Lung lesion identification method and device, and electronic device
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- NEUSOFT MEDICAL SYST CO LTD
- Filing Date
- 2026-02-26
- Publication Date
- 2026-06-23
AI Technical Summary
Existing technologies are insufficient for comprehensively and accurately identifying chronic obstructive pulmonary disease (COPD) lesions, especially in distinguishing between pulmonary arteries and pulmonary veins and in integrating multi-dimensional lesion information. Furthermore, their application costs are high, making it difficult to promote their use in primary healthcare institutions.
By accurately identifying pulmonary artery and pulmonary vein regions in CT scan images, quantifying vascular features, and fusing them with multidimensional lung parenchymal structural features, a pre-trained classifier is used for lesion assessment, simplifying data requirements.
It enables accurate identification of COPD lesions, reduces application costs, is suitable for promotion in primary healthcare institutions, and improves the comprehensiveness and accuracy of lesion assessment.
Smart Images
Figure CN122265702A_ABST
Abstract
Description
Technical Field
[0001] This disclosure relates to the technical field of medical imaging, such as a method, apparatus, and electronic device for identifying lung lesions. Background Technology
[0002] Lung diseases seriously affect people's health. Early and accurate identification and disease assessment are key to improving the prognosis of patients with lung lesions and reducing the disease burden. Imaging examinations, as non-invasive diagnostic tools, play an important role in the clinical assessment of lung lesions.
[0003] Taking chronic obstructive pulmonary disease (COPD) as an example, this disease is a respiratory disease characterized by chronic inflammatory lung lesions of the lung parenchyma, airways, and blood vessels. Computed tomography (CT) of the lungs, with its high-resolution imaging advantage, can clearly present typical pathological features of COPD, such as emphysema and airway structural changes. Quantitative analysis of these features can effectively assess the severity of the disease, providing objective evidence for diagnosis. With the continuous development of quantitative CT technology, the assessment of pulmonary vascular-related indicators has gradually become an important supplement to the assessment of COPD. Among these, quantitative characteristics such as the percentage of cross-sectional area of small pulmonary vessels (CSA%), vascular density, and distribution have been proven to be closely related to the pathogenesis and progression of COPD, providing new reference dimensions for disease diagnosis and classification.
[0004] Currently, several technologies have been developed to construct COPD (Chronic Obstructive Pulmonary Disease) identification models based on CT quantitative features. For example, some studies integrate quantitative parameters of emphysema, airways, and pulmonary vessels, and use machine learning algorithms such as Support Vector Machine (SVM) to develop diagnostic and grading models. Other studies combine demographic data, the percentage of pulmonary small vessel cross-sectional area, and tracheal quantitative parameters to identify individuals with pre-COPD conditions using deep learning models. In addition, technologies integrate clinical examinations, quantitative CT, and radiomics features, or construct airway measurement models based on CT, using machine learning algorithms to distinguish between healthy individuals and COPD patients.
[0005] However, these technologies still have significant limitations, making it difficult to meet the clinical need for accurate and efficient identification of COPD: First, some technologies focus only on single-dimensional quantitative parameters (such as focusing only on tracheal characteristics), failing to comprehensively cover key pathological changes related to emphysema, pulmonary vessels, etc., resulting in an incomplete assessment of COPD and a tendency to miss early, mild lesions; Second, although these technologies attempt to integrate multiple indicators, they often introduce complex data such as radiomics characteristics and clinical examination results. The standardized extraction and quality control of this data require strict operation by professionals, which not only significantly increases the application cost of the model but also reduces its feasibility for promotion in primary healthcare institutions; Third, the assessment of pulmonary vessels by these technologies is mostly at the overall level, lacking a clear distinction between pulmonary arteries and pulmonary veins and targeted quantitative analysis, failing to fully explore the potential correlation between changes in arterial and venous vascular structures and COPD, thus limiting the accuracy of COPD identification.
[0006] Therefore, how to construct a COPD lesion identification scheme that can comprehensively and accurately quantify the characteristics of pulmonary arteriovenous vessels and efficiently integrate multi-dimensional lesion information, and solve the problems of incomplete assessment, high application cost, and insufficient accuracy of existing technologies, has become a technical problem that urgently needs to be solved in this field.
[0007] It should be noted that the information disclosed in the background section above is only used to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention
[0008] To provide a basic understanding of some aspects of the disclosed embodiments, a brief summary is given below. This summary is not intended as a general commentary, nor is it intended to identify key / important components or describe the scope of protection of these embodiments, but rather as a prelude to the detailed description that follows.
[0009] This disclosure provides a method, device, and electronic device for identifying lung lesions. By clearly distinguishing pulmonary arteries and veins and quantifying their characteristics, and integrating multi-dimensional lesion indicators of emphysema and airways, the comprehensiveness and accuracy of lung lesion identification are improved. At the same time, no complex data support is required, which simplifies the process and reduces application costs.
[0010] According to a first aspect of this disclosure, a method for identifying lung lesions is provided, comprising: Acquire scan images of the patient's lungs and identify the pulmonary artery and pulmonary vein regions within the scan images; Obtain the size parameters of each blood vessel in the pulmonary artery and pulmonary vein regions, and calculate the vascular characteristics of each vascular region based on the size parameters of each blood vessel in the pulmonary artery and pulmonary vein regions. The vascular features of each vascular region are fused with the pre-acquired lung parenchymal structural features of the patient's lungs to obtain fused features; The fused features are input into a pre-trained classifier, which is then used to identify the patient's lung lesion assessment results.
[0011] In some embodiments, identifying the pulmonary artery region and pulmonary vein region of the lung in the scanned image includes: The pulmonary artery and pulmonary vein regions of the lungs are identified in the scanned images, and the pulmonary artery and pulmonary vein regions are merged to obtain a preliminary vascular merging region. Morphological closure operations are used to smooth the preliminary vascular merging region, eliminating boundary blurring and hollow pixels, and obtaining the target vascular merging region.
[0012] In some embodiments, obtaining the size parameters of each vessel in the pulmonary artery and pulmonary vein regions includes: Identify each blood vessel based on the bifurcation structure of the blood vessels in the pulmonary artery and pulmonary vein regions; For each blood vessel, multiple measurement points are selected in the blood vessel according to the preset sampling step size; The diameter and cross-sectional area of each blood vessel were obtained at various measurement points.
