Method and system for evaluating consistency of multiple / time blood vessel segmentation results

Abstract

The invention provides a method and system for evaluating consistency of multiple / secondary blood vessel segmentation results, and the method comprises the steps: reading original data of a blood vessel data set and multiple / secondary blood vessel segmentation result matrixes, and randomly selecting two blood vessel segmentation result matrixes to calculate a union set, an intersection set and a symmetry difference; selecting any point in the symmetric difference, simulating a path diffusion process by using a convolution kernel to perform iterative growth, obtaining a spherical set under the constraint of a union set range and a given geodesic distance, and calculating an intersection of the spherical set and a union set of the blood vessel segmentation result matrixes and an intersection of the intersection and two / time blood vessel segmentation result matrixes; performing calculation and statistical analysis on each point to obtain a statistical result, and mapping the statistical result back to the original data and two / time blood vessel segmentation result matrixes; and thus, a statistical result of any two / time blood vessel segmentation result matrixes is obtained, and a visual processing result is output. The method can be used for measuring the performance of the blood vessel segmentation algorithm or comparing the differences among the blood vessel segmentation algorithms, and accelerates the development and verification process of the blood vessel segmentation algorithm.

Description

Technical Field

[0001] The present invention relates to the technical field of medical image processing, and in particular to a method and system for evaluating the consistency of multiple / sub-blood vessel segmentation results. Background Art

[0002] Vascular morphological changes are associated with a variety of pathological features, so vascular segmentation is crucial for biomedical image analysis. However, due to the complexity of vascular structure, anatomical differences between patients, scarcity of annotated public datasets, and variability in image quality, achieving accurate vascular segmentation is still challenging. From the perspective of the development path of fully automatic vascular segmentation methods, high-quality vascular segmentation result evaluation methods are important and urgently need to be developed, because judging the quality of annotation / segmentation results is an unavoidable sub-problem in the vascular segmentation process. Whether it is manual annotation or other developed vascular segmentation methods, their annotation / segmentation results need to be objectively compared, verified, and evaluated to promote the development of vascular segmentation methods.

[0003] A classic approach is to compare the consistency / difference between the vascular segmentation results predicted by a certain algorithm and the "gold standard (ground truth)" data set manually annotated by experts, so as to measure the performance of the vascular segmentation algorithm. The current objective evaluation methods for annotation / segmentation results still mainly focus on characterizing and comparing the consistency or inconsistency of multiple / time segmentation results as a whole. Typical representatives of this type of global evaluation methods / indicators include: Dice Similarity Coefficient (DSC), Intersection over Union (IoU), Hausdorff Distance (HD) and F1 score. Taking the intersection-over-union ratio as an example, set X is the first / second segmentation result obtained from a certain blood vessel data, in which 120 voxels are marked as blood vessels, and set Y is the second / second segmentation result obtained using the same or different segmentation method, in which 100 voxels are marked as blood vessels. There are 80 overlapping voxels in the areas marked as blood vessels in sets X and Y, then |X∪Y|=140, |X∩Y|=80, IoU=|X∩Y| / |X∪Y|=80 / 140=0.571. Here, IoU=0.571 describes the overall overlap between sets X and Y, that is, the global consistency / inconsistency of the two / second blood vessel segmentation results. Taking the Hausdorff distance (HD) as an example, this indicator reflects the maximum value of the Euclidean distance between the "gold standard" set and the segmentation algorithm prediction set in the entire 3D vascular segmentation result. This means that although this indicator can tell us the maximum deviation between the predicted result and the gold standard result, it does not provide the specific local spatial location where this deviation occurs or the spatial distribution of the deviation. This type of global evaluation method is intuitive, but is more susceptible to subtle errors in thicker vascular branches, but is insensitive to obvious errors in thinner vascular branches. But in fact, in the task of vascular segmentation, it is more difficult and important to accurately segment thinner vascular branches than thicker vascular branches.

