Ultrasonic image-based bladder cancer contouring method and system
By employing multi-perspective analysis and adaptive adjustment methods, the problem of inaccurate bladder cancer contour delineation caused by respiratory movements and changes in bladder fullness was solved, thus improving the accuracy and stability of the delineation.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- THE PEOPLES HOSPITAL SHAANXI PROV
- Filing Date
- 2026-05-19
- Publication Date
- 2026-07-10
Smart Images
Figure CN122368036A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of image processing technology, specifically to a method and system for outlining bladder cancer based on ultrasound images. Background Technology
[0002] This invention relates to a method and system for bladder cancer contouring based on ultrasound imaging, which involves the process of detecting, segmenting, and contouring bladder cancer using ultrasound imaging technology. Bladder cancer is a common urinary system tumor, and early diagnosis is crucial for patient survival. Ultrasound imaging, as a non-invasive, real-time, safe, and economical examination method, is widely used in the detection and monitoring of bladder cancer. The ultrasound-based bladder cancer contouring method aims to provide physicians with more accurate information by precisely delineating the boundaries of the cancerous area, thus assisting in diagnostic and treatment decisions.
[0003] Currently, bladder cancer delineation mainly relies on dynamic contrast-enhanced ultrasound combined with three-dimensional reconstruction technology. This involves analyzing tumor angiogenesis patterns and differences in bladder wall layer structure to determine invasion depth and delineate lesion boundaries. However, organ displacement due to respiratory movements and tissue deformation caused by dynamic changes in bladder fullness often blurs the edges of tumor lesions due to tissue folding, leading to contour drift errors in lesion delineation. This is particularly true for muscle-invasive tumors, increasing the risk of false positives in bladder cancer lesion delineation during three-dimensional reconstruction, resulting in low accuracy of bladder cancer contour delineation results. Summary of the Invention
[0004] To address the issue of low accuracy in existing methods for outlining bladder cancer contours, the present invention aims to provide a method and system for outlining bladder cancer contours based on ultrasound imaging. The specific technical solution adopted is as follows: In a first aspect, the present invention provides a method for contour delineating bladder cancer based on ultrasound imaging, the method comprising the following steps: Acquire bladder ultrasound and cystography images of the target person from different perspectives within the current time period; Based on the morphological distribution, length distribution, and grayscale distribution of blood vessels in each region of each frame of bladder ultrasound image under each viewpoint, the muscle layer invasion depth factor is obtained; based on the positional changes of pixels on the same edge line in adjacent frames of bladder ultrasound image under the same viewpoint, the boundary confidence of each region in bladder ultrasound image under each viewpoint is determined; based on the muscle layer invasion depth factor and the boundary confidence of each region in cystography image under each viewpoint, the deviation error factor of each region in each viewpoint is obtained. Based on the changes in bladder cavity volume, the distance between the tumor region and the detrusor muscle bundle texture region, and the consistency between the overall displacement direction of the tumor and the main direction of the local muscle bundle in each frame of bladder ultrasound images from each viewpoint, a traction error sensitivity factor is obtained. By combining the boundary confidence of each region in bladder ultrasound images from each viewpoint, the difference in boundary confidence between each viewpoint and the same region in bladder ultrasound images from other viewpoints, and the traction error sensitivity factor, the cumulative value of registration error for each region in each viewpoint is obtained. By combining the correlation between the deviation error factor and the cumulative value of the registration error of the same area from different perspectives of the target personnel within the current time period, and the sensitivity factor of the traction error, the delineation error coefficient of each area under each perspective is obtained. The initial edge preservation parameters were adjusted using the delineation error coefficient to obtain the bladder cancer contour delineation result.
[0005] Preferably, determining the boundary confidence of each region in the bladder ultrasound images from each viewpoint based on the positional changes of pixels on the same edge line within each region in adjacent frames of bladder ultrasound images from the same viewpoint includes: Extract the spatial coordinates of the center point of each edge line in adjacent frames of bladder ultrasound images under the candidate viewpoint, and calculate the Euclidean distance between the center points of any two edge lines in adjacent frames of bladder ultrasound images; the edge line with the smallest Euclidean distance between the center points is taken as the same edge line. The mean Euclidean distance between all pixels on the same edge line in each region of each frame of bladder ultrasound image under the candidate view and the center point of the region is recorded as the first feature distance of each edge line of each region in each frame of bladder ultrasound image under the candidate view. The ratio between the first feature distance of each edge line of each region in each frame of bladder ultrasound image under the candidate view and the average first feature distance of the same edge line of the same region in all frames of bladder ultrasound image under the candidate view is used as the confidence factor of each edge line of each region in each frame of bladder ultrasound image under the candidate view; the average value of the confidence factors of all edge lines of each region in each frame of bladder ultrasound image under the candidate view is determined as the boundary confidence of the corresponding region. The candidate viewpoint can be any viewpoint.
[0006] Preferably, the step of obtaining the deviation error factor for each region under each viewpoint based on the muscle layer invasion depth factor and the boundary confidence in each region of the cystography image from each viewpoint includes: The least squares method combined with polynomial fitting is used to obtain the fitted line corresponding to each edge line in the region to be analyzed in the cystography image under the candidate view, and the goodness of fit between the edge line and its corresponding fitted line is obtained; the edge lines in the region to be analyzed in the cystography image under the candidate view represent the vascular structure. Based on the number of inflection points on the edge line of the region to be analyzed in the cystography image under the candidate view and the goodness of fit, the muscle layer invasion depth factor of the region to be analyzed in the cystography image under the candidate view is obtained. The number of inflection points is positively correlated with the muscle layer invasion depth factor, and the goodness of fit is negatively correlated with the muscle layer invasion depth factor. The product of the muscle layer infiltration depth factor and the boundary confidence in the cystography image under the candidate view is determined as the deviation error factor of the region to be analyzed in the cystography image under the candidate view. The region to be analyzed is any area of the target person's bladder.
[0007] Preferably, the step of obtaining the muscle layer invasion depth factor based on the morphological distribution, length distribution, and grayscale distribution of blood vessels in each region of each frame of bladder ultrasound image under each viewpoint includes: The ratio between the actual length of the vessel centerline and the Euclidean distance between the two ends of the vessel is used as the vessel path tortuosity index. Obtain the direction vector of the line connecting two adjacent points along the centerline of the blood vessel, and obtain the standard deviation of the average angle of all such direction vectors; Construct a graph structure using the branch points and endpoints of blood vessels as nodes; calculate the entropy value of the distance between nodes in the graph structure; On bladder ultrasound images, a grayscale profile of pixels is extracted along a direction perpendicular to the center line of the blood vessel. The average gradient amplitude of all pixels on the profile is obtained, and the reciprocal of the average gradient amplitude is used as the blur gradient of the blood vessel wall. By combining the vascular path tortuosity index, the standard deviation, the entropy value, and the vascular wall fuzzy gradient, the muscle layer infiltration depth factor is obtained.
