Quantitative analysis system for airway changes in chronic obstructive pulmonary disease based on three-dimensional reconstruction

Abstract

This invention discloses a quantitative analysis system for airway changes in chronic obstructive pulmonary disease (COPD) based on three-dimensional reconstruction, belonging to the field of medical image processing and computer-aided diagnostic technology. The system includes an image data acquisition module, an airway tree reconstruction module, a morphological analysis module, a small airway function assessment module, a disease progression prediction module, and a result output module. It ultimately generates a comprehensive report containing three-dimensional visualized images, morphological parameters, small airway function assessment results, and disease progression prediction results. This system achieves precise reconstruction of the entire lung airway down to the 6th and 7th grade bronchi, high-precision measurement of airway morphological parameters, indirect assessment of small airway function via single inspiratory CT, and intelligent prediction of disease progression, providing comprehensive imaging support for early diagnosis, disease assessment, individualized treatment, and prognosis of COPD.

Description

Technical Field

[0001] This invention belongs to the field of medical image processing and computer-aided diagnosis technology, specifically relating to a quantitative analysis system for airway changes in chronic obstructive pulmonary disease based on three-dimensional reconstruction. Background Technology

[0002] Chronic obstructive pulmonary disease (COPD) is a globally prevalent respiratory disease, and early diagnosis and accurate assessment are crucial for clinical treatment. Traditional diagnosis relies primarily on pulmonary function tests, but these depend on patient cooperation, provide only a holistic assessment, and are not sensitive to early-stage and small airway lesions. Chest CT can directly visualize airway and lung parenchymal lesions, but current CT quantitative analysis techniques have significant shortcomings: first, airway reconstruction depth is limited, often only reaching the fourth-order bronchi, making it difficult to cover critical small airways; second, measurement errors for parameters such as airway wall thickness are large, resulting in insufficient accuracy; third, small airway function assessment often requires dual-phase CT scans, which involve high radiation and registration difficulties; fourth, there is a lack of longitudinal follow-up and disease progression prediction capabilities; and fifth, the visualization and interactivity of results are poor, failing to meet the precise clinical needs. Therefore, there is an urgent need for a COPD quantitative analysis system that can accurately reconstruct the entire lung airway, quantify with high precision, assess small airway function with a single CT scan, and predict disease progression. Summary of the Invention

[0003] To address the aforementioned issues, this invention proposes a quantitative analysis system for airway changes in chronic obstructive pulmonary disease (COPD) based on three-dimensional reconstruction. This system enables precise reconstruction of the entire lung airway down to the 6th and 7th grade bronchi, high-precision measurement of airway morphological parameters, indirect assessment of small airway function via single inspiratory CT scan, and intelligent prediction of disease progression. It provides comprehensive imaging support for early diagnosis, disease assessment, individualized treatment, and prognosis of COPD.

[0004] To achieve the above objectives, the present invention adopts the following technical solution: A quantitative analysis system for airway changes in chronic obstructive pulmonary disease based on three-dimensional reconstruction, characterized by comprising: The image data acquisition module is used to acquire the patient's chest CT image data; The airway tree reconstruction module is connected to the image data acquisition module and is used to segment the airway in CT images, extract the airway mask and centerline, and realize the three-dimensional reconstruction of the whole lung airway tree down to the 6th or 7th level bronchi based on the multi-scale feature fusion algorithm to generate a three-dimensional airway tree model. The morphological analysis module, connected to the airway tree reconstruction module, is used to calculate the airway wall thickness, airway inner diameter, and wall area percentage of each airway branch based on the three-dimensional model of the airway tree, and generate an airway morphological parameter dataset. The small airway function assessment module is connected to the image data acquisition module and the morphological analysis module. It is used to calculate the small airway function status index and generate a small airway function assessment report based on the lung parenchyma density distribution characteristics, ventilation heterogeneity characteristics and airway morphological parameters. The disease progression prediction module is connected to the morphological analysis module and the small airway function assessment module. It is used to calculate the rate of change of parameters based on follow-up data at multiple time points, output the disease progression trend through the disease progression prediction model, and generate future disease progression prediction results. The results output module is connected to the airway tree reconstruction module, morphological analysis module, small airway function assessment module, and disease progression prediction module to generate a comprehensive report that includes three-dimensional visualization images, morphological parameters, small airway function assessment results, and disease progression prediction results.

[0005] Preferably, the airway tree reconstruction module includes: The airway segmentation unit is used to preprocess the chest CT image data, and to segment the airway using a three-dimensional convolutional neural network based on deep learning to extract the airway mask. The centerline extraction unit is used to extract the airway centerline based on the airway mask and to determine the topology of the airway tree using a skeletonization algorithm. The multi-scale reconstruction unit is used to perform three-dimensional reconstruction of the airway tree based on the airway centerline and airway mask using a multi-scale feature fusion algorithm, generating a three-dimensional airway tree model containing the 0th level main trachea to the 6th or 7th level bronchus.

[0006] Preferably, the multi-scale feature fusion algorithm used by the multi-scale reconstruction unit includes: Multi-scale downsampling is performed on the airway mask to generate a first-scale airway feature map, a second-scale airway feature map, and a third-scale airway feature map. Extract local morphological features and global topological features from airway feature maps at various scales; Local morphological features and global topological features at various scales are fused to generate a fused feature map; Based on the fused feature map and airway centerline, a three-dimensional model of the airway tree is reconstructed.

[0007] Preferably, the morphological analysis module includes: The airway wall thickness calculation unit is used to extract multiple cross-sections along the centerline direction for each airway branch in the three-dimensional model of the airway tree, calculate the airway wall thickness of each cross-section based on the full width half height method, and generate airway wall thickness distribution data. The airway inner diameter calculation unit is used to calculate the airway inner diameter of each airway branch based on multiple cross-sections, and generate airway inner diameter distribution data. The wall area percentage calculation unit is used to calculate the wall area percentage of each airway branch based on airway wall thickness distribution data and airway inner diameter distribution data.

[0008] Preferably, the full width at half height method used in the airway wall thickness calculation unit includes: In the 3D model of the airway tree, extract the cross-section of the airway along the direction perpendicular to the centerline of the airway; Density distribution curves are obtained radially on the cross-section of the airway. Determine the maximum and minimum values ​​of the density distribution curve, and calculate the half-height value; Locate the two half-height points on the density distribution curve; Calculate the distance between the two half-height points as the airway wall thickness.