[0004] Although the need to understand the spatial distribution of vessel segmentation errors has always existed, only a few methods can partially meet this need. For example, a new function to measure the geometric topological loss of the segmentation result calculates the radial distance of the voxel based on the tubular centerline and Euclidean distance, giving a greater weight to the voxels close to the centerline, achieving a more granular segmentation error characterization / control. Further improvements are made on this method, and the vessel cross-sectional size and voxel radial distance are calculated based on the vessel boundary and Euclidean distance, giving a greater weight to thinner vessel branches and voxels close to the vessel boundary, so that the thin vessel branches are also better segmented. However, the accurate determination of the vessel centerline and vessel boundary in a complex vascular network is a challenge in itself, especially in medium and small vessels (the cross section of the vessel has fewer voxels). In addition, the calculation of the centerline and edge in the above methods all rely on the referenced "gold standard" dataset, assuming that the "gold standard" dataset is accurate. But in fact, the difference in annotation personnel and annotation methods will make the "gold standard" dataset very easy to change, further limiting the practicality of such methods. If there are large differences in the results of multiple / sub-segmentations, the center lines and boundaries are significantly different, which may also make this method inapplicable.

[0005] The current vascular segmentation algorithm still has some factual defects and deficiencies. First, the error has different effects on blood vessels of different sizes. For example, when there are 10 voxel errors in blood vessels of different sizes at the same time, the error of small and medium blood vessels will affect the accuracy of the segmentation results more than that of large blood vessels, that is, the weight of the error impact is unbalanced in blood vessels of different sizes. Secondly, many vascular segmentation algorithms have been developed in the field of image processing, but the horizontal comparison between different algorithms is mainly based on the above-mentioned global / overall evaluation of the consistency of multiple / sub-segmentation results. This leads to users knowing that there are errors between multiple / sub-segmentation results, but they are not clear about the distribution of errors with blood vessel size. Finally, fully describing the distribution of errors in vascular segmentation results with changes in blood vessel size is an important basis for vascular image processing. However, the current small number of local methods for describing segmentation errors still cannot fully meet the needs, and rely on the technical solutions of the center line and blood vessel boundary of the "gold standard" data set, and the complexity of the solution is relatively high. Summary of the invention

[0006] Purpose of the invention: The first purpose of the present invention is to provide a method for evaluating the consistency of multiple / sub-vessel segmentation results that can quantify the differences between multiple / sub-vessel segmentation results of the same vascular data set; the second purpose is to provide a system for evaluating the consistency of multiple / sub-vessel segmentation results corresponding to the above-mentioned evaluation method.

[0007] Technical solution: A method for evaluating the consistency of multiple / sub-vessel segmentation results, comprising the following steps:

[0008] (1) reading the original data and multiple / sub-vascular segmentation result matrices of the same vascular data set, and selecting any two / sub-vascular segmentation result matrices s1 and s2;

[0009] (2) Calculate the union V, intersection S and symmetric difference Δs of s1 and s2;

[0010] (3) Select any point p∈Δs and perform iterative growth in a way that the convolution kernel simulates the path diffusion process. Under the geometric constraints of the two / times blood vessel segmentation result matrices s1 and s2 and the given geodesic distance r, calculate the spherical set B. The center of the spherical set B is point p, and the radius of the sphere is the geodesic distance r. Calculate the intersection B of the spherical set B and the union V of s1 and s2. V , and calculate B V The intersection B with s1 and s2 s1 , B s2 ;

[0011] (4) Repeat step (3) and output each point p and its corresponding B V , B s1 , B s2 , perform statistical analysis to obtain the proportion of blood vessels of different sizes and the statistical results of the difference between the two / sub-blood vessel segmentation result matrices s1 and s2, output the statistical results and map them back to the original data and the two / sub-blood vessel segmentation result matrices s1 and s2;

[0012] (5) Repeat the above steps to obtain statistical results of any two blood vessel segmentation result matrices among the multiple blood vessel segmentation result matrices, and output the visualization processing results for the statistical results of any two blood vessel segmentation result matrices.

[0013] Specifically, the blood vessel segmentation result matrix is ​​a two-dimensional or three-dimensional binary image matrix, in which element 1 indicates that the voxel point belongs to the blood vessel region, and element 0 indicates that the voxel point does not belong to the blood vessel region.