[0008] Preferably, the step of obtaining the cumulative registration error value for each region under each viewpoint by comprehensively considering the boundary confidence of each region in bladder ultrasound images from various perspectives, the difference in boundary confidence between each viewpoint and the same region in bladder ultrasound images from other views, and the traction error sensitivity factor, includes: Based on the difference between the boundary confidence of the region to be analyzed in the bladder ultrasound images under the candidate view and all other viewpoints, as well as the boundary confidence of the region to be analyzed in the bladder ultrasound images under the candidate viewpoint, the three-dimensional error accumulation factor of the region to be analyzed in the bladder ultrasound images under the candidate viewpoint is obtained. By combining the traction error sensitivity factor corresponding to the bladder ultrasound image under the candidate viewpoint and the three-dimensional error accumulation factor, the registration error accumulation value of the region to be analyzed under the candidate viewpoint is obtained.
[0009] Preferably, the step of combining the traction error sensitivity factor corresponding to the bladder ultrasound image under the candidate view and the three-dimensional error accumulation factor to obtain the registration error accumulation value of the region to be analyzed under the candidate view includes: determining the product of the traction error sensitivity factor corresponding to the bladder ultrasound image under the candidate view and the three-dimensional error accumulation factor as the registration error accumulation value of the region to be analyzed under the candidate view.
[0010] Preferably, the traction error sensitivity factor is obtained based on the changes in bladder cavity volume, the distance between the tumor region and the detrusor muscle bundle texture region, and the consistency between the overall displacement direction of the tumor and the main direction of the local muscle bundles in each frame of bladder ultrasound images from each viewpoint. This includes: The maximum anteroposterior diameter, maximum lateral diameter, and maximum vertical diameter of the bladder cavity are obtained in each frame of cystography image under candidate viewpoints. The volume of the bladder cavity is obtained based on the maximum anteroposterior diameter, maximum lateral diameter, and maximum vertical diameter. The change index of the bladder cavity is obtained based on the change in the volume of the bladder cavity in two adjacent frames of bladder ultrasound images. Based on the position of the center point of the tumor region in adjacent frames of cystography images under the candidate view, the displacement vector of the tumor center is obtained; principal component analysis is used to determine the principal orientation vector of the muscle bundles in the local region of each frame of cystography images under the candidate view; the cosine similarity between the average direction of the displacement vector of all the tumor centers under the candidate view and the average direction of the principal orientation vector of all the principal orientation vectors under the candidate view is calculated. The normalized value of the Euclidean distance between the tumor center and the nearest detrusor muscle bundle texture region is denoted as the first normalization result. By combining the change index, the cosine similarity, and the first normalization result, the traction error sensitivity factor is obtained.
[0011] Preferably, the correlation between the deviation error factor and the cumulative value of the registration error of the same area under different viewpoints of the target personnel within the current time period, and the pull error sensitivity factor, are used to obtain the delineation error coefficient of each area under each viewpoint, including: Calculate the Pearson correlation coefficient between the deviation error factor and the cumulative registration error of the region to be analyzed under the candidate viewpoints within the current time period; Calculate the sum of the deviation error factor and the cumulative value of the registration error for the region to be analyzed under the candidate viewpoint at the current moment; The product of the sum and the Pearson correlation coefficient is determined as the delineation error coefficient of the region to be analyzed under the candidate viewpoint.
[0012] Preferably, the step of adjusting the initial edge preservation parameters using the delineation error coefficient to obtain the bladder cancer contour delineation result includes: Calculate the sum of constant 1 and the delineation error coefficient, and multiply the sum by the initial edge preservation parameter as the adjusted edge preservation parameter; The contours of bladder cancer lesions were extracted using the U-Net model and adjusted edge-preserving parameters. The volume reconstruction algorithm based on Marching Cubes obtains the outline of the bladder cancer model by reconstructing grayscale values through interpolation.
[0013] Secondly, the present invention provides a bladder cancer contouring system based on ultrasound imaging, the system comprising: The data acquisition module is used to acquire bladder ultrasound images and cystography images of the target person from different perspectives within the current time period; The first evaluation module is used to obtain the muscle layer invasion depth factor based on the morphological distribution, length distribution, and grayscale distribution of blood vessels in each region of each frame of bladder ultrasound image under each viewpoint; to determine the boundary confidence of each region in bladder ultrasound image under each viewpoint based on the positional changes of pixels on the same edge line in adjacent frames of bladder ultrasound image under the same viewpoint; and to obtain the deviation error factor of each region under each viewpoint based on the muscle layer invasion depth factor and the boundary confidence of each region in cystography image under each viewpoint. The second evaluation module is used to obtain the traction error sensitivity factor based on the volume change of the bladder cavity in each frame of bladder ultrasound images under each viewpoint, the distance between the tumor region and the detrusor muscle bundle texture region, and the consistency between the overall displacement direction of the tumor and the main direction of the local muscle bundle. The module also integrates the boundary confidence of each region in bladder ultrasound images under each viewpoint, the difference between the boundary confidence of the same region in bladder ultrasound images under each viewpoint and other viewpoints, and the traction error sensitivity factor to obtain the cumulative registration error value of each region under each viewpoint. The delineation error coefficient determination module is used to comprehensively consider the correlation between the deviation error factor and the cumulative value of the registration error of the same area under different perspectives of the target personnel in the current time period and the pull error sensitivity factor to obtain the delineation error coefficient of each area under each perspective. The contour drawing module is used to adjust the initial edge preservation parameters using a drawing error coefficient to obtain the contour drawing result of bladder cancer.
[0014] The present invention has at least the following beneficial effects: This invention acquires bladder ultrasound images from different perspectives and sequentially calculates the muscle layer invasion depth factor, boundary confidence, and deviation error factor for each region under each perspective. Simultaneously, it combines changes in bladder volume, the distance between the tumor and the detrusor muscle bundle texture, and the consistency of displacement direction to obtain a traction error sensitivity factor, thereby obtaining the cumulative registration error value for each region under each perspective. By integrating the correlation between the deviation error factor and the cumulative registration error value of the same region under different perspectives, as well as the traction error sensitivity factor, it obtains the delineation error coefficient for each region under each perspective. This coefficient is then used to adaptively adjust the initial edge preservation parameters, ultimately obtaining the bladder cancer contour delineation result. The method provided in this embodiment dynamically analyzes the error sources of tumor boundaries under different physiological states and imaging conditions from multiple perspectives and dimensions, constructing an error compensation and parameter adjustment mechanism. This overcomes the problems of contour blurring and delineation drift caused by respiratory motion, changes in bladder filling, and tissue deformation, improving the accuracy and robustness of bladder cancer contour delineation. Attached Figure Description
[0015] 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.
[0016] Figure 1 This is a flowchart of a method for outlining bladder cancer based on ultrasound images, provided in an embodiment of the present invention. Detailed Implementation
[0017] To further illustrate the technical means and effects adopted by the present invention to achieve the intended purpose, the following detailed description of the bladder cancer contour delineation method and system based on ultrasound imaging proposed in accordance with the present invention is provided in conjunction with the accompanying drawings and preferred embodiments.
[0018] 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.