[0009] Preferably, the small airway function assessment module includes: The lung parenchyma density analysis unit is used to extract lung parenchyma regions based on chest CT image data, calculate lung parenchyma density distribution histogram and density mean, and identify low-density and high-density regions. The ventilation nonuniformity assessment unit is used to segment lung segments from chest CT image data, calculate the density standard deviation and density variation coefficient of each lung segment, and generate ventilation nonuniformity index. The small airway function calculation unit is used to calculate small airway function status indicators based on lung parenchyma density distribution histograms, ventilation non-uniformity indices, and airway morphology parameter datasets.

[0010] Preferably, the method for calculating the small airway function status index used by the small airway function calculation unit includes: Extract the mean airway wall thickness and mean airway inner diameter of the 5th to 7th grade bronchi; Calculate the percentage of volume in the lung parenchyma that is below the threshold density. Calculate the weighted average of the ventilation nonuniformity index; Based on the mean airway wall thickness, mean airway inner diameter, volume percentage of areas below the threshold density, and weighted average, the functional status indicators of small airways are determined.

[0011] Preferably, the disease progression prediction module includes: The longitudinal data comparison unit is used to receive the airway morphology parameter dataset and small airway function assessment report at the first time point, as well as the airway morphology parameter dataset and small airway function assessment report at the second time point, and to calculate the rate of change of airway parameters. The function decline trend analysis unit is used to analyze the trends of airway wall thickness change, airway inner diameter change, and small airway function change based on the rate of change of airway parameters. The progression prediction unit is used to input the trends of airway wall thickness change, airway internal diameter change, and small airway function change into the disease progression prediction model to generate future disease progression prediction results.

[0012] Preferably, the disease progression prediction model is trained based on historical patient data, and the training method includes: We collected multi-time-point airway morphology parameters and small airway function assessment data from multiple patients with chronic obstructive pulmonary disease. The rate of change of airway parameters was extracted as input features, and the disease progression was extracted as labels. A disease progression prediction model is trained using machine learning algorithms, enabling the model to predict future disease progression based on the rate of change in airway parameters.

[0013] Preferably, the result output module includes: The visualization generation unit is used to generate a 3D visualization image of the airway tree based on the 3D model of the airway tree, and to use different colors to mark airway branches at different levels and to highlight abnormal airway areas. The report generation unit integrates airway morphology parameter data, small airway function assessment reports, and disease progression prediction results to generate a comprehensive analysis report containing charts and text descriptions. The interactive operation unit is used to receive user operation commands and respond to the operation commands to rotate, scale, slice, and annotate parameters of the airway tree 3D visualization image.

[0014] By adopting the above technical solution, the present invention has the following beneficial effects: 1. This invention achieves accurate three-dimensional reconstruction of the entire lung airway tree from level 0 main bronchus to levels 6 and 7 bronchi by developing an airway tree reconstruction algorithm based on multi-scale feature fusion. The reconstruction accuracy reaches 97.5%. Compared with existing technologies that can only reconstruct airway branches up to about level 4, this invention significantly expands the depth and completeness of airway tree reconstruction. The complete airway tree reconstruction not only includes the small airway branches that are crucial for the early diagnosis of COPD, but also provides a complete anatomical basis for subsequent morphological parameter measurements, avoiding analytical blind spots caused by incomplete airway tree reconstruction.

[0015] 2. This invention achieves high-precision measurement of key morphological parameters such as airway wall thickness, airway inner diameter, and wall area percentage by employing an improved full-width half-height method and multi-section averaging technology. The measurement accuracy reaches 0.1 mm, a significant improvement compared to the 0.3 mm or higher error of existing technologies. This high-precision morphological parameter measurement can more sensitively detect early changes in airway remodeling, providing more reliable quantitative indicators for early diagnosis and disease grading of COPD. Furthermore, this invention systematically measures and statistically analyzes all branches of the airway tree, rather than measuring only a few representative airways, significantly improving the comprehensiveness and representativeness of the analysis results.

[0016] 3. This invention innovatively proposes a method for indirectly assessing small airway function based on lung parenchyma density and ventilation heterogeneity characteristics. Small airway function assessment requires only a single inspiratory phase CT scan, avoiding the limitation of traditional PRM methods which require paired inspiratory and expiratory phase scans. By analyzing the proportion of low-density regions and the lung segment density variation coefficient in the lung parenchyma density distribution histogram, combined with morphological parameters of small airways from grade 5 to 7, the calculated small airway function index showed a correlation of 0.89 with pulmonary function tests, validating the effectiveness of the method. This innovation significantly reduces patient radiation dose and examination time, while providing a feasible solution for small airway function assessment for patients unable to cooperate with expiratory phase scans.

[0017] 4. This invention establishes a disease progression prediction function based on longitudinal follow-up data. By comparing airway morphological parameters and small airway function indicators at multiple time points, the rate of parameter change and functional decline trend are calculated. A disease progression prediction model trained using machine learning algorithms predicts the future trajectory of the disease. Clinical application validation shows that the prediction model achieves an accuracy rate of 82% in predicting COPD progression, and can identify high-risk patients with rapid disease progression in advance, providing an important basis for timely adjustment of treatment plans and improvement of prognosis.

[0018] 5. This invention provides intuitive 3D visualization and interactive operation functions through the results output module. Clinicians can rotate, scale, and slice the airway tree 3D model at any angle. Different levels of airway branches are marked with different colors, abnormal airway areas are highlighted, and morphological parameters can be directly labeled on the corresponding airway branches. The comprehensive analysis report integrates information from multiple aspects such as airway tree structure, morphological parameters, small airway function assessment, and disease progression prediction, presented in a combination of charts and text, significantly improving the convenience and practicality of clinical application.

[0019] In summary, the system of this invention provides a comprehensive imaging tool for the accurate diagnosis, disease assessment, treatment planning, and prognosis prediction of COPD through innovative technologies such as precise three-dimensional reconstruction of the whole lung airway tree, high-precision morphological parameter measurement, indirect assessment of small airway function in a single scan, and prediction of disease progression. It promotes the transformation of COPD from traditional pulmonary function diagnosis to precise imaging assessment and has important clinical application value and broad prospects for promotion. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the overall system architecture of the present invention; Figure 2 This is a schematic diagram of the airway tree reconstruction module of the present invention; Figure 3 This is a flowchart illustrating the multi-scale feature fusion algorithm of the present invention; Figure 4 This is a schematic diagram of the morphological analysis module of the present invention; Figure 5 This is a schematic diagram of the small airway function assessment module of the present invention; Figure 6 This is a schematic diagram of the workflow of the disease progression prediction module of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0022] like Figures 1 to 6 As shown, this invention provides a quantitative analysis system for airway changes in chronic obstructive pulmonary disease based on three-dimensional reconstruction, including an image data acquisition module 1, an airway tree reconstruction module 2, a morphological analysis module 3, a small airway function assessment module 4, a disease progression prediction module 5, and a result output module 6. The modules are interconnected through data interfaces to enable data flow and processing.