[0014] Specifically, in step (3), iterative growth in a manner of simulating a path diffusion process with a convolution kernel includes: applying the convolution kernel to the union V of the vascular segmentation result matrix by iteratively performing a two-dimensional or three-dimensional convolution operation, and in each iteration, using the convolution kernel as a weight filter to simulate the diffusion of the vascular region, and the convolution iteration stops until the set sphere radius r is reached. The above-mentioned convolution kernel is a kernel function whose weight decreases as the sphere radius r increases.

[0015] Specifically, in step (3), the geometric constraints of the two / sub-vascular segmentation result matrices and the geodesic distance include: the growable area of ​​the convolution kernel iteration is limited to the union V of the two / sub-vascular segmentation result matrices s1 and s2, and the sphere radius r is the shortest path length from one point to another in the vascular area in the vascular segmentation result, and the shortest path length is constrained according to the geodesic distance.

[0016] Specifically, in step (4), the statistical analysis includes: calculating the set B corresponding to each point p V , B s1 , B s2 , and count the different values ​​of B V , B s1 , B s2 The number of points p corresponding to the modulus of the set B V The modulus is the size of the blood vessel at the corresponding point p, and the set B s1 , B s2 The modulus of is the difference between the two / sub-vascular segmentation result matrices s1 and s2 at point p.

[0017] Specifically, the visualization process includes: taking set B V The modulus and the corresponding number of points p are used as the horizontal and vertical coordinates to draw the proportion of blood vessels of different sizes; V The number of points p after weighting by the modulus and the corresponding given weight factor is used as the horizontal and vertical coordinates to draw the distribution diagram of the difference between the two / times blood vessel segmentation result matrices with respect to the blood vessel size. The above weight factor is B s1 , B s2 The absolute value of the difference in modulus divided by the corresponding B V The numerical value obtained by the modulus; according to the coordinates of each point p, the parts where the two / sub-vascular segmentation result matrices have differences are traced and marked, and visualized in a two-dimensional image or a three-dimensional model.

[0018] The present invention also provides a system for evaluating the consistency of multiple / sub-vascular segmentation results, comprising:

[0019] Data input module: used to read the original data of the same blood vessel data set and multiple / sub-blood vessel segmentation result matrices, and select any two / sub-blood vessel segmentation result matrices s1 and s2 from the multiple / sub-blood vessel segmentation result matrices;

[0020] Matrix processing module: used to calculate the union V, intersection S and symmetric difference Δs of s1 and s2;

[0021] Convolution iteration module: select any point p∈Δs, and iterate the growth by simulating the path diffusion process with the convolution kernel. Under the geometric constraints of the two / times blood vessel segmentation result matrices s1 and s2 and the given geodesic distance r, calculate the spherical set B, the center of which is point p and the radius of which is the geodesic distance r. Calculate the intersection B of the spherical set B and the union V of s1 and s2 V , and calculate B V The intersection B with s1 and s2 s1 , B s2 ;

[0022] Statistical analysis module: used to repeatedly run the convolution iteration module and output each point p and its corresponding B V , B s1 , B s2 , perform statistical analysis to obtain the proportion of blood vessels of different sizes and the statistical results of the difference between the two / sub-blood vessel segmentation result matrices s1 and s2, output the statistical results and map them back to the original data and the two / sub-blood vessel segmentation result matrices s1 and s2;

[0023] Visualization output module: used to repeatedly run the above submodules to obtain the statistical results of any two / times of the multiple / times blood vessel segmentation result matrices, and output the visualization processing results for the statistical results of any two / times of the blood vessel segmentation result matrices.

[0024] Beneficial effects: Compared with the prior art, the present invention has the following significant effects:

[0025] (1) The evaluation method provided by the present invention can evaluate the difference of multiple / multiple segmentation results with finer granularity based on the local blood vessel area / volume, characterize the spatial distribution of the error of blood vessel segmentation results with the size of blood vessels, and provide a complementary solution / theory for global evaluation methods such as Dice similarity coefficient (DSC), intersection over union (IoU), Hausdorff distance (HD) and F1 score. First, the statistical results of this method can reflect the proportion of blood vessels of different sizes in a certain blood vessel data, which is helpful for the developers of blood vessel segmentation algorithms or users in the field of clinical medicine to clearly understand the characteristics of the blood vessel size proportion of the segmented blood vessel network, such as recognizing the relative proportion of large, medium and small blood vessels. Second, this method can clarify the size of the differences in blood vessel segments of different sizes of multiple segmentation results, such as clarifying which part of the blood vessel size of large, medium and small blood vessels has the largest error during the segmentation process. The provision of this new statistical data has important theoretical value for accurately evaluating algorithm performance and subsequent algorithm development.