[0013] In some embodiments, vascular features include individual vessel features of each vessel; vascular features of each vascular region are calculated based on the size parameters of each vessel in the pulmonary artery and pulmonary vein regions, including at least one of the following: For each blood vessel, the maximum blood vessel diameter, minimum blood vessel diameter, and average blood vessel diameter are determined based on the blood vessel diameter at each measurement point. For each blood vessel, the average cross-sectional area of the blood vessel is determined based on the cross-sectional area of the blood vessel at each measurement point. For each blood vessel, the length of the blood vessel is determined based on the sampling step size and the number of measurement points. For each blood vessel, the blood vessel volume is determined based on the sampling step length, the number of measurement points, and the cross-sectional area of the blood vessel at each measurement point.
[0014] In some embodiments, vascular features include pulmonary microvascular features; calculating the vascular features of each vascular region based on the size parameters of each vessel in the pulmonary artery and pulmonary vein regions includes: Based on the size parameters of each vessel in the pulmonary artery and pulmonary vein regions, different subclasses of pulmonary microvessels are defined in each vascular region; For each subclass of pulmonary microvessels in each vascular region, the pulmonary microvessel characteristics of each subclass are calculated based on the size parameters of each pulmonary microvessel in the same subclass.
[0015] In some embodiments, the lung lesion identification method further includes: calculating the minimum straight-line distance between the blood vessel and the pleura for each blood vessel; The vascular characteristics of each vascular region are calculated based on the size parameters of each vessel in the pulmonary artery and pulmonary vein regions. The calculation also includes defining a subclass of pulmonary microvessels in each vascular region based on the minimum straight-line distance between each vessel in the pulmonary artery and pulmonary vein regions and the pleura.
[0016] In some embodiments, pulmonary microvascular characteristics include pulmonary microvascular cross-sectional area percentage, pulmonary microvascular volume, and pulmonary microvascular volume percentage.
[0017] In some embodiments, the vascular features of each vascular region are fused with pre-acquired lung parenchymal structural features of the patient's lungs to obtain fused features, including: The vascular features of each vascular region, as well as the pre-acquired lung parenchymal structural features of the patient's lungs, are used as candidate features. Calculate the similarity between each candidate feature, and determine the importance of each candidate feature based on the similarity between them; Target features are selected based on the importance of each candidate feature; The selected target features are fused to obtain the fused features.
[0018] According to a second aspect of this disclosure, a lung lesion identification device is provided, comprising: The image preprocessing module is configured to: acquire scan images of the patient's lungs and identify the pulmonary artery and pulmonary vein regions in the scan images; The vascular feature acquisition module is configured to: acquire the size parameters of each blood vessel in the pulmonary artery region and the pulmonary vein region, and calculate the vascular features of each vascular region based on the size parameters of each blood vessel in the pulmonary artery region and the pulmonary vein region; The feature fusion module is configured to fuse the vascular features of each vascular region with the pre-acquired lung parenchymal structural features of the patient's lungs to obtain fused features; The lung lesion identification module is configured to input fused features into a pre-trained classifier and use the classifier to identify the patient's lung lesion assessment results.
[0019] According to a third aspect of this disclosure, an electronic device is provided, including a processor and a memory storing program instructions, the processor being configured to execute the lung lesion identification method provided in the first aspect of this disclosure when the program instructions are executed.
[0020] The lung lesion identification method, apparatus, and electronic device provided in this disclosure can achieve the following technical effects: First, the pulmonary artery and pulmonary vein regions are accurately identified in the patient's lung scan images, achieving a clear distinction between pulmonary arteries and veins. Based on this, the size parameters of each vessel are acquired, and the vascular features of the corresponding region are calculated. This fully explores the potential correlation between changes in arteriovenous vascular structure and lung lesions, improving the accuracy of lesion identification. By fusing arteriovenous vascular features with pre-acquired lung parenchymal structural features in a multi-dimensional manner, key pathological changes related to lung lesions are comprehensively covered, ensuring a comprehensive lesion assessment and avoiding the omission of early, minor lesions. Simultaneously, this method focuses on vascular, emphysema, and airway-related features derived from lung scan images, eliminating the need for complex data such as radiomics features and clinical examination results. This simplifies the feature acquisition process, reduces reliance on professional quality control, and thus lowers the model application cost, contributing to its feasibility for promotion in primary healthcare institutions. Finally, a pre-trained classifier outputs lung lesion assessment results, achieving efficient fusion of multi-dimensional lesion information, thereby obtaining accurate and reliable lung lesion assessment results.
[0021] The above general description and the description below are exemplary and illustrative only and are not intended to limit this disclosure. Attached Figure Description
[0022] One or more embodiments are illustrated by way of example with reference to the accompanying drawings. These illustrations and drawings do not constitute a limitation on the embodiments. Elements having the same reference numerals in the drawings are shown as similar elements. The drawings are not to be scaled. And wherein: Figure 1 This is a schematic flowchart of a lung lesion identification method provided in an embodiment of this disclosure; Figure 2 This is a schematic flowchart of another lung lesion identification method provided in this embodiment of the disclosure; Figure 3 This is a schematic flowchart of another lung lesion identification method provided in this embodiment of the disclosure; Figure 4 This is a schematic flowchart of another lung lesion identification method provided in this embodiment of the disclosure; Figure 5 This is a schematic diagram of a lung lesion identification device provided in an embodiment of this disclosure; Figure 6 This is a schematic diagram of an electronic device provided in an embodiment of this disclosure. Detailed Implementation
[0023] To provide a more detailed understanding of the features and technical content of the embodiments of this disclosure, the implementation of the embodiments of this disclosure will be described in detail below with reference to the accompanying drawings. The accompanying drawings are for illustrative purposes only and are not intended to limit the embodiments of this disclosure. In the following technical description, for ease of explanation, several details are used to provide a full understanding of the disclosed embodiments. However, one or more embodiments may still be implemented without these details. In other cases, well-known structures and devices may be simplified in their depiction to simplify the drawings.
[0024] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this disclosure are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate for the embodiments of this disclosure described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion.
[0025] Unless otherwise stated, the term "multiple" means two or more.
[0026] In this embodiment of the disclosure, the character " / " indicates that the objects before and after it are in an "or" relationship. For example, A / B means: A or B.
[0027] The term "and / or" describes an association between objects, indicating that three relationships can exist. For example, A and / or B means: A or B, or A and B.