[0023] The image data acquisition module 1 is used to acquire chest CT image data (i.e., chest computed tomography image data) of patients with chronic obstructive pulmonary disease. In one embodiment of the present invention, the image data acquisition module 1 supports the acquisition of DICOM format image data from various CT devices, including mainstream CT devices such as Siemens, GE, and Philips. The preferred CT scan parameters are tube voltage of 120kV, tube current of 40-80mA, slice thickness of 0.625-1.25mm, reconstruction interval of 0.625-1.25mm, and matrix of 512×512. The scan is performed with breath-holding at the end of inspiration to ensure full lung expansion and airway expansion. The image data acquisition module 1 performs format parsing and preprocessing on the acquired DICOM data, extracting the image matrix and scan parameter information to provide standardized input data for subsequent analysis.

[0024] Reference Figure 2 The airway tree reconstruction module 2 is connected to the image data acquisition module 1 and is used to perform airway segmentation processing on chest CT image data, extract airway masks and airway centerlines, and realize three-dimensional reconstruction of the whole lung airway tree based on a multi-scale feature fusion algorithm. The airway tree reconstruction module 2 includes an airway segmentation unit, a centerline extraction unit, and a multi-scale reconstruction unit.

[0025] The airway segmentation unit employs a deep learning-based 3D convolutional neural network to segment the airway from CT image data. In one embodiment of the invention, the airway segmentation network uses a 3D U-Net architecture. The encoder part includes four downsampling layers, each containing two 3×3×3 convolutional layers and one 2×2×2 max-pooling layer. The decoder part includes four upsampling layers, each containing one 2×2×2 transposed convolutional layer and two 3×3×3 convolutional layers. Features are transferred between the encoder and decoder via skip connections. The network input is the 3D data volume of the CT image, and the output is an airway probability map. A binary airway mask is obtained through threshold segmentation. To improve segmentation accuracy, the airway segmentation network uses a hybrid loss function during training, which consists of Dice loss and Focal Loss, calculated as follows: L total = λ1L Dice + λ2 L Focal , Among them, L total For the total loss, L Dice For Dice's loss, L Focal The focus loss is λ1 and λ2, which are weighting coefficients. The Dice loss calculation formula is: , Where N is the total number of voxels, p iLet g be the predicted probability of the i-th voxel. i For the true label of the i-th voxel, The smoothing term is set to 1 to prevent the denominator from being 0, and i is the voxel number. The formula for calculating the focal loss is: , Where, α i λ1 is the class weight coefficient for the i-th voxel, used to balance positive and negative samples, and γ is the focusing parameter used to reduce the weight of easily classified samples. In a preferred embodiment, λ1 is 0.6, λ2 is 0.4, and α... i The value for airway voxels is 0.7, the value for background voxels is 0.3, and the value for γ is 2.0. This hybrid loss function considers both the overall segmentation overlap and pays more attention to the difficult-to-classify small airway branches, significantly improving the segmentation accuracy of small airways.

[0026] The airway segmentation network was trained using a dataset containing CT images of 500 COPD patients, with each image having its airway mask manually annotated by a professional radiologist as the gold standard. Training employed the Adam optimizer with an initial learning rate of 0.0001, 200 training epochs, and a batch size of 2. On the validation set, the airway segmentation network achieved a Dice coefficient of 0.92 and a detection rate of over 85% for small airway branches at levels 6 and 7.

[0027] The centerline extraction unit extracts the airway centerline based on the airway mask and uses a skeletonization algorithm to determine the topology of the airway tree. In one embodiment of the invention, the centerline extraction first performs a three-dimensional distance transformation on the airway mask, calculates the shortest distance from each airway voxel to the airway boundary, and then uses a skeletonization algorithm based on the maximum distance field to extract the airway centerline. Specifically, starting from the tracheal initiation point, it extends outward along the path with the maximum value of the distance field, and continues to extend along each branch when encountering a bifurcation point until all airway ends are reached. The extracted centerline contains the complete topology of the airway tree, and each centerline point records its three-dimensional coordinate position and local airway radius (equal to the distance transformation value of that point). To improve the smoothness and accuracy of the centerline, a B-spline curve is used to fit and smooth the centerline while keeping the positions of the airway bifurcation points unchanged.

[0028] Based on the extracted airway centerlines, the centerline extraction unit performs hierarchical labeling of the airway tree. According to airway anatomical nomenclature rules, the main trachea (trachea) is level 0, the left and right main bronchi are level 1, the lobar bronchus is level 2, the segmental bronchus is level 3, the subsegmental bronchus is level 4, and so on up to levels 6 and 7. Airway grading is automatically completed by analyzing the topological relationships of the centerlines. Starting from the main trachea, the system traverses outwards along the centerline, and at each bifurcation point, the sub-branch is one level higher than the parent branch. The grading results are stored in the airway centerline data structure, providing a basis for subsequent morphological parameter grading statistics.

[0029] Reference Figure 3 The multi-scale reconstruction unit is used to perform 3D reconstruction of the airway tree based on the airway centerline and airway mask, employing a multi-scale feature fusion algorithm to generate a 3D airway tree model containing the 0th-level main trachea to the 6th-7th-level bronchi. The core idea of ​​the multi-scale feature fusion algorithm is to extract local morphological features and global topological features of the airway at different scales, and then fuse the features at each scale to enhance the reconstruction effect of small airway branches.

[0030] The specific process of the multi-scale feature fusion algorithm is as follows. First, the airway mask is downsampled at multiple scales to generate a first-scale airway feature map, a second-scale airway feature map, and a third-scale airway feature map. The first scale maintains the original resolution and can capture the local morphological features of small airway branches; the second scale is downsampled by 2 times to capture the airway morphology and connectivity at a medium scale; and the third scale is downsampled by 4 times to capture the overall topological structure and main branch orientation of the airway tree. In a preferred embodiment, the voxel size of the original CT image is 0.7mm × 0.7mm × 0.7mm. The first scale maintains this resolution, the second scale is downsampled to 1.4mm × 1.4mm × 1.4mm, and the third scale is downsampled to 2.8mm × 2.8mm × 2.8mm.

[0031] Secondly, local morphological features and global topological features are extracted from airway feature maps at each scale. Local morphological features are extracted through 3D morphological operations, including dilation, erosion, opening, and closing operations, to enhance the boundary continuity and internal filling of the airways. Global topological features are obtained through connectivity analysis and skeleton extraction, characterizing the branching structure and connections of the airway tree. At the first scale, local morphological features preserve the detailed morphology of small airways; at the second scale, they smooth the surface of medium-sized airways; and at the third scale, global topological features ensure the overall structural integrity of the airway tree.