[0026] (2) The evaluation method provided by the present invention has significant practical value in vascular image post-processing and radiology practice. This method can be used to evaluate the quality of manual annotation of vascular images. Since the experience and habits of each annotator are different, it is necessary to evaluate the reliability of the quality of vascular annotation. By evaluating the consistency of multiple / times of manual vascular annotation results, a better gold standard data set can be further improved and created. It is particularly important to describe the spatial distribution of the error of vascular segmentation results with the thickness of the blood vessels. This method can also realize the error tracing in the segmented image and the original vascular image, and the user can intuitively understand the spatial distribution of the difference between multiple / times of segmentation results. Similarly, the comparison between the results obtained by different vascular segmentation algorithms also requires this method, so as to clarify the error between algorithms or between the segmentation results of a certain algorithm and the gold standard, so as to facilitate more targeted research and development. For example, in this method, a higher loss function weight is set for the vascular position that is less consistent with the gold standard data set, thereby improving the accuracy of the segmentation algorithm.

[0027] (3) The method for calculating the thickness of blood vessels in the present invention is based on the local blood vessel area / volume and geodesic distance constraints, rather than on the blood vessel centerline / blood vessel boundary and Euclidean distance. At the same time, the characterization of local errors uses weighted calculations of blood vessel thickness based on local blood vessel volume, solving the problem that errors of the same number of voxels have a greater negative impact on smaller blood vessels. Based on the local blood vessel area / volume and geodesic distance constraints, it can be ensured that the calculation of blood vessel thickness is based on multiple / times results at the same time, rather than relying on the centerline or blood vessel boundary of a single result, thereby avoiding related calculation errors. This also means that the present method does not require that one of the two / times blood vessel segmentation results to be compared is a gold standard data set or limit the closeness / similarity of multiple / times segmentation results. Even if there are large differences between the two / times segmentation results, the present method can accurately and intuitively reflect this situation. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 4 is a flowchart of a method for evaluating the consistency of multiple / time blood vessel segmentation results according to Example 1 of the present invention.

[0029] Figure 2 4 is a schematic diagram of the principle of the method for evaluating the consistency of multiple / sub-blood vessel segmentation results according to Example 1 of the present invention.

[0030] Figure 3 This is a visualization result diagram of the method for evaluating the consistency of multiple / sub-blood vessel segmentation results according to Example 2 of the present invention. DETAILED DESCRIPTION

[0031] The present invention is further described below in conjunction with the accompanying drawings.

[0032] Example 1

[0033] See also Figure 1 As shown, this embodiment takes two / times blood vessel segmentation results as an example to illustrate and provides a method for evaluating the consistency of multiple / times blood vessel segmentation results, including the following steps:

[0034] (1) Read the original data of the same blood vessel dataset and two / times blood vessel segmentation result matrices s1 and s2.

[0035] The blood vessel segmentation result matrix is ​​a two-dimensional or three-dimensional binary image matrix, that is, a 01 matrix. Element 1 in the binary image matrix indicates that the voxel point belongs to the blood vessel region, and element 0 indicates that the voxel point does not belong to the blood vessel region.

[0036] (2) Calculate the union of s1 and s2 V = s1∪s2 and the intersection s interest =s1∩s2 and symmetric difference Δs=V∩(s interest )^c=(s1∪s2)\(s1∩s2).

