, user posture data is collected in real time using sensor devices (such as accelerometers and gyroscopes). Respiratory monitoring equipment (such as a respiratory belt or chest sensor) is used to obtain the user's respiratory rate data. Time series analysis is performed on the collected user posture data to calculate changes in the user's posture at different time points. User posture change data is generated to describe the user's motion trajectory and posture changes during the scan. Signal processing and analysis are performed on the user's respiratory rate data to calculate the respiratory cycle and intensity. User respiratory intensity data is generated to describe the user's breathing pattern and intensity changes during the scan. The user posture change data and respiratory intensity data are combined to assess the impact of user motion on CT scan images. User motion impact assessment data is generated to describe the potential impact of user motion on different image regions. Based on the user motion impact assessment data, the disturbance of various parts of the CT image affected by motion is analyzed. User motion part disturbance data is generated to identify image regions significantly affected by motion. The user motion part disturbance data is used to process pulmonary airway structure segmentation images and pulmonary vascular structure segmentation images to identify and segment regions affected by motion. Generate an image of the respiratory tract associated artifacts, identifying areas containing motion artifacts. Perform artifact optimization on the image of the respiratory tract associated artifacts, using image restoration and filtering techniques to reduce the impact of artifacts. Generate an artifact-optimized image of the respiratory tract segmentation, providing clearer and more accurate image data.

[0045] Preferably, step S34 includes the following steps: Step S341: extracting a time series image sequence from the pulmonary airway structure segmentation image and the pulmonary vascular structure segmentation image to obtain CT image time series data; performing respiratory phase division on the CT image time series data to generate respiratory phase division data; Step S342: performing lung motion vector field analysis on the lung airway structure segmentation image and the lung vascular structure segmentation image using the respiratory phase division data to generate a lung motion vector field; performing motion pixel matching on the lung airway structure segmentation image and the lung vascular structure segmentation image using the lung motion vector field to generate motion matching pixels; Step S343: performing motion displacement region segmentation on the lung airway structure segmentation image and the lung blood vessel structure segmentation image according to the motion matching pixels to generate a respiratory tract associated artifact portion image; Step S344: performing respiratory motion region masking correction on the respiratory tract segmentation artifact part image based on the lung motion vector field to generate a respiratory tract segmentation artifact optimized image.

[0046] In an embodiment of the present invention, multi-frame CT scans are performed on segmented images of the lung airway structure and segmented images of the lung vascular structure to obtain images at different times, forming a time-series image sequence. CT image time-series data is obtained, containing time information and image data for each frame. Based on the user's respiratory rate data, the CT image time-series data is divided into respiratory phases, dividing the entire respiratory cycle into several phases. Respiratory phase segmentation data is generated, with each phase corresponding to a set of CT image frames, identifying different stages of the respiratory cycle. Using the respiratory phase segmentation data, motion analysis is performed on the segmented images of the lung airway structure and the segmented images of the lung vascular structure to calculate the motion vectors of the lungs at each phase during respiration. A lung motion vector field is generated, describing the motion direction and amplitude of each image pixel at different phases. Using the lung motion vector field, pixel-level matching is performed on the segmented images of the lung airway structure and the segmented images of the lung vascular structure to identify corresponding moving pixels. Motion-matched pixel data is generated, containing matching relationships at different phases. Based on the motion-matched pixel data, motion displacement region segmentation is performed on the segmented images of the lung airway structure and the segmented images of the lung vascular structure to identify regions affected by motion. Generate an image of respiratory tract artifacts and identify image regions containing motion artifacts. Based on the lung motion vector field, perform mask correction on the image of respiratory tract artifacts, utilizing respiratory motion patterns and vector field data to repair and optimize the image. Generate an optimized image of respiratory tract segmentation artifacts, reducing the impact of motion artifacts and improving image quality.