[0032] Then, the local morphological features and global topological features at each scale are fused to generate a fused feature map. Feature fusion employs a weighted fusion strategy, with the fusion weights adaptively adjusted according to the level of the airway branches. For large airways from level 0 to level 3, the global topological features at the third scale are given a higher weight (0.5), the local morphological features at the second scale are given a medium weight (0.3), and the local morphological features at the first scale are given a lower weight (0.2), ensuring the overall structural stability of the large airways. For small airways from level 4 to level 7, the local morphological features at the first scale are given a higher weight (0.6), the local morphological features at the second scale are given a medium weight (0.3), and the global topological features at the third scale are given a lower weight (0.1), ensuring that the detailed morphology of the small airways is effectively preserved. The fused feature map integrates the advantages of each scale, maintaining both the overall topological structure of the airway tree and the local morphological details of the small airway branches.

[0033] Finally, based on the fused feature map and airway centerline, the Marching Cubes algorithm was used to reconstruct the 3D surface model of the airway tree. The Marching Cubes algorithm extracts the airway region boundaries from the fused feature map as triangular mesh surfaces, generating the surface mesh data for the 3D airway tree model. The reconstructed 3D airway tree model includes all branches from the main trachea (level 0) to the bronchi (levels 6 and 7). The airway surface is smooth and continuous, and the branch connections are clear, providing an accurate geometric basis for subsequent morphological parameter measurements.

[0034] In one embodiment of the present invention, airway tree reconstruction was performed on CT images of 100 COPD patients, and the reconstruction results were compared with airway trees manually drawn by professional radiologists. Statistical results showed that the airway tree reconstructed by the method of the present invention achieved a Dice coefficient of over 0.98 for airways from level 0 to 4, and a Dice coefficient of over 0.85 for airways from level 5 to 7, with an overall reconstruction accuracy of 97.5%. Compared with traditional airway tree reconstruction methods based on region growing, the multi-scale feature fusion algorithm of the present invention improved the detection rate of small airway branches by approximately 30% and the assessment of the overall integrity of the airway tree by approximately 25%.

[0035] Reference Figure 4 The morphological analysis module 3 is connected to the airway tree reconstruction module 2 and is used to calculate the airway wall thickness, airway inner diameter, and wall area percentage of each airway branch based on the three-dimensional model of the airway tree, generating an airway morphological parameter dataset. The morphological analysis module 3 includes an airway wall thickness calculation unit, an airway inner diameter calculation unit, and a wall area percentage calculation unit.

[0036] The airway wall thickness calculation unit is used to calculate the airway wall thickness distribution data for each airway branch. The airway wall thickness is measured using a modified Full Width at Half Maximum (FWHM) method, which determines the inner and outer boundaries of the airway wall based on the density distribution curve on the airway cross-section. Specifically, multiple airway cross-sections are extracted vertically along the airway centerline; preferably, a cross-section is extracted every 1 mm along the centerline to ensure sufficient sampling of the airway wall thickness. On each airway cross-section, a density distribution curve is obtained radially from the centerline point. This density distribution curve shows a trend from low-density airway lumen (air, approximately -1000 HU) to high-density airway wall (soft tissue, approximately 40 HU) and then to intermediate-density lung parenchyma (approximately -850 HU).

[0037] The half-height method is used to determine the positions of the inner and outer boundaries of the airway wall on the density distribution curve. First, the maximum value HU of the density distribution curve is identified. max (Peak density corresponding to the airway wall) and minimum HU min (Corresponding to the density of the airway lumen). Among them, HU max It usually appears in the center of the airway wall, HU min It usually appears in the center of the airway lumen. Then, the half-height value (HU) is calculated. half The calculation formula is: , On the density distribution curve, find the first density value equal to HU from the lumen outwards. half The point corresponds to the inner boundary of the airway wall; continue searching outwards for the second density value equal to HU. half The point corresponds to the outer boundary of the airway wall. The distance from the inner boundary to the outer boundary of the airway wall is the airway wall thickness in that direction. The above measurement process is repeated along multiple radial directions (preferably 36 directions, one direction every 10°) of the airway cross-section to obtain the airway wall thickness distribution on that cross-section. The airway wall thickness of this cross-section is taken as the average of the airway wall thicknesses in all radial directions, calculated using the following formula: , Among them, WT section The thickness of the airway wall in this cross-section, M is the number in the radial direction (preferably 36), WT j Let be the airway wall thickness in the j-th radial direction, where j is the radial direction number. The average airway wall thickness of an airway branch is obtained by averaging the airway wall thicknesses across all cross-sections along its centerline. The calculation formula is: , Among them, WT branchWT represents the average airway wall thickness of this airway branch, K represents the number of cross-sections of this airway branch, and WT represents the total number of cross-sections. section,k Let be the airway wall thickness at the k-th cross-section, where k is the cross-section number.

[0038] To further improve measurement accuracy, this invention employs a multi-cross-section averaging technique. At least 20 cross-sections are extracted along the centerline of each airway branch for measurement. The measurement results from all cross-sections are then statistically analyzed, and outliers are removed before the average value is taken as the airway wall thickness for that airway branch. Outlier removal uses a box plot method, considering measurements exceeding 1.5 times the interquartile range as outliers and removing them. Multi-cross-section averaging effectively reduces the impact of measurement errors from individual cross-sections, significantly improving the stability and repeatability of airway wall thickness measurement. In one embodiment of this invention, the airway wall thickness of 10 COPD patients was repeatedly measured, with a coefficient of variation of less than 5% and a measurement accuracy of 0.1 mm.

[0039] The airway inner diameter calculation unit is used to calculate the airway inner diameter distribution data for each airway branch. The airway inner diameter is measured based on the position of the inner boundary of the airway wall. On each airway cross-section, the area enclosed by the inner boundary of the airway wall is the airway lumen, and the equivalent diameter of the lumen is the airway inner diameter. The equivalent diameter of the lumen is obtained by calculating the lumen area and then converting it; the calculation formula is: , Among them, D section Let A be the inner diameter of the airway in this cross-section. lumen Let A be the area of ​​the tube cavity, and π be the mathematical constant pi. Tube cavity area A lumen The airway diameter is obtained by counting voxels within the region enclosed by the inner boundary of the airway wall and multiplying by the voxel volume. The average airway diameter of the airway branch is obtained by averaging the airway inner diameters across all cross-sections along the centerline of the airway branch; the calculation formula is as follows: , Among them, D branch Let D be the average airway inner diameter of this airway branch, K be the number of cross-sections of this airway branch, and D be the number of cross-sections of this airway branch. section,k Let be the airway inner diameter at the k-th cross-section. Multi-cross-section averaging and outlier removal techniques are also used to improve measurement accuracy.