[0019] The following description, in conjunction with the accompanying drawings, details the specific scheme of the bladder cancer contour delineation method and system based on ultrasound imaging provided by this invention.
[0020] Example of a method for bladder cancer contour delineation based on ultrasound imaging: This embodiment proposes a method for contour delineating bladder cancer based on ultrasound imaging, such as... Figure 1 As shown, the bladder cancer contour delineation method based on ultrasound imaging in this embodiment includes the following steps: Step S1: Obtain bladder ultrasound images and cystography images of the target person from different perspectives within the current time period.
[0021] Before examining the target individual, instruct them to drink 500-800ml of water one hour prior to the examination to maintain a moderately full bladder, avoiding overfilling which could lead to excessive stretching of the bladder wall or shrinkage after emptying. Simultaneously, administer ultrasound contrast agent via the antecubital vein to enhance the contrast between the tumor and surrounding tissues and obtain dynamic blood flow information. Use a multi-frequency convex array ultrasound probe (2-5MHz), adjusting the center frequency to 3.5MHz to balance penetration depth and resolution. Set the dynamic range to 60-70dB, optimizing the gain until the posterior bladder wall is clearly visible. During the examination, the target individual should be in a supine position. Use a multi-frequency convex array ultrasound probe (2-5MHz), adjusting the center frequency to 3.5MHz to balance penetration depth. Depth and resolution are set with a dynamic range of 60-70dB, and gain optimized until the posterior bladder wall is clearly visible. After applying coupling gel, the probe is placed above the pubic symphysis in the lower abdomen, with moderate pressure applied to push away intestinal gas interference and avoid excessive compression that could deform the bladder. The probe is moved at a constant speed along the sagittal and transverse planes, covering the entire bladder region (from top to neck), continuously acquiring bladder ultrasound images from different angles during the current time period. The module has a built-in image quality assessment unit that automatically detects severe blurring, artifacts, or interruptions in bladder wall continuity, prompting the operator to rescan. Cystography images from each angle are acquired simultaneously, and the acquired bladder ultrasound and cystography images are stored. The current time period is the set of all historical moments with a time interval less than or equal to the preset duration, plus the current moment. The preset duration is set by the implementer according to specific circumstances, and will not be elaborated further here.
[0022] Thus, this embodiment has acquired multiple frames of bladder ultrasound images and multiple frames of cystography images from each viewpoint of the target person within the current time period. There is a one-to-one correspondence between the bladder ultrasound images and cystography images acquired in this embodiment; that is, at each acquisition time, from the same viewpoint, one bladder ultrasound image and one cystography image of equal size were acquired. It should be noted that all data acquired in this embodiment was obtained with full authorization.
[0023] Step S2: Based on the morphological distribution, length distribution, and grayscale distribution of blood vessels in each region of each frame of bladder ultrasound image under each viewpoint, obtain the muscle layer invasion depth factor; based on the positional changes of pixels on the same edge line in adjacent frames of bladder ultrasound image under the same viewpoint, determine the boundary confidence of each region in bladder ultrasound image under each viewpoint; based on the muscle layer invasion depth factor and the boundary confidence of each region in cystography image under each viewpoint, obtain the deviation error factor of each region under each viewpoint.
[0024] Next, this embodiment first uses the U-Net model to pre-segment the bladder cancer region, optimizing the computational load in subsequent processing. Specifically, historical medical ultrasound image data of different bladder cancer patients are obtained from an image database, and the bladder cancer tumor region in each image is manually identified. The labeled medical ultrasound image data is used as the training dataset for the U-Net model. For each ultrasound image in the bladder cancer image database, the U-Net image segmentation model is used to initially segment the tumor lesion, automatically extracting the tumor contour pre-segmentation information, that is, dividing the ultrasound image into one or more regions, which will not be elaborated further here. Since there is a one-to-one correspondence between the bladder ultrasound images and cystography images acquired in this embodiment, the cystography images can also be segmented based on the sub-region segmentation results in the ultrasound images.
[0025] Due to the real-time nature of ultrasound image acquisition, the morphology of tumors within the patient's bladder undergoes slight changes with time, body position, and bladder fullness. The core of contour correction lies in achieving precise and stable tracking of bladder cancer lesion boundaries within a complex and ever-changing anatomical environment. Therefore, to improve the accuracy of contour delineation results, it is necessary to construct a technical chain of dynamic lesion contour compensation and intelligent edge decision-making, embedding a channel attention window to achieve precise dynamic boundary correction of U-Net pre-segmentation results.
[0026] Considering that muscle-invasive bladder cancer has blurred boundaries due to its interweaving with the bladder wall muscle layer and adipose tissue, and is easily affected by factors such as respiratory movements and changes in bladder filling, thus increasing delineation errors, it is necessary to classify the lesion type beforehand when quantifying deviations in the bladder cancer tumor lesion outline to ensure the accuracy of the quantification results. Muscle-invasive bladder cancer invades the muscle layer of the bladder wall, resulting in tortuous blood vessels and arteriovenous fistulas in the region, appearing "worm-like," while non-muscle-invasive bladder cancer has regular blood vessel branching, appearing "tree-like."
[0027] The edge lines of each region in each cystography and bladder ultrasound image were extracted using the Canny edge detection algorithm. The Canny edge detection algorithm is existing technology and will not be elaborated upon further here.
[0028] Current methods for quantifying the degree of muscle layer invasion based on single vascular morphological features (such as goodness of fit and number of inflection points) can preliminarily distinguish the degree of vascular tortuosity, but they fail to establish a clear quantitative relationship between vascular morphological abnormalities and tumor cell muscle layer invasion behavior; at the same time, they do not consider the dynamic evolution and spatial heterogeneity of vascular morphology, making it difficult to accurately characterize the continuity and local differences of the invasion process.
[0029] This embodiment will now be described using one perspective as an example. Other perspectives can be processed using the methods provided in this embodiment.
[0030] Specifically, any viewpoint is designated as a candidate viewpoint.
[0031] For ultrasound contrast imaging from candidate viewpoints, a Frangi filter based on the Hessian matrix is used to enhance tubular structures, and vascular networks within the pre-segmented tumor region are extracted through adaptive threshold segmentation and skeleton refinement. Tumor cells typically advance in a "finger-like" or "clump-like" manner, compressing and invading the spaces between basal fiber bundles, forcing newly formed blood vessels supplying the tumor to meander through dense normal tissue, resulting in abnormally tortuous and elongated vascular pathways.
[0032] Based on the above characteristics, the ratio between the actual length of the vessel centerline and the Euclidean distance between the two ends of the vessel is used as the vascular path tortuosity index. The vascular path tortuosity index directly quantifies the proportion of the path that the vessel adds to adapt to and bypass the resistance of the tissue microenvironment. A larger vascular path tortuosity index indicates that the tortuosity of the vessel is a direct morphological reflection of the mechanical resistance encountered during the process of propulsive invasion.
