A dynamic monitoring system for pulmonary nodules in a dose CT scan

Abstract

This invention relates to the field of image registration technology, specifically to a dynamic monitoring system for pulmonary nodules using dose-dependent CT scans. The invention obtains an edge cluster encompassing all lung images based on the positional distances between corresponding contour points on edge lines in all lung images; it obtains multiple sets of isotopic contour points based on the positional distribution of all contour points in different lung images; it obtains the referenceability of each set of isotopic contour points by considering the positional relationship between each set of isotopic contour points and other sets of isotopic contour points with the same number of angles on the lung images, as well as the fluctuation characteristics of the corresponding angles of each set of isotopic contour points, and selects anchor points; it constructs a relative positional model of the nodule region in each lung image based on the positional distribution of anchor points and nodule regions in each lung image; and it achieves lung image registration for dynamic monitoring of pulmonary nodules. This invention improves the accuracy of dynamic monitoring of pulmonary nodules by flexibly registering the patient's lung images.

Description

Technical Field

[0001] This invention relates to the field of image registration technology, and more specifically to a dynamic monitoring system for lung nodules in dose-CT scans. Background Technology

[0002] In the dynamic monitoring of the lungs, low-dose CT is considered a core technology due to its low radiation dose and high sensitivity to lung nodules. By tracking changes in the morphology, volume, and density of nodules through multiple LDCT scans, the malignancy risk of nodules can be effectively assessed, providing key evidence for clinical decision-making.

[0003] In existing technologies, image registration is performed on multiple lung CT scans to observe the development of lung nodules. However, in the dynamic monitoring scenario of lung LDCT, global alignment based on translation and rotation cannot compensate for local deformation caused by lung respiratory motion or patient position differences. The registration error in the lung base region, which is significantly affected by diaphragmatic motion, is still large, resulting in poor dynamic monitoring of lung nodules. Summary of the Invention

[0004] To address the technical problem of large registration errors and poor dynamic monitoring of pulmonary nodules caused by local deformation due to lung respiration or body position differences, the present invention aims to provide a dynamic monitoring system for pulmonary nodules using dose-CT scanning. The specific technical solution adopted is as follows:

[0005] This invention proposes a dynamic monitoring system for pulmonary nodules using dose-CT scanning, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it performs the following steps:

[0006] Acquire multiple lung images from dose-dependent CT scans, the lung images including nodular regions;

[0007] For any lung image, based on the morphological characteristics of the lung image, multiple contour points are obtained for each edge line; based on the positional distribution characteristics of the contour points on each edge line, the angle between each contour point on each edge line and the contour points of different orders is obtained.

[0008] Based on the positional distances between all contour points corresponding to the edge lines in all lung images, all edge lines are clustered to obtain multiple edge line clusters. For any edge line cluster containing all lung images, multiple sets of corresponding contour points are obtained based on the positional distribution of all contour points in different lung images. Based on the positional relationship between each set of corresponding contour points and other sets of corresponding contour points with the same number of angles on the lung images, as well as the fluctuation characteristics of the corresponding angles of each set of corresponding contour points, the referenceability of each set of corresponding contour points is obtained.

[0009] Anchor points are selected based on the referability of corresponding contour points for each number of angles; and a relative position model of nodule regions in each lung image is constructed based on the positional distribution of anchor points and nodule regions in each lung image.

[0010] Based on the morphological differences in the relative positions of nodule regions in different lung images, lung image registration is achieved, enabling dynamic monitoring of lung nodules.

[0011] Furthermore, the method for obtaining the contour points includes:

[0012] For any lung image, an edge detection algorithm is used to obtain multiple edge lines in the lung image. Each edge line is sampled to obtain multiple contour points for each edge line.

[0013] Furthermore, the method for obtaining the order angle includes:

[0014] For any edge line, the inverse cosine function value of the position vector between each contour point and each preceding and following contour point of each order is obtained in turn as the turning angle between each contour point and the preceding and following contour points of different orders.