[0040] The wall area percentage calculation unit is used to calculate the wall area percentage for each airway branch. Wall area percentage (WA%) is an important indicator for assessing the degree of airway remodeling, defined as the percentage of the airway wall cross-sectional area to the total airway cross-sectional area. The airway wall cross-sectional area is the area enclosed by the outer boundary of the airway wall minus the area enclosed by the inner boundary of the airway wall; the total airway cross-sectional area is the area enclosed by the outer boundary of the airway walls. The formula for calculating the wall area percentage is: , Where WA% is the percentage of wall area, A wall A is the cross-sectional area of ​​the airway wall. total A is the total cross-sectional area of ​​the airway. outer A is the area enclosed by the outer boundary of the airway wall. lumen The wall area is the lumen area. Calculate the percentage of wall area on each airway cross-section, and then average the percentage of wall area across all cross-sections of that airway branch to obtain the average percentage of wall area for that airway branch.

[0041] The morphological analysis module 3 measures the aforementioned morphological parameters for all airway branches (levels 0 to 7) in the 3D airway tree model, generating an airway morphological parameter dataset. This dataset is stored in tabular form, with each row corresponding to an airway branch and containing information such as airway level, airway location (left or right lung, specific lobe segment), mean airway wall thickness, mean airway diameter, and mean wall area percentage. Furthermore, the morphological analysis module 3 calculates statistical measures of airway morphological parameters for the entire lung and each lobe segment, including the mean airway wall thickness, mean airway diameter, and mean wall area percentage for each airway level, as well as the standard deviation, median, upper quartile, and lower quartile of these parameters, providing clinicians with comprehensive statistical analysis results.

[0042] In one embodiment of the present invention, airway morphological parameters were measured in 50 COPD patients with different GOLD grades. Statistical results showed that as COPD severity increased (from GOLD I to GOLD IV), airway wall thickness gradually increased, airway diameter gradually decreased, and the percentage of wall area gradually increased. The mean airway wall thickness for GOLD I patients was 1.2 ± 0.2 mm, the mean airway diameter was 4.5 ± 0.8 mm, and the mean wall area percentage was 45 ± 5%; the mean airway wall thickness for GOLD IV patients was 1.8 ± 0.3 mm, the mean airway diameter was 3.2 ± 0.6 mm, and the mean wall area percentage was 62 ± 7%. These morphological parameters were significantly correlated with pulmonary function indicators (FEV1%pred) (correlation coefficient r = -0.76, p < 0.001), validating the clinical value of morphological parameters in COPD assessment.

[0043] Reference Figure 5 The small airway function assessment module 4 is connected to the image data acquisition module 1 and the morphological analysis module 3. It is used to extract lung parenchyma density distribution characteristics and ventilation heterogeneity characteristics based on chest CT image data, calculate small airway function status indicators by combining them with airway morphological parameter datasets, and generate a small airway function assessment report. The small airway function assessment module 4 includes a lung parenchyma density analysis unit, a ventilation heterogeneity assessment unit, and a small airway function calculation unit.

[0044] The lung parenchyma density analysis unit is used to extract lung parenchyma regions based on chest CT image data, calculate the lung parenchyma density distribution histogram and density mean, and identify low-density and high-density regions. First, the lung parenchyma region is segmented from the CT image data using an automatic lung parenchyma segmentation algorithm, excluding structures such as airways, blood vessels, and the chest wall. Lung parenchyma segmentation employs a threshold-based and morphological operation-based method, setting the lung parenchyma density range to -1024 HU to -200 HU. Connectivity analysis is performed on voxels within this density range to extract the left and right lung regions. Then, morphological closure operations are used to fill lung cavities, and boundary smoothing operations are used to optimize lung boundaries.

[0045] For the segmented lung parenchyma regions, the density values ​​of each voxel are statistically analyzed, and a histogram of lung parenchyma density distribution is plotted. The horizontal axis of the histogram represents the density value (HU), and the vertical axis represents the number of voxels or volume percentage. The mean density of the lung parenchyma is calculated. The calculation formula is: , in, V represents the mean density of lung parenchyma. lung HU is the total volume of lung parenchyma. v This represents the density value of voxel v. The mean density of lung parenchyma reflects the overall density level of the lung parenchyma. In COPD patients, due to emphysema leading to lung parenchyma destruction and increased air content, the mean density of lung parenchyma is usually lower than that of normal individuals.

[0046] Identify low-density areas (LAA) and high-density areas (HAA) within the lung parenchyma. Low-density areas are defined as voxels with a density value below -950 HU, reflecting the presence and severity of emphysema. High-density areas are defined as voxels with a density value above -700 HU, reflecting inflammation, fibrosis, or increased vascularity in the lung parenchyma. The volume percentage of low-density areas (LAA%) and high-density areas (HAA%) are calculated using the following formula: , , Where LAA% is the volume percentage of the low-density region, V LAA V represents the volume of the low-density region, HAA% represents the volume percentage of the high-density region, and V represents the volume of the low-density region. HAA V is the volume of the high-density region. lung This represents the total volume of lung parenchyma. LAA% is a classic indicator for assessing the severity of emphysema and is closely related to decreased lung function. HAA% reflects the degree of inflammation and fibrosis in the lung parenchyma and is usually elevated during acute exacerbations of COPD.

[0047] The ventilation heterogeneity assessment unit is used to segment lung data from chest CT images, calculate the standard deviation of density and coefficient of variation for each lung segment, and generate a ventilation heterogeneity index. Lung segmentation employs an automatic segmentation algorithm based on anatomical standards, dividing the left lung into 8 segments and the right lung into 10 segments, for a total of 18 segments. Within each lung segment, the standard deviation σ of lung parenchyma density is calculated. segment and coefficient of variation CV segment The calculation formula is: , , Where, σ segment V represents the standard deviation of density in that lung segment. segment HU represents the volume of that lung segment. v Let v be the density value of voxel v. The mean density of this lung segment, CV segment This represents the density variation coefficient for that lung segment. The density standard deviation and coefficient of variation reflect the uniformity of density distribution within a lung segment; higher values ​​indicate more uneven ventilation. In COPD patients, small airway dysfunction leads to insufficient ventilation in some lung segments, resulting in gas retention, manifested as uneven density distribution within the lung segment, with increased density standard deviation and coefficient of variation.

[0048] Calculate the whole lung ventilation heterogeneity index VH global , defined as the weighted average of the density variation coefficients of all lung segments, is calculated using the following formula: , Among them, VH global V is an indicator of whole-lung ventilation heterogeneity. segment,i Let CV be the volume of the i-th lung segment. segment,i Let VH be the density variation coefficient of the i-th lung segment, and the summation covers 18 lung segments. global It comprehensively reflects the degree of uneven ventilation throughout the lungs and is an important indicator for assessing small airway dysfunction.