[0033] The direction vector of the line connecting two adjacent points along the centerline of each blood vessel segment is obtained, and the standard deviation of the average angle of all direction vectors within the entire blood vessel segment is calculated. A larger standard deviation indicates more frequent and drastic changes in the local direction of blood vessel movement, reflecting the morphological manifestation of cells groping their way forward within the microscopic tissue spaces. The angle of the direction vector is the angle between the direction vector and a preset direction. In this embodiment, the preset direction is horizontal to the right; however, in specific applications, the implementer can set this direction according to the specific circumstances.
[0034] During tumor cell invasion, proteases and other substances are secreted, which directly degrade the extracellular matrix and vascular basement membrane of the bladder wall muscle layer, destroying the original orderly structure, leading to uncontrolled angiogenesis and the formation of a chaotic and disordered vascular network.
[0035] The vascular network is abstracted as a graph structure, with branch points and endpoints of blood vessels as nodes. The entropy value of the distance between nodes in the graph structure is calculated. The larger the entropy value, the more random and disordered the branching pattern of the blood vessels is, indicating that the hierarchical order of the normal vascular tree structure has been lost, directly corresponding to the state of severe destruction of the underlying tissue structure by the tumor. Here, the distance between nodes is the actual distance between corresponding positions (branch points or endpoints) of blood vessels.
[0036] On a bladder ultrasound cross-sectional image, grayscale profiles of pixels are extracted along a direction perpendicular to the vessel centerline. In normal vessels, the interface between the vessel wall and surrounding tissue is clear, with a sharp change in grayscale at the boundary and a large gradient value. However, in destructive and invasive areas, due to accompanying inflammatory cell infiltration and tissue edema, the tumor-vessel interface becomes blurred, and the grayscale change is gradual. The average gradient amplitude of all pixels on the profile is obtained, and the reciprocal of the average gradient amplitude is taken as the vessel wall blur gradient. A larger vessel wall blur gradient indicates a more blurred interface, suggesting more severe tissue damage and inflammatory response in that area.
[0037] Furthermore, by combining the calculated vascular path tortuosity index, standard deviation, entropy value, and vascular wall blur gradient, the muscle layer invasion depth factor is obtained. Specifically, the average value of all vascular path tortuosity indices, the average value of all standard deviations, the average value of all entropy values, and the average value of the vascular wall blur gradient in bladder ultrasound images are calculated. The sum of these four average values is taken as the muscle layer invasion depth factor. The larger the muscle layer invasion depth factor, the more irregular and invasive the boundary is histologically. Therefore, when the image is disturbed by breathing, filling, etc., its inherent ambiguity and instability of the contour are higher, resulting in a larger real-time delineation deviation error.
[0038] Extract the spatial coordinates of the center point of each edge line in adjacent frames of bladder ultrasound images under the candidate viewpoint, and calculate the Euclidean distance between the center points of any two edge lines in adjacent frames of bladder ultrasound images; the edge line with the smallest Euclidean distance between the center points is taken as the same edge line.
[0039] The average Euclidean distance between all pixels on the same edge line in each region of each frame of bladder ultrasound image under the candidate view and the center point of the region is recorded as the first feature distance of each edge line in each region of each frame of bladder ultrasound image under the candidate view. The first feature distance reflects the distance between the edge and the center of the lesion. The larger the value, the closer the edge may be to the cancer boundary. It is used to remove the influence of artifacts such as calcification on the contour delineation of the region. It should be noted that there is a corresponding first feature distance for each edge line in each region of each frame of bladder ultrasound image under the candidate view.
[0040] True bladder cancer tumors have clear and consistent boundaries, and are less affected by the filling state; while false positive boundaries are unstable due to the filling state or external factors, resulting in morphological fluctuations between adjacent frames.
[0041] For any given region, the average first characteristic distance of all edge lines within that region in all frames of bladder ultrasound images under the candidate viewpoint is calculated based on the first characteristic distance of each edge line in that region in each frame of bladder ultrasound images under the candidate viewpoint. This average first characteristic distance is denoted as the mean first characteristic distance. The mean first characteristic distance reflects the structural stability of the edges within that region; the smaller the value, the more stable the edge structure. Therefore, the ratio between the first characteristic distance of each edge line within that region in each frame of bladder ultrasound images under the candidate viewpoint and the mean first characteristic distance of the same edge line in the same region across all frames of bladder ultrasound images under the candidate viewpoint is used as the confidence factor for each edge line in that region in each frame of bladder ultrasound images under the candidate viewpoint. The larger the value, the greater the likelihood that the edge of that region is the true edge of bladder cancer.
[0042] Using the above method, the confidence factor of each region and each edge line in each frame of bladder ultrasound image under the candidate view can be obtained.
[0043] The average confidence factor of all edge lines in each region of each frame of bladder ultrasound image under the candidate viewpoint is determined as the boundary confidence of each region in each frame of bladder ultrasound image under the candidate viewpoint. Using the above method, the boundary confidence of each region in each frame of bladder ultrasound image under the candidate viewpoint can be obtained.
[0044] The following embodiment will still use the candidate viewpoint as an example for explanation. Other viewpoints can be processed using the method provided in this embodiment.
[0045] Any region of the target person's bladder is designated as the region to be analyzed.
[0046] Edge lines within the region to be analyzed in cystography images from candidate viewpoints are used to characterize vascular structures. A least-squares method combined with polynomial fitting is used to obtain the fitted straight line corresponding to each edge line within the region to be analyzed in the cystography images from candidate viewpoints. The goodness of fit between the edge line and its corresponding fitted straight line is obtained. The smaller the goodness of fit value, the greater the deviation between the fitted straight line and the actual tumor edge, and the more consistent it is with the vascular characteristics of invasive bladder cancer.
[0047] Furthermore, based on wavelet transform, inflection points on each edge line are identified, and the number of inflection points on the edge lines within the region to be analyzed in the cystography images from candidate views is counted. A larger number of inflection points indicates a more tortuous vascular structure corresponding to that edge line. Based on the number of inflection points on the edge lines within the region to be analyzed in the cystography images from candidate views and the goodness of fit, a muscle layer invasion depth factor for the region to be analyzed in the cystography images from candidate views is obtained. The number of inflection points is positively correlated with the muscle layer invasion depth factor, while the goodness of fit is negatively correlated with the muscle layer invasion depth factor. In this embodiment, the ratio of the number of inflection points on each edge line within the region to be analyzed in the cystography images from candidate views to the corresponding goodness of fit is used as the muscle layer invasion coefficient for each edge line within the region to be analyzed in the cystography images from candidate views. A larger muscle layer invasion depth coefficient indicates that the vascular structure more closely matches the tortuous, "worm-like" structural characteristics, meaning that the vascular structure is more likely to be located in the area infiltrated by the tumor and the bladder wall muscle layer. It should be noted that the goodness of fit in this embodiment will not be 0.
[0048] The average value of the muscle invasion coefficient of all edge lines in the region to be analyzed in the cystography image under the candidate view is used as the muscle invasion depth factor of the region to be analyzed in the cystography image under the candidate view. The larger the muscle invasion depth factor, the deeper the infiltration between the tumor lesion and the bladder wall muscle layer may be, which means that the contour delineation error caused by external factors is greater.