[0028] The term "correspondence" can refer to an association or binding relationship. The correspondence between A and B means that there is an association or binding relationship between A and B.
[0029] This disclosure provides an electronic device, which can be a computer, terminal, server, or other device with computing capabilities. The electronic device is equipped with a pre-trained classifier and can identify the patient's lung lesion assessment results based on the classifier. The lesion assessment results are used to characterize whether the patient has lung diseases such as COPD, asthma, bronchiectasis, pneumonia, or pulmonary hypertension.
[0030] In conjunction with the electronic device (hereinafter referred to as the device) provided in the embodiments of this disclosure, the embodiments of this disclosure provide a method for identifying lung lesions, such as... Figure 1 As shown, methods for identifying lung lesions include: S101, the device acquires scan images of the patient's lungs and identifies the pulmonary artery and pulmonary vein regions in the scan images.
[0031] In this embodiment, the scanned image of the patient's lungs can be a CT scan image obtained by scanning the patient's lungs with a CT scanner. Medical image segmentation algorithms (such as threshold-based segmentation, region growing, deep learning segmentation models, etc.) are used to identify the pulmonary artery and pulmonary vein regions from the scanned image, taking into account the differences in imaging features between the pulmonary artery and pulmonary vein. During the identification process, the parameter settings of the segmentation algorithm can be optimized by incorporating clinical anatomical rules to improve the accuracy of region identification and avoid overlap or omission of arterial and venous regions.
[0032] S102, the device acquires the size parameters of each blood vessel in the pulmonary artery region and the pulmonary vein region, and calculates the vascular characteristics of each blood vessel region based on the size parameters of each blood vessel in the pulmonary artery region and the pulmonary vein region.
[0033] In this embodiment, the vascular size parameters may include parameters related to diameter, cross-sectional area, and length. For each vessel in the pulmonary artery and pulmonary vein regions, the vascular characteristics of each vascular region are calculated by statistically analyzing the vascular size parameters. By obtaining the vascular size parameters of each vessel in the pulmonary artery and pulmonary vein regions, accurate and comprehensive basic morphological data can be provided for vascular feature calculation, supporting statistical analysis of single vessel morphological characteristics and regional vascular summary characteristics, resulting in more accurate vascular features and achieving the quantification of pulmonary arterial and venous morphology.
[0034] S103, the device fuses the vascular features of each vascular region with the pre-acquired lung parenchymal structural features of the patient's lungs to obtain fused features.
[0035] In this embodiment of the disclosure, the structural features of lung parenchyma include at least one of the following: emphysema features, airway features, pulmonary nodule features, and pulmonary mass features. Emphysema features are quantitative characteristics characterizing emphysema lesions in the lungs, including the average CT value, lowest CT value, volume of the emphysema region, percentage of the total lung volume occupied by the emphysema region, and lobar distribution ratio of the emphysema region. Airway features are quantitative characteristics characterizing the structural morphology and physiological state of the lung airways, including airway diameter, airway wall thickness, airway cross-sectional area, airway bifurcation angle, airway length, and airway patency parameters. Basic lung parenchyma features are quantitative characteristics characterizing the basic pathological changes of lung parenchyma tissue, excluding emphysema features, including the volume and proportion of the pulmonary consolidation region. Pulmonary nodule and mass features are quantitative characteristics characterizing pulmonary nodules and masses, including quantitative parameters such as the long diameter, short diameter, volume, and morphology of the nodules or masses.
[0036] In this embodiment, lung parenchymal structural features can be extracted from scanned images of a patient's lungs using mature algorithms. Taking CT scans as an example, candidate regions for emphysema are initially marked using a preset threshold screening algorithm. The segmentation results are optimized through region growing and morphological opening operations. Finally, quantitative features such as the average CT value, lowest CT value, volume, percentage of total lung volume, and lobe distribution ratio of the emphysema region are calculated. The airway signal is enhanced using a preset enhancement algorithm. The ATN (Adaptive Threshold Navigation) algorithm iteratively segments the bronchi at each level from the trachea's origin along the airway. After correcting the segmentation results through morphological closing operations, the airway is digitally labeled according to its bifurcation structure. Measurement points are selected for each airway at preset intervals, and then the diameter, airway wall thickness, cross-sectional area, bifurcation angle, length, patency parameters, and other related quantitative features of each airway are calculated.
[0037] Here, by integrating arterial and venous vascular features with pre-acquired lung parenchymal structural features in a multi-dimensional manner, key pathological changes related to lung lesions are comprehensively covered.
[0038] S104, the device inputs the fused features into a pre-trained classifier, which is then used to identify the patient's lung lesion assessment results.
[0039] In the embodiments of this disclosure, the classifier may be a zero-inflated regression model; the classifier may be a machine learning classifier, such as a support vector machine (SVM) and a random forest; the classifier may be a deep learning classifier, such as a lightweight convolutional neural network (CNN) or a multilayer perceptron (MLP).
[0040] The classifier training follows a standardized process of "data preparation - model training - optimization and validation" to ensure reliable training results. The specific steps are as follows: Collect a dataset of clinically labeled lung CT images, including CT images of patients with lung lesions and healthy individuals; perform binary classification labeling on the samples according to clinical diagnostic criteria (such as pulmonary function test results and symptom presentation) (labeled as "disease" or "healthy"), and confirm inconsistent labels through expert consultation. Standardize the aforementioned fusion features (normalize to the [0,1] interval or standardize to a distribution with a mean of 0 and a variance of 1) to eliminate differences in feature dimensions; randomly partition the dataset into training, validation, and test sets in a 7:2:1 ratio; initialize machine learning classifiers with default parameters (e.g., C=1.0 for SVM, number of trees in random forest=200); initialize deep learning classifiers with He or Xavier parameters to avoid gradient vanishing. The binary classification task employs the cross-entropy loss function, with the optimizer selected as either Adam (learning rate 0.0001~0.001) or SGD (learning rate 0.001~0.01, momentum 0.9). Model parameters are updated via backpropagation. Grid search or Bayesian optimization methods are used to optimize key hyperparameters (such as the kernel function and C value of SVM, the number of trees in random forest, and the learning rate and number of hidden layer neurons in deep learning), with validation set accuracy as the core evaluation metric. The test set is used to evaluate model performance, with core metrics including accuracy, sensitivity, specificity, and AUC (referring to existing technical standards, AUC must be ≥0.85). Confusion matrix analysis is used to analyze the causes of classification errors. If the missed diagnosis rate for a specific class is too high, oversampling (for minority class samples) or a weighted loss function can be used for adjustment. The model with the best performance on the validation set is selected as the final deployment model, and the model parameter files (e.g., .pth, .pkl format) are saved for rapid retrieval in subsequent clinical scenarios.