[0049] The small airway function calculation unit is used to calculate small airway functional status indices based on lung parenchyma density distribution histograms, ventilation heterogeneity indices, and airway morphology parameter datasets. The small airway functional status index (SAD) is defined as a dimensionless numerical value that comprehensively reflects the degree of small airway dysfunction, ranging from 0 to 100, with higher values ​​indicating more severe small airway dysfunction. score The calculation uses a multi-parameter weighted fusion method, and the calculation formula is as follows: , Among them, SAD score This is an indicator of small airway functional status; LAA% is the volume percentage of low-density regions; VH global WT is an indicator of whole-lung ventilation heterogeneity. small D represents the average airway wall thickness (in mm) for small airways from level 5 to level 7. small The average airway diameter (in mm) for small airways from level 5 to level 7 is represented by w1, w2, w3, and w4, which are weighting coefficients. These weighting coefficients are determined through correlation analysis with pulmonary function test indicators. In the preferred embodiment, w1 = 0.3, w2 = 0.35, w3 = 0.2, and w4 = 0.15. This formula comprehensively considers the degree of emphysema (LAA%) and ventilation heterogeneity (VH). global Small airway wall thickening (WT) small ) and small airway stenosis (D small The information from these four aspects comprehensively reflects the functional status of the small airways.

[0050] Module 4 of the small airway function assessment module generates a small airway function assessment report. The report includes a lung parenchyma density distribution histogram, percentage of low-density areas, indicators of whole-lung ventilation heterogeneity, small airway function status indicators, and clinical interpretations and recommendations for each indicator. The report is presented in a combined text and graphic format. The lung parenchyma density distribution histogram is displayed as a bar chart, with low-density areas marked in red on the 3D lung parenchyma model and lung segments with high ventilation heterogeneity marked in blue, facilitating a clear understanding of the small airway function assessment results by clinicians.

[0051] In one embodiment of the present invention, small airway function was assessed in 80 COPD patients, and a correlation analysis was performed between small airway function status indicators and small airway function indicators (maximum mid-expiratory flow rate %pred) from pulmonary function tests. The results showed a significant negative correlation between small airway function status indicators and MMEF%pred (correlation coefficient r = -0.89, p < 0.001), meaning that higher small airway function status indicators correlated with lower MMEF%pred and more severe small airway dysfunction. This validates the effectiveness and clinical value of the small airway function assessment method of the present invention. Compared to the PRM method, which requires paired inspiratory and expiratory phase scans, the method of the present invention requires only a single inspiratory phase scan, significantly reducing the patient's radiation dose (from approximately 4 mSv to approximately 2 mSv) and examination time (from approximately 10 minutes to approximately 5 minutes), while avoiding examination failures due to patient inability to cooperate with expiratory phase scanning.

[0052] Reference Figure 6 The disease progression prediction module 5 is connected to the morphology analysis module 3 and the small airway function assessment module 4. It receives airway morphology parameter datasets and small airway function assessment reports from multiple time points, calculates the rate of change of airway parameters and the trend of functional decline, and generates future disease progression prediction results based on the disease progression prediction model. The disease progression prediction module 5 includes a longitudinal data comparison unit, a functional decline trend analysis unit, and a progression prediction unit.

[0053] The longitudinal data comparison unit receives airway morphology parameter datasets and small airway function assessment reports from the first time point (baseline) and the second time point (follow-up), calculating the rate of change of airway parameters. In clinical practice, COPD patients typically undergo CT follow-up examinations every 6 or 12 months to assess disease progression. The longitudinal data comparison unit registers the three-dimensional airway tree models from the two time points to ensure accurate matching of corresponding airway branches, and then calculates the amount and rate of change of morphological parameters for each airway branch.

[0054] The formulas for calculating the rate of change of airway wall thickness ΔWT, the rate of change of airway inner diameter ΔD, and the rate of change of wall area percentage ΔWA% are as follows: , , , Where ΔWT is the rate of change of airway wall thickness, WT follow-up WT represents the airway wall thickness at the follow-up time point. baseline ΔD represents the airway wall thickness at the baseline time point, and ΔD represents the rate of change of airway diameter. follow-up D represents the airway diameter at the follow-up time point. baselineΔWA% represents the airway diameter at the baseline time point, and ΔWA% represents the percentage change in wall area. follow-up WA% represents the percentage of wall area at the follow-up time point. baseline This represents the percentage of wall area at the baseline time point. A positive rate of change in airway wall thickness indicates increased airway wall thickening, a negative rate of change in airway diameter indicates increased airway narrowing, and a positive percentage change in wall area indicates progress in airway remodeling.

[0055] Similarly, the change in small airway functional status index ΔSAD is calculated. score The calculation formula is: , Wherein, ΔSAD score SAD is a measure of change in small airway functional status. score,follow-up SAD is an indicator of small airway functional status at follow-up time points. score,baseline This is a small airway function status indicator at the baseline time point. A positive change in the small airway function status indicator indicates a worsening of small airway dysfunction.

[0056] The functional decline trend analysis unit is used to analyze trends in airway wall thickness, airway diameter, and small airway function based on the rate of change of airway parameters. If the patient has follow-up data at three or more time points, the functional decline trend analysis unit uses linear regression or exponential fitting methods to fit the curves of airway parameter changes over time, assessing the rate and acceleration of parameter changes. For example, linear fitting is used to measure changes in airway wall thickness over time. ,in Let t be the airway wall thickness at time t, WT0 be the initial airway wall thickness, k be the annual growth rate of airway wall thickness, and t be time (in years). The annual growth rate k obtained by fitting the data can quantify the rate of airway remodeling.

[0057] The progression prediction unit is used to input the trends of airway wall thickness change, airway internal diameter change, and small airway function change into the disease progression prediction model to generate future disease progression prediction results. The disease progression prediction model is trained using a machine learning algorithm; in one embodiment of this invention, a Support Vector Machine (SVM) or Random Forest algorithm is used. The training dataset contains multi-timepoint CT image data and pulmonary function test data of 200 COPD patients, with a follow-up time span of 2 to 5 years. Input features include baseline airway morphological parameters (airway wall thickness, airway internal diameter, wall area percentage), small airway functional status indicators, and lung parenchyma density parameters (…). ), and the rate of change of airway parameters at the first year of follow-up (ΔWT, ΔD, ΔWA%, ΔSAD) scoreThe output label represents the disease progression category over the next 1 to 2 years, defined as stable, slow progression, and rapid progression. The classification criterion is the rate of decline of FEV1%pred: less than 30 mL / year is stable, 30-60 mL / year is slow progression, and greater than 60 mL / year is rapid progression.