[0049] The product of the muscle layer infiltration depth factor and the boundary confidence of the region to be analyzed in the cystography image under the candidate view is determined as the deviation error factor of the region to be analyzed in the cystography image under the candidate view. The larger the deviation error factor, the more the segmentation accuracy needs to be optimized.
[0050] Using the above method, the deviation error factor for each region under each viewpoint can be obtained.
[0051] Step S3: Based on the volume change of the bladder cavity in each frame of bladder ultrasound images under each viewpoint, the distance between the tumor region and the detrusor muscle bundle texture region, and the consistency between the overall displacement direction of the tumor and the main direction of the local muscle bundle, the traction error sensitivity factor is obtained; by combining the boundary confidence of each region in bladder ultrasound images under each viewpoint, the difference between the boundary confidence of the same region in bladder ultrasound images under each viewpoint and other viewpoints, and the traction error sensitivity factor, the cumulative value of registration error for each region under each viewpoint is obtained.
[0052] Automatic segmentation of cystography images to extract the bladder cavity region.
[0053] The following explanation will still use the candidate perspective as an example. Other perspectives can be processed using the method provided in this embodiment.
[0054] Specifically, the maximum anteroposterior diameter, maximum lateral diameter, and maximum vertical diameter of the bladder cavity are obtained in each frame of cystography image under candidate viewpoints. The volume of the bladder cavity is obtained based on these three dimensions. Specifically, the product of the maximum anteroposterior diameter, maximum lateral diameter, and maximum vertical diameter is taken as the volume of the bladder cavity. A change index corresponding to the bladder cavity is obtained based on the change in bladder cavity volume between two adjacent frames of bladder ultrasound images. Specifically, the difference between the volume of the bladder cavity in the later frame and the volume of the bladder cavity in the earlier frame is calculated. The ratio of this difference to the acquisition time interval between these two frames of bladder ultrasound images is calculated, and this ratio is normalized. The normalized result is used as the change index corresponding to the bladder cavity, which is the quantitative index of the real-time filling / emptying rate of the bladder. In this embodiment, the maximum-minimum value normalization method is used for data normalization. In other implementations, other existing normalization methods can also be used.
[0055] Based on the position of the center point of the tumor region in adjacent frames of cystography images under candidate viewpoints, the displacement vector of the tumor center is obtained. This displacement vector is obtained by subtracting the coordinates of the center point of the tumor region in the previous frame of cystography images from the coordinates of the center point of the tumor region in the latter frame of cystography images. This vector is used to characterize the instantaneous movement direction and amplitude of the tumor caused by bladder wall traction.
[0056] In each frame of cystography image under candidate viewpoints, directional filtering (such as Gabor filter banks) is used to identify and enhance the linear texture features of detrusor muscle fibers. Principal component analysis is used to determine the principal orientation vector of muscle bundles in local regions of each frame of cystography image under candidate viewpoints. The cosine similarity between the average direction of the displacement vector of all tumor centers under candidate viewpoints and the average direction of all principal orientation vectors under candidate viewpoints is calculated. The closer this value is to 1, the more consistent the tumor displacement direction is with the muscle bundle traction direction. The normalized value of the Euclidean distance between the tumor center and the nearest detrusor muscle bundle texture region is recorded as the first normalization result.
[0057] The deformation and displacement patterns of the bladder caused by changes in bladder filling can be divided into two categories based on its anatomical relationship with the detrusor muscle. The first category is the lateral attachment type, where the tumor base is attached to the side of the detrusor muscle bundle at a relatively far distance. When the bladder filling changes, the tumor is mainly subjected to indirect and non-uniform traction, resulting in significant deformation of some of its edges. The second category is the central infiltration type, where the tumor base is located in the main detrusor muscle bundle. As a mechanical whole, changes in bladder filling will cause the tumor to undergo large-scale translational and homogeneous changes, with the displacement vectors of all its edges being relatively consistent. Furthermore, combining the calculated bladder cavity change index, cosine similarity, and first normalization result, the traction error sensitivity factor is obtained. The formula for calculating the traction error sensitivity factor can be expressed as: in, Indicates the sensitivity factor to pulling error. Indicators representing changes in the bladder cavity. This represents the first normalization result. This represents the cosine similarity.
[0058] The larger the traction error sensitivity factor, the more severe the impact of the dynamic changes of the bladder on the spatial position of the tumor contour. When the two-dimensional section acquired at this moment is integrated into the three-dimensional reconstruction, a larger registration inconsistency error will be introduced.
[0059] Three-dimensional reconstruction of bladder cancer involves acquiring two-dimensional cross-sectional images (such as sagittal, transverse, and coronal planes) at different spatial locations using an ultrasound probe, and then performing data registration and modeling. However, due to the influence of respiratory motion, the displacement differences of tissues at different spatial locations at the same time can lead to inconsistencies in the displacement between ultrasound cross-sectional images, resulting in "ghosting" artifacts during the three-dimensional reconstruction process.
[0060] Traditional U-Net algorithms cannot effectively distinguish between real tumor infiltration and blurred boundaries caused by ghosting, and are prone to misclassifying artifact regions as extensions of tumor infiltration. Since 3D reconstruction requires the fusion of multiple frames of images, the single-frame registration error caused by respiratory motion is amplified after multiple frames are superimposed, causing the tumor outline to gradually deviate from the real boundary, forming a cumulative error over time.
[0061] In deep breathing mode, the diaphragm's movement amplitude increases, and the difference in image displacement of bladder cancer lesions under different spatial perspectives becomes more obvious, exacerbating the "ghosting" effect in the three-dimensional reconstruction process and producing a greater cumulative registration error in the delineation of the tumor lesion outline in the final three-dimensional model.
[0062] Furthermore, during the patient's deep breathing, numerous false-positive boundaries were observed in the ultrasound images of the pre-segmented area of the bladder cancer lesion. The proportion of false-positive boundaries varied significantly across different viewpoints, which could introduce substantial registration errors into the current viewpoint during multi-view image registration. This could exacerbate the propagation of overall registration errors, potentially affecting the accuracy of subsequent image analysis.
[0063] The following embodiment will still use the candidate viewpoint as an example for explanation. Other viewpoints can be processed using the method provided in this embodiment.
[0064] Based on the difference in boundary confidence between the candidate viewpoint and all other viewpoints in bladder ultrasound images of the region to be analyzed, and the boundary confidence of the region to be analyzed in bladder ultrasound images of the candidate viewpoint, the three-dimensional error accumulation factor of the region to be analyzed in the bladder ultrasound images of the candidate viewpoint is obtained. In this embodiment, a specific formula for calculating the three-dimensional error accumulation factor is given. The three-dimensional error accumulation factor of the region to be analyzed in bladder ultrasound images of the candidate viewpoint can be expressed as: in, This represents the three-dimensional error accumulation factor of the region to be analyzed in the bladder ultrasound images from the candidate viewpoint. This indicates the boundary confidence of the region to be analyzed in the bladder ultrasound image from the candidate viewpoint. Indicates the number of viewpoints. Indicates the other first-order views besides the candidate perspectives. Boundary confidence of the region to be analyzed in bladder ultrasound images from various perspectives. Indicates the absolute value sign.