[0047] As an example of the present invention, refer to Figure 4 As shown, in this example, step S4 includes: Step S41: performing abnormality detection on the respiratory tract segmentation artifact optimized image to generate lesion area positioning data; performing lesion area segmentation on the respiratory tract segmentation artifact optimized image based on the lesion area positioning data to generate a respiratory tract lesion area segmentation image; Step S42: performing three-dimensional coordinate mapping on the pulmonary airway structure segmentation image, the pulmonary vascular structure segmentation image, and the respiratory tract lesion area segmentation image to generate three-dimensional point cloud data of the segmentation structure; Step S43: annotating and quantifying the segmented structure three-dimensional point cloud data to generate respiratory tract segmented structure quantitative data; reconstructing the segmented structure three-dimensional point cloud data using the respiratory tract segmented structure quantitative data to generate a respiratory tract structure segmented three-dimensional image.

[0048] In an embodiment of the present invention, the respiratory tract segmentation artifact optimized image is analyzed by utilizing image processing and machine learning techniques to detect abnormal areas in the image. Abnormality detection methods include but are not limited to edge detection, texture analysis, and deep learning models. Lesion area positioning data is generated to identify the image area where the lesion exists. Based on the lesion area positioning data, the respiratory tract segmentation artifact optimized image is finely segmented to extract the specific lesion area. Segmentation algorithms such as region growing, K-means clustering, or a deep learning segmentation network are used to segment the lesion area. A respiratory tract lesion area segmentation image is generated to clearly display the location and morphology of the lesion. Three-dimensional coordinate mapping is performed on the lung airway structure segmentation image, the lung blood vessel structure segmentation image, and the respiratory tract lesion area segmentation image to convert the two-dimensional image data into point cloud data in three-dimensional space. A registration algorithm is used during the mapping process to ensure the alignment and consistency of different images. Three-dimensional point cloud data of the segmented structure is generated, including the spatial location and structural information of the lung airways, blood vessels, and lesion areas. The point cloud data contains the three-dimensional coordinates and corresponding attribute information of each structure. The three-dimensional point cloud data of the segmented structure is annotated and quantified, and the volume, surface area, and other geometric properties of each structure are measured. The annotation and quantification methods include voxel counting, surface mesh generation, and morphological analysis. Generate quantitative data of the respiratory tract segmentation structure, providing detailed structural properties and measurement results. Using the quantitative data of the respiratory tract segmentation structure, the three-dimensional point cloud data of the segmented structure is reconstructed to generate a complete three-dimensional image. During the reconstruction process, surface reconstruction algorithms such as Marching Cubes or Poisson Surface Reconstruction are used to generate a high-quality three-dimensional model. Generate a three-dimensional image of the respiratory tract structure segmentation, and intuitively display the three-dimensional morphology of each respiratory tract structure and its lesion area.

[0049] Preferably, step S41 includes the following steps: Step S411: performing pixel brightness detection on the respiratory tract segmentation artifact optimized image to obtain respiratory tract pixel brightness data; performing pixel clustering on the respiratory tract pixel brightness data to generate respiratory tract pixel brightness cluster data; Step S412: confirming the range of suspicious lesion points in the respiratory tract segmentation artifact optimized image based on the respiratory tract pixel brightness cluster data, and generating respiratory tract pixel suspicious lesion point range data; Step S413: Performing a connected domain analysis on the range data of suspicious lesion points in the respiratory tract pixels to generate respiratory tract abnormal point range data; performing a sliding window scan on the respiratory tract segmentation artifact optimized image based on the respiratory tract abnormal point range data to generate lesion area positioning data.