[0058] The disease progression prediction model was trained using a 10-fold cross-validation method, achieving a classification accuracy of 82% on the validation set, a sensitivity of 87% for identifying rapidly progressing patients, and a specificity of 80%. This indicates that the disease progression prediction model can accurately predict the disease progression trend in COPD patients, providing important reference for clinicians to adjust treatment plans. For patients predicted to have rapid progression, clinicians can consider increasing the intensity of drug therapy, increasing the frequency of follow-up visits, or implementing comprehensive interventions such as pulmonary rehabilitation to slow disease progression.

[0059] The Disease Progression Prediction Module 5 generates a report predicting future disease progression. The report includes trend graphs of airway parameters and small airway function, predictions of disease progression categories, prediction confidence levels, and individualized treatment recommendations. The report uses line graphs to show the trends of each parameter over time and bar charts to display the predicted probability distribution of disease progression categories, providing clinicians with intuitive information on disease progression prediction.

[0060] The results output module 6 is connected to the airway tree reconstruction module 2, morphological analysis module 3, small airway function assessment module 4, and disease progression prediction module 5. It is used to generate a comprehensive analysis report that includes a 3D visualization image of the airway tree, airway morphological parameter data, small airway function assessment report, and disease progression prediction results. The results output module 6 includes a visualization generation unit, a report generation unit, and an interactive operation unit.

[0061] The visualization generation unit generates 3D visualizations of the airway tree based on the 3D model. Different colors are used to label airway branches at different levels, and abnormal airway areas are highlighted. The 3D visualization of the airway tree employs volumetric or surface rendering techniques to display the complete structure of the airway tree in 3D space. To facilitate the identification of airway branches at different levels, a color coding scheme is used: red for level 0 main trachea, orange for level 1 main bronchus, yellow for level 2 lobar bronchus, green for level 3 segmental bronchus, and a gradient from blue to purple for levels 4 to 7 small airways. Abnormal airway areas with wall thickness exceeding the normal range or significantly narrowed airway diameters are highlighted on the visualization image, for example, using a flashing effect or a warning symbol next to the airway, to remind clinicians to pay close attention.

[0062] On a 3D visualization of the airway tree, numerical values ​​of airway morphological parameters can be overlaid. For each airway branch, the airway wall thickness, internal diameter, and percentage of wall area are labeled next to its location. A heatmap is used to display the degree of abnormality of these parameters: parameters within the normal range are shown in green, mild abnormalities in yellow, moderate abnormalities in orange, and severe abnormalities in red. This visualization method intuitively combines quantitative parameters with anatomical structures, significantly improving clinicians' understanding of lesion distribution and severity.

[0063] The report generation unit integrates airway morphology parameter data, small airway function assessment reports, and disease progression prediction results to generate a comprehensive analysis report containing charts and text descriptions. The comprehensive analysis report uses a structured format and includes the following sections.

[0064] The first part contains basic patient information, including patient name, gender, age, smoking history, COPD history, and current symptoms. The second part presents the airway tree reconstruction results, including a 3D visualization of the airway tree, the total number of airway branches, statistics on the number of branches at each level, and an assessment of the integrity of the airway tree reconstruction. The third part analyzes airway morphological parameters, including statistical tables and trend charts of average airway wall thickness, average airway diameter, and average wall area percentage for each level of airway, a list and location markings of abnormal airway branches, and a comparative analysis with normal reference ranges. The fourth part assesses small airway function, including a histogram of lung parenchymal density distribution, percentage of low-density areas, indicators of whole-lung ventilation heterogeneity, small airway functional status indicators, and a severity rating of small airway dysfunction (normal, mild, moderate, severe). The fifth part predicts disease progression (if follow-up data is available), including trend charts of airway parameter changes, trend charts of small airway function changes, prediction results for disease progression categories, prediction confidence levels, and individualized treatment recommendations. Part VI contains comprehensive conclusions and recommendations, which are automatically generated by the system based on the analysis results. For example, the patient's airway tree reconstruction is complete, the walls of the small airways from level 5 to level 7 are significantly thickened, the airway diameter is significantly narrowed, the small airway function status index is 68 points (moderate impairment), the disease is predicted to progress rapidly in the next year, and it is recommended to strengthen drug treatment and increase the frequency of follow-up.

[0065] The comprehensive analysis report uses a combination of text and graphics. The charts and graphs adopt a professional medical imaging report style, and the text descriptions are concise and clear, conforming to the reading habits of clinicians. The report supports exporting to PDF format for easy archiving and printing.

[0066] The interactive unit receives user commands and responds by rotating, scaling, slicing, and annotating the 3D visualization of the airway tree. Users can drag the mouse to rotate the 3D airway tree model at any angle, zoom in and out using the scroll wheel, and select specific airway branches to view their detailed morphological parameters. The interactive unit also supports slice viewing; users can specify the position and orientation of the slice plane on the 3D model, and the system automatically generates a cross-sectional image of the airway on that plane, displaying the density distribution of the airway walls, lumen, and surrounding lung parenchyma, facilitating detailed observation of the airway's morphological features. Furthermore, users can annotate and comment on airway areas of interest using the interactive unit, such as marking a specific airway branch as a focus and adding a note indicating significant airway narrowing and recommending bronchoscopy. These annotations and comments are automatically saved and displayed in the report.

[0067] In one embodiment of this invention, the system was clinically validated in the respiratory department of a tertiary hospital, analyzing CT image data from 150 COPD patients. Clinicians' satisfaction with the comprehensive analysis report generated by the system showed that 85% of the doctors considered the information provided in the report comprehensive and valuable, and 90% of the doctors believed that the 3D visualization of the airway tree and interactive operation functions significantly improved diagnostic efficiency. Compared with traditional manual CT image interpretation, using the system of this invention improved the phenotypic classification accuracy of COPD patients from 58% to 75%, the treatment response prediction accuracy from 65% to 82%, and the average diagnosis time from 30 minutes to 15 minutes. These clinical application data fully validate the practicality and effectiveness of the system of this invention in the precision diagnosis and treatment of COPD.