[0041] The lung lesion identification method provided in this disclosure first accurately identifies the pulmonary artery and pulmonary vein regions in the patient's lung scan image, achieving a clear distinction between pulmonary arteries and veins. Based on this, the size parameters of each vessel are obtained and the vascular features of the corresponding region are calculated, fully exploring the potential correlation between changes in arteriovenous vascular structure and lung lesions, thus improving the accuracy of lesion identification. By fusing arteriovenous vascular features with pre-acquired lung parenchymal structural features in a multi-dimensional manner, the method comprehensively covers key pathological changes related to lung lesions, ensuring the comprehensiveness of lesion assessment and avoiding the omission of early, minor lesions. Simultaneously, this method focuses on vascular, emphysema, and airway-related features derived from lung scan images, eliminating the need to introduce complex data such as radiomics features and clinical examination results. This simplifies the feature acquisition process, reduces reliance on professional quality control, and thus lowers the model application cost, contributing to its feasibility for promotion in primary healthcare institutions. Finally, a pre-trained classifier outputs the lung lesion assessment results, achieving efficient fusion of multi-dimensional lesion information, thereby obtaining accurate and reliable lung lesion assessment results.
[0042] In this embodiment of the disclosure, identifying the pulmonary artery region and pulmonary vein region of the lung in the scanned image includes: identifying the pulmonary artery region and pulmonary vein region of the lung in the scanned image, merging the pulmonary artery region and pulmonary vein region to obtain a preliminary vascular merging region; and using morphological closure operation to smooth the preliminary vascular merging region, eliminating boundary blurring and hollow pixels, to obtain the target vascular merging region.
[0043] First, an initial vascular merging region is formed by merging arterial and venous areas, fully covering the pulmonary vascular system and avoiding feature omissions caused by vascular segmentation assessment. Then, morphological closure operation, through the logic of "expansion before erosion", effectively fills the tiny voids in the initial merging region, eliminates boundary blurring and spiculated pixels, and preserves the integrity of the main structure and branches of the blood vessels while improving the edge regularity of the vascular region. This reduces errors in subsequent vascular size parameter measurement and feature calculation, ensuring the accuracy and reliability of vascular feature quantification results. It provides high-quality vascular region data support for multi-dimensional feature fusion and final accurate identification of lung lesions.
[0044] This disclosure provides another method for identifying lung lesions, such as... Figure 2 As shown, methods for identifying lung lesions include: S201, the device acquires scan images of the patient's lungs and identifies the pulmonary artery and pulmonary vein regions in the scan images.
[0045] S202, the device merges the pulmonary artery region and the pulmonary vein region to obtain a preliminary vascular merging region.
[0046] S203, the device uses morphological closure operation to smooth the preliminary vascular merging area, eliminate boundary blurring and hollow pixels, and obtain the target vascular merging area.
[0047] Because the boundaries between pulmonary arteries and pulmonary veins in the pulmonary vascular system are often blurred, partially overlapping, or excessively separated, directly using them for subsequent parameter measurements can easily lead to errors. Therefore, it is necessary to first merge the segmented pulmonary artery and pulmonary vein regions to obtain a preliminary vascular merged region containing the complete pulmonary vascular structure. To address potential issues such as blurred boundaries and tiny void pixels in the preliminary vascular merged region, a morphological closure operation is further employed for smoothing optimization. This operation follows the logic of "dilation before erosion," that is, filling voids and connecting discrete small vascular fragments within the region through a dilation operation (⊕), and then performing an erosion operation (⊕). The blood vessel boundaries are contracted to their original contours to eliminate boundary redundancy caused by expansion. Ultimately, through these steps, a target vascular merging region with clear boundaries, intact structure, and no void noise is obtained.
[0048] In this embodiment of the disclosure, the mathematical expression for the target vascular merging region is as follows: Region_vessel=((Region_artery∪Region_vein)⊕B) B; In the above mathematical expression, Region_vessel represents the target vascular merging region, Region_artery represents the pulmonary artery region, Region_vein represents the pulmonary vein region, and ⊕ indicates the expansion operation. B represents the erosion operation, and B represents the structural element.
[0049] S204, the device acquires the size parameters of each blood vessel in the pulmonary artery region and the pulmonary vein region, and calculates the vascular characteristics of each blood vessel region based on the size parameters of each blood vessel in the pulmonary artery region and the pulmonary vein region.
[0050] S205, the device fuses the vascular features of each vascular region with the pre-acquired lung parenchymal structural features of the patient's lungs to obtain fused features.
[0051] S206, the device inputs the fused features into a pre-trained classifier, which is then used to identify the patient's lung lesion assessment results.
[0052] In some embodiments, obtaining the size parameters of each blood vessel in the pulmonary artery region and the pulmonary vein region includes: identifying each blood vessel based on the bifurcation structure of the blood vessels in the pulmonary artery region and the pulmonary vein region; selecting multiple measurement points in the blood vessel for each blood vessel according to a preset sampling step size; and obtaining the blood vessel diameter and cross-sectional area at each measurement point of each blood vessel.
[0053] Based on the natural vascular bifurcation anatomy within the pulmonary artery and pulmonary vein regions, and using the starting and ending points of these bifurcations as the defining criteria, each vessel within these regions is segmented and independently identified. A unique morphological identifier is established for each vessel, clarifying the definition criteria for a single vessel and ensuring that each vessel is independently traceable, with dimensional parameter acquisition proceeding independently without interference and with clear boundaries, avoiding confusion of parameters from different vessels. For each independently identified vessel, multiple measurement points are evenly selected along the vessel's course at a preset equidistant sampling step size (e.g., 1 mm). The sampling step size can be flexibly adjusted according to the resolution of the scanned image and the vessel diameter, ensuring that the measurement points comprehensively cover the entire proximal, mid-, and distal segments of the vessel, guaranteeing the completeness and representativeness of the dimensional parameter acquisition and avoiding parameter deviations caused by local sampling. At each preset measurement point of each vessel, medical image quantification analysis algorithms are used to extract vessel diameter-related data and cross-sectional contour data at that location, accurately obtaining the vessel diameter and cross-sectional area corresponding to each measurement point, forming a dataset of full-segment dimensional parameters for each vessel, providing data support for subsequent calculations of vessel features.