[0065] The greater the difference between the boundary confidence of the region to be analyzed in the bladder ultrasound images of the candidate view and all other viewpoints, and the greater the boundary confidence of the region to be analyzed in the bladder ultrasound images of the candidate view, the greater the difference between the tumor boundary in the candidate view and other viewpoints. This may lead to a larger error during registration, thus affecting the overall registration accuracy. In this case, the three-dimensional error accumulation factor is larger.
[0066] Furthermore, the product of the traction error sensitivity factor corresponding to the bladder ultrasound image under the candidate view and the three-dimensional error accumulation factor of the image of the region to be analyzed in the bladder ultrasound image under the candidate view is determined as the registration error accumulation value of the region to be analyzed under the candidate view. The larger the registration error accumulation value, the more registration error accumulates in the region, and more algorithm optimization is needed to improve the segmentation accuracy.
[0067] Thus, by using the above method, the cumulative registration error value of each region under each viewpoint can be obtained.
[0068] Step S4: By combining the correlation between the deviation error factor and the cumulative value of the registration error of the same area under different perspectives of the target personnel in the current time period and the traction error sensitivity factor, the delineation error coefficient of each area under each perspective is obtained.
[0069] For the region to be analyzed from candidate perspectives, the deviation error factors of the region to be analyzed for the target personnel within the current time period are sorted according to time sequence to obtain a deviation error factor sequence. Simultaneously, the cumulative registration error values of the region to be analyzed for the target personnel within the current time period are sorted according to time sequence to obtain a cumulative registration error value sequence. The Pearson correlation coefficient between the deviation error factor sequence and the cumulative registration error value sequence is calculated and used as the Pearson correlation coefficient between the deviation error factors and the cumulative registration error values of the region to be analyzed from candidate perspectives within the current time period. This coefficient measures the linear correlation between the real-time deviation error of the pre-segmentation contour of bladder cancer tumor lesions and the registration error. The closer the value is to 1, the more positively correlated the real-time deviation error and the registration error are, suggesting that the delineation error of the pre-segmentation contour of bladder cancer tumor lesions is due to the real-time deviation of the pre-segmentation contour exacerbating the "ghosting" effect in the 3D reconstruction process, thus affecting the accuracy and precision of the final segmentation contour, and further impacting the effectiveness of diagnosis and treatment planning.
[0070] Calculate the sum of the deviation error factor and the cumulative value of the registration error of the region to be analyzed under the candidate view at the current time. The product of this sum and the Pearson correlation coefficient is determined as the delineation error coefficient of the region to be analyzed under the candidate view. The larger the delineation error coefficient, the greater the deviation between the pre-segmentation contour and the real tumor boundary, indicating that the segmentation accuracy is low and the accuracy of the pre-segmentation result is poor.
[0071] Using the above method, the delineation error coefficient of each region under each viewpoint can be obtained.
[0072] Step S5: Adjust the initial edge preservation parameters using the delineation error coefficient to obtain the bladder cancer contour delineation result.
[0073] In image segmentation using the U-Net network model, features are first extracted from the input image using a convolutional neural network. The U-Net network structure consists of an encoder and a decoder. The encoder gradually compresses the features of the input image, while the decoder restores the compressed features to the original image size. During this process, U-Net combines shallow features from the encoder with deep features from the decoder through skip connections to preserve more detailed information.
[0074] However, due to organ displacement caused by respiratory movements and tissue deformation caused by dynamic changes in bladder fullness, the edges of tumor lesions are often blurred due to tissue folding, which in turn causes the contour delineation error of the lesion infiltration boundary to drift, resulting in the loss of details in the lesion contour delineation.
[0075] Based on this, the U-Net model can effectively extract detailed information in images by fusing image features from different perspectives and accurately predict pixel-level segmentation results at the boundaries.
[0076] The sum of the constant 1 and the delineation error coefficients is calculated, and the product of this sum and the initial edge preservation parameters is used as the adjusted edge preservation parameters. The delineation error coefficients for each region may differ from different viewpoints; therefore, the adjusted edge preservation parameters for each region may also differ from different viewpoints. It should be noted that the initial edge preservation parameters are obtained during the training of the U-Net model.
[0077] This embodiment adjusts the edge preservation parameters, which is equivalent to optimizing the U-Net model. By combining the optimized U-Net model with an adaptive edge preservation strategy, the model can more accurately depict the boundaries of bladder cancer tumor lesions. During image registration, rigid or non-rigid registration techniques using feature point matching are employed to accurately align image data from different viewpoints.
[0078] Subsequently, a volume reconstruction algorithm based on Marching Cubes was used to reconstruct a continuous three-dimensional tumor volume model by interpolating gray values, and the outline of bladder cancer was extracted.
[0079] Finally, the outline of the bladder cancer was displayed using a 3D visualization tool, and quantitative analysis was performed to extract key parameters such as tumor volume and surface area.
[0080] This embodiment acquires bladder ultrasound images from different perspectives and sequentially calculates the muscle layer invasion depth factor, boundary confidence, and deviation error factor for each region under each perspective. Simultaneously, it combines changes in bladder volume, the distance between the tumor and the detrusor muscle bundle texture, and the consistency of displacement direction to obtain a traction error sensitivity factor, thereby obtaining the cumulative registration error value for each region under each perspective. By comprehensively considering the correlation between the deviation error factor and the cumulative registration error value of the same region under different perspectives, as well as the traction error sensitivity factor, it obtains the delineation error coefficient for each region under each perspective. This coefficient is then used to adaptively adjust the initial edge preservation parameters, ultimately obtaining the bladder cancer contour delineation result. The method provided in this embodiment dynamically analyzes the error sources of tumor boundaries under different physiological states and imaging conditions from multiple perspectives and dimensions, constructs an error compensation and parameter adjustment mechanism, overcomes the problems of contour blurring and delineation drift caused by respiratory motion, changes in bladder filling, and tissue deformation, and improves the accuracy and robustness of bladder cancer contour delineation.
[0081] Example of a bladder cancer contouring system based on ultrasound imaging: An embodiment of the present invention provides a bladder cancer contour delineation system based on ultrasound images, which may include a data acquisition module, a first evaluation module, a second evaluation module, a delineation error coefficient determination module, and a contour delineation module.