[0068] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A quantitative analysis system for airway changes in chronic obstructive pulmonary disease based on three-dimensional reconstruction, characterized in that, include: The image data acquisition module is used to acquire the patient's chest CT image data; The airway tree reconstruction module is connected to the image data acquisition module and is used to segment the airway in CT images, extract the airway mask and centerline, and realize the three-dimensional reconstruction of the whole lung airway tree down to the 6th or 7th level bronchi based on the multi-scale feature fusion algorithm to generate a three-dimensional airway tree model. The morphological analysis module, connected to the airway tree reconstruction module, is used to calculate the airway wall thickness, airway inner diameter, and wall area percentage of each airway branch based on the three-dimensional model of the airway tree, and generate an airway morphological parameter dataset. The small airway function assessment module is connected to the image data acquisition module and the morphological analysis module. It is used to calculate the small airway function status index and generate a small airway function assessment report based on the lung parenchyma density distribution characteristics, ventilation heterogeneity characteristics and airway morphological parameters. The disease progression prediction module is connected to the morphological analysis module and the small airway function assessment module. It is used to calculate the rate of change of parameters based on follow-up data at multiple time points, output the disease progression trend through the disease progression prediction model, and generate future disease progression prediction results. The results output module is connected to the airway tree reconstruction module, morphological analysis module, small airway function assessment module, and disease progression prediction module to generate a comprehensive report that includes three-dimensional visualization images, morphological parameters, small airway function assessment results, and disease progression prediction results.

2. The quantitative analysis system for airway changes in chronic obstructive pulmonary disease based on three-dimensional reconstruction as described in claim 1, characterized in that, The airway tree reconstruction module includes: The airway segmentation unit is used to preprocess the chest CT image data, and to segment the airway using a three-dimensional convolutional neural network based on deep learning to extract the airway mask. The centerline extraction unit is used to extract the airway centerline based on the airway mask and to determine the topology of the airway tree using a skeletonization algorithm. The multi-scale reconstruction unit is used to perform three-dimensional reconstruction of the airway tree based on the airway centerline and airway mask using a multi-scale feature fusion algorithm, generating a three-dimensional airway tree model containing the 0th level main trachea to the 6th or 7th level bronchus.

3. The quantitative analysis system for airway changes in chronic obstructive pulmonary disease based on three-dimensional reconstruction as described in claim 2, characterized in that, The multi-scale feature fusion algorithm used in the multi-scale reconstruction unit includes: Multi-scale downsampling is performed on the airway mask to generate a first-scale airway feature map, a second-scale airway feature map, and a third-scale airway feature map. Extract local morphological features and global topological features from airway feature maps at various scales; Local morphological features and global topological features at various scales are fused to generate a fused feature map; Based on the fused feature map and airway centerline, a three-dimensional model of the airway tree is reconstructed.

4. The quantitative analysis system for airway changes in chronic obstructive pulmonary disease based on three-dimensional reconstruction as described in claim 1, characterized in that, The morphological analysis module includes: The airway wall thickness calculation unit is used to extract multiple cross-sections along the centerline direction for each airway branch in the three-dimensional model of the airway tree, calculate the airway wall thickness of each cross-section based on the full width half height method, and generate airway wall thickness distribution data. The airway inner diameter calculation unit is used to calculate the airway inner diameter of each airway branch based on multiple cross-sections, and generate airway inner diameter distribution data. The wall area percentage calculation unit is used to calculate the wall area percentage of each airway branch based on airway wall thickness distribution data and airway inner diameter distribution data.

5. The quantitative analysis system for airway changes in chronic obstructive pulmonary disease based on three-dimensional reconstruction as described in claim 4, characterized in that, The full-width half-height method used in the airway wall thickness calculation unit includes: In the 3D model of the airway tree, extract the cross-section of the airway along the direction perpendicular to the centerline of the airway; Density distribution curves are obtained radially on the cross-section of the airway. Determine the maximum and minimum values ​​of the density distribution curve, and calculate the half-height value; Locate the two half-height points on the density distribution curve; Calculate the distance between the two half-height points as the airway wall thickness.

6. The quantitative analysis system for airway changes in chronic obstructive pulmonary disease based on three-dimensional reconstruction as described in claim 1, characterized in that, The small airway function assessment module includes: The lung parenchyma density analysis unit is used to extract lung parenchyma regions based on chest CT image data, calculate lung parenchyma density distribution histogram and density mean, and identify low-density and high-density regions. The ventilation nonuniformity assessment unit is used to segment lung segments from chest CT image data, calculate the density standard deviation and density variation coefficient of each lung segment, and generate ventilation nonuniformity index. The small airway function calculation unit is used to calculate small airway function status indicators based on lung parenchyma density distribution histograms, ventilation non-uniformity indices, and airway morphology parameter datasets.

7. The quantitative analysis system for airway changes in chronic obstructive pulmonary disease based on three-dimensional reconstruction as described in claim 6, characterized in that, The small airway function calculation unit uses the following methods to calculate the small airway function status index: Extract the mean airway wall thickness and mean airway inner diameter of the 5th to 7th grade bronchi; Calculate the percentage of volume in the lung parenchyma that is below the threshold density. Calculate the weighted average of the ventilation nonuniformity index; Based on the mean airway wall thickness, mean airway inner diameter, volume percentage of areas below the threshold density, and weighted average, the functional status indicators of small airways are determined.

8. The quantitative analysis system for airway changes in chronic obstructive pulmonary disease based on three-dimensional reconstruction as described in claim 1, characterized in that, The disease progression prediction module includes: The longitudinal data comparison unit is used to receive the airway morphology parameter dataset and small airway function assessment report at the first time point, as well as the airway morphology parameter dataset and small airway function assessment report at the second time point, and to calculate the rate of change of airway parameters. The function decline trend analysis unit is used to analyze the trends of airway wall thickness change, airway inner diameter change, and small airway function change based on the rate of change of airway parameters. The progression prediction unit is used to input the trends of airway wall thickness change, airway internal diameter change, and small airway function change into the disease progression prediction model to generate future disease progression prediction results.

9. The quantitative analysis system for airway changes in chronic obstructive pulmonary disease based on three-dimensional reconstruction as described in claim 8, characterized in that, The disease progression prediction model is trained based on historical patient data, and the training method includes: We collected multi-time-point airway morphology parameters and small airway function assessment data from multiple patients with chronic obstructive pulmonary disease. The rate of change of airway parameters was extracted as input features, and the disease progression was extracted as labels. A disease progression prediction model is trained using machine learning algorithms, enabling the model to predict future disease progression based on the rate of change in airway parameters.

10. The quantitative analysis system for airway changes in chronic obstructive pulmonary disease based on three-dimensional reconstruction as described in claim 1, characterized in that, The result output module includes: The visualization generation unit is used to generate a 3D visualization image of the airway tree based on the 3D model of the airway tree, and to mark different levels of airway branches with different colors and to highlight abnormal airway areas. The report generation unit integrates airway morphology parameter data, small airway function assessment reports, and disease progression prediction results to generate a comprehensive analysis report containing charts and text descriptions. The interactive operation unit is used to receive user operation commands and respond to the operation commands to rotate, scale, slice, and annotate parameters of the airway tree 3D visualization image.

Images