[0015] The extreme points in the turning angle curve formed by the turning angles between each contour point and the preceding and following contour points of different orders are obtained. The order corresponding to the first extreme point and all previous orders are used to form a reference order range. The turning angles between each contour point on each edge line and the preceding and following contour points of different orders within the reference order range are obtained as the order angles.

[0016] Furthermore, the method for obtaining the edge line cluster includes:

[0017] Obtain the DTW distance of the coordinate sequence of all corresponding contour points between the edge lines, and use it as the positional distance;

[0018] Based on the positional distances between all contour points corresponding to the edge lines in all lung images, DBSCAN clustering is performed on all edge lines to obtain multiple edge line clusters.

[0019] Furthermore, the method for obtaining the co-position contour points includes:

[0020] For any edge cluster containing all lung images, obtain the coordinate sequence of all contour points in different lung images and perform DTW matching to obtain the matching contour points of each contour point. Then, the matching contour points of each contour point in all lung images constitute each set of co-position contour points.

[0021] Furthermore, the method for obtaining the referability includes:

[0022] The mean positional distance between each group of corresponding contour points and other groups of corresponding contour points with the same number of angles on different lung images is obtained as a positional reference coefficient.

[0023] The degree of fluctuation of the sequence formed by the corresponding order angles of each set of contour points is obtained as the order angle fluctuation.

[0024] The ratio of the positional reference coefficient to the stage fluctuation is obtained as the reference value for each group of co-located contour points.

[0025] Furthermore, the method for obtaining the anchor point includes:

[0026] Select the preset number of points with the highest reference value among the corresponding contour points of each angle level as anchor points.

[0027] Furthermore, the method for obtaining the relative position model includes:

[0028] For any lung image, anchor points and nodule regions are used as nodes to form the graph structure. Delaunay triangulation is used to establish connection edges between nodes within the preset node connection range.

[0029] Based on the length characteristics of the connecting lines between different nodes, the intermediary centrality of the connecting edges between corresponding nodes in the graph structure is obtained;

[0030] Obtain the betweenness centrality of the edge connecting two nodes in the graph structure, normalize it, and calculate the sum of the normalized result and the preset bifurcation judgment coefficient as the topological importance of the edge connecting the corresponding nodes; mark the topological importance of the corresponding edge in the graph structure as a relative position model.

[0031] Furthermore, the method for obtaining the betweenness centrality includes:

[0032] The first quantity is the number of pixels with the smallest value on the edge connecting nodes in the graph structure; the second quantity is the number of pixels with the smallest value on the edge connecting other nodes between nodes; the ratio of the first quantity to the second quantity is used as the betweenness centrality of the edge connecting nodes in the graph structure.

[0033] Furthermore, the process of registering lung images based on the morphological differences in the relative position models of nodule regions in different lung images includes:

[0034] Based on the difference in topological importance between each node in each lung image and other nodes in the first lung image, and the position vector of the corresponding node, the local displacement vector of each node in each lung image is used. The difference in topological importance and the position vector are both positively correlated with the local displacement vector.

[0035] The continuous deformation field of the local displacement vector of each node is obtained using the TPS interpolation algorithm.

[0036] The thin-plate spline interpolation algorithm is used to map pixels from multiple lung images onto the first lung image based on a continuous deformation field, thereby achieving registration of the corresponding lung images.

[0037] The present invention has the following beneficial effects:

[0038] This invention obtains multiple contour points on each edge line of each lung image based on the shape characteristics of each lung image, reflecting the anatomical structure; based on the positional distribution characteristics of the contour points on each edge line, it obtains the angle between each contour point on each edge line and contour points of different orders before and after, quantifying the local edge direction and capturing subtle deformations; based on the positional distances between all contour points corresponding to the edge lines in all lung images, it clusters all edge lines to obtain multiple edge line clusters, initially screening out reliable edge regions that can be used for registration; for any edge line cluster containing all lung images, based on the positional distribution of all contour points in different lung images, it obtains multiple sets of isotopic contour points, screening out registration and dynamic... The key control points for tracking are identified. Based on the positional relationship between each group of corresponding contour points and other groups of corresponding contour points with the same number of angles on lung images, and the fluctuation characteristics of the corresponding angles of each group of corresponding contour points, the referenceability of each group of corresponding contour points is obtained. Anchor points are obtained based on the referenceability of all contour points at each order, improving registration accuracy and reducing computational complexity. Based on the positional distribution of anchor points and nodule regions in each lung image, a relative position model of the nodule region in each lung image is constructed, clearly demonstrating the interaction between the nodule and surrounding tissues. Based on the morphological differences in the relative position models of nodule regions in different lung images, lung image registration is achieved, enabling dynamic monitoring of lung nodules. This invention improves the accuracy of dynamic monitoring of lung nodules by performing flexible registration of patients' lung images. Attached Figure Description

[0039] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0040] Figure 1 This is a flowchart illustrating an implementation method of a dose-CT scan-based dynamic monitoring system for lung nodules, as provided in one embodiment of the present invention.

[0041] Figure 2 This is a flowchart illustrating a reference acquisition method provided in one embodiment of the present invention. Detailed Implementation

[0042] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a dose-CT scan-based dynamic monitoring system for lung nodules proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0043] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0044] The following description, in conjunction with the accompanying drawings, details the specific scheme of a dose-CT scan-based dynamic monitoring system for lung nodules provided by this invention.

[0045] Please see Figure 1 The diagram illustrates a flowchart of an implementation method for a dose-CT scan-based dynamic monitoring system for lung nodules, according to an embodiment of the present invention. The method specifically includes:

[0046] Step S1: Acquire multiple lung images from dose CT scans, including nodular regions.

[0047] In embodiments of the present invention, a pulmonary nodule refers to a focal shadow or increased density with a diameter of less than or equal to 2 cm on a chest X-ray or computed tomography (CT) image, usually round or oval in shape. For subsequent dynamic monitoring of the nodule, multiple lung images of the patient need to be acquired. First, the patient's lungs are scanned multiple times using a low-dose CT scanner. Low-risk nodules are usually examined by CT every 6-12 months, while high-risk nodules require more frequent monitoring and multiple lung images are acquired.

[0048] It should be noted that, in order to ensure the image quality of subsequent image processing, all lung images obtained during the dynamic observation of the patient are preprocessed. In one embodiment of the present invention, the preprocessing method is to process the lung images based on a deep learning denoising model to output images of conventional dose quality and eliminate grayscale differences between different devices or reconstruction algorithms. The deep learning denoising model can be RED-CNN or Noise2Noise, etc. The specific means are well known to those skilled in the art and will not be described in detail here.

[0049] It should be noted that, in one embodiment of the present invention, the method for obtaining the nodule region includes: identifying the nodule region in a lung image using a semantic segmentation neural network, inputting a preprocessed lung image, and outputting the nodule region in the lung image. The specific semantic segmentation neural network is a well-known technique to those skilled in the art and will not be described in detail here.

[0050] Step S2: For any lung image, based on the morphological characteristics of the lung image, obtain multiple contour points for each edge line; based on the positional distribution characteristics of the contour points on each edge line, obtain the angle between each contour point on each edge line and the contour points of different orders.

[0051] Lung images contain information not only about nodular areas but also about other structural areas of the lung. By analyzing contour points, edge information can be extracted to reflect the shape characteristics of the lung images, which helps to analyze the deformation characteristics of tissues in lung images. For any lung image, multiple contour points of each edge line can be obtained based on the morphological characteristics of the lung images.

[0052] Preferably, in one embodiment of the present invention, the method for obtaining contour points includes:

[0053] For any lung image, an edge detection algorithm is used to obtain multiple edge lines in the lung image. Each edge line is sampled to obtain multiple contour points for each edge line.