[0050] In an embodiment of the present invention, brightness data for respiratory tract pixel segments is obtained by performing brightness detection on each pixel in the respiratory tract segmentation artifact-optimized image and recording its brightness value. A specific brightness detection method, such as grayscale analysis or color space-based brightness calculation, is used. Cluster analysis is performed on the respiratory tract pixel brightness data to group pixels with similar brightness characteristics. Clustering algorithms, such as K-means and DBSCAN, are used to process the pixel brightness data to generate respiratory tract pixel brightness cluster data and identify regions with different brightness characteristics. Based on the respiratory tract pixel brightness cluster data, the respiratory tract segmentation artifact-optimized image is analyzed to determine the possible range of lesions. A threshold segmentation method or a statistically based anomaly detection algorithm is used to identify pixel regions with abnormal brightness characteristics. Suspected lesion range data for respiratory tract pixels is generated to identify the pixel ranges where lesions may be present. Connected domain analysis is performed on the suspected lesion range data for respiratory tract pixels to identify interconnected pixel groups and generate respiratory tract abnormality range data. A connected component labeling algorithm, such as the Flood Fill algorithm, is used to process the pixel data to determine the range and morphology of each connected domain. Based on the respiratory tract abnormality range data, a sliding window scan is performed on the respiratory tract segmentation artifact-optimized image to precisely locate the lesion area. Using an appropriately sized sliding window, the image is progressively scanned to detect the distribution of abnormalities within each window. This generates lesion area localization data, identifying the specific location and boundaries of the lesion area.

[0051] In this specification, a respiratory image-based segmentation system is provided for executing the above-mentioned respiratory image-based segmentation method. The respiratory image-based segmentation system includes: The airway structure segmentation module is used to obtain an initial respiratory image; extract lung airway image features from the initial respiratory image to obtain lung airway feature data; and segment the respiratory image into lung airway structure regions based on the lung airway feature data to generate a lung airway structure segmentation image. A vascular structure segmentation module is used to construct an airway exclusion mask based on the lung airway structure segmentation image to obtain the airway exclusion mask; perform binary division of the respiratory image using the airway exclusion mask to generate an initial vascular region image; perform vascular tree analysis on the initial vascular region image to generate respiratory vascular tree data; perform vascular structure region image segmentation on the initial vascular region image using the respiratory vascular tree data to generate a lung vascular structure segmentation image; An artifact optimization module is used to obtain user posture data and user respiratory rate data; perform user motion impact assessment based on the user posture data and user respiratory rate data to generate user motion impact assessment data; perform image artifact optimization on the lung airway structure segmentation image and the lung vascular structure segmentation image using the user motion impact assessment data to generate a respiratory segmentation artifact optimized image; The three-dimensional reconstruction module is used to detect abnormalities in the respiratory tract segmentation artifact optimized image and generate lesion area positioning data; perform lesion area segmentation on the respiratory tract segmentation artifact optimized image based on the lesion area positioning data to generate a respiratory tract lesion area segmentation image; perform three-dimensional image reconstruction on the lung airway structure segmentation image, the lung vascular structure segmentation image and the respiratory tract lesion area segmentation image to generate a respiratory tract structure segmentation three-dimensional image.

[0052] The beneficial effect of the present invention is that by acquiring an initial image of the respiratory tract and performing lung airway image feature extraction, characteristic data related to the respiratory tract can be extracted for subsequent analysis and processing. Based on the lung airway feature data, the respiratory tract image is segmented into a lung airway structure region image to generate a lung airway structure segmentation image. This helps to accurately locate and segment the airway structure of the respiratory tract. Based on the lung airway structure segmentation image, an airway exclusion mask is constructed to exclude the airway from the respiratory tract image so as to more accurately analyze and process other structures. By performing a vascular tree analysis on the initial vascular region image, vascular tree data of the respiratory tract can be extracted to provide vascular structure information for subsequent analysis and processing. Based on the respiratory vascular tree data, the initial vascular region image is segmented into a vascular structure region image to generate a lung vascular structure segmentation image. This helps to accurately describe the vascular structure of the respiratory tract. By acquiring user posture data and respiratory rate data, the user's movement is evaluated to understand the degree of influence of the user's movement state on the respiratory tract image. Based on the user motion impact assessment data, the image artifact optimization is performed on the lung airway structure segmentation image and the lung blood vessel structure segmentation image to reduce the artifacts caused by user motion and improve the image quality. The respiratory part segmentation artifact optimized image is detected for abnormalities, the lesion area is located, and a respiratory lesion area segmentation image is generated. This helps to detect and locate the lesion area of ​​the respiratory tract. By performing three-dimensional image reconstruction on the lung airway structure segmentation image, the lung blood vessel structure segmentation image and the respiratory lesion area segmentation image, a three-dimensional image of the respiratory tract structure segmentation can be generated to provide more comprehensive respiratory tract structure information. Therefore, the present invention improves the accuracy and simplicity of the respiratory tract image segmentation method by performing feature extraction, structure segmentation optimization, motion impact assessment, lesion area detection and three-dimensional reconstruction technology on the respiratory tract CT image.