[0082] The data acquisition module is used to acquire bladder ultrasound images and cystography images of the target person from different perspectives within the current time period. The first evaluation module is used to obtain the muscle layer invasion depth factor based on the morphological distribution, length distribution, and grayscale distribution of blood vessels in each region of each frame of bladder ultrasound image under each viewpoint; to determine the boundary confidence of each region in bladder ultrasound image under each viewpoint based on the positional changes of pixels on the same edge line in adjacent frames of bladder ultrasound image under the same viewpoint; and to obtain the deviation error factor of each region under each viewpoint based on the muscle layer invasion depth factor and the boundary confidence of each region in cystography image under each viewpoint. The second evaluation module is used to obtain the traction error sensitivity factor based on the volume change of the bladder cavity in each frame of bladder ultrasound images under each viewpoint, the distance between the tumor region and the detrusor muscle bundle texture region, and the consistency between the overall displacement direction of the tumor and the main direction of the local muscle bundle. The module also integrates the boundary confidence of each region in bladder ultrasound images under each viewpoint, the difference between the boundary confidence of the same region in bladder ultrasound images under each viewpoint and other viewpoints, and the traction error sensitivity factor to obtain the cumulative registration error value of each region under each viewpoint. The delineation error coefficient determination module is used to comprehensively consider the correlation between the deviation error factor and the cumulative value of the registration error of the same area under different perspectives of the target personnel in the current time period and the pull error sensitivity factor to obtain the delineation error coefficient of each area under each perspective. The contour drawing module is used to adjust the initial edge preservation parameters using a drawing error coefficient to obtain the contour drawing result of bladder cancer.
[0083] It should be understood that the structural block diagram and modules of the ultrasound-based bladder cancer contouring system can be implemented in various ways. For example, in some embodiments, the system and its modules can be implemented in hardware, software, or a combination of both. The hardware portion can be implemented using dedicated logic; the software portion can be stored in memory and executed by an appropriate instruction execution system, such as a microprocessor or dedicated hardware. Those skilled in the art will understand that the above-described methods and systems can be implemented using computer-executable instructions and / or included in processor control code, for example, on a carrier medium such as a disk, CD, or DVD-ROM, a programmable memory such as read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. The systems and modules of this specification can be implemented not only with hardware circuits such as very large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, or programmable hardware devices such as field-programmable gate arrays, programmable logic devices, etc., but also with software, for example, executed by various types of processors, or with a combination of the above-described hardware circuits and software (e.g., firmware).
[0084] For more details about the above modules, please refer to other parts of this manual; they will not be repeated here.
[0085] In other embodiments, a bladder cancer contouring device based on ultrasound imaging is also provided, including a memory and a processor. The memory stores executable program code, and the processor calls and runs the executable program code from the memory, causing the device to perform the aforementioned bladder cancer contouring method based on ultrasound imaging. Specifically, the device may be a chip, component, or module. The chip may include a connected processor and memory; wherein the memory stores instructions, and when the processor calls and executes the instructions, the chip can perform the bladder cancer contouring method based on ultrasound imaging provided in the above embodiments.
[0086] In other embodiments, a computer program product is also provided, which, when run on a computer, causes the computer to perform the aforementioned steps to implement the bladder cancer contour delineation method based on ultrasound images provided in the above embodiments.
[0087] In other embodiments, a computer-readable storage medium is also provided, which stores computer program code. When the computer program code is run on a computer, it causes the computer to perform the above-described method steps to implement the bladder cancer contour delineation method based on ultrasound images provided in the above embodiments.
[0088] The systems, electronic devices, computer program products, and computer-readable storage media provided are all used to execute the corresponding methods provided above. Therefore, the beneficial effects they can achieve can be referred to the beneficial effects of the corresponding methods provided above, and will not be repeated here.
[0089] It should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A method for contour delineating bladder cancer based on ultrasound imaging, characterized in that, The method includes the following steps: Acquire bladder ultrasound and cystography images of the target person from different perspectives within the current time period; Based on the morphological distribution, length distribution, and grayscale distribution of blood vessels in each region of each frame of bladder ultrasound image under each viewpoint, the muscle layer invasion depth factor is obtained; based on the positional changes of pixels on the same edge line in adjacent frames of bladder ultrasound image under the same viewpoint, the boundary confidence of each region in bladder ultrasound image under each viewpoint is determined; based on the muscle layer invasion depth factor and the boundary confidence of each region in cystography image under each viewpoint, the deviation error factor of each region in each viewpoint is obtained. Based on the changes in bladder cavity volume, the distance between the tumor region and the detrusor muscle bundle texture region, and the consistency between the overall displacement direction of the tumor and the main direction of the local muscle bundle in each frame of bladder ultrasound images from each viewpoint, a traction error sensitivity factor is obtained. By combining the boundary confidence of each region in bladder ultrasound images from each viewpoint, the difference in boundary confidence between each viewpoint and the same region in bladder ultrasound images from other viewpoints, and the traction error sensitivity factor, the cumulative value of registration error for each region in each viewpoint is obtained. By combining the correlation between the deviation error factor and the cumulative value of the registration error of the same area from different perspectives of the target personnel in the current time period, and the sensitivity factor of the traction error, the delineation error coefficient of each area under each perspective is obtained. The initial edge preservation parameters were adjusted using the delineation error coefficient to obtain the bladder cancer contour delineation result.
2. The method for bladder cancer contour delineation based on ultrasound imaging according to claim 1, characterized in that, The step of determining the boundary confidence of each region in the bladder ultrasound images from each viewpoint based on the positional changes of pixels on the same edge line within each region in adjacent frames of bladder ultrasound images under the same viewpoint includes: Extract the spatial coordinates of the center point of each edge line in adjacent frames of bladder ultrasound images under the candidate viewpoint, and calculate the Euclidean distance between the center points of any two edge lines in adjacent frames of bladder ultrasound images; the edge line with the smallest Euclidean distance between the center points is taken as the same edge line. The mean Euclidean distance between all pixels on the same edge line in each region of each frame of bladder ultrasound image under the candidate view and the center point of the region is recorded as the first feature distance of each edge line of each region in each frame of bladder ultrasound image under the candidate view. The ratio between the first feature distance of each edge line of each region in each frame of bladder ultrasound image under the candidate view and the average first feature distance of the same edge line of the same region in all frames of bladder ultrasound image under the candidate view is used as the confidence factor of each edge line of each region in each frame of bladder ultrasound image under the candidate view; the average value of the confidence factors of all edge lines of each region in each frame of bladder ultrasound image under the candidate view is determined as the boundary confidence of the corresponding region. The candidate viewpoint can be any viewpoint.
3. The method for bladder cancer contour delineation based on ultrasound imaging according to claim 2, characterized in that, The deviation error factor for each region under each viewpoint is obtained based on the muscle layer invasion depth factor and the boundary confidence level in each region of the cystography image from each viewpoint, including: The least squares method combined with polynomial fitting is used to obtain the fitted line corresponding to each edge line in the region to be analyzed in the cystography image under the candidate view, and the goodness of fit between the edge line and its corresponding fitted line is obtained; the edge lines in the region to be analyzed in the cystography image under the candidate view represent the vascular structure. Based on the number of inflection points on the edge line of the region to be analyzed in the cystography image under the candidate view and the goodness of fit, the muscle layer invasion depth factor of the region to be analyzed in the cystography image under the candidate view is obtained. The number of inflection points is positively correlated with the muscle layer invasion depth factor, and the goodness of fit is negatively correlated with the muscle layer invasion depth factor. The product of the muscle layer infiltration depth factor and the boundary confidence in the cystography image under the candidate view is determined as the deviation error factor of the region to be analyzed in the cystography image under the candidate view. The region to be analyzed is any area of the target person's bladder.