[0054] It should be noted that, in the embodiments of the present invention, edge lines can be sampled by methods such as uniform sampling, random sampling, or interval sampling. The specific edge detection and sampling are technical means well known to those skilled in the art, and will not be described in detail here.

[0055] Because the anatomical structure of the lungs exhibits fractal characteristics, the positional distribution of these contour points can be analyzed to quantify their topological invariance across different order contour points. First-order angles reflect local curvature, while higher-order angles reflect anatomical structures over a relatively larger area, which is more helpful in distinguishing between fixed and dynamic anatomical regions. Based on the positional distribution characteristics of contour points along each edge line, the order angles between each contour point on each edge line and subsequent contour points of different orders can be obtained.

[0056] Preferably, in one embodiment of the present invention, the method for obtaining the order angle includes:

[0057] For any edge line, the inverse cosine function value of the position vector between each contour point and each preceding and following contour point of each order is obtained in turn as the turning angle between each contour point and the preceding and following contour points of different orders.

[0058] Make an example, with the first Using the nth contour point as the center, and utilizing the coordinates of the contour points in the sequence, calculate the difference between the coordinates of each contour point and the coordinates of the previous contour point, and use this difference as the position vector to obtain the nth contour point. The contour point and the first order after the th contour point The vector of positions between contour points, denoted as ; Get the The contour points and the first order before the first... The position vector between contour points, as... ; obtained the first The inverse cosine function value of the position vector between each contour point and the first-order preceding and following contour points. As the first The turning angle between each contour point and its corresponding first-order front and rear contour points ,but Indicates the first The turning angle between the first contour point and the second-order front and rear contour points By analyzing each order sequentially, we obtain the... The turning angle between each contour point and its corresponding Nth-order front and rear contour points .

[0059] The extreme points in the turning angle curve formed by the turning angles between each contour point and the preceding and following contour points of different orders are obtained. The order corresponding to the first extreme point and all previous orders are used to form a reference order range. The turning angles between each contour point on each edge line and the preceding and following contour points of different orders within the reference order range are obtained as the order angles.

[0060] It should be noted that, in one embodiment of the present invention, the steering angle curve contains front and rear steering angles that are greater than or less than each steering angle, and the corresponding steering angle is taken as the extreme point; in other embodiments of the present invention, the extreme points of the steering angle curve can also be obtained by Newton's method or the like, and the specific means are well known to those skilled in the art, and will not be described in detail here.

[0061] Step S3: Based on the positional distances between all contour points corresponding to the edge lines in all lung images, cluster all edge lines to obtain multiple edge line clusters; for any edge line cluster containing all lung images, obtain multiple sets of co-located contour points based on the positional distribution of all contour points in different lung images; based on the positional relationship between each set of co-located contour points and other sets of co-located contour points with the same number of angles on the lung images, and the fluctuation characteristics of the angles corresponding to each set of co-located contour points, obtain the referenceability of each set of co-located contour points.

[0062] By analyzing the positional distances between all corresponding contour points on different edge lines, the spatial differences between corresponding contour points on different edge lines are reflected. Clustering can group edge lines with similar positional distributions into one category, which is more likely to analyze cases of stable and changing edge lines. Based on the positional distances between all corresponding contour points on all lung images, all edge lines are clustered to obtain multiple edge line clusters.

[0063] Preferably, in one embodiment of the present invention, the method for obtaining edge line clusters includes:

[0064] Obtain the DTW distance of the coordinate sequence of all corresponding contour points between the edge lines, and use it as the positional distance;

[0065] Based on the positional distances between all contour points corresponding to the edge lines in all lung images, DBSCAN clustering is performed on all edge lines to obtain multiple edge line clusters.

[0066] It should be noted that, in the embodiments of the present invention, the DTW distance is used to measure the positional distance of the coordinate sequence formed by all contour points between edge lines. The larger the DTW distance, the larger the positional distance, the smaller the possibility of representing edges at the same position, and the smaller the possibility of clustering into a feature cluster. The specific DTW distance method is a technical means well known to those skilled in the art, and will not be described in detail here.