[0053] The present invention is therefore intended to be illustrative and non-restrictive in all respects, with the scope of the invention being defined by the appended claims rather than the foregoing description, and all changes that come within the meaning and range of equivalents of the application documents are intended to be embraced therein.

[0054] The foregoing description is intended only to provide specific embodiments of the present invention, which will enable those skilled in the art to understand and implement the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not intended to be limited to the embodiments shown herein, but is to be construed in the widest possible manner consistent with the principles and novel features disclosed herein.

Claims

1. A respiratory tract image segmentation method, characterized in that: The following steps are involved: Step S1: Acquire an initial respiratory tract image; Extracting lung airway image features from the initial respiratory tract image to obtain lung airway feature data; Performing lung airway structure region image segmentation on the respiratory image according to the lung airway feature data to generate a lung airway structure segmentation image; Step S2: constructing an airway exclusion mask based on the lung airway structure segmentation image to obtain an airway exclusion mask; performing binary segmentation on the respiratory image using the airway exclusion mask to generate an initial vascular region image; performing vascular tree analysis on the initial vascular region image to generate respiratory vascular tree data; performing vascular structure region image segmentation on the initial vascular region image using the respiratory vascular tree data to generate a lung vascular structure segmentation image; Step S3: Acquire user posture data and user respiratory rate data; perform user motion impact assessment based on the user posture data and user respiratory rate data to generate user motion impact assessment data; perform image artifact optimization on the lung airway structure segmentation image and the lung vascular structure segmentation image using the user motion impact assessment data to generate a respiratory segmentation artifact optimized image; Step S4: performing abnormality detection on the respiratory tract segmentation artifact optimized image to generate lesion area positioning data; performing lesion area segmentation on the respiratory tract segmentation artifact optimized image according to the lesion area positioning data to generate a respiratory tract lesion area segmentation image; performing three-dimensional image reconstruction on the lung airway structure segmentation image, the lung vascular structure segmentation image and the respiratory tract lesion area segmentation image to generate a respiratory tract structure segmentation three-dimensional image.

2. The respiratory tract image segmentation method according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: Scan the user's respiratory tract using a CT machine to obtain an initial respiratory tract image; Step S12: performing image standardization on the initial respiratory tract image to obtain a standard respiratory tract image, wherein the image preprocessing includes image denoising, image brightness enhancement, and image smoothing; Step S13: extracting lung airway image features from the standard respiratory tract image to obtain lung airway feature data; Step S14: performing lung airway structure region image segmentation on the standard respiratory image according to the lung airway feature data to generate a lung airway structure segmentation image.

3. The respiratory tract image segmentation method according to claim 2, wherein: Step S14 includes the following steps: Step S131: performing airway pixel category probability calculation on the standard airway image to obtain lung airway pixel category data; performing initial airway lung region segmentation on the standard airway image based on the lung airway pixel category data to obtain an initial airway lung region image; Step S132: removing small-area impurities from the initial respiratory tract and lung region image using cell tissue morphology theory to obtain an impurity-removed image of the respiratory tract and lung region; Step S133: performing 3D voxelization on the impurity-removed image of the respiratory tract lung region to generate respiratory tract lung region voxels; confirming the center points of the respiratory tract lung region voxels to obtain the main airway center points of the respiratory tract lung region; performing lung airway routing search on the respiratory tract lung region voxels using the main airway center points of the respiratory tract lung region to generate lung airway routing search data; Step S134: extracting airway branch points from the respiratory tract lung region impurity removal image using the lung airway route search data to obtain respiratory tract lung bronchial path data; performing lung bronchial topology analysis on the respiratory tract lung bronchial path data to generate respiratory tract lung bronchial topology data; Step S135: performing lung airway structure region image segmentation on the respiratory tract lung region impurity removal image according to the respiratory tract lung bronchial topological structure data and the main airway center point of the respiratory tract lung region to generate a lung airway structure segmentation image.