[0054] This disclosure provides another method for identifying lung lesions, such as... Figure 3 As shown, methods for identifying lung lesions include: S301, the device acquires scan images of the patient's lungs and identifies the pulmonary artery and pulmonary vein regions in the scan images.
[0055] S302, the device identifies each blood vessel based on the bifurcation structure of the blood vessels in the pulmonary artery and pulmonary vein regions.
[0056] For the pulmonary artery and pulmonary vein regions, each vessel is numerically labeled based on its bifurcation structure. Specifically, each vessel has a unique numerical label corresponding to the segment from the start of the bifurcation to the end of the next bifurcation. For example, the nth vessel in the pulmonary artery region corresponds to the unique numerical label 'n'.
[0057] The S303 device selects multiple measurement points in each blood vessel according to a preset sampling step size.
[0058] For a blood vessel n, multiple measurement points are selected within the vessel according to a preset sampling step size. Sampling step size. The value can be determined according to the actual design requirements. Taking 1mm as an example, it can be selected from blood vessel n. There are 10 measurement points.
[0059] S304, the device obtains the vessel diameter and cross-sectional area at each measurement point in each vessel.
[0060] S305, the device calculates the vascular characteristics of each vascular region based on the size parameters of each blood vessel in the pulmonary artery region and the pulmonary vein region.
[0061] S306, the device fuses the vascular features of each vascular region with the pre-acquired lung parenchymal structural features of the patient's lungs to obtain fused features.
[0062] S307: The device inputs the fused features into a pre-trained classifier, which is then used to identify the patient's lung lesion assessment results.
[0063] In this embodiment of the disclosure, the vascular features include the individual vascular features of each vascular vessel, and the individual vascular features include at least one of the following: maximum vascular diameter, minimum vascular diameter, average vascular diameter, average vascular cross-sectional area, vascular length, and vascular volume.
[0064] In this embodiment of the disclosure, the vascular characteristics of each vascular region are calculated based on the size parameters of each blood vessel in the pulmonary artery region and the pulmonary vein region, including at least one of the following: for each blood vessel, the maximum blood vessel diameter, minimum blood vessel diameter, and average blood vessel diameter are determined based on the blood vessel diameter at each measurement point; for each blood vessel, the average blood vessel cross-sectional area is determined based on the blood vessel cross-sectional area at each measurement point; for each blood vessel, the blood vessel length is determined based on the blood vessel sampling step size and the number of measurement points; for each blood vessel, the blood vessel volume is determined based on the blood vessel sampling step size, the number of measurement points, and the blood vessel cross-sectional area at each measurement point.
[0065] Taking the pulmonary artery region's vessel n as an example, the calculation formulas for various single vessel characteristics are as follows: Maximum blood vessel diameter: , It is the nth in the blood vessels The diameter of the blood vessel at each measurement point. This represents the number of measurement points in blood vessel n. Minimum blood vessel diameter: , It is the nth in the blood vessels The diameter of the blood vessel at each measurement point. This represents the number of measurement points in blood vessel n. Average vessel diameter: , It is the nth in the blood vessels The diameter of the blood vessel at each measurement point. This represents the number of measurement points in blood vessel n. Cross-sectional area of blood vessels: , It is the nth in the blood vessels The cross-sectional area of the blood vessel at each measurement point This represents the number of measurement points in blood vessel n. Blood vessel length: , It is the sampling step size. This represents the number of measurement points in blood vessel n. Blood vessel volume: , It is the nth in the blood vessels The cross-sectional area of the blood vessel at each measurement point It is the sampling step size. It represents the number of measurement points in blood vessel n.
[0066] In some embodiments, vascular features include pulmonary microvascular features, which include at least one of pulmonary microvascular cross-sectional area percentage, pulmonary microvascular volume, and pulmonary microvascular volume percentage.
[0067] In some embodiments, calculating the vascular characteristics of each vascular region based on the size parameters of each vessel in the pulmonary artery region and the pulmonary vein region includes: defining different subclasses of pulmonary microvessels in each vascular region based on the size parameters of each vessel in the pulmonary artery region and the pulmonary vein region; and for each subclass of pulmonary microvessels in each vascular region, calculating the pulmonary microvessel characteristics of each subclass of pulmonary microvessels based on the size parameters of each pulmonary microvessel in the same subclass.
[0068] In some embodiments, the lung lesion identification method further includes: calculating the minimum straight-line distance between the blood vessel and the pleura for each blood vessel. Calculating the vascular characteristics of each vascular region based on the size parameters of each blood vessel in the pulmonary artery and pulmonary vein regions further includes: defining a subclass of pulmonary small vessels in each vascular region based on the minimum straight-line distance between each blood vessel in the pulmonary artery and pulmonary vein regions and the pleura.
[0069] In this embodiment, based on the size parameters of each vessel in the pulmonary artery and pulmonary vein regions, a multi-dimensional threshold rule is used to accurately define different subclasses of small pulmonary vessels in each vessel region, ensuring the scientific rigor and flexibility of the definition criteria. The definition dimensions can be used individually or in combination: in the vessel diameter dimension, vessels with diameters below preset thresholds (Td1), such as 1.5mm or 2mm, can be classified into a specific subclass of small pulmonary vessels; in the vessel cross-sectional area dimension, multiple threshold levels can be set (e.g., T1=5mm², T2=10mm²), defining vessels with cross-sectional areas below T1 or between T1 and T2 into different subclasses; in the pleural distance dimension, vessels with a minimum straight-line distance to the pleura ≤10mm or 15mm (Tp1) can be grouped into one subclass; a comprehensive definition rule can also be used, i.e., simultaneously satisfying the requirement that the vessel cross-sectional area T1 ≤ area i ≤T2, vessel diameter iVessels with a distance ≤Td1 and a pleural distance ≤Tp1 are defined as a certain subclass, and the thresholds can be flexibly adjusted according to clinical needs, scanning equipment accuracy, or lung lobe / vessel location.
[0070] Subsequently, for each subclass of pulmonary microvessels defined in each vascular region, multidimensional pulmonary microvessel characteristics are calculated based on the size parameters of each pulmonary microvessel within the same subclass.