[0067] The lung edges in all images exhibit almost identical shapes in the same areas, representing regions without significant pathological variations. If a region contains similar areas across all lung images, it facilitates registration of identical lung images. For any cluster of edge lines encompassing all lung images, multiple sets of isotopic contour points are obtained based on the positional distribution of all contour points in different lung images.

[0068] Preferably, in one embodiment of the present invention, the method for obtaining co-located contour points includes:

[0069] For any edge cluster containing all lung images, obtain the coordinate sequence of all contour points in different lung images and perform DTW matching to obtain the matching contour points of each contour point. Then, the matching contour points of each contour point in all lung images constitute each set of co-position contour points.

[0070] It should be noted that DTW finds the optimal alignment path through dynamic programming, matching the coordinate sequences of corresponding contour points between different lung images to determine the matching contour point for each contour point.

[0071] Based on the positional relationship between each group of locative contour points and other groups of locative contour points with the same number of angles on lung images, as well as the fluctuation characteristics of the corresponding angles of each group of locative contour points, the referenceability of each group of locative contour points is obtained.

[0072] Preferably, in one embodiment of the present invention, the method for obtaining referability can be found in [reference needed]. Figure 2 It illustrates a flowchart of a referential acquisition method, including:

[0073] Step S201: Obtain the average positional distance between each group of co-located contour points and other groups of co-located contour points with the same number of angles on different lung images, and use it as a positional reference coefficient.

[0074] By combining and analyzing other contour points with the same number of relative angles, the greater the positional distance, the more likely it is to be the analysis of different contour points, thus avoiding the clustering of contour points and facilitating a more comprehensive analysis of lung location.

[0075] Step S202: Obtain the degree of fluctuation of the sequence formed by the corresponding order angles of each set of contour points, as the order angle fluctuation.

[0076] It should be noted that, in one embodiment of the present invention, the standard deviation is used to process all sequences to represent the degree of fluctuation. The larger the standard deviation, the greater the degree of fluctuation and the greater the trend of contour change of each group of corresponding contour points. The smaller the standard deviation, the smaller the degree of fluctuation and the more consistent the trend of contour change of each group of corresponding contour points. In other embodiments of the present invention, the degree of fluctuation can be represented by variance and range, etc. The specific means are well known to those skilled in the art and will not be described in detail here.

[0077] Step S203: Obtain the ratio of the position reference coefficient to the step angle fluctuation as the referenceability of each set of co-located contour points. In one embodiment of the present invention, the formula for referenceability is expressed as:

[0078] :

[0079] in, Indicates the first Referenceability of co-located contour points; Indicates the first The mean positional distance between a group of corresponding contour points and other groups of corresponding contour points with the same number of angles on different lung images, i.e., the positional reference coefficient; Indicates the first The degree of fluctuation of the sequence formed by the corresponding order angles of the same contour points is called the order angle fluctuation.

[0080] In the formula for referenceability, Add 0.01 to avoid the formula being meaningless when the denominator is 0; the first... The smaller the fluctuation of the sequence formed by the corresponding angles of the same set of contour points, the smaller the change of the sequence formed by the corresponding angles of each set of contour points, that is, the more consistent the edge direction, the greater the reference value; the second The greater the average distance between the corresponding contour points in one group and other corresponding contour points in other groups with the same number of angles on different lung images, the greater the distance between different corresponding contour points, the wider the distribution of contour points, the more lung features are represented, and the greater the reference value.

[0081] Step S4: Based on the referability of the corresponding contour points under each number of angles, select anchor points; based on the positional distribution of anchor points and nodule regions in each lung image, construct a relative position model of nodule regions in each lung image.

[0082] In multiple lung images of a patient, the conditions of each scan result in complex variations between the images. Images less affected by CT scans are more likely to act as anchor points, exhibiting relative stability in their position within the lung images and showing little change over multiple CT scans. These anchor points are more likely to reflect anatomical landmarks of the lungs and are thus selected. Anchor points are further selected based on the referenceability of corresponding contour points for each number of angles.