4. The respiratory tract image segmentation method according to claim 1, characterized in that: Step S2 includes the following steps: Step S21: constructing an airway exclusion mask based on the lung airway structure segmentation image to obtain an airway exclusion mask; Step S22: eliminating the influence of the airway position on the standard airway image using the airway exclusion mask, thereby generating a first residual image of the airway; Step S23: performing blood vessel pixel category probability calculation on the first residual image of the respiratory tract to obtain blood vessel pixel category data; performing binary division on the first residual image of the respiratory tract based on the blood vessel category data to generate an initial blood vessel region image; Step S24: Extract the vascular centerline of the initial vascular region image to obtain vascular centerline data; perform vascular tree analysis on the initial vascular region image based on the vascular centerline data to generate respiratory vascular tree data; and segment the vascular structure region image of the initial vascular region image using the respiratory vascular tree data to generate a pulmonary vascular structure segmentation image.

5. The respiratory tract image segmentation method according to claim 4, characterized in that: The vascular tree analysis of the initial vascular region image based on the vascular centerline data includes: Perform data preprocessing on the vascular centerline data to generate a standard vascular centerline coordinate sequence, wherein the data preprocessing includes data cleaning, data missing value filling and data standardization; Performing topological analysis on the initial vascular region image according to the standard vascular centerline coordinate sequence to generate vascular region topology data; using a tree structure method to connect the vascular region topology data according to the standard vascular centerline coordinate sequence to generate a vascular site network; Performing a blood vessel diameter size analysis on the blood vessel network to generate blood vessel diameter size data; comparing the blood vessel diameter size data with a preset blood vessel diameter size threshold; and when the blood vessel diameter size data is greater than or equal to the preset blood vessel diameter size threshold, marking the corresponding blood vessel in the blood vessel network as main blood vessel data; When the blood vessel diameter size data is smaller than the preset blood vessel diameter size threshold, the corresponding blood vessel in the blood vessel area network is marked as branch vessel data; the initial blood vessel area image is analyzed for blood vessel tree nodes using the trunk vessel data and branch vessel data to generate respiratory tract blood vessel tree data.

6. The respiratory tract image segmentation method according to claim 1, characterized in that: Step S3 includes the following steps: Step S31: Acquire user posture data and user breathing rate data; Step S32: performing posture change analysis on the user posture data to generate user posture change data; performing breathing intensity analysis on the user breathing frequency data to generate user breathing intensity data; Step S33: performing a user motion impact assessment based on the user posture change data and the user breathing intensity data to generate user motion impact assessment data; performing a CT image part disturbance analysis on the user motion impact assessment data to obtain user motion part disturbance data; Step S34: performing motion displacement region segmentation on the lung airway structure segmentation image and the lung vascular structure segmentation image based on the user motion part disturbance data to generate a respiratory tract associated artifact part image; performing image artifact optimization on the respiratory tract segmentation artifact part image to generate a respiratory tract part segmentation artifact optimized image.