[0071] The volume of pulmonary microvessels in each subclass was calculated using the following formula: ; ; ; .
[0072] In the above formula, It refers to the volume of a pulmonary microvessel within a subclass of pulmonary microvessels with a cross-sectional area smaller than T1. It refers to the volume of a pulmonary microvessel within a subclass of pulmonary microvessels whose cross-sectional area falls between T1 and T2. It refers to the volume of a pulmonary microvessel within a subclass of pulmonary microvessels with a diameter smaller than Td1. It is the volume of a pulmonary microvessel in a subclass of pulmonary microvessels located less than Tp1 from the pleura.
[0073] Taking the pulmonary artery region as an example, the percentage of pulmonary microvessel volume is the proportion of pulmonary microvessel volume in the pulmonary artery blood volume. The pulmonary microvessel volume of each subtype of pulmonary microvessels is calculated using the following formula: ; ; ; .
[0074] In the above formula, It is the pulmonary artery blood volume. It refers to the volume of a pulmonary microvessel within a subclass of pulmonary microvessels with a cross-sectional area smaller than T1. It is the percentage of pulmonary small vessel volume in a subclass of pulmonary small vessels with a cross-sectional area smaller than T1. It refers to the volume of a pulmonary microvessel within a subclass of pulmonary microvessels whose cross-sectional area falls between T1 and T2. It is the percentage of pulmonary small vessel volume in a subclass of pulmonary small vessels with a cross-sectional area between T1 and T2. It refers to the volume of a pulmonary microvessel within a subclass of pulmonary microvessels with a diameter smaller than Td1. It is the percentage of pulmonary small vessel volume in a subclass of pulmonary small vessels with a vessel diameter smaller than Td1. It refers to the volume of a pulmonary microvessel within a subclass of pulmonary microvessels located less than Tp1 from the pleura. It is the percentage of pulmonary small vessel volume in a subclass of pulmonary small vessels that are less than Tp1 from the pleura.
[0075] In some embodiments, the vascular features of each vascular region are fused with the pre-acquired lung parenchymal structural features of the patient's lungs to obtain fused features, including: using the vascular features of each vascular region and the pre-acquired lung parenchymal structural features of the patient's lungs as candidate features; calculating the similarity between each candidate feature, determining the importance of each candidate feature based on the similarity between each candidate feature; selecting target features according to the importance of each candidate feature; and fusing the selected target features to obtain fused features.
[0076] By calculating the similarity between candidate features, the system accurately distinguishes between "low similarity" and "high similarity" feature combinations. For low similarity scenarios, candidate features are directly fused to retain complete information. For high similarity scenarios, the importance of candidate features to the classification task is further quantified, enabling the precise removal of redundant features with "high similarity but low contribution (importance)," thus avoiding information waste and reducing redundant interference. Furthermore, by selecting target features based on importance, the system retains key quantitative details of single-dimensional lesions (such as the volume of small blood vessels in specific subclasses of lungs and the density features of local emphysema) and integrates the synergistic relationships between different lesion dimensions (such as the correlation between the cross-sectional area of blood vessels and the emphysema of the corresponding lung lobe), significantly improving the information density and representativeness of the fused features.
[0077] This disclosure provides another method for identifying lung lesions, such as... Figure 4 As shown, methods for identifying lung lesions include: S401, the device acquires scan images of the patient's lungs and identifies the pulmonary artery and pulmonary vein regions in the scan images.
[0078] S402, the device acquires the size parameters of each blood vessel in the pulmonary artery region and the pulmonary vein region, and calculates the vascular characteristics of each blood vessel region based on the size parameters of each blood vessel in the pulmonary artery region and the pulmonary vein region.
[0079] S403, the device uses the vascular features of each vascular region, as well as the pre-acquired lung parenchymal structural features of the patient's lungs, as candidate features.
[0080] S404, the device calculates the similarity between each candidate feature and determines the importance of each candidate feature based on the similarity between them.
[0081] S405, the device selects the target feature based on the importance of each candidate feature.
[0082] S406, the device fuses the selected target features to obtain fused features.
[0083] In this embodiment, mature statistical learning methods are employed to calculate the similarity between candidate features, accurately identifying redundant information. Considering the needs of clinical image feature assessment, methods such as variance filtering (judging information discriminability through feature variance), cosine similarity (calculating the similarity between feature vectors), and collinearity testing can be used to quantitatively evaluate the pairwise similarity and overall similarity distribution of all candidate features in the initial candidate feature set, identifying high-similarity feature combinations and providing a basis for subsequent redundant removal. Subsequently, based on the feature similarity evaluation results and the requirements of the classification task, the importance of each candidate feature is determined. Mature quantitative methods such as linear model coefficients, Gini importance, and SHAP values are used to calculate the contribution (i.e., importance) of each candidate feature to the lung lesion identification task. Among high-similarity feature combinations, it is crucial to distinguish the differences in importance among the features to avoid mistakenly deleting high-value information due to feature similarity. Then, all candidate features are sorted from high to low importance to form a candidate feature priority sequence. Finally, differential screening is performed based on the importance of each candidate feature, iteratively removing redundant features to obtain the target features. If the similarity assessment results show low similarity between features (no obvious redundancy), the initial candidate feature set is directly used as the target feature. If the similarity is high (redundancy exists), the following iterative screening process is followed: From the sorted priority sequence of candidate features, the first feature that meets the high similarity criteria (e.g., collinearity test value ≥ 10) and has relatively low importance is removed first. The importance of the remaining features after removal is recalculated and sorted, and the first high-similarity feature is removed again. The above iterative process is repeated until the similarity between features drops to a preset low level (no high-similarity feature combinations) or the number of features reaches a preset fixed value. Finally, the target features that are both highly representative and low in redundancy are retained. Finally, the selected target features are fused to obtain the fused features.
[0084] S407, the device inputs the fused features into a pre-trained classifier, which is then used to identify the patient's lung lesion assessment results.
[0085] Combination Figure 5 As shown, this embodiment of the present disclosure provides a lung lesion recognition device 500, which includes an image preprocessing module 501, a vascular feature acquisition module 502, a feature fusion module 503, and a lung lesion recognition module 504.
[0086] The image preprocessing module 501 is configured to acquire scan images of the patient's lungs and identify the pulmonary artery and pulmonary vein regions of the lungs in the scan images.