[0037] (3) Select any point p∈Δs, perform two-dimensional or three-dimensional convolution operations iteratively, and apply the convolution kernel to the union of the vascular segmentation result matrix V=s1∪s2. In each iteration, the convolution kernel is used as a weight filter to simulate the diffusion of the vascular area. When the values ​​of some areas exceed 1 during the convolution iteration, these values ​​will be reset to 1 to ensure that the results remain within the valid binary range. The convolution iteration stops until the set spherical radius r is reached. Define and calculate the spherical set B(p,r), the center of which is point p and the spherical radius is the geodesic distance r. Calculate the intersection B of the spherical set B and the union V of s1 and s2 V (p,r), and calculate B V The intersection of (p,r) and s1 and s2

[0038] The growable region of the above convolution kernel iteration is limited to the union V = s1 ∪ s2 of the two / times blood vessel segmentation result matrices s1 and s2. The sphere radius r is the shortest path length from one point to another in the blood vessel region in the blood vessel segmentation result. The shortest path length is constrained by the geodesic distance rather than the Euclidean distance constraint. Due to the high density of blood vessels, it is easy to "break through" one blood vessel from another during the iterative growth process using the Euclidean distance constraint, resulting in erroneous growth. By iteratively growing in a way that simulates the path diffusion process with the convolution kernel under the geometric constraints of the blood vessel segmentation results and the geodesic distance, the entire blood vessel network can be gradually "explored" while maintaining the coherence of the blood vessel structure.

[0039] In this embodiment, the convolution kernel is a kernel function whose weight decreases as the radius r of the sphere increases, and has a "bell-shaped" characteristic that changes with distance. It can be 2D or 3D, and is suitable for two-dimensional or three-dimensional blood vessel segmentation results / matrices s1, s2, respectively.

[0040] (4) Repeat step (3) and output each point p and its corresponding B V (p,r),B s1 (p,r),B s2 (p, r), perform statistical analysis and calculate the B corresponding to each point p V (p,r),B s1 (p,r),B s2 The modulus of (p,r) || B V (p,r)||、||B s1 (p,r)||、||B s2 (p,r)||, and count the different values ​​of ||B V (p,r)||、||B s1 (p,r)||、||B s2 (p,r)||The number of points p corresponding to ||B V (p,r)|| is the blood vessel size at the corresponding point p, ||B s1 (p,r)||、||B s2 (p,r)|| is the difference between the two / sub-vascular segmentation result matrices s1 and s2 at point p. Output the above statistical results and map them back to the original data and the two / sub-vascular segmentation result matrices s1 and s2, and output the visualization results.

[0041] The visualization results can be divided into two levels. The first level: based on ||B V (p,r)|| (arranged in ascending or descending order) and the corresponding number of points p are used as the horizontal and vertical coordinates to draw a plot. This sub-graph reflects the number of blood vessels of different sizes in the union of the two / times of the tested blood vessel segmentation results V=s1∪s2, that is, the proportion of blood vessels of different sizes in the blood vessel bundle is obtained; with ||B V The number of points p after weighting (p,r)|| and the corresponding given weight factor is used as the horizontal and vertical coordinates to plot the difference between the two / times blood vessel segmentation result matrices with respect to the blood vessel size. The weight factor here is ||B s1 (p,r)||、||B s2 The absolute value of the difference between (p,r)|| is divided by the corresponding ||B V (p,r)||, that is, |(||B s1 (p,r)||-||B s2 (p,r)||)| / ||B V(p,r)||, this weight factor describes the difference between the two / sub-segmentation results relative to the local blood vessel size where point p is located.

[0042] The second level: according to the coordinates of each point p, the different parts of the two / sub-vascular segmentation result matrices s1 and s2 are traced and marked, so as to analyze the difference between the two / sub-segmentation results more intuitively, and finally visualize them in a two-dimensional image or a three-dimensional model.