7. The respiratory tract image segmentation method according to claim 6, characterized in that: Step S34 includes the following steps: Step S341: extracting a time series image sequence from the pulmonary airway structure segmentation image and the pulmonary vascular structure segmentation image to obtain CT image time series data; performing respiratory phase division on the CT image time series data to generate respiratory phase division data; Step S342: performing lung motion vector field analysis on the lung airway structure segmentation image and the lung vascular structure segmentation image using the respiratory phase division data to generate a lung motion vector field; performing motion pixel matching on the lung airway structure segmentation image and the lung vascular structure segmentation image using the lung motion vector field to generate motion matching pixels; Step S343: performing motion displacement region segmentation on the lung airway structure segmentation image and the lung blood vessel structure segmentation image according to the motion matching pixels to generate a respiratory tract associated artifact portion image; Step S344: performing respiratory motion region masking correction on the respiratory tract segmentation artifact part image based on the lung motion vector field to generate a respiratory tract segmentation artifact optimized image.

8. The respiratory tract image segmentation method according to claim 1, characterized in that: Step S4 includes the following steps: Step S41: performing abnormality detection on the respiratory tract segmentation artifact optimized image to generate lesion area positioning data; performing lesion area segmentation on the respiratory tract segmentation artifact optimized image based on the lesion area positioning data to generate a respiratory tract lesion area segmentation image; Step S42: performing three-dimensional coordinate mapping on the pulmonary airway structure segmentation image, the pulmonary vascular structure segmentation image, and the respiratory tract lesion area segmentation image to generate three-dimensional point cloud data of the segmentation structure; Step S43: annotating and quantifying the segmented structure three-dimensional point cloud data to generate respiratory tract segmented structure quantitative data; reconstructing the segmented structure three-dimensional point cloud data using the respiratory tract segmented structure quantitative data to generate a respiratory tract structure segmented three-dimensional image.

9. The respiratory tract image segmentation method according to claim 8, characterized in that: Step S41 includes the following steps: Step S411: performing pixel brightness detection on the respiratory tract segmentation artifact optimized image to obtain respiratory tract pixel brightness data; performing pixel clustering on the respiratory tract pixel brightness data to generate respiratory tract pixel brightness cluster data; Step S412: confirming the range of suspicious lesion points in the respiratory tract segmentation artifact optimized image based on the respiratory tract pixel brightness cluster data, and generating respiratory tract pixel suspicious lesion point range data; Step S413: Performing a connected domain analysis on the range data of suspicious lesion points in the respiratory tract pixels to generate respiratory tract abnormal point range data; performing a sliding window scan on the respiratory tract segmentation artifact optimized image based on the respiratory tract abnormal point range data to generate lesion area positioning data.

10. A respiratory tract image-based segmentation system, characterized in that: For executing the respiratory tract image-based segmentation method according to claim 1, the respiratory tract image-based segmentation system comprises: The airway structure segmentation module is used to obtain an initial respiratory image; extract lung airway image features from the initial respiratory image to obtain lung airway feature data; and segment the respiratory image into lung airway structure regions based on the lung airway feature data to generate a lung airway structure segmentation image. A vascular structure segmentation module is used to construct an airway exclusion mask based on the lung airway structure segmentation image to obtain the airway exclusion mask; perform binary division of the respiratory image using the airway exclusion mask to generate an initial vascular region image; perform vascular tree analysis on the initial vascular region image to generate respiratory vascular tree data; perform vascular structure region image segmentation on the initial vascular region image using the respiratory vascular tree data to generate a lung vascular structure segmentation image; An artifact optimization module is used to obtain user posture data and user respiratory rate data; perform user motion impact assessment based on the user posture data and user respiratory rate data to generate user motion impact assessment data; perform image artifact optimization on the lung airway structure segmentation image and the lung vascular structure segmentation image using the user motion impact assessment data to generate a respiratory segmentation artifact optimized image; The three-dimensional reconstruction module is used to detect abnormalities in the respiratory tract segmentation artifact optimized image and generate lesion area positioning data; perform lesion area segmentation on the respiratory tract segmentation artifact optimized image based on the lesion area positioning data to generate a respiratory tract lesion area segmentation image; perform three-dimensional image reconstruction on the lung airway structure segmentation image, the lung vascular structure segmentation image and the respiratory tract lesion area segmentation image to generate a respiratory tract structure segmentation three-dimensional image.

Images