[0087] The vascular feature acquisition module 502 is configured to: acquire the size parameters of each blood vessel in the pulmonary artery region and the pulmonary vein region, and calculate the vascular features of each vascular region based on the size parameters of each blood vessel in the pulmonary artery region and the pulmonary vein region.
[0088] The feature fusion module 503 is configured to fuse the vascular features of each vascular region with the pre-acquired lung parenchymal structural features of the patient's lungs to obtain fused features.
[0089] The lung lesion identification module 504 is configured to input fused features into a pre-trained classifier and use the classifier to identify the patient's lung lesion assessment results.
[0090] The lung lesion identification device 500 provided in this embodiment first accurately identifies the pulmonary artery and pulmonary vein regions in the patient's lung scan image, achieving a clear distinction between pulmonary arteries and veins. Based on this, it acquires the size parameters of each vessel and calculates the vascular features of the corresponding region, fully exploring the potential correlation between changes in arteriovenous vascular structure and lung lesions, thus improving the accuracy of lesion identification. By fusing arteriovenous vascular features with pre-acquired lung parenchymal structural features in multiple dimensions, it comprehensively covers key pathological change-related indicators of lung lesions, ensuring the comprehensiveness of lesion assessment and avoiding the omission of early, minor lesions. Simultaneously, this method focuses on vascular, emphysema, and airway-related features derived from lung scan images, eliminating the need to introduce complex data such as radiomics features and clinical examination results. This simplifies the feature acquisition process, reduces reliance on professional quality control, and thus lowers the model application cost, contributing to its feasibility for promotion in primary healthcare institutions. Finally, it outputs lung lesion assessment results through a pre-trained classifier, achieving efficient fusion of multi-dimensional lesion information, thereby obtaining accurate and reliable lung lesion assessment results.
[0091] In some embodiments, the image preprocessing module 501 is configured to: The pulmonary artery and pulmonary vein regions of the lungs are identified in the scanned images, and the pulmonary artery and pulmonary vein regions are merged to obtain a preliminary vascular merging region. Morphological closure operations are used to smooth the preliminary vascular merging region, eliminating boundary blurring and hollow pixels, and obtaining the target vascular merging region.
[0092] In some embodiments, the vascular feature acquisition module 502 is configured to: Identify each blood vessel based on the bifurcation structure of the blood vessels in the pulmonary artery and pulmonary vein regions; For each blood vessel, multiple measurement points are selected in the blood vessel according to the preset sampling step size; The diameter and cross-sectional area of each blood vessel were obtained at various measurement points.
[0093] In some embodiments, vascular features include individual vascular features of each vessel; the vascular feature acquisition module 502 is configured to implement at least one of the following: For each blood vessel, the maximum blood vessel diameter, minimum blood vessel diameter, and average blood vessel diameter are determined based on the blood vessel diameter at each measurement point. For each blood vessel, the average cross-sectional area of the blood vessel is determined based on the cross-sectional area of the blood vessel at each measurement point. For each blood vessel, the length of the blood vessel is determined based on the sampling step size and the number of measurement points. For each blood vessel, the blood vessel volume is determined based on the sampling step length, the number of measurement points, and the cross-sectional area of the blood vessel at each measurement point.
[0094] In some embodiments, vascular features include pulmonary microvascular features; the vascular feature acquisition module 502 is configured to: Based on the size parameters of each vessel in the pulmonary artery and pulmonary vein regions, different subclasses of pulmonary microvessels are defined in each vascular region; For each subclass of pulmonary microvessels in each vascular region, the pulmonary microvessel characteristics of each subclass are calculated based on the size parameters of each pulmonary microvessel in the same subclass.
[0095] In some embodiments, the vascular feature acquisition module 502 is configured to: calculate the minimum straight-line distance between the vascular vessel and the pleura for each vascular vessel; and define a subclass of pulmonary microvessels in each vascular region based on the minimum straight-line distance between each vascular vessel and the pleura in the pulmonary artery region and the pulmonary vein region.
[0096] In some embodiments, pulmonary microvascular characteristics include pulmonary microvascular cross-sectional area percentage, pulmonary microvascular volume, and pulmonary microvascular volume percentage.
[0097] In some embodiments, the feature fusion module 503 is configured to: The vascular features of each vascular region, as well as the pre-acquired lung parenchymal structural features of the patient's lungs, will be used as candidate features. Calculate the similarity between each candidate feature, and determine the importance of each candidate feature based on the similarity between them; Target features are selected based on the importance of each candidate feature; The selected target features are fused to obtain the fused features.
[0098] Combination Figure 6 As shown, this embodiment of the disclosure provides an electronic device 600, which includes a processor 601 and a memory 602. Optionally, the electronic device 600 may further include a communication interface 603 and a bus 604. The processor 601, communication interface 603, and memory 602 can communicate with each other via the bus 604. The communication interface 603 can be used for information transmission. The processor 601 can call logical instructions in the memory 602 to execute the lung lesion identification method of the above embodiment.
[0099] Furthermore, the logic instructions in the aforementioned memory 602 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium.
[0100] The memory 602, as a computer-readable storage medium, can be used to store software programs and computer-executable programs, such as program instructions / modules corresponding to the methods in the embodiments of this disclosure. The processor 601 executes functional applications and data processing by running the program instructions / modules stored in the memory 602, thereby implementing the lung lesion identification method in the above embodiments.
[0101] The memory 602 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the terminal device. Furthermore, the memory 602 may include high-speed random access memory and may also include non-volatile memory.
[0102] This disclosure provides a computer-readable storage medium storing computer-executable instructions configured to perform the above-described lung lesion identification method.
[0103] The technical solutions of this disclosure can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes one or more instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in this disclosure. The aforementioned storage medium can be a non-transitory storage medium, such as a USB flash drive, external hard drive, read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk, etc., and other media capable of storing program code.