[0083] Preferably, in one embodiment of the present invention, the method for obtaining the anchor point includes:

[0084] Select the preset number of points with the highest reference value among the corresponding contour points of each angle level as anchor points.

[0085] It should be noted that, in one embodiment of the present invention, the preset number is 10% of the total number of contour points; in other embodiments of the present invention, the size of the preset number can be set according to specific circumstances, and will not be limited or elaborated here.

[0086] Since the anchor point has a relatively stable relative position in lung images and usually does not change significantly due to multiple CT scans, a relative position model of the nodule region in each lung image is constructed based on the positional distribution of the anchor point and nodule region in each lung image.

[0087] Preferably, in one embodiment of the present invention, the method for obtaining the relative position model includes:

[0088] For any lung image, anchor points and nodule regions are used as nodes to form the graph structure. Delaunay triangulation is used to establish connection edges between nodes within the preset node connection range.

[0089] It should be noted that, in one embodiment of the present invention, a U-Net neural network is used to perform multi-label segmentation on CT lung images, outputting the peripheral zones of the hilar region and subpleural region within 2 cm of the mediastinum; for nodes in the hilar region, a range where the Euclidean distance between nodes is less than or equal to 15 mm is used as the preset node connection range; for nodes in the peripheral zone, a range where the Euclidean distance between nodes is less than or equal to 30 mm is used as the preset node connection range. In other embodiments of the present invention, the size of the preset node range can be set according to specific circumstances, and is not limited or elaborated here.

[0090] Based on the length characteristics of the connecting lines between different nodes, the intermediary centrality of the connecting edges between corresponding nodes in the graph structure is obtained;

[0091] Preferably, in one embodiment of the present invention, the method for obtaining the betweenness centrality includes:

[0092] The first quantity is the number of pixels with the smallest value on the edge connecting nodes in the graph structure; the second quantity is the number of pixels with the smallest value on the edge connecting other nodes between nodes; the ratio of the first quantity to the second quantity is used as the betweenness centrality of the edge connecting nodes in the graph structure.

[0093] Obtain the betweenness centrality of the edge connecting two nodes in the graph structure, normalize it, and calculate the sum of the normalization result and the preset judgment coefficient as the topological importance of the edge connecting the corresponding nodes; mark the topological importance of the corresponding edge in the graph structure as a relative position model.

[0094] In one embodiment of the present invention, the formula for topological importance for any two nodes is expressed as:

[0095] ;

[0096] in, Indicates the topological importance of the edges connecting nodes; This indicates the centrality of the edges connecting nodes in the graph structure. This represents the maximum degree of betweenness centrality of the edges connecting all nodes in the graph structure. This represents the preset bifurcation judgment coefficient.

[0097] In the formula for topological importance, This represents the ratio of the intermediate centrality of the edges connecting nodes in the graph structure to the maximum intermediate centrality of all edges connecting nodes in the graph structure. In other words, it normalizes the intermediate centrality of the edges connecting nodes. The larger the ratio, the greater the intermediate centrality of the edges connecting nodes, the more other edges connecting nodes pass through, and the greater the topological importance of the edges connecting nodes.

[0098] It should be noted that, in one embodiment of the present invention, the method for obtaining the preset fork judgment coefficient is as follows: if the connecting line between nodes is at a fork, the preset fork judgment coefficient is 1; otherwise, the preset fork judgment coefficient is 0.

[0099] Step S5: Based on the morphological differences in the relative position models of nodule regions in different lung images, register the lung images and dynamically monitor the lung nodules.

[0100] Because local areas will shift in the images due to changes in body position and respiration, the relative positions of different lung images will also change to some extent. By analyzing the positional distribution between nodes in different lung images, we can analyze the relative shift in lung images.