[0043] Combine the following Figure 2 For a more intuitive explanation, Figure 2 Each sub-image in the figure adopts a two-dimensional cross-sectional schematic diagram. Please refer to sub-image a. For the original data and two blood vessel segmentation results s1 and s2 of the same blood vessel data set (red is the original data, green is the blood vessel segmentation result s1, and blue is the blood vessel segmentation result s2), sub-image a is partially enlarged to obtain sub-image b. It can be seen that the difference between the two blood vessel segmentation results is reflected in the incomplete matching of the segmentation result space. The following sets are defined and calculated: the union of s1 and s2 V = s1∪s2 (the blue part of sub-image c), the intersection s interest =s1∩s2 (gray part of sub-graph d) and symmetric difference Δs=V∩(s interest )^c=(s1∪s2)\(s1∩s2) (the yellow part of sub-figure e), under the geometric constraints of V=s1∪s2 and geodesic distance, each point p of the symmetric difference Δs is traversed in the form of convolution iteration, and the corresponding spherical set B(p,r) and the intersection B of the spherical set and the union V=s1∪s2 of the two blood vessel segmentation results are calculated. V (p,r), then ||B V (p,r)|| depicts the size of the blood vessels (gray area in sub-figure f), and sub-figure g provides a more intuitive ||B V (p,r)|| is an example of describing the size of blood vessels. The corresponding ||B is obtained at different points p1 and p2. V (p1,r)||、||B V (p2,r)|| characterizes the different vessel sizes at these two points.

[0044] This embodiment illustrates a consistency analysis method for two / sub-vessel segmentation result matrices. For more than two / sub-vessel segmentation result matrices, by repeating the above steps, the statistical results and visualization results of any two / sub-vessel segmentation result matrices in the multiple / sub-vessel segmentation result matrices can be obtained.

[0045] Example 2

[0046] The following describes a method for evaluating the consistency of multiple / sub-blood vessel segmentation results provided in Example 1 with reference to a specific example.

[0047] In this embodiment, the original data of the blood vessel data set is a 3D magnetic resonance liver vascular imaging data. Figure 3 As shown, sub-image a is the original data, using a two-dimensional screenshot to show a local portion of the blood vessel segmentation results s1 and s2 superimposed on a certain layer of the original image; sub-image b is the blood vessel segmentation result s1, and sub-image c is the blood vessel segmentation result s2, both of which are displayed in three dimensions, showing the 3D blood vessel structures of the two result matrices s1 and s2 manually annotated by two independent individuals. In this embodiment, the difference between the two result matrices s1 and s2 is intentional to reflect the recognition performance of the method for blood vessel segments with obvious segmentation differences.

[0048] Set the sphere radius r = 10 voxels, and obtain the statistical results and visualization results through the method flow of Example 1. Figure 3 The subgraph d of ||B V (p,r)|| is the horizontal axis (put all ||B V (p,r)|| in ascending order), and two visualization results are shown at the same time, one is to take different values ​​of ||B V (p,r)|| The number of time points p is the vertical coordinate. At this time, the height of each bar reflects the proportion of blood vessels of different thicknesses in the vascular tree, that is, the dark histogram in sub-graph d. From this, we can know that small blood vessels account for the largest proportion in the vascular tree (dark red), while medium-sized blood vessels (dark yellow) and large-sized blood vessels (dark green) account for a smaller proportion. The so-called division of large, medium, and small blood vessels can be arbitrarily defined according to specific needs. The second is to reflect the difference in the size of blood vessel segments of different sizes in the two segmentation results. At this time, the vertical axis is the number of points p in a certain interval multiplied by |(||B s1 (p,r)||-||B s2 (p,r)||)| / ||B V (p,r)||, which is the light-colored histogram in the figure. In this embodiment, the error of the small blood vessel part is the largest, which is manifested by the highest light red bar, which corresponds to the fact that s1 in this embodiment obviously segmented more small blood vessels than s2. When such a significantly high value appears, it indicates that there is a significant difference in the two segmentation results of the blood vessel segment in this size range. The height of the light yellow and light green columns is almost close to 0, which indicates that the difference between the large and medium-sized blood vessel segments in the two segmentation results is very small.