[0104] The foregoing description and accompanying drawings fully illustrate embodiments of this disclosure to enable those skilled in the art to practice them. Other embodiments may include structural, logical, electrical, procedural, and other changes. The embodiments represent only possible variations. Individual components and functions are optional unless explicitly required, and the order of operation may vary. Parts and features of some embodiments may be included in or replace parts and features of other embodiments. Moreover, the terminology used in this application is for describing embodiments only and is not intended to limit the claims. As used in the description of embodiments and claims, the singular forms “a,” “an,” and “the” are intended to equally include the plural forms unless the context clearly indicates otherwise. Similarly, the term “and / or” as used in this application means including one or more of the associated listed items and all possible combinations thereof. Additionally, when used in this application, the term "comprise" and its variations "comprises" and / or "comprising" refer to the presence of stated features, integrals, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components, and / or groups thereof. Without further limitations, an element defined by the phrase "comprises a..." does not exclude the presence of other identical elements in the process, method, or apparatus that includes said element. In this document, each embodiment may focus on the differences from other embodiments, and similar or identical parts between embodiments can be referred to mutually. For methods, products, etc., disclosed in the embodiments, if they correspond to the method section disclosed in the embodiments, the relevant parts can be referred to the description of the method section.
[0105] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the embodiments of this disclosure. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.
[0106] The methods and products disclosed in the embodiments herein (including but not limited to devices and equipment) can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For instance, the division of units may be merely a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces, and the indirect coupling or communication connection of devices or units may be electrical, mechanical, or other forms. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to implement this embodiment according to actual needs. In addition, the functional units in the embodiments of this disclosure may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.
[0107] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions marked in the blocks may occur in a different order than that shown in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. In the descriptions corresponding to the flowcharts and block diagrams in the accompanying drawings, the operations or steps corresponding to different blocks may also occur in a different order than disclosed in the description, and sometimes there is no specific order between different operations or steps. For example, two consecutive operations or steps may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. Each block in a block diagram and / or flowchart, and combinations of blocks in a block diagram and / or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.
Claims
1. A method for identifying lung lesions, characterized in that, include: Acquire scan images of the patient's lungs and identify the pulmonary artery and pulmonary vein regions within the scan images; Obtain the size parameters of each blood vessel in the pulmonary artery and pulmonary vein regions, and calculate the vascular characteristics of each vascular region based on the size parameters of each blood vessel in the pulmonary artery and pulmonary vein regions. The vascular features of each vascular region are fused with the pre-acquired lung parenchymal structural features of the patient's lungs to obtain fused features; The fused features are input into a pre-trained classifier, which is then used to identify the patient's lung lesion assessment results.
2. The lung lesion identification method according to claim 1, characterized in that, The pulmonary artery and pulmonary vein regions of the lungs were identified in the scanned images, including: The pulmonary artery and pulmonary vein regions of the lungs are identified in the scanned images, and the pulmonary artery and pulmonary vein regions are merged to obtain a preliminary vascular merging region. Morphological closure operations are used to smooth the preliminary vascular merging region, eliminating boundary blurring and hollow pixels, and obtaining the target vascular merging region.
3. The lung lesion identification method according to claim 1, characterized in that, Obtain the dimensional parameters of each vessel in the pulmonary artery and pulmonary vein regions, including: Identify each blood vessel based on the bifurcation structure of the blood vessels in the pulmonary artery and pulmonary vein regions; For each blood vessel, multiple measurement points are selected in the blood vessel according to the preset sampling step size; The diameter and cross-sectional area of each blood vessel were obtained at various measurement points.
4. The lung lesion identification method according to claim 3, characterized in that, Vascular features include the individual vessel features of each vessel; vascular features for each vascular region are calculated based on the size parameters of each vessel in the pulmonary artery and pulmonary vein regions, including at least one of the following: For each blood vessel, the maximum blood vessel diameter, minimum blood vessel diameter, and average blood vessel diameter are determined based on the blood vessel diameter at each measurement point. For each blood vessel, the average cross-sectional area of the blood vessel is determined based on the cross-sectional area of the blood vessel at each measurement point. For each blood vessel, the length of the blood vessel is determined based on the sampling step size and the number of measurement points. For each blood vessel, the blood vessel volume is determined based on the sampling step length, the number of measurement points, and the cross-sectional area of the blood vessel at each measurement point.
5. The lung lesion identification method according to claim 3, characterized in that, Vascular features include pulmonary microvascular features; vascular features for each vascular region are calculated based on the size parameters of each vessel in the pulmonary artery and pulmonary vein regions, including: Based on the size parameters of each vessel in the pulmonary artery and pulmonary vein regions, different subclasses of pulmonary microvessels are defined in each vascular region; For each subclass of pulmonary microvessels in each vascular region, the pulmonary microvessel characteristics of each subclass are calculated based on the size parameters of each pulmonary microvessel in the same subclass.
6. The lung lesion identification method according to claim 5, characterized in that, Also includes: For each blood vessel, calculate the minimum straight-line distance between the blood vessel and the pleura; The vascular characteristics of each vascular region are calculated based on the size parameters of each vessel in the pulmonary artery and pulmonary vein regions. The calculation also includes defining a subclass of pulmonary microvessels in each vascular region based on the minimum straight-line distance between each vessel in the pulmonary artery and pulmonary vein regions and the pleura.
7. The lung lesion identification method according to claim 5, characterized in that, Characteristics of pulmonary microvessels include the percentage of pulmonary microvessel cross-sectional area, pulmonary microvessel volume, and pulmonary microvessel volume percentage.
8. The method for identifying lung lesions according to claim 1, characterized in that, The vascular features of each vascular region are fused with the pre-acquired lung parenchymal structural features of the patient's lungs to obtain fused features, including: The vascular features of each vascular region, as well as the pre-acquired lung parenchymal structural features of the patient's lungs, are used as candidate features. Calculate the similarity between each candidate feature, and determine the importance of each candidate feature based on the similarity between them; Target features are selected based on the importance of each candidate feature; The selected target features are fused to obtain the fused features.
9. A lung lesion identification device, characterized in that, include: The image preprocessing module is configured to: acquire scan images of the patient's lungs and identify the pulmonary artery and pulmonary vein regions in the scan images; The vascular feature acquisition module is configured to: acquire the size parameters of each blood vessel in the pulmonary artery region and the pulmonary vein region, and calculate the vascular features of each vascular region based on the size parameters of each blood vessel in the pulmonary artery region and the pulmonary vein region; The feature fusion module is configured to fuse the vascular features of each vascular region with the pre-acquired lung parenchymal structural features of the patient's lungs to obtain fused features; The lung lesion identification module is configured to input fused features into a pre-trained classifier and use the classifier to identify the patient's lung lesion assessment results.
10. An electronic device comprising a processor and a memory storing program instructions, characterized in that, The processor is configured to execute the lung lesion identification method as described in any one of claims 1 to 8 when running the program instructions.