[0101] Preferably, in one embodiment of the present invention, the method for obtaining the local displacement vector includes:

[0102] Based on the difference in topological importance between each node in each lung image and other nodes in the first lung image, and the position vector of the corresponding node, the local displacement vector of each node in each lung image is used. The difference in topological importance and the position vector are both positively correlated with the local displacement vector.

[0103] In one embodiment of the present invention, the formula for the local displacement vector is expressed as:

[0104] ;

[0105] in, This indicates that each lung image is relative to the first lung image in terms of its position. Local displacement vectors of each node; Indicates the first A relative position model corresponding to lung images; Indicates the first The relative position model of the lung images relative to the first Fibib image; Indicates the first The first lung image and the second lung image The difference in average topological importance between each node and other nodes; Indicates the first The first lung image and the second lung image A vector formed by the coordinates of each node.

[0106] In the formula for the local displacement vector, for the first... In the first lung image, the lung image is compared to the second lung image. The greater the difference in average topological importance between a node and other nodes, the greater the relative distance between node distributions, the greater the relative deviation, and the larger the coordinate vector is adjusted, the larger the local displacement vector of the corresponding node will be.

[0107] The TPS interpolation algorithm is used to obtain the continuous deformation field of the local displacement vector of each node; the thin plate spline interpolation algorithm is used to map the pixels in multiple lung images to the first lung image based on the continuous deformation field, thereby achieving the registration of the corresponding lung images.

[0108] The area and size of each nodule are extracted from all registered lung images and arranged in chronological order. The area and size of the nodules are then predicted using the ARIMA algorithm, enabling dynamic monitoring of the lung nodules. The specific methods used are well-known to those skilled in the art and will not be elaborated upon here.

[0109] In summary, this invention obtains an edge cluster encompassing all lung images based on the positional distances between corresponding contour points across all edge lines in all lung images; it obtains multiple sets of isotopic contour points based on the positional distribution of all contour points in different lung images; it obtains the referenceability of each set of isotopic contour points and filters out anchor points based on the positional relationship between each set of isotopic contour points and other sets of isotopic contour points with the same number of angles on the lung images, as well as the fluctuation characteristics of the corresponding angles of each set of isotopic contour points; it constructs a relative positional model of the nodule region in each lung image based on the positional distribution of anchor points and nodule regions in each lung image; and it achieves lung image registration for dynamic monitoring of lung nodules. This invention improves the accuracy of dynamic monitoring of lung nodules by performing flexible registration of patient lung images.

[0110] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0111] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A dynamic monitoring system for pulmonary nodules using dose-CT scanning, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it performs the following steps: Acquire multiple lung images from dose-dependent CT scans, the lung images including nodular regions; For any lung image, based on the morphological characteristics of the lung image, multiple contour points are obtained for each edge line; based on the positional distribution characteristics of the contour points on each edge line, the angle between each contour point on each edge line and the contour points of different orders is obtained. Based on the positional distances between all contour points corresponding to the edge lines in all lung images, all edge lines are clustered to obtain multiple edge line clusters. For any edge line cluster containing all lung images, multiple sets of corresponding contour points are obtained based on the positional distribution of all contour points in different lung images. Based on the positional relationship between each set of corresponding contour points and other sets of corresponding contour points with the same number of angles on the lung images, as well as the fluctuation characteristics of the corresponding angles of each set of corresponding contour points, the referenceability of each set of corresponding contour points is obtained. Anchor points are selected based on the referability of corresponding contour points for each number of angles; and a relative position model of nodule regions in each lung image is constructed based on the positional distribution of anchor points and nodule regions in each lung image. Based on the morphological differences in the relative positions of nodule regions in different lung images, lung image registration is achieved to dynamically monitor lung nodules. The method for obtaining the order angle includes: For any edge line, the inverse cosine function value of the position vector between each contour point and each preceding and following contour point of each order is obtained in turn as the turning angle between each contour point and the preceding and following contour points of different orders. The extreme points in the turning angle curve formed by the turning angles between each contour point and the preceding and following contour points of different orders are obtained. The order corresponding to the first extreme point and all previous orders are used to form a reference order range. The turning angles between each contour point on each edge line and the preceding and following contour points of different orders within the reference order range are obtained as the order angles.