[0049] The difference between the two segmentation results is traced back and marked to the corresponding s1, s2 and original data to obtain an intuitive display of the difference between the two segmentation results. Figure 3As shown in sub-figure e of , a two-dimensional screenshot is used to show the difference between the two segmentation result matrices superimposed on a local part of a certain layer of the original image, where the red area represents the difference in the small-sized blood vessels, the yellow area represents the difference in the medium-sized blood vessels, and the green area represents the difference in the large-sized blood vessels. Figure 3 As shown in sub-figure f, the difference distribution of two results / matrices s1 and s2 manually annotated by two independent individuals is displayed in three dimensions. The red area represents the difference of small-sized blood vessels, the yellow area represents the difference of medium-sized blood vessels, and the green area represents the difference of large-sized blood vessels. From sub-figures e and f, it can be seen that the difference between small-sized blood vessels is the most obvious, which is consistent with Figure 3 The quantitative data in sub-figure d reflect that the result that the error in the small blood vessel part in this embodiment is the largest is consistent.

[0050] Example 3

[0051] This embodiment provides a system for evaluating the consistency of multiple / sub-vessel segmentation results corresponding to the method for evaluating the consistency of multiple / sub-vessel segmentation results described in Embodiment 1, including:

[0052] Data input module: used to read the original data of the same blood vessel data set and multiple / sub-blood vessel segmentation result matrices, and select any two / sub-blood vessel segmentation result matrices s1 and s2 from the multiple / sub-blood vessel segmentation result matrices;

[0053] Matrix processing module: used to calculate the union V, intersection S and symmetric difference Δs of s1 and s2;

[0054] Convolution iteration module: select any point p∈Δs, and iterate the growth by simulating the path diffusion process with the convolution kernel. Under the geometric constraints of the two / times blood vessel segmentation result matrices s1 and s2 and the given geodesic distance r, calculate the spherical set B, the center of which is point p and the radius of which is the geodesic distance r. Calculate the intersection B of the spherical set B and the union V of s1 and s2 V , and calculate B V The intersection B with s1 and s2 s1 , B s2 ; Statistical analysis module: used to repeatedly run the convolution iteration module and output each point p and its corresponding B V , B s1 , B s2 , perform statistical analysis to obtain the proportion of blood vessels of different sizes and the statistical results of the difference between the two / sub-blood vessel segmentation result matrices s1 and s2, output the statistical results and map them back to the original data and the two / sub-blood vessel segmentation result matrices s1 and s2;

[0055] Visualization output module: used to repeatedly run the above submodules to obtain the statistical results of any two / times of the multiple / times blood vessel segmentation result matrices, and output the visualization processing results for the statistical results of any two / times of the blood vessel segmentation result matrices.

Claims

1. A method for evaluating the consistency of multiple / sub-vessel segmentation results, It is characterized in that The following steps are involved: (1) Read the original data and multiple / sub-vascular segmentation result matrices of the same vascular data set, and select two / sub-vascular segmentation result matrices s from the multiple / sub-vascular segmentation result matrices 1 、s 2 ; (2) Calculate s 1 、s 2 The union V, intersection S and symmetric difference Δs of; (3) Select any point p∈Δs and iteratively grow it by simulating the path diffusion process with the convolution kernel. In the two / times blood vessel segmentation result matrix s 1 、s 2 Under the geometric constraints of the given geodesic distance r, the spherical set B is calculated. The center of the spherical set B is point p, and the radius of the sphere is the geodesic distance r. The spherical set B and s are calculated. 1 、s 2 The intersection B of the union V of V , and calculate B V With s 1 、s 2 The intersection of s1 , B s2 ; (4) Repeat step (3) and output each point p and its corresponding B V , B s1 , B s2 , perform statistical analysis to obtain the proportion of blood vessels of different sizes and the two / second blood vessel segmentation result matrix s 1 、s 2 The statistical results of the difference parts are output and mapped back to the original data and the two / times blood vessel segmentation result matrix s. 1 、s 2 middle; (5) Repeat the above steps to obtain statistical results of any two blood vessel segmentation result matrices among the multiple blood vessel segmentation result matrices, and output the visualization processing results for the statistical results of any two blood vessel segmentation result matrices.

2. The method for evaluating the consistency of multiple / sub-vessel segmentation results according to claim 1, Features: The blood vessel segmentation result matrix is ​​a two-dimensional or three-dimensional binary image matrix, in which element 1 indicates that the voxel point belongs to the blood vessel region, and element 0 indicates that the voxel point does not belong to the blood vessel region.