4. The method for bladder cancer contour delineation based on ultrasound imaging according to claim 3, characterized in that, The muscle layer invasion depth factor is obtained based on the morphological distribution, length distribution, and grayscale distribution of blood vessels in each region of each frame of bladder ultrasound image from each viewpoint, including: The ratio between the actual length of the vessel centerline and the Euclidean distance between the two ends of the vessel is used as the vessel path tortuosity index. Obtain the direction vector of the line connecting two adjacent points along the centerline of the blood vessel, and obtain the standard deviation of the average angle of all such direction vectors; Construct a graph structure using the branch points and endpoints of blood vessels as nodes; calculate the entropy value of the distance between nodes in the graph structure; On bladder ultrasound images, a grayscale profile of pixels is extracted along a direction perpendicular to the center line of the blood vessel. The average gradient amplitude of all pixels on the profile is obtained, and the reciprocal of the average gradient amplitude is used as the blur gradient of the blood vessel wall. By combining the vascular path tortuosity index, the standard deviation, the entropy value, and the vascular wall fuzzy gradient, the muscle layer infiltration depth factor is obtained.
5. The method for bladder cancer contour delineation based on ultrasound imaging according to claim 1, characterized in that, The cumulative registration error value for each region in bladder ultrasound images from various perspectives is obtained by combining the boundary confidence of each region in bladder ultrasound images from different perspectives, the difference in boundary confidence between each perspective and the same region in bladder ultrasound images from other perspectives, and the traction error sensitivity factor. This includes: Based on the difference between the boundary confidence of the region to be analyzed in the bladder ultrasound images under the candidate view and all other viewpoints, as well as the boundary confidence of the region to be analyzed in the bladder ultrasound images under the candidate viewpoint, the three-dimensional error accumulation factor of the region to be analyzed in the bladder ultrasound images under the candidate viewpoint is obtained. By combining the traction error sensitivity factor corresponding to the bladder ultrasound image under the candidate viewpoint and the three-dimensional error accumulation factor, the registration error accumulation value of the region to be analyzed under the candidate viewpoint is obtained.
6. The method for bladder cancer contour delineation based on ultrasound imaging according to claim 5, characterized in that, The step of combining the traction error sensitivity factor corresponding to the bladder ultrasound image under the candidate view and the three-dimensional error accumulation factor to obtain the registration error accumulation value of the region to be analyzed under the candidate view includes: determining the product of the traction error sensitivity factor corresponding to the bladder ultrasound image under the candidate view and the three-dimensional error accumulation factor as the registration error accumulation value of the region to be analyzed under the candidate view.
7. The method for bladder cancer contour delineation based on ultrasound imaging according to claim 1, characterized in that, The traction error sensitivity factor is obtained based on the changes in bladder cavity volume, the distance between the tumor region and the detrusor muscle bundle texture region, and the consistency between the overall displacement direction of the tumor and the main direction of the local muscle bundles in each frame of bladder ultrasound images from each viewpoint. This includes: The maximum anteroposterior diameter, maximum lateral diameter, and maximum vertical diameter of the bladder cavity are obtained in each frame of cystography image under candidate viewpoints. The volume of the bladder cavity is obtained based on the maximum anteroposterior diameter, maximum lateral diameter, and maximum vertical diameter. The change index of the bladder cavity is obtained according to the change in the volume of the bladder cavity in two adjacent frames of bladder ultrasound images. Based on the position of the center point of the tumor region in adjacent frames of cystography images under the candidate view, the displacement vector of the tumor center is obtained; principal component analysis is used to determine the principal orientation vector of the muscle bundles in the local region of each frame of cystography images under the candidate view; the cosine similarity between the average direction of the displacement vector of all the tumor centers under the candidate view and the average direction of the principal orientation vector of all the principal orientation vectors under the candidate view is calculated. The normalized value of the Euclidean distance between the tumor center and the nearest detrusor muscle bundle texture region is denoted as the first normalization result. By combining the change index, the cosine similarity, and the first normalization result, the traction error sensitivity factor is obtained.
8. The method for bladder cancer contour delineation based on ultrasound imaging according to claim 3, characterized in that, The correlation between the deviation error factor and the cumulative value of the registration error of the same area under different perspectives of the target personnel within the current time period, and the pull error sensitivity factor, are used to obtain the delineation error coefficient of each area under each perspective, including: Calculate the Pearson correlation coefficient between the deviation error factor and the cumulative registration error of the region to be analyzed under the candidate viewpoints within the current time period; Calculate the sum of the deviation error factor and the cumulative value of the registration error for the region to be analyzed under the candidate viewpoint at the current moment; The product of the sum and the Pearson correlation coefficient is determined as the delineation error coefficient of the region to be analyzed under the candidate viewpoint.
9. The method for bladder cancer contour delineation based on ultrasound imaging according to claim 1, characterized in that, The process of adjusting the initial edge preservation parameters using a delineation error coefficient to obtain the bladder cancer contour delineation result includes: Calculate the sum of constant 1 and the delineation error coefficient, and multiply the sum by the initial edge preservation parameter as the adjusted edge preservation parameter; The contours of bladder cancer lesions were extracted using the U-Net model and adjusted edge-preserving parameters. The volume reconstruction algorithm based on Marching Cubes obtains the outline of the bladder cancer model by reconstructing grayscale values through interpolation.
10. A bladder cancer contouring system based on ultrasound imaging, the system being used to perform the method of claim 1, characterized in that, The system includes: The data acquisition module is used to acquire bladder ultrasound images and cystography images of the target person from different perspectives within the current time period; The first evaluation module is used to obtain the muscle layer invasion depth factor based on the morphological distribution, length distribution, and grayscale distribution of blood vessels in each region of each frame of bladder ultrasound image under each viewpoint; to determine the boundary confidence of each region in bladder ultrasound image under each viewpoint based on the positional changes of pixels on the same edge line in adjacent frames of bladder ultrasound image under the same viewpoint; and to obtain the deviation error factor of each region under each viewpoint based on the muscle layer invasion depth factor and the boundary confidence of each region in cystography image under each viewpoint. The second evaluation module is used to obtain the traction error sensitivity factor based on the volume change of the bladder cavity in each frame of bladder ultrasound images under each viewpoint, the distance between the tumor region and the detrusor muscle bundle texture region, and the consistency between the overall displacement direction of the tumor and the main direction of the local muscle bundle. The module also integrates the boundary confidence of each region in bladder ultrasound images under each viewpoint, the difference between the boundary confidence of the same region in bladder ultrasound images under each viewpoint and other viewpoints, and the traction error sensitivity factor to obtain the cumulative registration error value of each region under each viewpoint. The delineation error coefficient determination module is used to comprehensively consider the correlation between the deviation error factor and the cumulative value of the registration error of the same area under different perspectives of the target personnel in the current time period and the pull error sensitivity factor to obtain the delineation error coefficient of each area under each perspective. The contour drawing module is used to adjust the initial edge preservation parameters using a drawing error coefficient to obtain the contour drawing result of bladder cancer.