2. The dynamic monitoring system for lung nodules using dose-CT scanning according to claim 1, characterized in that, The method for obtaining the contour points includes: For any lung image, an edge detection algorithm is used to obtain multiple edge lines in the lung image. Each edge line is sampled to obtain multiple contour points for each edge line.

3. The dynamic monitoring system for pulmonary nodules using dose-controlled CT scanning according to claim 1, characterized in that, The method for obtaining the edge line cluster includes: Obtain the DTW distance of the coordinate sequence of all corresponding contour points between the edge lines, and use it as the positional distance; Based on the positional distances between all contour points corresponding to the edge lines in all lung images, DBSCAN clustering is performed on all edge lines to obtain multiple edge line clusters.

4. The dynamic monitoring system for pulmonary nodules using dose-CT scanning according to claim 1, characterized in that, The method for obtaining the co-position contour points includes: For any edge cluster containing all lung images, obtain the coordinate sequence of all contour points in different lung images and perform DTW matching to obtain the matching contour points of each contour point. Then, the matching contour points of each contour point in all lung images constitute each set of co-position contour points.

5. The dynamic monitoring system for pulmonary nodules using dose-CT scanning according to claim 1, characterized in that, The methods for obtaining referability include: The mean positional distance between each group of corresponding contour points and other groups of corresponding contour points with the same number of angles on different lung images is obtained as a positional reference coefficient. The degree of fluctuation of the sequence formed by the corresponding order angles of each set of contour points is obtained as the order angle fluctuation. The ratio of the positional reference coefficient to the stage fluctuation is obtained as the reference value for each group of co-located contour points.

6. The dynamic monitoring system for pulmonary nodules using dose-CT scanning according to claim 1, characterized in that, The method for obtaining the anchor point includes: Select the preset number of points with the highest reference value among the corresponding contour points of each angle level as anchor points.

7. The dynamic monitoring system for pulmonary nodules using dose-CT scanning according to claim 1, characterized in that, The method for obtaining the relative position model includes: For any lung image, anchor points and nodule regions are used as nodes to form the graph structure. Delaunay triangulation is used to establish connection edges between nodes within the preset node connection range. Based on the length characteristics of the connecting lines between different nodes, the intermediary centrality of the connecting edges between corresponding nodes in the graph structure is obtained; Obtain the betweenness centrality of the edge connecting two nodes in the graph structure, normalize it, and calculate the sum of the normalized result and the preset bifurcation judgment coefficient as the topological importance of the edge connecting the corresponding nodes; mark the topological importance of the corresponding edge in the graph structure as a relative position model.

8. A dynamic monitoring system for pulmonary nodules using dose-controlled CT scanning according to claim 7, characterized in that, The method for obtaining the betweenness centrality includes: The first quantity is the number of pixels with the smallest value on the edge connecting nodes in the graph structure; the second quantity is the number of pixels with the smallest value on the edge connecting other nodes between nodes; and the ratio of the first quantity to the second quantity is the betweenness centrality of the edge connecting nodes in the graph structure.

9. A dynamic monitoring system for pulmonary nodules using dose-dependent CT scanning according to claim 7, characterized in that, The method of registering lung images based on the morphological differences in the relative position models of nodule regions in different lung images includes: Based on the difference in topological importance between each node in each lung image and other nodes in the first lung image, and the position vector of the corresponding node, the local displacement vector of each node in each lung image is used. The difference in topological importance and the position vector are both positively correlated with the local displacement vector. The continuous deformation field of the local displacement vector of each node is obtained using the TPS interpolation algorithm. The thin-plate spline interpolation algorithm is used to map pixels from multiple lung images onto the first lung image based on a continuous deformation field, thereby achieving registration of the corresponding lung images.

Images