3. The method for evaluating the consistency of multiple / sub-vessel segmentation results according to claim 1, Features: In the step (3), iterative growth is performed in a manner of simulating the path diffusion process by using a convolution kernel, including: applying the convolution kernel to the union V of the vascular segmentation result matrix by iteratively performing a two-dimensional or three-dimensional convolution operation, and in each iteration step, using the convolution kernel as a weight filter to simulate the diffusion of the vascular area, and the convolution iteration stops until the set sphere radius r is reached.

4. The method for evaluating the consistency of multiple / sub-vessel segmentation results according to claim 1, Features: The convolution kernel is a kernel function whose weight decreases as the radius r of the sphere increases.

5. The method for evaluating consistency of multiple / sub-vessel segmentation results according to claim 1, Features: In the step (3), the geometric constraints of the two / sub-vascular segmentation result matrices and the geodesic distance include: the growable region of the convolution kernel iteration is limited to the two / sub-vascular segmentation result matrices s 1 、s 2 The radius r of the sphere is the shortest path length from one point to another in the vascular area in the vascular segmentation result, and the shortest path length is constrained according to the geodesic distance.

6. The method for evaluating the consistency of multiple / sub-vessel segmentation results according to claim 1, Features: In step (4), the statistical analysis includes: calculating the set B corresponding to each point p V , B s1 , B s2 , and count the different values ​​of B V , B s1 , B s2 The number of points p corresponding to the modulus of the set B V The modulus is the size of the blood vessel at the corresponding point p, and the set B s1 , B s2 The modulus is the two / sub-vessel segmentation result matrix s at point p 1 、s 2 The differences between them.

7. The method for evaluating consistency of multiple / sub-vessel segmentation results according to claim 6, Features: The visualization process includes: taking a set B V The modulus and the corresponding number of points p are used as the horizontal and vertical coordinates to draw the proportion of blood vessels of different sizes; V The number of points p weighted by the modulus and the corresponding given weight factor is used as the horizontal and vertical coordinates to plot the difference between the two / times blood vessel segmentation result matrices with respect to the distribution of blood vessel size.

8. The method for evaluating the consistency of multiple / sub-vessel segmentation results according to claim 7, Features: The weight factor is B s1 , B s2 The absolute value of the difference in modulus divided by the corresponding B V The value obtained modulo .

9. The method for evaluating consistency of multiple / sub-vessel segmentation results according to claim 6, Features: The visualization processing includes: tracing and marking the parts of the two / sub-blood vessel segmentation result matrices corresponding to each point p where there are differences according to the coordinates of the points, and visually presenting them in a two-dimensional image or a three-dimensional model.

10. A system for evaluating the consistency of multiple / sub-vessel segmentation results. It is characterized in that include: Data input module: used to read the original data of the same vascular data set and multiple / sub-vascular segmentation result matrices, and select two / sub-vascular segmentation result matrices from multiple / sub-vascular segmentation result matrices 1 、s 2 ; Matrix processing module: used to calculate s 1 、s 2 The union V, intersection S and symmetric difference Δs of; Convolution iteration module: select any point p∈Δs, and iterate the growth by simulating the path diffusion process with the convolution kernel, and then generate the segmentation matrix s of two / times blood vessels. 1 、s 2 Under the geometric constraints of the given geodesic distance r, the spherical set B is calculated. The center of the spherical set B is point p, and the radius of the sphere is the geodesic distance r. The spherical set B and s are calculated. 1 、s 2 The intersection B of the union V of V , and calculate B V With s 1 、s 2 The intersection of s1 , B s2 ; Statistical analysis module: used to repeatedly run the convolution iteration module and output each point p and its corresponding B V , B s1 , B s2 , perform statistical analysis to obtain the proportion of blood vessels of different sizes and the two / second blood vessel segmentation result matrix s 1 、s 2 The statistical results of the difference parts are output and mapped back to the original data and the two / times blood vessel segmentation result matrix s. 1 、s 2 middle; Visualization output module: used to repeatedly run the above submodules to obtain the statistical results of any two / times of the multiple / times blood vessel segmentation result matrices, and output the visualization processing results for the statistical results of any two / times of the blood vessel segmentation result matrices.

Images